Академический Документы
Профессиональный Документы
Культура Документы
Ioannis A. Kougioumtzoglou
Edoardo Patelli
Siu-Kui Au
Editors
Encyclopedia of
Earthquake
Engineering
1 3Reference
Encyclopedia of Earthquake Engineering
Michael Beer
Ioannis A. Kougioumtzoglou
Edoardo Patelli • Siu-Kui Au
Editors
Encyclopedia of
Earthquake Engineering
v
Acknowledgments
The Editors greatly appreciate the effort made by the section editors, the
contributions of the authors related to work of high standards, as well as
the determination of the reviewers to preserve high quality levels. Further, the
Editors would like to thank Springer for its guidance and leadership on
making this great project a reality.
vii
About the Editors
Edoardo Patelli Institute for Risk & Uncertainty and Centre for Engineer-
ing Sustainability, Liverpool, UK
Edoardo Patelli, Ph.D., obtained his degrees in nuclear engineering and
Ph.D. in radiation science and technology from the Politecnico di Milano,
Italy. He worked at the Institute for Engineering Mechanics, University of
Innsbruck, Austria. He is currently a lecturer in uncertainty and engineering at
the Institute for Risk and Uncertainty at the University of Liverpool, UK, and
the head of computational technology for the EPSRC Centre for Doctoral
Training in Risk and Uncertainty Quantification. Patelli is also an honorary
member of the National Tsing Hua University, Taiwan.
ix
x About the Editors
Siu-Kui Au Institute for Risk & Uncertainty and Centre for Engineering
Dynamics, Liverpool, UK
Siu-Kui Au, Ph.D., is a chartered civil engineer in Hong Kong and obtained
his B.Eng. and M.Phil. from the Hong Kong University of Science and
Technology and Ph.D. from the California Institute of Technology (USA).
He is currently a professor of uncertainty, reliability and risk at the Institute
for Risk and Uncertainty. Before joining Liverpool he has held faculty
positions at the Nanyang Technological University (Singapore) and the City
University of Hong Kong. He was a visiting professor at the Tokyo City
University (Japan) and Wuhan University (China).
Section Editors
xi
xii Section Editors
xv
xvi Contributors
Ilan Kelman Institute for Risk and Disaster Reduction and Institute for
Global Health, University College London, London, England and Norwegian
Institute of International Affairs, Oslo, Norway
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis,
Fig. 1 Conventional structure and the response under the seismic excitation
PED
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis,
Fig. 2 Structure with passive energy dissipation devices (PED)
different from buildings as far as their structural Structural Control (Passive, Active, Semi-
behavior is concerned. The former, in most cases, active, Hybrid), Corresponding Control
are mechanisms, while the latter are construc- Devices, and Practical Applications
tions with a high degree of redundancy. More- Structural control systems fall into four basic
over, in the control of electromechanical devices, categories: passive, active, semi-active, and
the loading is known a priori and is usually har- hybrid control (Soong 1990; Soong and Spencer
monic, while for buildings the earthquake load- 2002). These structural control systems are
ing is unknown and contains a multitude of presented in Figs. 1, 2, 3, 4, and 5 below.
frequencies. Thus, there is a need of selecting
among existing algorithms those that are suitable Passive Control
for the control of buildings and then modifying Passive control devices are devices that do not
them accordingly. Along these lines, a common require power to operate. Examples of passive
feature of all control strategies that are presented devices are base isolation, tuned mass dampers
here is that they are applied to structures excited (TMD), tuned liquid dampers (TLD), metallic
at their base by incoming earthquake waves. yield dampers, viscous fluid dampers, and
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis 3
PED
friction dampers. They dissipate energy using an actuator that moves the mass as needed to
the motion of the structure to produce relative increase the amount of damping and the opera-
movement within the control device or to alter tional frequency range of the device. The first
the dynamic properties of the structure implementation of this control method, and of
(damping, natural frequencies), so that the earth- active control in general, was performed in 1989
quake action will be minimized. Since they do in the Kyobashi Seiwa building, Tokyo, Japan, by
not inject energy into the system, they are stable Kajima Corporation (Kobori et al. 1991). Other
devices. Another advantage of such devices is applications of such devices include the Applause
their low maintenance requirements and the fact Tower (Hankyu Chayamachi building) in Osaka,
that they are unaffected by potential interrup- Japan; the Riverside Sumida Central Tower, in
tions in power supply. These systems are well Tokyo, Japan; the Nanjing Communication
understood and well accepted by the engineering Tower, in Nanjing, China; and the Shin-Jei build-
community as a means for mitigating the effects ing in Taipei, Taiwan (Spencer and Nagarajaiah
of dynamic loadings, such as strong earthquakes 2003).
and high winds. However, such passive devices Active control devices require considerable
have the limitation of not being able to adapt to amount of external power to operate actuators
structural changes and to different earthquake that supply a control force to the structure. Such
excitations. Active, semi-active, and hybrid power may not always be available during seis-
control systems aim at addressing these mic events. Another drawback is that due to their
shortcomings. capacity to add energy to the system, they may
destabilize it. Cost and maintenance of such sys-
Active Control tems is also significantly higher than that of pas-
Active control strategies have been developed in sive devices. On the other hand, they are more
the 1990s, (Soong 1990; Housner et al. 1997); effective than passive devices because of their
they operate by using external energy supplied by ability to adapt to different loading conditions
actuators to impart forces on the structure. The and to control different modes of vibration.
appropriate control action is determined based on Housner et al. (1997) point out the importance
measurements of the structural response. Active of system integration in the design and develop-
control devices include the active tendon system ment of active control systems. Not only is it
(Abdel-Rohman and Leipholz 1983), the active necessary to consider the individual components
bracing system (Reinhorn et al. 1989), and the of a control system, but the system as a whole
active tuned mass damper (Abdel-Rohman and must be understood, including the structure, con-
Leipholz 1983). trol devices, sensors, and computer control sys-
The most famous active control device is the tem. Błachowski (2007) uses model-based
active mass damper (AMD), which uses predictive control to reduce the vibration for
a mass–spring–damper system combined with guyed mast.
4 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
variable-orifice valve
Load
ER/MR Fluid
Load
Controllable Valve
Accumulator
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis,
Fig. 6 Variable-orifice damper and controllable fluid damper
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis,
Fig. 7 Friction semi-active device (a) and semi-active tuned liquid column damper (b)
control devices are electrorheological (ER) and Semi-active tuned mass dampers are similar
magnetorheological (MR) fluids. They consist to TMDs, but with the capability of varying their
of dielectric polarizable (electrorheological, level of damping. They are mainly used for A
ER fluids) or magnetically polarized wind vibration reduction. Another type of
(magnetorheological, MR fluids) particles semi-active TMD is the semi-active variable
suspended in an oil medium. They have the ability stiffness tuned mass damper (SAIVS-TMD),
to reversibly change from viscous fluids to semi- where the stiffness is also controllable. Their
solids with controllable yield strength in millisec- performance is similar to that of AMDs but
onds, with the application of an electrical or with less power consumption.
magnetic field, respectively. This property makes Variable stiffness control devices have the
them ideal for use in controllable dampers. The ability to modify the structure’s stiffness and
advantage of controllable fluid devices is that they therefore its natural frequency to avoid resonant
contain no moving parts other than the piston, conditions. These systems have been studied by
which makes them very reliable and very easy to Kobori et al. (1993). They are installed in bracing
maintain. Moreover, they require low power to systems, and with opening or closing a valve,
operate. they allow the connection between the brace
The discovery of both ER and MR fluids dates and beam, thus changing the building stiffness
back to the late 1940s (Winslow 1947). ER fluid and therefore its frequency, to avoid resonance
dampers have been developed, modeled, and with the incoming earthquake. Their energy oper-
tested for civil engineering applications ation is very low, and they are designed so that in
(Erhrogott and Marsi 1992, 1993; Makris the case of power failure, the connection is auto-
et al. 1995). Work on MR devices have been matically closed and the structure’s stiffness is
done by Spencer et al. (1997), Soong and Spencer increased.
(2002), Spencer and Nagarajaiah (2003), Carlson The first full-scale application of semi-active
et al. (1995), and Dyke et al. (1996c–f). control was the installation of variable stiffness
Other semi-active devices use the force gen- devices on both sides of the Kajima Technical
erated by surface friction to dissipate energy in Research Institute.
a structural system.
Other types of semi-active control devices use Hybrid Control
the dynamic motion of a sloshing fluid or Hybrid control refers either to a combination of
a column of fluid to reduce the response of passive and active systems or, more commonly,
a structure. These liquid dampers are the evolu- to a combination of passive with semi-active
tion of passive tuned sloshing dampers (TSD) and systems, aiming at lowering the forces required
tuned liquid column dampers (TLCD). The TSD by active or semi-active systems, respectively.
uses the liquid in a sloshing tank to add damping One such device is the hybrid mass damper
to the structural system. Similarly, in a TLCD the (HMD), which combines tuned mass dampers
moving mass is a column of liquid, which is with active actuators. The actuator force is only
driven by the vibrations of the structure. These used to increase efficiency and robustness to
passive systems are not very effective for varying changes in structural dynamic characteristics.
loading conditions. To improve their effective- Also in the category of hybrid mass dampers is
ness, a semi-active device based on the passive the active–passive composite tuned mass damper
TSD has been proposed, in which the length of (APTMD) developed by Ohrui et al. (1994) and
the sloshing tank, thus also the properties of the named DUOX. This device is composed of an
device, and therefore its natural frequency can be active mass damper mounted on a tuned mass
changed. Similarly, in semi-active devices based damper, Fig. 8. During structural motion, the
on a TLCD, a variable orifice within the liquid mass of the AMD is driven in the opposite direc-
column is used, or the cross section of the tion of the TMD, therefore magnifying the
sloshing tank is changed. motion of the passive device. When the building
6 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
which controls natural vibration period and hys- €ðtÞ þ CU_ ðtÞ þ KUðtÞ ¼ MEag ðtÞ þ Ef FðtÞ
MU
teretic damping. Active control is accomplished
(1)
by the placement of viscous damping devices at A
the base of the structure to provide velocity- where M, C, K denote the mass, damping, and
dependent damping, which controls the overall stiffness matrixes of the structure, respectively, E,
building displacements. Ef are the location matrix for the earthquake and the
Structures equipped with hybrid mass control forces on the structure, and F(t) is the con-
dampers are the Kansai International Airport in trol force matrix which is applied to the structure.
Osaka, Japan; the Mitsubishi Heavy Industry in In the state space approach, Eq. 1 can be
Yokohama, Japan; and the RIHGA Royal Hotel written as follows:
in Hiroshima, Japan. An interesting device can be
found in the Shinjuku Park Tower consisting of
X_ ðtÞ ¼ AXðtÞ þ Bg ag ðtÞ þ Bf FðtÞ
a V-shaped HMD developed by the (2)
Ishikawajima-Harima Heavy Industries. This YðtÞ ¼ CXðtÞ þ DFðtÞ þ v
device has an easily adjustable fundamental
period. The matrixes X, A, Bg, Bf are given by
An application of hybrid control carried out
U O I
by Lin et al. (2007) included a series of large- X¼ ,A¼ ,
scale experimental tests conducted on a mass U_
2n1 M K M1 C
1
2n2n
equipped with a hybrid controlled base isolation O O
Bg ¼ , Bf ¼
system, consisting of a rolling pendulum system E 2n1 M1 Ef 2n1
(RPS) and a 20-KN magnetorheological (MR) (3)
damper. The 12-t mass and its hybrid isolation
system were subjected to various intensities of The matrixes Υ, C, D, and v are the output
near-fault and far-fault earthquakes on a large states, the output matrix, the feed forward control
shake table. The results showed that a combina- force matrix, and the noise matrix, respectively. In
tion of rolling pendulum system and an adjust- the case where the output variables are the same
able MR damper can provide robust vibration with the states of the system and there is no appli-
control for large civil engineering structures cation of the control forces to the output variables,
that need protection from a wide range of seis- the matrixes C, D are the identity and zero matrix,
mic events. respectively. The noise matrix depends on the
A benchmark smart base-isolated eight-story sensor that is used to measure the response of
building structure has been presented by the system. The above equation can be solved by
Narasimhan et al. (2003), similar to existing any numerical technique for differential equations,
buildings in Los Angeles, California. The base like an explicit Runge–Kutta formula, the
isolation system includes both linear and Dormand–Prince pair, Bogacki–Shampine, and
nonlinear bearings and control devices. Design Adams–Bashforth–Moulton PECE solver.
and implementation of active semi-active and The continuous solution of Eq. 3 is
hybrid systems can also be found in the work of
(Chu et al. 2005; Yi et al. 2001).
ðt
Aðt-tO Þ
xðtÞ ¼ e xO þ eAðt-tÞ BF FðtÞ þ Bg ag ðtÞ
to
Modeling
þBp pðtÞ dt
Continuous and Discrete Control (4)
The equation of motion of a controlled structural
system with n degrees of freedom ui, subjected to Equation 4 is applied assuming that the displace-
an earthquake excitation ag, is given by Eq. 1: ment or velocity and control force are continuous
8 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
A Dt
functions of time. This does not apply to a real Xjþ1 ¼ eAj Dt A1
j e j I BF Kf Xj
control situation, where the control force is calcu- A Dt
lated by observed values of displacement, velocity, þ A1j e j I Bg ag, j þ Bp pj (10)
or acceleration at discrete time intervals. For dis-
crete description, the total time is divided into small The output response is obtained by starting
intervals t0-t1, t1-t2, .... tn-tf, with time interval Dt. from time to, when the response is known, and
An approximate solution of Eq. 3 between two calculating subsequent time points. The critical
points of time tj and tj+1 = tj + Dt is obtained issue here is the determination of time Dt and the
making the following substitutions in the above distribution of stiffness and feedback parameters
continuous solution Eq. 4 and editing the expo- kv and kd.
nential integral:
Linear and Nonlinear Control
t ¼ tjþ1 to ¼ tj In Eq. 1, the change of the material properties
change during loading leads to changes in the
X tj ¼ Xj FðtÞ ¼ F tj ¼ Fj (5) stiffness matrix. Then, the differential equations
pðtÞ ¼ p tj ¼ pj ag ðtÞ ¼ ag tj ¼ ag j become nonlinear:
Xjþ1 ¼ eADt Xj þ A1 eADt I BF Fj þ Bg ag, j þ Bp pj € ðtÞ þ CU_ ðtÞ þ Fs ðUðtÞÞ ¼ MEag ðtÞ þ Ef satFðt td Þ
MU
(6) (11)
The above equation provides an estimate of In this case the nonlinearity originates from
the response at time tj+1 based on values of the the structure and is described as material
response at previous time tj. The first term in the nonlinearity.
right-hand side represents the free oscillation When the equation of motion is formulated in
response (transient state), while the other terms the deformed configuration to account for the
provide the response to load during time Dt structure’s flexibility and associated large dis-
(steady state). The above approach can be applied placements, then equation of motion (1) also
to adaptive systems (term in civil structures: becomes nonlinear; the nonlinearity also origi-
material nonlinearity), wherein the stiffness and nates from the structure, but now it is described
damping constants vary with time, while the mass as geometric nonlinearity.
is maintained constant. In that case matrix A is When the control force, F, is a linear function
altered over time but remains constant during the of the response of the structure, then the above
interval Dt. In this case Eq. 6 becomes equations of motion (1) are linear differential
A Dt equations, and the control is said to be linear.
Xjþ1 ¼ eAj Dt Xj þ A1
j e j I BF Fj þ Bg ag, j þ Bp pj When the control force is not a linear function,
(7) then Eq. 1 are nonlinear differential equations,
and the nonlinearity originates from the control
where force.
Thus, the source of nonlinearity could be
kðtÞ ¼ k tj ¼ kj AðtÞ ¼ A tj ¼ Aj cðtÞ ¼ c tj ¼ cj either from the structure or from the control
(8) force. In Table 1 below, the possible cases are
presented.
The feedback control force in discrete form is
Practical Considerations
Fj ¼ Kf , j Xj , tj t tjþ1
Over the past few decades, various control algo-
(9)
Kf , j ¼ kd tj kv tj rithms and control devices have been developed,
modified, and investigated by various groups of
Replacing into Eq. 7 yields researchers. Several well-established algorithms
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis 9
Actively and Semi-actively Controlled Structures motion as a delay differential equation with sat-
Under Seismic Actions: Modeling and Analysis, uration effects. The expected negative influence
Table 1 Linear and nonlinear cases of analysis
of those parameters should be considered in the A
Structure design process. Thus, there is a need to take them
Control Nonlinear into account in the numerical simulations before
force Linear Material Geometric
the installation of the control system on the real
Linear Linear Nonlinear Nonlinear
building.
structure, structure, structure,
linear linear linear The equation of motion (1) of a controlled
control control control structural system considering time delay and sat-
Nonlinear Linear Nonlinear Nonlinear uration becomes
structure, structure, structure,
nonlinear nonlinear nonlinear € ðtÞ þ CU_ ðtÞ þ KUðtÞ ¼ MEag ðtÞ þ Ef satFðt td Þ
control control control MU
(12)
in control engineering have been introduced to satF is the saturated control force matrix which is
control structures. While many of these structural applied to the structure with time delay td and is
control strategies have been successfully applied, given as
technological problems and challenges relating to
time delay, saturation capacity effects, cost, reli-
F ð t td Þ Fðt td Þ < Fallowable
satFðt td Þ ¼
ance on external power, and mechanical com- Fallowable Fðt td Þ > Fallowable
plexity and reliability during the life of the (13)
structure have delayed their widespread use, and
relatively few actual structures are equipped with Fallowable is the maximum capacity of the control
control systems. device. In the state space approach, Eq. 12 can be
Another practical issue that influences the written as follows:
effectiveness and the reliability of the proposed
X_ ðtÞ ¼ AXðtÞ þ Bg ag ðtÞ þ Bf satFðtd tÞ
control algorithms is the effect of the position of
control forces. The selection of locations of con-
trol forces influences the location matrix, Bf of YðtÞ ¼ CXðtÞ þ DFðtd tÞ þ v
the control force in the differential equation of (14)
motion of controlled structure. Thus, this effect
can be investigated numerically by parametric This equation can be solved by the technique
variation of the location matrix. of delay differential equation, or one can use the
Other practical effects include following transformation
control–structure interaction, actuator dynamics, ð
and digital control implementation. The reliabil- ZðtÞ ¼ XðtÞ þ eAðZþtd Þ Bf Fðt þ ZÞdZ (15)
ity of applied semi-active structural control sys-
tems and practical applications and verification
for active and semi-active vibration control of Then:
buildings in Japan have been studied by
Ikeda (2009). Z_ ðtÞ ¼ AZðtÞ þ Bg ag ðtÞ þ BðAÞFðtÞ
(16)
BðAÞ ¼ eAtd Bf
Time Delay–Saturation Capacity
Two practical issues that influence the effective- Since the entire control process involves mea-
ness and the reliability of the proposed control suring response data, computing control forces
algorithms are time delay and saturation of the through an appropriate algorithm, transmitting
control force. Those parameters come into con- data and signals to actuators, and activating the
sideration by solving the differential equation of actuators to a specified level of force, time delays
10 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
arise and cannot be avoided. The problem of time freedom systems. Casciati et al. (2006) have
delay in the active control of structural systems taken into consideration the time-delay effect
has been investigated from many scientists and solving numerically delayed differential equa-
engineers. The stability of the structure could be tions. All of these studies demonstrate how
lost due to time delay, and two ways of time- important the issue of time delay is in structural
delay compensation can be followed. In the first control and how it may result in a degradation of
the gain matrix is redesigned considering the the control performance and may even drive the
presence of time delay, while in the second controlled structure to become unstable. Most
low-pass filters are used to filter the velocity studies show that time delays influence nega-
measurements from the frequency components tively the control system; therefore, they should
of the high-order modes. In the first case, the be kept small compared to the fundamental
structure could remain unstable when using con- period of vibration of the system and should, if
trol moments as control actions, while in the possible, be eliminated.
second a number of vibration modes can be con- The second important practical problem is the
trolled and compensated for time delay, but the saturation of the control force. Actuator satura-
higher-order modes remain uncontrolled. Time tion occurs when the force which is given by the
delay can be compensated with Pade approxima- control algorithm is larger than its designed peak
tions. The allowable time delay is related with capacity. Failure to account for this nonlinear
natural period and feedback gain. The maximum effect can decrease the efficiency of the control
allowable time delay is decreased with decrease system and possibly drive the structure to become
in natural period of the structure, as well as with unstable. Most control algorithms are linear,
increase in active damping. Under earthquake assuming that there is no limit in the magnitude
excitations, simulation results for the response of the control force. However, maximum capac-
of multi-degree of freedom structures indicated ity of the control devices is limited. Therefore,
that the degradation of the control performance designing controllers to account for the bounded
due to fixed time delay is significant when time nature of the devices is desirable.
delay is close to a critical value. The time-delay The two issues of time delay and saturation of
problem is more serious for structures with the control device are, in most cases, considered
closely spaced vibration modes. and studied separately. However, in the applica-
In optimal control of linear systems, time tion of real control systems, these two issues act
delay is considered at the very beginning of the simultaneously. Pnevmatikos and Gantes (2011)
control design, and no approximations and esti- investigated a combined effect of the nonlinear
mations are made in the control system. Thus, the phenomena of bounded capacity of the actuators
system performance and stability can be and time delay of the system, acting simulta-
guaranteed. Instability in the response might neously during the control process, on the sys-
occur only if a system with time delay is con- tems response. They proposed limits for pair of
trolled by an optimal controller that was designed time delay and saturation capacity that can be
with no consideration of time delay. used in the design process of controlled
For pole assignment algorithm, through varied structures.
location of the controlled poles, the control sys-
tem shows variable performance. However, the Spillover Effect
locations of the controlled pole pairs should be With the discretization of a continuous system to
carefully specified and checked according to the a finite degree of freedom system, some informa-
characteristics of the system. Analytical expres- tion is lost since a real physical system has infi-
sions of limiting values of time delay for single nite natural mode shapes and frequencies, while
degree of freedom systems were derived by the discrete systems contain some of them. Those
Connor (2003); however, such expressions were frequencies that are not included in the discrete
very difficult to obtain for multi-degree of model are called remaining frequencies or
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis 11
residual modes. If flexible structure is modeled, (noncentralized) if the control system is man-
there is a danger that the control based on aged by several computers that take the input
a reduced model is destabilized due to the high from some specific sensors and give the A
remaining modes which are not included in the command output to some actuators. It is thus
model. Balas (1978) defined and studied the possible that a noncentralized system could be
above phenomenon, which is called spillover collocated or that a centralized system could be
effect. Meirovich (1990) investigated observa- non-collocated.
tion spillover (measurements at some points,
without having the total picture of the response)
and stated that this can be really dangerous for the Analysis
controlled structures.
Control Strategies and Algorithms
Controllability: Observability Several well-established algorithms in control
Controllability of system deals with the number engineering have been introduced to control
of control positions and degrees of freedom of structures, such as optimal control, LQR or
the structure. When the number of control posi- LQG, pole assignment, sliding mode control, H2
tions is equal to the number of degrees of free- and H1, fuzzy control, and many others. The
dom of the structure, then full control of the most suitable algorithms for structural applica-
system is achieved. In that case, the building tion and the practical considerations that should
performs a rigid body motion, following the be taken into account are described by Soong
imposed ground motion, without relative dis- (1990) and Casciati et al. (2006).
placements between the floors. It was also
shown that with reduced number of control Optimal Control, LQR, or LQG
forces, positioned at appropriate locations, Research in structural control has focused on
which is a more realistic choice for real build- a variety of control algorithms based on different
ings, the response can be reduced to control design criteria. Some algorithms origi-
a satisfactory level. Observability is associated nate from direct applications of optimal control
with measurement positions in the structure to theory. Some others, however, are specifically
degrees of freedom of the structure. When the proposed for civil engineering structural control
measurement positions are equal to the number applications.
of degrees of freedom of the structure, then full- The control force, F, can be applied directly or
state feedback is achieved. Otherwise, feedback indirectly to the structure. The way in which the
with an observer is performed. control force is calculated is determined from the
control algorithm that is used. If the control force
Collocated and Not Collocated Control/ is calculated by linear state feedback:
Centralized and Decentralized Control
A definition of collocation and centralization is _ U
given by Casciati et al. (2006). A control system F ¼ G1 U G2 U ¼ ½G1 G2 _ ¼ GX
U
is collocated when the force generated by an (17)
actuator at a point of the structure is measured
by a force sensor at the same location, in other G is the gain matrix, which is calculated
words when the actuator and sensor are according to the desired poles of the controlled
connected exactly at the same location. Other- system. Replacing the force F into Eq. 1 or 2, the
wise, control is non-collocated. controlled system can be described by
A control system is centralized if it is managed
by a unique computer that receives the inputs €ðtÞ þ ðEf G2 þ CÞU_ ðtÞ
MU
from all sensors and gives the command output (18)
to all actuators. The system is decentralized þ ðEf G1 þ KÞUðtÞ ¼ MEag ðtÞ
12 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
Wavelet
analysis
In civil structures it is unrealistic to expect that A control algorithm which addresses this dif-
the state vector can be fully measured. The case ficulty is pole assignment (pole placement) algo-
that a few degrees of freedom are measured and rithm. Pole placement algorithms have been
used for the calculation of the control force is studied extensively in the general control litera-
called output control, in contrast to the full-state ture, while its applications in structural control
control where all degrees of freedom of the sys- have been investigated by Martin and soong
tem are measured. Suitable control design tech- (1976), Leonard (1990), Soong (1990), Utku
niques, like observers with linear quadratic (1998), and Preumont (2002).
Gaussian (LQG) control, have been developed A procedure of on-line selection of poles in
for output feedback and random disturbances. such a way that first resonance is avoided and
Many researchers have studied and applied opti- secondly sufficient equivalent damping is
mal control in civil structures (Abdel-Rohman added, based on the specific characteristics of
and Leipholz 1983; Chang and Soong 1980; the incoming dynamic earthquake excitation,
Yang 1975). An evolutionary control of damaged have been proposed by Pnevmatikos and Gantes
systems using a rehabilitative, modified LQR (2010a). This procedure drives the poles to their
algorithm has been proposed by Attard and optimum location and does not need the poles to
Dansby (2008). be predefined and constant during the application
of dynamic loading. Numerical simulations show
Pole Placement Algorithm that sufficient reduction of the response, in terms
It should be note that, if the external excitation is of both displacement and acceleration, can be
ignored or set to zero in the derivation of the achieved for all examined earthquakes with rea-
Riccati equation, the provided control law is not sonable amount of required equivalent control
optimal. In order to include the excitation in the force. The procedure is shown schematically in
Riccati equation, a priori knowledge of the load- Fig. 10. The effectiveness to the response of the
ing history is required. This is generally not structure of the control strategy is shown in
possible for excitations such as earthquakes, Fig. 11. It is shown that both the displacement
wind, or waves, which are common in structural and the acceleration are reduced one order of
engineering applications. magnitude.
14 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
-0.04
-0.06
-0.08
-0.1
0 5 10 15 20 25 30
Time (s)
20
15
10
Acceleration 3rd (m/sec2)
-5
-10
-15
-20
0 5 10 15 20 25 30
Time (s)
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
8 9
H2/H1 Control Algorithm u ð
u <1 1 =
The objective of the H2 or H1 control algorithms u
jjHjj2 ¼ ttrace HðjoÞH ðjoÞdo (27)
is to design a controller K that minimizes the H2 :2p ;
1
or H1 norm of the closed-loop transfer function
matrix, H, from the disturbance to the output More details regarding the use of control H2
vector. By definition, the H2 norm of a stable and LQR methods for civil engineering applica-
transfer function matrix is tions can be found in Zacharenakis et al. (2001).
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis 15
VSD
Choice of type II
(Close valve)
Yes
VSD
Choice of type I No
f1II >fH
(Open valve)
af,e
0.1
0.09
0.08
0.07
0.06
0.05
FFT or 0.04
Wavelet analysis 0.03 ap
0.02
0.01
0
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
fLfH
Actively and Semi-actively Controlled Structures equipped with active variable stiffness systems and
Under Seismic Actions: Modeling and Analysis, subjected to earthquake actions
Fig. 12 The integrated control strategy for buildings
of the incoming signal, the variable stiffness Fuzzy control defines fuzzy rules to determine
device, by connecting or disconnecting the brac- actions to be taken, based on the measured struc-
ing system, alters the stiffness of the structure and ture’s responses. The advantages of fuzzy
consequently changes its dynamic characteristics controller are simple algorithms, suitable for
in real time, avoiding resonance between the con- real-time control, no need for information
trolled structure and the applied signal. Thus, the on structural and vibration characteristics, and
structural safety and serviceability against a robust system in terms of performance and
extreme dynamic excitation are enhanced. The implementation.
general scheme of the proposed algorithm is Fuzzy control uses experience instead of dif-
shown in Fig. 12. ferential equations to determine desirable control
actions. Fuzzy control consists of implementa-
Fuzzy Logic Control tion of rules, in the format of IF. . .THEN
A very useful application of fuzzy logic in civil statements, which relate the input variables to
engineering is fuzzy control. Fuzzy logic, intro- the control action. The process begins by first
duced by Zadeh (1965), has to do with the way defining membership functions to classify the
that the human brain deals with concepts such as inputs and outputs using linguistic terms.
uncertainty, vagueness, and imprecision. Bool- In structural control applications, the inputs are
ean logic determines whether an argument defi- usually a combination of structural responses:
nitely belongs or not to some set, while fuzzy displacement, velocity, and acceleration. The
logic considers the idea of partial truths, and input values are then converted to fuzzy values,
determines the degree of membership of an argu- using the membership functions. This step is
ment to a fuzzy set. referred to as fuzzification. The next step is
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis 17
named decision making and consists of using In the clipped-optimal control algorithm, the
predetermined rules to correlate the fuzzy inputs control forces fc is given by the following
values to fuzzy outputs. Finally, in the last step, expression: A
the fuzzy output values are defuzzified, that is,
they are converted to values that can be used as ym
fc ¼ L1 Kc ðsÞL (31)
control actions. fm
The rules and the membership functions are
important to the effectiveness of the fuzzy con- where L{} is the Laplace transform, Kc(s) is
troller. Several methods are available for the cre- a linear optimal controller obtained from H2 or
ation of such rules, including logical reasoning, LQG strategies because of the stochastic nature
experiments and simulations, and learning from of earthquake ground motions (any other control
examples. Fuzzy control rules can be created law can also be chosen), ym is the measured
based on data obtained by an LQG controller. structural response vector, and fm is the measured
Genetic algorithms can be tuned to the member- control force vector.
ship functions in order to improve the results. Because the force generated in the semi-
Genetic algorithms can also be used to simulta- active device is dependent on the local responses
neously determine the optimal rules and member- of the structural system, the desired optimal con-
ship functions. trol force, fc, cannot always be produced by the
Fuzzy theory can be applied to determine the control device. Only the control voltage, vi, can
desired control force to be applied by the actua- be directly controlled to increase or decrease the
tor. Neural network performance function selec- force produced by the device. Thus, a force feed-
tion can be used in structural control (Casciati back loop is incorporated to induce the semi-
et al. 1993). active device to generate approximately the
Fuzzy logic has been used to vary the mechan- desired optimal control force fc. To clip the
ical properties of the control device. It has been active control law to the semi-active one and to
applied for the reduction of longitudinal bridge generate approximately the desired optimal con-
displacement due to earthquake or traffic loading trol force, the command signal is selected as
with variable dampers or control of structural follows.
vibrations with hybrid systems composed of When the semi-active device is providing
base isolation and semi-active dampers. Fuzzy the desired optimal force fi = fc, the voltage
control is also used to regulate a structure with applied to the damper should remain at the pre-
an MR damper system (Choi Kang-Min sent level. If the magnitude of the force produced
et al. 2004, Zhou et al. 2003). by the device fi is smaller than the magnitude
of the required target force fc and the two
Clip Optimal Control forces have the same sign, the voltage applied
The clip optimal control has two parts: the first to the current driver is increased to the maximum
part consists of designing a control law assuming level so as to increase the force produced
that an ideal device is installed to the building. In by the device to match the desired control
the second part, a controller is designed allowing force. Otherwise, the commanded voltage is set
the semi-active damper to develop the force that to zero.
the control law calculated. In the first part, any The above procedure that regulates the applied
control law can be chosen. voltage has the following mathematical
Clipped-optimal control strategy has been expression:
proposed by Dyke et al. (1996) to control
a single MR damper. The control algorithm was vi ¼ Vmax Hððfci fi Þfi Þ (32)
extended to control multiple MR devices, and the
performance of this algorithm has been experi- where Vmax is the maximum voltage to the current
mentally verified. driver and H( ) is the Heaviside step function
18 Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis
which is a discontinuous function whose value is Since the damping values now are not limited
zero for negative argument and one for positive to this state alone and can take any value within
argument. this state, a continuous version of skyhook con-
In clipped-optimal control algorithm, the trol can be achieved. The choice of damping is
command voltage takes two values, either zero now given as follows:
or the maximum value. In some situations when
the dominant frequencies of the system under z_ b vrel 0 c¼a
control are low, large changes in the forces (36)
z_ b vrel < 0 c ¼ low state
applied to the structure may result in high local
acceleration values. This behavior is dependent
on the time lag in the generation of the control where
voltage. In Yoshida (2003), a modification to the
original clipped-optimal control algorithm is a ¼ maxflow state, minðg z_ b , high stateÞg
proposed to reduce this effect. In the modified (37)
version of the control algorithm, the control
voltage can be any value between 0 and Vmax. In the case of ground hook control, the damper
The control voltage, vci, is determined using is adjusted to its high or low state depending on
a linear relationship between the applied voltage the product of the relative velocity across the
and the maximum force of MR damper. The damper and the absolute velocity of lower mass
modified clipped-optimal control algorithm is attached to that damper. Contrary to skyhook
given as control, this control law is driving the lower
vibration mass to a reduced vibration.
vi ¼ Vci Hððfci fi Þfi Þ (33) The mathematical expression of the ground
hook control is as follows:
where
z_ t vrel 0 c ¼ high state
(38)
mf ci for fci fmax z_ t vrel > 0 c ¼ low state
Vci ¼ (34)
Vmax for fci fmax
The continuous ground hook control strategy
Skyhook/Ground Hook Control is similar to the sky hook control.
On skyhook control, the damper is controlled by
two damping values. The choice between high or
low state is made using the following law: based Summary
on the product of the relative velocity urel across
the damper and the absolute velocity of the sys- Representative control devices and their applica-
tem body mass attached to that damper, tions have been presented. Then, mathematical
a damping value is chosen as follows: formulations and simulation of controlled struc-
tures have been illustrated. Control algorithms
z_ b vrel 0 c ¼ high state suitable for civil structure inspired by the classi-
(35)
z_ b vrel 0 c ¼ low state cal control theory were presented. Optimal con-
trol, LQR, pole assignment, sliding mode control,
The skyhook control is an on–off strategy. H2, fuzzy control, clipped-optimal control, and
When the relative velocity of the damper is pos- skyhook/ground hook control algorithms were
itive, the force of the damper acts to pull on the briefly described. It was shown that with the
system body mass, while when the relative veloc- help of modern technology, it is possible to con-
ity of the damper is negative, the force of the trol civil structures in order to protect them from
damper pushes the body mass. earthquake excitations.
Actively and Semi-actively Controlled Structures Under Seismic Actions: Modeling and Analysis 19
approximation (Sorenson and Alspach 1971). However, an over-fitted model leads to poor pre-
However, this class of methods is often compu- dictions because the model parameters depend
tationally very demanding, especially for online too much on the detail of the data, and conse-
updating. Furthermore, the conditional mean of quently the measurement noise and the model-
the state vector is not available in some cases. ing error play an important role in the data
Another class of methods attempts to assign fitting. Therefore, in order to select a suitable
different weights, which are some heuristic model class for identification purposes, it is nec-
function of the data, to different data points essary to penalize a complicated model albeit
(Durovic and Kovacevic 1999). Methods of the quantification of this penalty is a nontrivial
this class require tuning of the threshold param- task. This was first recognized by H. Jeffreys
eters. However, performance is deteriorated who did pioneering work on the application of
with improper choice of the weights due to the Bayesian methods (Jeffreys 1961). In the pre-
difficulty in the choice of the thresholds. There- sent context, the selected class of models should
fore, this class of methods should be utilized agree closely with the observed behavior of the
with special attention. system but otherwise be as simple as possible.
In this chapter, an outlier-resistant extended In recent years, the Bayesian approach to model
Kalman filter (OR-EKF) is presented for robust class selection has been further developed by
online structural parametric identification using showing that the evidence for each model class
dynamic response data, contaminated with out- provided byte data automatically enforces
liers in addition to Gaussian noise. In this algo- a quantitative expression of principle of model
rithm, a novel outlier detection algorithm is parsimony or of Ockham’s razor (Gull 1988).
embedded into the EKF. It is capable for robust As a result, no ad hoc penalty is needed as in
online estimation of structural parameters using some of the earlier work on model class selec-
outlier-contaminated dynamic response data. tion. Applications in civil engineering include
Instead of definite judgment on the outlierness the damage detection (Lam et al. 2006), selec-
of a data point, the OR-EKF provides its outlier tion of nonlinear hysteretic models (Muto and
probability, which is the extension of the concept Beck 2008), soil compressibility (Yan et al.
in Yuen and Mu (2012). Data points with outlier 2009), ambient effects to structural modal
probability over 0.5 will be regarded as suspi- parameters (Yuen and Kuok 2010b), and seismic
cious data points, and they will be discarded for attenuation relationships (Yuen and Mu 2011).
the identification purpose. In contrast to other In this chapter, a recently developed recursive
existing outlier detection criteria that require algorithm is introduced for Bayesian online
some subjective threshold (e.g., normalized model class selection.
residual larger than 2.5), the outlier probability In section “Structural Parametric Identifica-
threshold of 0.5 is intuitive in the presented tion by Extended Kalman Filter,” online struc-
approach. tural parametric identification using the EKF will
The third issue to be introduced is on the be briefly reviewed. In section “Online Identifi-
online model class selection. The usual approach cation of Noise Parameters,” an online identifica-
in system identification is to find the best/optimal tion algorithm for the noise parameters in the
model in a prescribed class of models, e.g., class EKF is introduced. Then, in section “Outlier-
of shear building models with uncertain inter- Resistant Extended Kalman Filter,” an online
story stiffnesses. This problem is commonly outlier detection algorithm is presented, and it is
referred to as parametric identification. The embedded into the EKF. This algorithm allows
more general problem of model class selection for robust structural identification in the presence
has not been explored as intensively as paramet- of possible outliers. In section “Online Bayesian
ric identification. It is well known that a more Model Class Selection,” a recursive Bayesian
complicated model class often fits the data better model class section method is presented for non-
than one which has fewer adjustable parameters. parametric identification problems.
Advances in Online Structural Identification 23
Ak ¼
@y
yðtÞ¼yk
2 3 20 13 2
3
x_
Nd
0Nd Nd
INd
0N d Nj
d
yðtÞ ¼ 4M1 r x, x;
6 7
_ w 5 þ 4 M1 T 5 f ðtÞ 6
7
¼6 1 @r
1 @r
M1 @r 7
dt 4 M M 5 (8)
0Nj 1 0Nj 1 @x
@ x_
@w
0N N
0N N d
dI
j d j Nj
(3)
To simplify the notation, the rate vector and input and the vector hk is given by
distribution matrix are defined as follows:
2 3 hk ¼ rjyðtÞ¼yk Ak yk (9)
2 3 0N d N f
x_
6 7
s 4M1 r x, x;
_ w 5 ; B 4 M1 T 5 (4) The variable d in Ak in Eq. 8 is chosen as
0Nj 1 0Nj Nf a small positive number to prevent singularity
of the matrix Ak;Ia, and 0a b denote the a a
where 0a b denotes the a b zero matrix. identity matrix and the a b zero matrix,
The structural response is sampled at No DOFs respectively.
with time step Dt, and the measurement at the kth
time step can be represented by the observation State Estimation and Parametric
vector: Identification by Kalman Filter
The augmented state vector y can be updated by
zk ¼ Cd yk þ nk (5) the Kalman filter. First, the linearized state-space
24 Advances in Online Structural Identification
equation in Eq. 7 can be discretized to a difference process (Sorensen and Sacks 1971). Here, it is
equation: employed to track the time-varying structural and
noise parameters. The underlying concept of fad-
ykþ1 ¼ Ad,k yk þ Bd,k f k þ hd, k (10) ing memory filtering is to discount the contribu-
tion of the past data gradually for adaptive
where yk+1 = y((k + 1)Dt), Ad,k = exp(Ak Dt),
identification of the time-varying parameters.
Bd,k ¼ A1 Ad, k I2Nd þN’ B, f k ¼ f ðkDtÞ , and This can be accomplished by assigning an appro-
k
hd, k ¼ A1
k Ad, k I2Nd þN’ hk . priate weighting, which is called the fading factor
With the measurement Dk {z1, z2, . . ., zk} m(>1), to enlarge the estimated covariance
defined in Eq. 5, the predicted state at the (k + 1)th matrix at every time step so that the information
time step (yk + 1|k E[yk + 1|z1, z2, . . ., zk]) can be from the past data will fade out gradually.
estimated by taking the conditional expected In particular, the covariance matrix of the filtered
value of Eq. 10: augmented state in Eq. 14 is modified as follows:
ykþ1jk ¼ Ad, k ykjk þ hd, k (11) Sy, kþ1jkþ1 ¼ m I2N d þNj Gkþ1 Cd Sy, kþ1jk
(16)
where ykjk is the updated state at the kth time step.
In addition, the covariance matrix of yk + 1jk is where m is the widely adopted exponential fading
factor (Sorensen and Sacks 1971)
Sy, kþ1jk ¼ Ad, k Sy, kjk ATd, k þ Bd, k Sf , k BTd, k 2
lnms
(12) m ¼ exp (17)
Nm
where Sf,k is the covariance matrix of the excita-
where m 1. The effect of this exponential
tion at the kth time step.
fading factor is to dilute the information of
With a new data point zk+1, the updated state
a data point with a factor of m2s, in terms of the
yk+1|k+1 and its covariance matrix Sy,k + 1|k + 1 can
contribution to the posterior variance, after
be obtained by using the following Kalman filter
every Nm time steps.
recursive formulae (Kalman and Bucy 1961):
ykþ1jkþ1 ¼ ykþ1jk þ Gkþ1 zkþ1 Cd ykþ1jk Online Identification of Noise
(13) Parameters
In this section, a Bayesian probabilistic approach
Sy, kþ1jkþ1 ¼ I2Nd þNj Gkþ1 Cd Sy, kþ1jk
is presented for online estimation of the noise
(14) parameters of the process noise and measurement
noise. Section “Parameterization and Bayesian
where Gk+1 is called the Kalman filter gain Formulation” introduces the parameterization of
(Kalman and Bucy 1961) the noise covariance matrices and the Bayesian
1 formulation. Thereafter, the online identification
Gkþ1 ¼ Sy, kþ1jk CTd Cd Sy, kþ1jk CTd þ Sn, k algorithm is presented in section “Online Estima-
(15) tion of Noise Parameters.”
as Sf,k = Sf,k (uf,k) and Sn,k+1 = Sn,k+1 (un,k+1), Given the measurement Dk, the one-step ahead
respectively. In other words, both matrices predictor for the noise parameters is taken as the
are time varying in order to allow online tracking. updated noise parameters of the previous time A
The uncertain noise parameter vector is given by step:
h iT
ukþ1jk ¼ ukjk (21)
ukþ1 ¼ uTf, k , uTn,kþ1 ℝNy (18)
On the other hand, the associated covariance
This parameterization allows for the tracking matrix is assumed to be the following:
problem for general nonstationary situations. In
the following, a concurrent procedure is intro- Su, kþ1jk ¼ mSu, kjk (22)
duced to estimate the noise parameter vector. As
a result, not only the optimal estimation can be where the fading factor m is defined in Eq. 17.
obtained but also the associated uncertainty can Here, it is assumed that the noise covariances are
be quantified. time slowly varying so the ensemble mean can be
When a new measurement zk+1 is available, approximately estimated by the temporal average
the noise parameters can be updated. Using the with certain time delay.
Bayes’ theorem, the posterior probability density On the other hand, the likelihood function is
function (PDF) of the noise parameter vector given by
given the measurement data set Dk+1 is given by
(Yuen 2010a)
pðzkþ1 jukþ1 , Dk Þ ¼ ð2pÞNo =2 jSz, kþ1jk j1=2
pðukþ1 jDk Þpðzkþ1 jukþ1, Dk Þ 1 T
pðukþ1 jDkþ1 Þ ¼ exp zkþ1 zkþ1jk S1 z, kþ1jk z kþ1 z kþ1jk
pðzkþ1 jDk Þ 2
(19) (23)
where p(uk+1|Dk) is the prior PDF of uk+1; where zk+1|k and Sz,k+1|k are readily obtained from
p(zk+1|uk+1,Dk) is the likelihood function; and Eqs. 11 and 12 given the measurement data set
p(zk+1|Dk) is a normalizing constant such that D k:
the integral of the posterior PDF over the entire
parameter space is unity. The prior PDF is zkþ1jk ¼ Cd Ad, k ykjk þ Cd hd, k (24)
approximated as a Gaussian distribution:
Sz, kþ1jk ¼ Cd Ad, k Sy, kjk ATd, k CTd þ Cd Bd, k Sf , kjk BTd, k CTd
ð2pÞðNy þNo Þ=2 1 T
pðukþ1 jDkþ1 Þ ¼ jSz, kþ1jk j 1=2
exp ukþ1 ukjk S1u, kjk
pðzkþ1 jDk ÞjmSu, kjk j1=2 2m
1 T (26)
ukþ1 ukjk zkþ1 zkþ1jk S1 z, kþ1jk zkþ1 zkþ1jk
2
26 Advances in Online Structural Identification
where zk+1|k and Sz,k+1|k are given by Eqs. 24 the posterior PDF without taking the terms that
and 25, respectively. The objective function do not depend on the noise parameters:
J(uk+1) is defined as the negative logarithm of
1 1 T
J ðukþ1 Þ ¼ lnjSz, kþ1jk j þ ukþ1 ukjk S1 u, kjk ukþ1 ukjk
2 2m
(27)
1 T
þ zkþ1 zkþ1jk S1 z, kþ1jk zkþ1 zkþ1jk
2
The updated noise parameter vector uk+1|k+1 can before solving directly this optimization prob-
be obtained by maximizing the posterior PDF lem. In this training process, a heuristic local
p(uk+1|Dk+1) in Eq. 26, and it is equivalent to search method is utilized. Thereafter,
minimizing the objective function J(uk+1) in a modified Newton’s method is used for the
Eq. 27: operating stage. Details of this estimation
scheme are presented as follows.
ukþ1jkþ1 ¼ arg min J ðukþ1 Þ (28)
ukþ1
Training Stage
Furthermore, the uncertainty of the updated noise At the beginning of the identification process, the
parameter vector uk+1|k+1 can be represented by Gaussian approximation is inaccurate due to the
its covariance matrix Su,k+1|k+1 which is equal to large prior uncertainty, i.e., large difference
the inverse of the Hessian matrix of the objective between the prior mean and the actual values,
function calculated at uk+1 = uk+1|k+1: and large prior variances. Therefore, solving
1 directly the optimization problem in Eq. 28 may
Su, kþ1jkþ1 ¼ HJ ukþ1jkþ1 (29) result in computational problems. Herein,
a training process is introduced to obtain
where the Hessian matrix is given by a preliminary solution before the long-term oper-
HJ ukþ1jkþ1 ¼ ∇J ðukþ1 Þ∇T jukþ1 ¼ukþ1jkþ1 , and it ating stage. During the training process,
can be computed using the finite difference a heuristic local search method is applied to the
method. optimization problem. This method provides
a generic tool for complicated optimization prob-
lems. The basic principle is to conduct the opti-
Online Estimation of Noise Parameters mization within a prescribed parameter set. By
In this subsection, a computationally efficient exhaustive search from this finite set, the optimal
algorithm is presented for online updating of (suboptimal) solution is the one that gives the
the noise parameter vector uk+1|k+1 and the asso- smallest objective function value. Specifically,
ciated covariance matrix Su,k+1|k+1. Starting with the candidate parameter set of the (k + 1)th time
an arbitrary initial condition u0|0 and arbitrarily step is defined as
positive definite matrix Su,0|0, the noise param- n
ðlÞ ðlÞ
eter vector and its associated covariance matrix Qkþ1 ukþ1 ℝNy : ykþ1 ¼ nykjk , l ¼ 1, . . . ,
can be updated based on Eqs. 28 and 29. As o
aforementioned, the noise parameter vector can N y , n f1=2, 1, 2g
be updated by solving the optimization problem (30)
in Eq. 28. However, due to the large prior uncer-
tainty and numerical considerations, a training where y(l) (l)
k + 1 and yk|k are the lth component of the
process is necessary for a preliminary solution noise parameter vector uk+1 and the lth
Advances in Online Structural Identification 27
one-step-ahead predictor zk + 1|k. Hereafter, the distribution with probability Qk+1. Then, the
superscript (s) is used to denote the sth compo- outlier probability for z(s)
k+1 can be defined as the
nent of a vector. Then, the normalized residual probability that no more than k+1 samples drawn
(prediction error) for z(s)
k + 1 can be defined from the normalized residuals among the (R(s) k )
(Myers and Tapley 1976): previous regular data points fall outside the inter-
k+1j, jek+1j). This probability can be
val (je(s) (s)
ðsÞ ðsÞ ðsÞ ðsÞ obtained by considering the total probability:
ekþ1 ¼ zkþ1 zkþ1jk =se, kþ1 (34)
X
ðsÞ ðsÞ
(s)
Po zkþ1 ¼ t¼0kþ1
C n Rk , t
where se,k + 1 is the standard deviation for the sth (37)
n RðsÞ t
measured channel of the difference between Qtkþ1 1 Qkþ1 ð k Þ
the measurement and the corresponding
one-step-ahead predictor. Since the probability
nðRk Þ!
ðsÞ
ðsÞ
model for the measurement noise is Gaussian, where C n Rk , t ¼
is the
ðsÞ
n Rk t !t!
the probability of a data point falling outside the
interval (|e(s) (s)
k + 1|, |ek + 1|) is
binomial coefficient. Note that introduction of
this moving window is to release the memory
ðsÞ
and computational burden in considering the
Qkþ1 ¼ 2F
ekþ1
(35)
entire history. On the other hand, due to possible
condition changes of the excitation and the
where F(
) is the cumulative distribution function underlying structure, it is suitable to consider
(CDF) of the standard Gaussian random variable. only the past data points in a reasonable neigh-
Then, a moving time window is introduced to borhood of the current time step. It is suggested to
include not more than Nw previous regular data use a value for Nw that corresponds to approxi-
points, and the set R(s)
k is introduced to include the mately 100 fundamental periods of the structure.
absolute normalized residuals of the regular data
points in this time window: Efficient Screening Criteria
n
In order to enhance the computational efficiency,
ðsÞ
ðsÞ
ðsÞ
Rk ¼
el
: zl is a regular data point, an efficient screening rule is introduced.
o (36) First, a data point is classified as a regular point
l ¼ maxð1, k N W Þ, . . . , k k+1| xL, a conservatively small bound
when |e(s)
(e.g., xL = 2). On the other hand, a data point
where n(R(s) k ) is used to denote the number of
is classified as an outlier when |e(s) k+1| xU.
k + 1 is a regular point, Rk + 1 = Rk [
If z(s) (s) (s)
the absolute value of the normalized residual of
the measurement lies within (xL, xU). This helps {|ek+1|} if n(Rk ) < NW. Otherwise, Rk+1 =
(s) (s) (s)
data points (ignoring the first one period of One can refer to Mu and Yuen (2014) for
data points), and sort them in the ascending numerical examples.
order. Then, the standard deviation s(s)e,1 can be
estimated as the value of the 68-percentile
point. This is a robust estimator of the standard Online Bayesian Model Class Selection
deviation for possible presence of outliers.
Bayesian model class selection is utilized for
Operating Process selecting the most plausible model class from
Compute the xU using Eq. 38 and initialize a set of NC dynamic model class candidates
0 = f, an empty set for s = 1, . . ., No.
R(s) C1 , C2 , . . . , CNC by considering their plausibility
P(Cj|D) conditional on the available set of
1. Calculate the one-step-ahead predictor zk+1|k dynamic measurement D (Beck and Yuen 2004;
by Eq. 33, the sth element of the normalized Yuen 2010b):
residual |e(s)
k+1| by Eq. 34.
2. P Cj p DjCj
k+1| xL (=2), zk + 1 will be classified
(a) If |e(s) (s) P Cj jD ¼ (39)
pðDÞ
as a regular data point.
k+1| xU, zk + 1 will be classified as an
(b) If |e(s) (s)
where the denominator p(D) is the normalizing
outlier.
constant; P(Cj) is the prior plausibility of model
(c) When xL < |e(s) k+1| < xU, calculate Qk+1
class Cj. In general, a noniformative prior can be
using Eq. 35 and count k + 1 from R(s) k .
used, i.e., P(Cj) = 1/NC, where j = 1,. . .,NC, and
Then, compute the outlier probability
p(D|Cj) is the evidence given by
k+1) using Eq. 37. If Po(zk+1) < 0.5,
Po(z(s) (s)
(s)
zk+1 will be classified as a regular point. ð
Otherwise, it will be concluded as p DjCj ¼ p Dju; Cj p ujCj du (40)
a suspicious data point, and it will be Yj
discarded for identification purpose.
3. According to the outlier detection result in where Yj denotes the parameter space of model
step (2), update the sets R(s) k+1, s = 1, . . ., No. class Cj. However, direct numerical computation
k+1 is an outlier, Rk+1 = Rk .
If z(s) (s) (s)
of this integral is in general computationally
30 Advances in Online Structural Identification
prohibitive unless the number of uncertain where Nj is the number of uncertain parameters of
parameters is very small (say 3 or less). To over- model class Cj and u^ is the parameter vector that
come this computational obstacle, an asymptotic maximizes the posterior PDF, which is propor-
expansion was developed for the globally identi- tional to the integrand on the right-hand side of
fiable cases (Beck and Yuen 2004): Eq. 40.
In the locally identifiable cases (Katafygiotis
1=2
^ Cj p ujC
p DjCj ð2pÞN j =2 p Dju; ^
^ j
12
1
N j =2 N i ^ ^
^
^ ^i
p Dju; Cj p ujCj ð2pÞ Si¼1 p Djui ; Cj p ui jCj
Hj ui
G u; u, Hj u (42)
where u^i is the ith local optimal point, i = 1, . . ., previous k time steps. The denominator p(zk+1|Dk)
1
^ ^i is a normalizing constant that does not depend on
Ni, and G u; ui , Hj u denotes the multi-
the model class. Note that the plausibility P(Cj|D0)
variate Gaussian distribution for the random vec- is deduced to the prior plausibility of the model
tor u with mean u ^i and covariance matrix class P(Cj). By using the theorem of total proba-
1 bility, the conditional evidence p(zk+1|Dk;Cj) can
Hj u^i . Then, the evidence is given as
be expressed as
ð
p DjCj ð2pÞNj =2 SNi¼1
i
p Dju^i ;Cj p zkþ1 jDk ; Cj ¼ p zkþ1 juk ;Dk ; Cj p uk jDk ;Cj duk
1=2 (43) Yj
p u^i jCj
Hj u ^i
(45)
k k
where p(zk+1|Dk;Cj) denotes the evidence of
model class Cj conditional on the data of the (46)
Advances in Online Structural Identification 31
where Nj denotes the number of uncertain param- Eq. 45 but it will be approximated by u*k uk|k.
eters in model class Cj and u*k = uk|k + 1 is the The maximum conditional likelihood of model
parameter vector that maximizes the integrand of class Cj evaluated at yk|k is given by A
p zkþ1 juk ; Dk ; Cj
1=2
ð2pÞNo =2
Cd Sy, kþ1jk CTd þ Sn, kþ1
1
T 1
(47)
exp zkþ1 Cd ykþ1jk Cd Sy, kþ1jk Cd þ Sn, kþ1
T
zkþ1 Cd ykþ1jk
2
The maximum posterior probability density of u*k the state vector, the structural parameters and
conditional on Dk for model class Cj is given by their associated uncertainty. This approach allows
for the tracking of nonstationary process noise
1=2
p uk jDk ; Cj ð2pÞNj =2
Su, kjk
(48) and measurement noise. It estimates the noise
parameters recursively for every time step. Then,
where Su,k|k denotes the covariance matrix of the the outlier-resistant extended Kalman filter
updated parameters. The matrix Hj(u*k ) denotes (OR-EKF) was presented for robust online outlier
the Hessian matrix of the negative natural loga- detection and structural parametric identification.
rithm of the integrand in Eq. 45 with respect to This method embeds an online outlier detection
the parameter vector, evaluated at u*k : algorithm into the EKF. It is capable for robust
estimation of structural parameters using outlier-
contaminated dynamic response data in an online
Hj uk ¼ ∇ln p zkþ1 juk ; Dk ; Cj ∇T juk ¼ukjk
X1 manner. In contrast to other existing robust
þ y, kjk
KF/EKF algorithms, the OR-EKF requires no
prior information of the outlier distribution
(49) model. Finally, the Bayesian model class selection
approach allows for nonparametric online struc-
where ∇ denotes the gradient operator with
tural identification. This recursive algorithm
respect to the parameter vector uk. The term
Summary Cross-References
This chapter presented several recent advances in ▶ Blind Identification of Output-Only Systems
online structural identification using the extended and Structural Damage via Sparse
Kalman filter (EKF). First, a Bayesian approach Representations
was introduced for online identification of the ▶ Model Class Selection for Prediction Error
noise parameters. This approach resolves the Estimation
divergence problem possibly encountered in the ▶ Nonlinear System Identification: Particle-
conventional EKF due to improper selection of Based Methods
the noise covariance matrices. Furthermore, the ▶ Stochastic Structural Identification from
presented approach ensures reliable estimation of Vibrational and Environmental Data
32 Advances in Online Structural Identification
Yuen KV, Hoi KI, Mok KM (2007) Selection of noise to normal conditions (dead loads, wind, micro-
parameters for Kalman filter. Earthq Eng Eng Vib tremors); and (c) the multiple-input nature of
6:49–56
Zhou J, Luecke RH (1994) Estimation of the covariances ambient excitation, ensuring that the response A
of the process noise and measurement noise for a linear includes the contribution of a certain number of
discrete dynamic system. Comput Chem Eng modes. It is indeed true that the response of
19(2):187–195 a historic building to ambient excitation is gener-
ally low, but this cannot be considered
a prohibitive issue as currently highly sensitive
and relatively inexpensive accelerometers are
Ambient Vibration Testing of available on the market.
Cultural Heritage Structures In addition, the popularity of operational
modal testing and analysis of civil engineering
Carmelo Gentile and Antonella Saisi structures has been favored also by the techno-
Department ABC, Politecnico di Milano, logical advances (i.e., the availability of data
Milan, Italy acquisition and storage systems, which are fully
computer based) and by the large number of
output-only modal identification techniques
Synonyms available in the literature (see, e.g., Magalhães
and Cunha 2011).
Operational modal analysis; Operational modal Furthermore, AVT and OMA seem ideal tools
testing to methodologically complement the investigations
currently carried out to assess the structural safety
of Cultural Heritage structures. Although general
Introduction rules, which can be applied to all historic construc-
tions, are very difficult to define, it is generally
Ambient vibration testing (AVT), long-term agreed (see, e.g., Binda et al. 2000) that the first
dynamic monitoring, and operational modal anal- phase of a correct diagnostic approach (i.e., the
ysis (OMA, i.e., the identification of modal param- evaluation of the current health state or perfor-
eters from ambient vibration data) of Cultural mance of the building) involves the collection of
Heritage structures are a rather recent topic, and all the essential information on the geometry of the
only a limited number of complete investigations building, its evolution from the origin to the present
are reported in the literature (Jaishi et al. 2003; state, the construction technologies, the character-
Bennati et al. 2005; Ivorra and Pallares 2006; istics of masonry texture, the mechanical charac-
Gentile and Saisi 2007, 2013; Pau and Vestroni terization of the materials, and the evaluation of
2008; Casarin and Modena 2008; Peña et al. 2010; their state of preservation.
Ramos et al. 2010; Aras et al. 2011; Oliveira The above documentary and experimental
et al. 2012). On the other hand, there is a growing information – as it is suggested also in current
interest on this topic since the preservation of Italian Guidelines for the seismic risk mitigation
Cultural Heritage is of primary concern in many of Cultural Heritage (DPCM 2011) – provide a
countries all over the world. 1st-level diagnosis, highlighting the overall state
AVT and OMA are especially suitable to his- of preservation, the presence of local defects and
toric structures for several reasons: (a) the easy vulnerabilities, as well as the need of possible
and fully nondestructive way of testing, repair interventions, especially when local issues
performed by measuring only the structural are detected; in this case, the intervention design
response under ambient excitation; (b) the sus- should be addressed by simplified local models
tainability of testing, which does not interfere (see, e.g., Giuffrè 1993; DPCM 2011).
with the normal use of the structure and does In principle, the previously collected knowl-
not induce additional loads rather than those due edge of the building should be synthesized in
34 Ambient Vibration Testing of Cultural Heritage Structures
a finite element (FE) model of the structure. The Other possible applications of ambient vibration-
FE model, in turn, should provide a 2nd-level based modal analysis in the field of historic struc-
diagnosis since it could be used for evaluating tures include periodic or continuous monitoring
the structural safety under service loads, in order to evaluate the effects of repair interven-
predicting the performance under exceptional tions or to perform dynamics-based damage
loads (such as earthquakes), and simulating the assessment (Ramos et al. 2010).
effects of structural modifications or repair The present entry, after a review of some
interventions. output-only modal identification techniques, pre-
However, FE modeling of historic structures is sents the application of ambient vibration-based
characterized by well-known issues: modal and structural identification to a historic
masonry tower. Subsequently, the role of AVT
(a) The correlation between the results of local and OMA in the preservation of Cultural Heritage
tests (which indeed provide the mechanical structures is exemplified in a further application
characterization of the materials) and quanti- (i.e., the long-term dynamic monitoring of a tower)
tative parameters to build up global structural aimed at (1) evaluating the effects of structural
capacity models is still an open issue (Binda modifications, (2) assessing the influence of envi-
et al. 2000). ronmental effects on natural frequencies, (3) iden-
(b) The structural model of a historic structure, tifying the evolution of damage mechanisms.
even when all the collected information is
accurately represented, continues to involve
significant uncertainties, e.g., in the material Modal Identification from Ambient
properties (and their distribution) as well as Vibration Data
in the boundary conditions. This aspect is
especially critical for complex historic build- As previously pointed out, a large number of
ings evolved in different phases. output-only modal identification techniques are
(c) FE models are often used, even in refined available in the literature, ranging from the sim-
nonlinear analyses, without experimental ple peak picking technique (PP, Bendat and
validation, and only occasionally the model Piersol 1993) to the more advanced frequency
validation is roughly performed by using few domain decomposition (FDD, Brincker
available local data (such as the stress level et al. 2000) and stochastic subspace identification
evaluated in few points through flat-jack (SSI, van Overschee and De Moor 1996). In this
tests). section, the PP, FDD, and SSI techniques are
briefly described.
Within this context, one possible key role of
AVT and OMA is to provide effective and accu- Peak Picking and Frequency Domain
rate validation of the FE model prior to its use in Decomposition
numerical analysis, as demonstrated in different The PP and FDD techniques work in frequency
studies on temples and monuments (Jaishi domain and are based on the evaluation of the
et al. 2003; Pau and Vestroni 2008), churches spectral matrix Gyy( f ) of the recorded responses:
(Casarin and Modena 2008; Ramos et al. 2010;
Gattulli et al. 2013), ancient palaces (Aras Gyy ð f Þ ¼ E Yð f ÞYH ð f Þ (1)
et al. 2011), towers, and minarets (Bennati
et al. 2005; Ivorra and Pallares 2006; Gentile where the vector Y(f) collects the responses in the
and Saisi 2007, 2013; Peña et al. 2010; Ramos frequency domain, the superscript H denotes the
et al. 2010; Oliveira et al. 2012). Hermitian transpose operation (i.e., complex
In some cases, AVT can possibly help also to conjugate matrix transpose), and E denotes
limit the number of on-site and laboratory tests, expected value. The diagonal terms of the matrix
which are time-consuming and cost-ineffective. Gyy( f ) are the (real valued) auto-spectral
Ambient Vibration Testing of Cultural Heritage Structures 35
densities (ASD), while the other terms are the the spectral matrix can be approximated, in the
(complex) cross-spectral densities (CSD). The neighborhood of a resonant frequency fk, as
spectral matrix Gyy( f ) is generally computed by A
using the modified periodogram method (Welch Gyy ð f k Þ ak fk fk H (6)
1967). According to this approach, an average is
made over each recorded signal, divided into where ak depends on the damping ratio, the nat-
M frames of 2n samples, where windowing and ural frequency, the modal participation factor,
overlapping are applied. and the excitation spectra. Equation 6 highlights
The principle of the two techniques is easiest that (a) each row or column of the spectral matrix
illustrated by recalling that any response vector at a natural frequency fk can be considered as an
y(t) can be expressed in modal coordinates q(t) as estimate of the mode shape fk at that frequency
and (b) the square root of the diagonal terms of
yðtÞ ¼ f1 q1 ðtÞ þ f2 q2 ðtÞ þ . . . ¼ FqðtÞ (2) the spectral matrix at a natural frequency fk can be
considered as an estimate of the mode shape fk at
where fi represents the i-th mode shape vector that frequency.
and F is the mode shape matrix. Hence, the The PP technique leads to reliable results pro-
correlation matrix Cyy(t) (see, e.g., Bendat and vided that the basic assumptions of low damping
Piersol 1993) of the responses and well-separated modes are satisfied; draw-
backs of the method are related to the difficulties
Cyy ðtÞ ¼ E yðt þ tÞyT ðtÞ (3) in identifying closely spaced modes and damping
ratios.
becomes Some refinements or variants of the classic PP
technique have been proposed in the literature.
Cyy ðtÞ ¼ FE qðt þ tÞqT ðtÞ FT ¼ FCqq ðtÞFT For example, Felber (1993) suggested obtaining a
(4) “global picture” of the eigenfrequencies by eval-
uating an appropriate function, called averaged
On the other hand, the spectral matrix may be normalized power spectral density (ANPSD).
defined as the Fourier transform of the correlation More specifically, only the diagonal elements of
matrix: the spectral matrix Gyy(f) are considered, normal-
ized, and averaged so that a unique function is
Gyy ð f Þ ¼ ℑ Cyy ðtÞ ¼ FGqq ð f ÞFH (5) obtained, summarizing the information on the
“modal” peaks contained in all the ASDs.
where Gqq( f ) is the spectral matrix of the modal The FDD technique (Brincker et al. 2000)
coordinates. It is worth underlining that since the involves the singular value decomposition
modal coordinates are uncorrelated, the matrix (SVD) of the spectral matrix at each frequency
Gqq( f ) is diagonal. and the inspection of the curves representing the
The more traditional approach to estimate the singular values, in order to identify the resonant
modal parameters of a structure is often called frequencies and to estimate the corresponding
peak picking method after its key step: the iden- mode shape using the information contained in
tification of the resonant frequencies as the peaks the singular vectors of the SVD.
of ASDs and CSDs. In fact, for a lightly damped The SVD of the spectral matrix at each fre-
structure subjected to a white-noise random exci- quency is given by
tation, both ASDs and CSDs reach a local maxi-
mum at the frequencies corresponding to the Gyy ð f Þ ¼ Uð f ÞSð f ÞUH ð f Þ (7)
system normal modes (Bendat and Piersol
1993); furthermore, if the assumption is intro- where the diagonal matrix S collects the real
duced of modes having well-separated frequen- positive singular values in descending order and
cies, it can be shown (see, e.g., Peeters 2000) that U is a complex matrix containing the singular
36 Ambient Vibration Testing of Cultural Heritage Structures
vectors as columns. The SVD is used for estimat- domain by inverse FFT yields a free decaying
ing the rank of Gyy at each frequency with the time domain function, which represents the
number of nonzero singular values being equal to autocorrelation function of the modal coordi-
the rank; if only one mode is important at a given nate. The natural frequency and the related
frequency fk, as it has to be expected for well- damping ratio are thus simply found by esti-
separated modes, the spectral matrix can be mating crossing times and employing the log-
approximated by a rank-one matrix: arithmic decrement method.
where Ca, Cv, and Cd are the output location A ¼ eAc Dt B ¼ ðA IÞA1
c Bc
matrices for accelerations, velocity, and displace- (16)
C ¼ Cc E ¼ Ec
ments, respectively. These matrices contain a lot A
of zeros and a few unit entries since they are
It is finally observed that the deterministic
formulated in order to assign the measured
model (Eq. 15) is not capable of exactly describ-
degrees of freedom. The vector y(t) can be written
ing real measurement data; consequently, Eq. 15
as
needs to be modified to account for both the
process noise wk ℜ2N due to disturbances
yðtÞ ¼ Cc xðtÞ þ Dc fðtÞ (12)
and modeling inaccuracies and the measurement
noise vk ℜL due to sensor inaccuracy:
Cc ¼ Cd Ca M1 K, Cv Ca M1 C1
Dc ¼ Ca M1 xkþ1 ¼ A xk þ B fk þ wk
(17)
(13) y k ¼ C xk þ D f k þ v k
By combining the state Eq. 9 and the observa- Furthermore, both wk and vk are unmeasurable
tion Eq. 12, the classical continuous-time state- vectors, assumed to be zero mean, white, and
space model is found: with covariance matrices:
x_ ðtÞ ¼ Ac xðtÞ þ Bc fðtÞ wp Q S
(14) E wTp vTp ¼ d
yðtÞ ¼ Cc xðtÞ þ Dc fðtÞ vp ST R pq
(18)
Since real measurements are taken at discrete-
time instants and in order to fit model Eq. 14 to It is very important to remark that an AVT
the measurements, this model needs to be provides information only on the vibration
converted into discrete time. Hence, assuming responses of a structure excited by unmeasured
a constant sampling period Dt and that the input inputs. Consequently, it is impossible to distin-
is piecewise constant over the sampling period, guish the input term fk from the noise terms wk
the continuous-time equations (Eq. 14) are and vk in Eq. 17. This results in the following
discretized and solved at all discrete-time instants discrete-time stochastic state-space model:
tk = kDt, obtaining the discrete-time state-space
model: xkþ1 ¼ A xk þ wk
(19)
yk ¼ C xk þ vk
xkþ1 ¼ A xk þ B fk
(15)
yk ¼ C xk þ D fk where the input is implicitly modeled by the noise
terms and the white noise assumption of these
where xk is the discrete-time state vector terms turns out to be essential: if the white noise
(containing displacements and velocities describ- assumption is violated because the input contains
ing the state of the system at time instant tk = some dominant frequency components, these
kDt), fk and yk are the sampled input and output frequency components cannot be distinguished
vectors, A is the discrete state matrix (dependent from the eigenfrequencies of the system (of the
on the mass, stiffness, and damping properties of state matrix A).
the structure), B is the discrete input matrix, C is The key step of the SSI techniques is the
the discrete output matrix (which maps the state estimation of the state-space matrices A and
vector into the measured output), and D is the C from the measured output yk. The estimation
direct transmission matrix. The matrices A, B, C, of A and C can be performed by using different
and D are related to their continuous-time coun- algorithms. The well-known data-driven SSI
terparts Eq. 10 and Eq. 13 by the following: algorithms are based on linear algebra theorems
38 Ambient Vibration Testing of Cultural Heritage Structures
(van Overschee and De Moor 1996) demonstrat- number of modes that are needed to accurately
ing that the state-space matrices can be calculated describe the structural response. However, with
from the knowledge of the block Hankel matrix the purpose of detecting weakly excited modes,
of the measurements, defined as it is often necessary to consider more than 2 N
values, which are what would be physically suf-
0 1
y0 y1
yj1 ficient. On the other hand, over-modeling leads
B y1 y2
yiþj C
B iþ1 C tion diagrams, where the modal parameters
@ ⋮ ⋮ ⋱ ⋮ A obtained for increasing model orders are collec-
y2i1 y2i
y2iþj2 tively represented. A physical mode is conceiv-
(20) ably identified when consistent frequencies,
damping, and mode shapes (classified as stable
where 2i and j are user-defined quantities, poles) are obtained for models of increasing
representing the number of output block rows order.
and the number of columns of matrix Hi,
respectively.
In Eq. 2, the block Hankel matrix of the mea- Ambient Vibration Testing and
surements is subdivided into two sub-matrices, Structural Identification of a Historic
named as Yp and Yf, which are usually referred to Masonry Tower
as past and future output block matrices. The
orthogonal projection of the row space of Yf onto Description of the Tower
the row space of Yp can be directly calculated The first investigated historic building is a bell
yielding to matrix Pi, which is known as projection tower (Fig. 1), about 37.0 m high and built in
matrix. The observability matrix Oi of the stonework masonry (Gentile and Saisi 2013).
system – which, in turn, allows to compute A and The tower – located in the small town of
C – can be estimated from the SVD of matrix Arcisate (northern Italy) – has a square cross
W1PiW2, where W1 and W2 are convenient weight section, with sides of 5.8 m, and is connected
matrices. The available SSI-data algorithms essen- to the church Chiesa Collegiata on the East
tially differ for the expressions of the weight matri- side and partly on the South side. The church,
ces (van Overschee and De Moor 1996). dedicated to St. Vittore, dates back to the
After estimating the model matrices, the fifteenth century and replaced a more ancient
modal parameters fi, zi, and fi of the structural church, built in the fourth century and modified
system are calculated from in the eleventh century. Probably the tower
foundation dates back to the late Roman age,
A ¼ CLC1 (21a) as well.
The first historic document concerning the
lnli
L ¼ diagðli Þ lci ¼ (21b) tower goes back to the sixteenth century and
Dt reports St. Carlo Borromeo’s request of access
c
c
l
real l modification. Seven orders of floors are present,
fi ¼ i
Bi ¼
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 1 (a) View of the bell tower of Chiesa Collegiata
(Arcisate, Varese); (b) crack patterns on the fronts of the tower (dimensions in m)
progressively along the height, from 135 cm at Ambient Vibration Testing Procedures and
the ground level up to 65 cm at the top level. Modal Identification
Although extensive visual inspections and few The dynamic tests were carried out in June 2007,
sonic tests generally indicate that the stone using a 16-channel data acquisition system and
masonry is relatively compact and of fairly WR 731A piezoelectric accelerometers (10 V/g
good execution, the masonry texture appears sensitivity and 0.5 g peak acceleration). Each
locally often highly disordered and characterized accelerometer was connected with a short cable
by the local presence of vertical joints. (1 m) to a WR P31 power unit/amplifier, provid-
The crack pattern (Fig. 1b) has been accu- ing the constant current needed to power the
rately surveyed also by using an aerial platform. accelerometer’s internal amplifier, signal ampli-
The tower exhibits long vertical cracks on every fication, and selective filtering.
side, most of them cutting the entire wall thick- The response of the tower was measured in
ness and passing through the keystones of the 15 selected points, belonging to 5 different cross
arch window openings. These cracks are mainly sections along the height of the building, according
distributed between the second/third order of the to the sensor layout illustrated in Fig. 2a. Figure 2b
tower. Many superficial cracks are also diffused, shows the mounting of accelerometers in the
particularly on the North and West fronts, which instrumented cross section at level +22.78 m.
are not adjacent to the church. The acceleration time histories induced by
The visual inspection highlighted that the ambient excitation were recorded for 3,600 s at
upper part of the tower, beneath the belfry level, a sampling frequency of 200 Hz. A sample of the
could be probably considered the most vulnera- acceleration time histories recorded during the
ble, due to the widespread mortar erosion and the test in the upper part of the tower is shown in
infilled openings (Fig. 1b), mainly on the East and Fig. 3: it should be noticed that very low level of
North fronts; in addition, the infillings are often ambient excitation was present during the tests,
not properly linked to the surrounding load- with the maximum recorded acceleration being
bearing masonry. always lower than 0.4 cm/s2.
40 Ambient Vibration Testing of Cultural Heritage Structures
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 2 (a) Sensor layout adopted in the dynamic tests
(dimensions in m), (b) mounting of the accelerometers at level +22.78 m
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 3 Typical acceleration time series measured at
the upper instrumented level
Ambient Vibration Testing of Cultural Heritage Structures 41
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 4 (a) First singular A
value (SV) curve and
identification of natural
frequencies (FDD), (b)
stabilization diagram and
automatic (A) identification
of natural frequencies (SSI)
The modal identification was performed using The results of OMA in terms of natural fre-
time windows of 3,600 s, in order to comply with quencies can be summarized through the plots of
the widely agreed recommendation of using an Fig. 4a, b, showing the first singular value (SV) of
appropriate duration of the acquired time window the spectral matrix and the stabilization diagrams
(ranging between 1,000 and 2,000 times the fun- obtained by applying the FDD and the SSI tech-
damental period of the structure; see, e.g., nique, respectively. The inspection of Fig. 4a
Cantieni 2005) to obtain accurate estimates of highlights that the FDD technique provides
the modal parameters from OMA techniques. In a clear indication of the tower modes through
fact, as already pointed out in section “Modal well-defined local maxima in the first SV; simi-
Identification from Ambient Vibration Data,” larly, Fig. 4b shows that the alignments of the
OMA methods assume that the excitation input stable poles in the stabilization diagram of the
is a zero mean Gaussian white noise, and this SSI method provide a clear indication of these
assumption is as closely verified as the length of modes, as well. Furthermore, Fig. 4a, b shows the
the acquired time window is longer. correspondence of the natural frequency esti-
The extraction of modal parameters from mates between the two techniques, with the res-
ambient vibration data was carried out using the onant peaks of Fig. 4a being placed practically at
FDD and the data-driven SSI techniques avail- the same frequencies of the alignments of stable
able in the commercial software ARTeMIS (SVS poles in Fig. 4b.
2012). Notwithstanding the very low level of As it had to be expected, the identified modes
ambient response (Fig. 3) that existed during the can be classified as bending and torsion. Figure 5
tests, the application of both techniques allowed shows the identified mode shapes (SSI tech-
to identify five vibration modes in the frequency nique): dominant bending (B) modes were iden-
range of 0–6 Hz. tified at 1.22 (B1), 1.28 (B2), 4.01 (B3), and
42 Ambient Vibration Testing of Cultural Heritage Structures
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 5 Vibration modes
identified from ambient
vibration data (SSI)
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 7 Vibration modes of A
the initial model
Hence, three steps of manual tuning were car- the first three modes. The discrepancy is conceiv-
ried out to establish a base FE model. In the base ably related to the simplified distribution of the
model, the following assumptions were adopted: model elastic properties.
As suggested by the mode shapes, the higher
1. An orthotropic elastic behavior was assumed bending modes depend on the elastic characteris-
for the stone masonry, with the average char- tics of the masonry in the upper part of the tower.
acteristics of the material being E = 3.00 GPa In addition, as pointed out in section “Description
and G13 = G23 = 0.45 GPa (corresponding to of the Tower,”, the upper region is characterized
a = G13/E = G23/E = 0.15). It is worth noting by a more evident mortar joint erosion and
that the assumed Young’s modulus was in changes of the masonry texture including wide
good agreement with the results of the tests infilled openings on the East and North fronts,
performed to characterize the masonry and beneath the belfry level.
reported in (Binda et al. 2012): more specifi- Hence, the distribution of Young’s modulus
cally, the average sonic velocity (P waves) was updated, and the tower was divided in two
varied between 1,620 and 2,240 m/s2, and the regions, with the masonry Young’s modulus
values measured by double flat-jack tests at being assumed as constant within each zone.
the base of the tower were slightly larger The two regions, denoted as I and II, correspond
than 3.00 GPa. to the lower five levels of the building (EI,
2. The effects of the connection between the h 26.0 m, including a large number of
tower and the church were accounted for passing-through cracks) and the upper part (EII,
through a series of linear (nodal) springs of h > 26.0 m, including infilled/repaired areas and
constant k. After tuning the parameter k in the belfry), respectively. The possible set of
a preselected interval (1 104 kN/m k updating parameters includes EI, EII, a, and k. It
10 104 kN/m), the value k = 4 104 kN/m should be noticed that assuming only one param-
was assumed since it tends to minimize the eter a means that the ratio between the shear
average difference between the measured and moduli in the two regions of the tower equals to
predicted modal frequencies. the ratio between Young’s moduli.
Subsequently, the optimal values of the
Figure 7 illustrates the dynamic characteristics updating structural parameters were determined.
of the base model and highlights that the mode Among the different classes of procedures avail-
shapes are fully consistent with the experimental able in the literature (see, e.g., Friswell and
results (Fig. 5). Furthermore, the correspondence Mottershead 1995), it was decided to evaluate
with the identified frequencies is fairly good for the updating parameters by minimizing the
44 Ambient Vibration Testing of Cultural Heritage Structures
Ambient Vibration Testing of Cultural Heritage Structures, Table 1 Structural parameters for FE model
identification
Structural parameter Base value Upper value Base value Optimal value
EI (height 26 m) (GPa) 2.0 4.0 3.0 2.97
EII (height > 26 m) (GPa) 1.0 3.0 2.0 1.60
a 0.110 0.190 0.150 0.172
k (kN/m) 1.0 104 10.0 104 4.0 104 8.30 104
difference between theoretical and experimental values; then each parameter is varied, one at time,
natural frequencies, through the simple proce- from the base value to upper and lower limit,
dure proposed in (Douglas and Reid 1982). respectively.
According to this approach, the dependence of It is further noticed that, in principle, the qua-
the natural frequencies of the model on the dratic approximation Eq. 22 is as better as the
unknown structural parameters Xk (k = 1, 2,. . ., base values are closer to the solution; hence, the
N) is approximated around the current values of accuracy and stability of the optimal estimates
Xk, by the following: should be carefully checked either by the com-
plete correlation with the experimental data or by
X
N repeating the procedure with new base values.
f i ðX1 , X2 , . . . , XN Þ ¼ Ai, k Xk þ Bi, k X2k þ Ci Table 1 summarizes the optimal estimates of
k¼1
the structural parameters, the base values, and the
(22) assumed lower and upper limits. By examining
the optimal values in Table 1, the following com-
where fi* represents the approximation of the i-th
ments can be made: (a) the optimal values of the
frequency of the FE model. Once the set of
elastic parameters in the lower region
approximating functions Eq. 22 has been
(EI = 2.97 GPa, GI13 = GI23 = a EI = 0.172
established, the structural parameters of the
2.97 ffi 0.51 GPa) are very similar to those of
model are evaluated by a least-square minimiza-
the base model; (b) the optimal estimates of the
tion of the difference between each fi* and its
elastic parameters in the upper region
experimental counterpart fiEXP:
(EII = 1.60 GPa, GII13 = GII23 = a
EII = 0.172 1.60 ffi 0.28 GPa) turned out to be
X
M
J¼ wi e2i (23a) lower than the ones in the lower part, reflecting the
i¼1 observed state of preservation in that region; and
(c) the stiffness of the springs becomes larger than
ei ¼ f EXP
i f i ðX1 , X2 , . . . , XN Þ (23b) in the base model to better fit the ratio between the
natural frequencies of the first two modes.
where wi is a weight constant. However, Eq. 22 Figure 8 shows the mode shapes of the
represents a reasonable approximation in a range, updated model, corresponding to the experimen-
around the “base” value of the structural param- tal ones (Fig. 5), and the correlation with the
eters XkB, limited by lower XkL and upper values measured modal behavior. It should be noticed
XkU (k = 1, 2, . . . , N); thus, the coefficients Aik, that the updated model represents an excellent
Bik, Ci are dependent on both the base value of the approximation of the real structure, with the max-
structural parameters and the range in which imum relative error between predicted and mea-
these parameters can vary. The coefficients Aik, sured modal frequencies being larger than 1 %
Bik, Ci are readily evaluated from (2 N + 1) finite only for mode B3.
element analyses (Douglas and Reid 1982), each Furthermore, also the correlation between
with a different choice of the parameters: the first mode shapes – estimated via the MAC
choice of the parameters corresponds to the base (Allemang and Brown 1983) – is very good for
Ambient Vibration Testing of Cultural Heritage Structures 45
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 8 Vibration modes of A
the optimal (updated)
model
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 9 Views of the Gabbia Tower (Mantua, Italy)
from South and East
the first two modes (with the MAC being larger Visual inspection of all main-bearing walls
than 0.96); for the higher modes, the MAC is in (Saisi et al. 2013) clearly indicated that the
the range 0.711–0.845 so that appreciable aver- upper part of the tower is characterized by
age differences are detected and again related to the presence of several discontinuities due to the
the simplified distribution of the model elastic historic evolution of the building, local lack
properties, which were held constant for large of connection, and extensive masonry decay.
regions of the tower. The poor state of preservation of the same region
was confirmed by the observed dynamic charac-
teristics (Saisi et al. 2013), and one local mode
Ambient Vibration Testing and Dynamic involving the upper part of the tower was clearly
Monitoring of a Historic Masonry Tower identified by applying different output-only tech-
niques to the response data collected for more
After the Italian earthquakes of May 2012, an than 24 h on the historic building.
extensive research program has been performed These results clearly highlighted the critical
to assess the state of preservation of the tallest situation of the upper part of the tower, pointing
historic tower in Mantua, Italy. The investigated out the need for structural interventions to be
tower (Fig. 9), about 54.0 m high and dating back carried out. With this motivation, and in order to
to the twelfth century, is known as the Gabbia allow for better indoor inspection of the tower-
Tower. bearing walls, a metal scaffolding and a light
46 Ambient Vibration Testing of Cultural Heritage Structures
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 10 Fronts and section of the Gabbia Tower
(dimensions in m)
wooden roof have been installed inside the tower. The original layout of the surrounding struc-
Hence, a second dynamic test was performed – tures is unknown. At present, the tower is part of
aimed at checking the possible effects of scaf- an important palace, evolved since the thirteenth
folding and wooden roof on the modal character- century, complicating the geometry of the struc-
istics of the structure – and a simple permanent ture and the mutual links.
dynamic monitoring system (including three Few historic documents are available on the
highly sensitive accelerometers and one temper- past interventions on the tower, but the observa-
ature sensor) was installed in the tower, with tion of the masonry texture reveals passing-
structural health monitoring and seismic early through discontinuities in the upper region,
warning purposes. which are conceivably related to the tower evo-
lution. Traces of past structures are visible on all
Description of the Tower and On-Site fronts, and the presence of merlon-shaped discon-
Inspections tinuities suggests modifications and further
The Gabbia Tower, about 54.0 m high, is built in adding at the top of the tower. Moreover, at
solid masonry bricks and has an almost square about 8.0 m from the top, a clear change of the
base; the load-bearing walls are about 2.4 m thick brick surface workmanship (the bricks of the
up to the upper levels (Fig. 10) where the thick- lower part are superficially scratched) could
ness of the masonry cross section decreases to reveal a first addition; in the same region concen-
about 0.7 m. The top part of the building has trated changes of the masonry texture reveal local
a two-level lodge, which hosted in the nineteenth repair.
century the observation and telegraph post. An accurate on-site survey of all fronts of the
A wooden staircase reached the lodge, but it is tower was firstly performed using a mobile plat-
no more practicable since several years due to the form. The visual inspection, omitting the upper
lack of maintenance. The inner access to the part of the tower (i.e., a portion about 8.0 m high),
tower was reestablished recently (October 2012) did not reveal evident structural damage but only
through provisional scaffoldings. superficial decay of the materials (mainly mortar
Ambient Vibration Testing of Cultural Heritage Structures 47
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 11 Instrumented cross sections and sensors
layout during the dynamic tests performed on August and November 2012 (dimensions in m)
joint erosion, due to the natural aging and the lack Ambient Vibration Testing Procedures and
of maintenance). Subsequent pulse sonic tests, Modal Identification
double flat jacks, and laboratory tests on sampled Two AVTs were conducted on the tower:
mortars and bricks confirm the soundness and the between 31/07/2012 and 02/08/2012 and on
compactness of the masonry until the height of 27/11/2012. It is worth recalling that the second
about 46.0 m. On the contrary, significant dam- test was performed after the installation of
ages were observed in the upper 8.0 m of the a metallic scaffolding and a wooden roof inside
tower; those damages are related to the the tower in order to check the possible effects of
abovementioned detachment of the several con- those additions on the dynamic characteristics of
struction phases and worsened by the natural the structure.
decay. More specifically, critical areas are the The response of the tower was measured in
infillings between the merlons, supported only 12 selected points, belonging to 4 different cross
by few courses of thin masonry due to the unusual sections along the height of the building,
layout of the scaffolding holes. according to the sensor layout shown in Fig. 11.
48 Ambient Vibration Testing of Cultural Heritage Structures
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 12 Vibration modes
generally identified during
the first AVT (SSI, 31/07/
2012, 21:00–22:00)
It is worth noting that the same cross sections N-E/S-W plane with slight components also in
were instrumented in the two tests, but in the the orthogonal N-W/S-E plane; (c) just one tor-
second survey the accelerometers were installed sion mode (T) was identified (Fig. 12d); and
on the inner side of load-bearing walls. (d) the last identified mode is local (L) and only
In both tests, the excitation was provided only involves deflections of the upper portion of the
by wind and micro-tremors. In the first test, accel- tower (Fig. 12e).
eration data were acquired for 28 h (between The presence of a local vibration mode pro-
16:00 and 23:00 of 31/07/2012 and from 9:00 of vides further evidence of the structural effect of
01/08/2012 to 6:00 of 02/08/2012), and a second the change in the masonry quality and morphol-
acquisition system was used to measure the tem- ogy observed on top of the tower during the visual
perature in three different points of the tower: on inspection. On the other hand, both visual inspec-
the S-W front both indoor and outdoor tempera- tion and operational modal analysis confirm the
ture were measured, whereas only the outdoor concerns about the seismic vulnerability of the
temperature was measured on the S-E front. It is building and explain the fall of small masonry
worth mentioning that the changes of outdoor pieces from the upper part of the tower, reported
temperature were very significant, ranging during the earthquake of May 29, 2012.
between 25
C and 55
C, whereas slight varia- Statistics of the modal frequencies identified
tions were measured by the indoor sensor between 31/07/2012 and 02/08/2012 are summa-
(29–30
C), due to the high thermal inertia of rized in columns (2)–(5) of Table 2 through the
the load-bearing walls. mean value, the standard deviation, and the
The modal identification was performed using extreme values of each modal frequency. It
3,600 s long time windows and applying the data- should be noticed that the natural frequencies of
driven SSI algorithm available in the Artemis all modes exhibit slight but clear variation, with
software (SVS 2012). the standard deviation ranging between 0.011 Hz
Figure 12 shows typical results of the first (mode B2) and 0.037 Hz (mode L1). The corre-
AVT, in terms of natural frequencies and mode lation analysis performed to investigate the pos-
shapes. It should be noticed that (a) two closely sible relationships between natural frequencies
spaced modes were identified around 1.0 Hz and and temperature (Saisi et al. 2013) clearly indi-
these modes (Fig. 12a, b) are dominant bending cated that the natural frequencies of the global
(B) and involve flexure in the two main planes of modes B1–B3 and T1 increase with increased
the tower, respectively; (b) the third mode temperature. This behavior, observed also in pre-
(Fig. 12c) involves dominant bending in the vious measurements on masonry structures
Ambient Vibration Testing of Cultural Heritage Structures 49
Ambient Vibration Testing of Cultural Heritage Structures, Table 2 Natural frequencies identified (SSI) in the
ambient vibration tests of Gabbia Tower
31/07/2012 – 02/08/2012 27/11/2012
A
Mode fave (Hz) sf (Hz) fmin (Hz) fmax (Hz) f (Hz)
(1) (2) (3) (4) (5) (6)
1 (B1) 0.981 0.018 0.957 1.014 0.918
2 (B2) 1.026 0.011 1.006 1.052 0.986
3 (B3) 3.891 0.025 3.857 3.936 3.887
4 (T1) 4.763 0.022 4.714 4.802 4.648
5 (L1) 6.925 0.037 6.849 6.987 –
6 (L10 ) – – – – 9.893
B bending mode, T torsion mode, L local mode
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 13 Vibration modes
identified during the second
AVT (SSI, 27/11/2012)
(Ramos et al. 2010; Gentile et al. 2012), can be modes B1–B3 did not exhibit significant
explained through the closure of superficial changes (see Figs. 12a–c and 13a–c). Hence,
cracks, minor masonry discontinuities, or mortar the metallic scaffolding and the wooden roof
gaps induced by the thermal expansion of mate- practically do not affect those modes.
rials. Hence, the temporary “compacting” of the 2. On the contrary, the mode shape T1 (Fig. 13d)
materials induces a temporary increase of stiff- now involves both torsion and bending. The
ness and modal frequencies, as well. identified frequency did not change apprecia-
As previously stated, the possible effects of bly with respect to the first dynamic survey,
scaffolding and wooden roof on the dynamic but the mode shape looks significantly differ-
characteristics of the tower were investigated in ent. The torsion component is still dominant in
the second AVT, performed on November the lower portion of the structure, while the
27, 2012 with the outdoor temperature being upper part is characterized by dominant bend-
almost constant (10–11
C). The results of this ing with significant components along the two
investigation in terms of identified natural fre- main planes of the tower. In other words, after
quencies’ main and mode shapes are shown in the installation of the wooden roof, mode
Fig. 13 and can be summarized as follows: shape T1 becomes mixture of previous
modes T1 (Fig. 12d, lower part of the struc-
1. Beyond the difference in terms of natural fre- ture) and L1 (Fig. 12e, upper part of the struc-
quency (that are conceivably related to the tem- ture). Furthermore, the previous mode L1 was
perature effects), the mode shapes of bending no more detected.
50 Ambient Vibration Testing of Cultural Heritage Structures
3. The previous local mode L1 (Fig. 12e) has database, with essential data records, to be used
been “replaced” by another local mode, with in the modal identification phase. The results of
higher frequency of 9.89 Hz, involving torsion modal identification herein presented were
of the upper part of the tower (Fig. 13e). obtained applying the SSI technique.
Figure 14a presents the evolution of the outdoor
As a further comment, it seems that especially temperature on the S-W front during the period
the wooden roof, even if very light, affects the from 07/01/2013 to 06/04/2013 and shows that the
dynamic characteristics of the upper part of the temperature changed between 2
C and 25
C
building: the roof acts as a mass directly with significant daily variations in sunny days.
connected to a highly inhomogeneous and The identification of the modal frequencies
weaker portion of the building, so that from the datasets collected during the same
a possible decrease of the natural frequency of period provided the frequency tracking shown
the previous local mode is generated. in Fig. 14b. The inspection of Fig. 14b firstly
suggests that the slight fluctuation of the natural
Long-Term Dynamic Monitoring and frequencies of global modes follows the temper-
Typical Results ature variation. In order to better explore the
Few weeks after the execution of the second temperature effect on the modal frequencies,
AVT, a simple dynamic monitoring system was Fig. 15 presents the first four natural frequencies
installed in the tower. The system is composed by of the tower plotted with respect to temperature,
a 4-channel data acquisition system (24-bit reso- along with linear best fit lines. The plots in Fig. 15
lution, 102 dB dynamic range, and anti-aliasing confirm what already observed in the first
filters) with three piezoelectric accelerometers dynamic survey: the natural frequencies of the
(WR model 731A, 10 V/g sensitivity, and global modes tend to increase with increased
0.50 g peak). The response of the tower is temperature almost linearly, as a consequence
measured in three points, belonging to the cross of the temporary increase of the local stiffness
section at the crowning level of the tower. Fur- due to the thermal expansion of materials.
thermore, a temperature sensor is installed on the The time evolution of the natural frequency of
S-W front, measuring the outdoor temperature. the upper mode, i.e., the local mode L10
The digitized data are transmitted to an indus- (Fig. 13e), deserves some concern because the
trial PC on site. A binary file, containing three trend of this modal frequency is very different
acceleration time series (sampled at 200 Hz) and from the others (Fig. 14b). More specifically, the
the temperature data, is created every hour, stored modal frequency exhibits more significant fluctu-
on the local PC, and transmitted to Politecnico di ations and clearly decreases in time, from an
Milano for subsequent data processing. initial value of about 10.0 Hz (07/01/2013) to
The continuous dynamic monitoring system a final value of about 9.0 Hz at the end of the
has been active since December 2012. The data analyzed period (06/04/2013).
files received from the monitoring system are A better inspection of Fig. 14b reveals that two
managed by a software developed in LabVIEW clear drops of the modal frequency took place:
and including the following tasks: (a) creation of (a) between 03/02/2013 and 04/02/2013 and
a database with the original data (in compact (b) between 14/03/2013 and 15/03/2013. These
format) for later developments; (b) preliminary drops divide the analyzed time period in three
preprocessing (i.e., de-trending, automatic recog- parts that are also easily identified by plotting the
nition and extraction of possible seismic events, modal frequency versus the measured outdoor
creation of one dataset per hour); (c) statistical temperature, as shown in Fig. 16. The
analysis of data, including the evaluation of aver- temperature-frequency plot of Fig. 16 highlights
aged acceleration amplitudes and temperature that the clouds of temperature-frequency points
trends; (d) low-pass filtering and decimation of corresponding to each of the three different periods
the each dataset; and (e) creation of a second are characterized by similar slope of the best fit
Ambient Vibration Testing of Cultural Heritage Structures 51
Ambient Vibration Testing of Cultural Heritage Structures, Fig. 14 Time evolution of (a) the outdoor temper-
ature measured on the S-W front, (b) the natural frequencies identified with the SSI technique
line, whereas the average frequency value signifi- frequencies, mode shapes, and damping ratios
cantly decreases. This behavior suggests the quick from ambient vibration data) to Cultural Heritage
progress of a damage mechanism, conceivably structures. Although the topic is rather recent, it is
related to the worsening of the connection between emerging as a subject of great importance in the
the wooden roof and the masonry walls, and modern approach to preservation of historic
confirms – once more – the poor structural condi- structures.
tion and the high seismic vulnerability of the upper AVT is a fully nondestructive test, especially
part of the tower, highlighting the urgent need for suitable to historic structures because it is
preservation actions to be performed. performed by just measuring the response in
operational conditions. In turn, the knowledge
of global parameters of the building, such as the
Summary modal parameters, provides essential information
to validate the numerical models currently used
This entry focuses on the application of ambient to quantitatively estimate the structural safety of
vibration testing (AVT) and operational modal Cultural Heritage structures or to design repair
analysis (OMA, i.e., the identification of natural interventions.
52 Ambient Vibration Testing of Cultural Heritage Structures
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 15 Correlation
between the identified
modal frequencies and the
outdoor temperature (S-W
front): (a) mode B1, (b)
mode B2, (c) mode B3, (d)
mode T1
Ambient Vibration
Testing of Cultural
Heritage Structures,
Fig. 16 Natural frequency
of local mode L1’ plotted
with respect to the outdoor
temperature
Ambient Vibration Testing of Cultural Heritage Structures 53
Furthermore, in some cases, AVT can possibly of temperature on the natural frequencies (i.e., the
help also to limit the number of on-site and lab- modal frequencies increase with increased tem-
oratory tests of materials, which are time- perature); and (c) the quick progress of a damage A
consuming and cost-ineffective. Other possible mechanism, involving the upper part of the tower,
applications of ambient vibration-based modal which is clearly identified through the remarkable
analysis in the field of historic structures are for fluctuations and the significant decrease (about
evaluation of the effects of repair interventions or 10 % in 3 months) of the natural frequency
in order to perform dynamics-based damage corresponding to a local mode.
assessment.
In the first part of this entry, some robust
techniques to perform OMA are reviewed, Cross-References
namely, the classical peak picking method
and the more recent frequency domain decom- ▶ Bayesian Operational Modal Analysis
position and data-driven stochastic subspace ▶ Masonry Structures: Overview
identification. ▶ Operational Modal Analysis in Civil
Subsequently, the application of these tech- Engineering: An Overview
niques is exemplified with reference to two his- ▶ Seismic Analysis of Masonry Buildings:
toric towers. In both cases, notwithstanding the Numerical Modeling
very low level of ambient vibrations that existed ▶ Seismic Vulnerability Assessment: Masonry
during the tests, AVT and OMA have proved to Structures
be effective tools for identifying the dynamic ▶ Stochastic Structural Identification from
characteristics of key vibration modes, provided Vibrational and Environmental Data
that appropriate and very sensitive acquisition
chain (capable of capturing the “interesting”
dynamics embedded in the noise) is used in the References
tests.
In the first case study (the bell tower of the Allemang RJ, Brown DL (1983) Correlation coefficient
church Chiesa Collegiata in Arcisate, Varese), for modal vector analysis. In: Proceedings of the 1st
a rational vibration-based methodology, devel- international modal analysis conference (IMAC-I),
Orlando, FL, USA
oped for the calibration of the numerical model, Aras F, Krstevska L, Altay G, Tashkov L (2011) Experi-
has been discussed as well. The presented mental and numerical modal analyses of a historical
methodology – involving systematic manual masonry palace. Constr Build Mater 25(1):81–91
tuning, sensitivity analysis, and a simple system Bendat JS, Piersol AG (1993) Engineering applications of
correlation and spectral analysis. Wiley Interscience,
identification algorithm – provided a linear elas- New York
tic model of the tower, representing an excellent Bennati S, Nardini L, Salvatore W (2005) Dynamic
approximation of the structure in its present con- behavior of a medieval masonry bell tower. II: mea-
dition (i.e., the calibrated model summarizes all surement and modeling of the tower motion. J Struct
Eng ASCE 131(11):1656–1664
the collected documentary and geometric and Binda L, Saisi A, Tiraboschi C (2000) Investigation pro-
experimental information and exhibits very cedures for the diagnosis of historic masonries. Constr
good agreement between predicted and measured Build Mater 14(4):199–233
modal parameters). Binda L, Condoleo P, Tiraboschi C, Rigamonti P (2012)
On-site investigation and crack monitoring of an
In the second case study, typical results from ancient bell-tower. In: Proceedings of the 14th inter-
long-term monitoring of the ambient vibration national conference on structural faults & repair
data collected on the Gabbia Tower in Mantua, (SF&R-2012), Edinburgh, Scotland
Italy (twelfth century), are summarized. The Brincker R, Zhang LM, Andersen P (2000) Modal identi-
fication from ambient responses using frequency
results clearly demonstrate that the modal identi- domain decomposition. In: Proceedings of the 18th
fication is capable of highlighting (a) the effects international modal analysis conference (IMAC-
of slight structural modifications; (b) the impact XVIII), San Antonio, TX, USA
54 Analysis and Design Issues of Geotechnical Systems: Flexible Walls
Brincker R, Ventura CE, Andersen P (2001) Damping Peña F, Lourenço PB, Mendes N, Oliveira DV
estimation by frequency domain decomposition. In: (2010) Numerical models for the seismic assessment
Proceedings of the 19th international modal analysis of an old masonry tower. Eng Struct 32(5):1466–1478
conference (IMAC-XIX), Orlando, FL, USA Ramos LF, Marques L, Lourenço PB, De Roeck G,
Cantieni R (2005) Experimental methods used in system Campos-Costa A, Roque J (2010) Monitoring histori-
identification of civil engineering structures. In: Pro- cal masonry structures with operational modal analy-
ceedings of the 1st international operational modal anal- sis: two case studies. Mech Syst Signal Process
ysis conference (IOMAC-2005), Copenaghen, Denmark 24(5):1291–1305
Casarin F, Modena C (2008) Seismic assessment of com- Saisi A, Gentile C, Cantini L (2013) Post-earthquake
plex historical buildings: application to Reggio Emilia assessment of a masonry tower by on-site inspection
Cathedral, Italy. Int J Archit Herit 2(3):304–327 and operational modal testing. In: Proceedings
Douglas BM, Reid WH (1982) Dynamic tests and system COMPDYN 2013, Kos, Greece
identification of bridges. J Struct Div ASCE SVS (2012) ARTeMIS extractor 2012. http://www.svibs.
108(10):295–312 com
DPCM (2011) Valutazione e riduzione del rischio sismico van Overschee P, De Moor B (1996) Subspace identifica-
del patrimonio culturale con riferimento alle NTC tion for linear systems: theory, implementation, appli-
14 gennaio 2008. G.U.R.I, Rome cations. Kluwer, Boston/London/Dordrecht
Felber AJ (1993) Development of a hybrid bridge evalua- Welch PD (1967) The use of fast Fourier transform for the
tion system. PhD thesis, University of British Columbia estimation of power spectra: a method based on time
Friswell M, Mottershead JE (1995) Finite element model averaging over short modified periodograms. IEEE
updating in structural dynamics. Kluwer, Boston/Lon- Trans Audio Electro-Acoust 15(2):70–73
don/Dordrecht
Gattulli V, Antonacci E, Vestroni F (2013) Field observa-
tions and failure analysis of the Basilica S. Maria di
Collemaggio after the 2009 L’Aquila earthquake. Eng
Fail Anal 34(X):715–734
Gentile C, Saisi A (2007) Ambient vibration testing of
Analysis and Design Issues of
historic masonry towers for structural identification Geotechnical Systems: Flexible Walls
and damage assessment. Constr Build Mater 21(6):
1311–1321 Amir M. Halabian
Gentile C, Saisi A (2013) Operational modal testing of
Department of Civil Engineering, Isfahan
historic structures at different levels of excitation.
Constr Build Mater 48:1273–1285 University of Technology, Isfahan, Iran
Gentile C, Saisi A, Cabboi A (2012) Dynamic monitoring
of a masonry tower. In: Proceedings of the 8th inter-
national conference on structural analysis of historical
constructions (SAHC-2012), Wroclaw, Poland
Synonyms
Giuffrè A (1993) Sicurezza e conservazione di centri
storici in area sismica: il caso Ortigia. Laterza, Bari Non-yielding walls; Retaining flexible walls; Soil
Ivorra S, Pallares FJ (2006) Dynamic investigation on thrust
a masonry bell tower. Eng Struct 25(5):660–667
Jaishi B, Ren WX, Zong ZH, Maskey PN (2003) Dynamic
and seismic performance of old multi-tiered temples in
Nepal. Eng Struct 25(14):1827–1839 Introduction
Magalhães F, Cunha A (2011) Explaining operational
modal analysis with data from an arch bridge. Mech
The effect of ground motion on rigid retaining
Syst Signal Process 25(5):1431–1450
Maia NMM, Silva JMM (1997) Theoretical and experi- walls is examined using the method proposed by
mental modal analysis. Research Studies Press, Mononobe and Okabe, which is based on the
Baldock, Hertfordshire, England Coulomb’s theory of static soil pressure. The anal-
Oliveira CS, Çakti E, Stengel D, Branco M (2012) Minaret
ysis and design issues for these walls are covered
behavior under earthquake loading: the case of histor-
ical Istanbul. Earthq Eng Struct Dyn 41:19–39 in a separate companion article. M-O method
Pau A, Vestroni F (2008) Vibration analysis and dynamic requires that the retaining walls can move freely
characterization of the Colosseum. Struct Control (slide or rotate) so that active or passive earth
Health Monit 15(8):1105–1121
pressures develop behind the wall. Nevertheless,
Peeters B (2000) System identification and damage detec-
tion in civil engineering structures. PhD thesis, there are many cases (such as basement walls)
Katholieke Universiteit Leuven where the free movement of the wall is fully or
Analysis and Design Issues of Geotechnical Systems: Flexible Walls 55
H Rigid Wall
Rigid Wall u=0
τxy=0
u=0
v=0
Rigid Boundary x, u
L
partially restrained, referred as non-yielding walls; by partial height earth-retaining concrete walls
furthermore, in situ retaining wall systems made and there were windows between the top of the
from sequential excavation process provide stabil- earth-retaining walls and the beams at the top of
ity and minimize movements of the adjacent the basement.
ground throughout their flexible facing. In the
last decades, a great deal of research work both Wood’s Method
in the analytical and in experimental areas has M-O method requires that the retaining walls can
been performed to evaluate the adequacy of the move freely (slide or rotate) so that active or
M-O method or to extend the method for specific passive earth pressures develop behind the wall.
applications such as non-yielding walls or flexible Nevertheless, there are many cases where the free
retaining structures. Discussion of the existing movement of the wall is fully or partially
approaches to evaluate the seismic soil pressure restrained (such as basement walls, massive grav-
on non-yielding walls and practical design ity walls embedded in rock-like formations, and
methods based on limit equilibrium concept and braced walls). Wood (1973) analyzed the
numerical algorithms are described below. response of a homogenous linear elastic soil
trapped between two rigid walls connected to
rigid base (Fig. 1). In this figure, u, v, and txy
Non-Yielding Walls are referred to the in-plane deformations and
shear stresses, respectively.
Although there are reports of damage and failure If both walls are spaced far apart, the pressures
on retaining walls due to earthquakes around the on one wall are not influenced by the presence of
world, the distress has been attributed to some the other. For low-frequency input motions with
form of soil or foundation failure, such as slope frequency less than half the fundamental fre-
instability or soil liquefaction (Al Atik and Sitar quency of the unrestrained backfill, Vs/4H (Vs is
2010). There have been a few reports of damage the soil shear wave velocity), the pseudo static
to building basement walls as a result of seismic conditions are governed (i.e., the dynamic ampli-
earth pressures in recent earthquakes including fication is negligible). For this range of frequen-
damages to basements in two recent earthquakes cies, wall pressures with plane strain assumption
in Turkey. Gur et al. (2009) reported that base- can be obtained from elastic solution for the case
ment damage occurred in a half-buried basement of a uniform, constant horizontal acceleration
of a school building during 1999 D€uzce earth- applied throughout the soil. The dynamic earth
quake; the half-buried basement was surrounded pressures obtained from this method must be
56 Analysis and Design Issues of Geotechnical Systems: Flexible Walls
L/H-1 L/H-10
υ-0.5 υ-0.4 υ-0.3 υ-0.2
1.0 1.0
0.8 0.8
HEIGHT Y/H
HEIGHT Y/H
0.6 0.6
υ-0.3
υ-0.2
0.4 0.4
υ-0.5
0.2 0.2
υ-0.4
0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.25 0.50 0.75 1.00 1.25 1.50
DIMENSIONESS NORMAL STRESS σ°x /γ H DIMENSIONESS NORMAL STRESS σ°x /γ H
Analysis and Design Issues of Geotechnical Systems: Flexible Walls, Fig. 2 Normal wall stresses from Wood’s
Solution (From Ebeling et al. 1992)
added to the static earth pressures to obtain the where ah is the amplitude of the harmonic base
total earth pressures during an earthquake. acceleration, and Fp and Fm are the dimensionless
Figure 2 presents two examples of the variation dynamic thrust and moment factors which can be
in the values for normalized horizontal stresses obtained from Fig. 3. The point of application the
with normalized elevations above the base of a dynamic thrust will be located at a height
wall, based on Wood’s solution (Wood 1973). An
L/H value of 1 (where L = the width of backfill DMeq
and H = the wall height) where value of 1 corre- heq ¼ (3)
DFeq
sponds to a narrow backfill placed within rigid
containment, and an L/H value of 10 represents a
Ostadan-White Approach
backfill of great width. The horizontal stresses
The solution by Wood (1973) commonly used
(sxO) at any vertical location Y, along the wall,
for critical facilities are, in fact, based on static
are normalized by the product of gH in Fig. 2,
“1g” loading of the soil-wall system and does not
where g is the soil unit weight.
include the wave propagation and amplification
With the assumption of smooth rigid walls, the
of motion. On the other hand, Wood’s solution is
dynamic thrust and dynamic overturning moment
mathematically complicated to apply in engi-
about the base of the wall are expressed as:
neering practice and is limited to harmonic
ah input motions. Employing the finite element
DFeq ¼ g H 2 Fp (1) technique, in a simplified method proposed by
g
Ostadan and White (1997) and Ostadan (2004)
which incorporates the main parameters affect-
ah
DMeq ¼ g H 3 Fm (2) ing the seismic soil pressure for buildings, the
g lateral seismic soil pressure on these structures
Analysis and Design Issues of Geotechnical Systems: Flexible Walls 57
15000
10000
5000
0
0 5 10 15 20
Frequency (Hz)
b 4.5
4
Vs = 1000 ft/s
3
Vs = 1500 ft/s
2.5 Vs = 2000 ft/s
1.5
0.5
0
0 0.25 0.5 0.75 1 1.25 1.5
Normalized Frequency
Analysis and Design Issues of Geotechnical Systems: Flexible Walls 59
0 0.2 0.4 0.6 0.8 1 1.2 the system and the input motion to the SDOF
1
system, the relationship in the form proposed by
M-O Method
Veletsos and Younan (1994) can be used to A
0.9 Wood, v=1/3
compute the total mass and the damping of the
Proposed Method
0.8
SDOF system. Therefore, the total mass is
obtained from
0.7
m ¼ 0:5r H 2 Cv (6)
0.6
0.5
take into account the Poisson’s ratio effect. Since
0.4 the original study performed for Poisson’s ratio
equals 1/3 (Ostadan (2004)), the pressure distri-
0.3 bution should be adjusted for different Poisson’s
ratios using a factor recommended by Veletsos
0.2 et al., Cn, defined as
0.1 2
Cv ¼ (7)
0
½ð2 nÞð1 nÞ0:5
Analysis and Design Issues of Geotechnical Systems: Study of the soil pressure transfer functions and
Flexible Walls, Fig. 6 Normalized pressure distributions
the free-field response motions for wall with
(Ostadan (2004))
height of 50 ft showed that spectral values at
the soil column frequency and at 30 % damping
have the best correlation with the forces com-
normalized height, y = Y/H (Y coordinate
defined in Fig. 4), as: puted directly from the SSI analysis (Ostadan
(2004)).
The computational steps of the proposed
pðyÞ ¼ 0:0015 þ 5:05y þ 15:84y2
method by Ostadan (2004) to evaluate the lateral
þ 28:25y3 þ 24:59y4 þ 8:14y5 (5) seismic soil pressures applied to the basement
walls as a type of the non-yielding wall can be
The total thrust applied to the wall is summarized as follows:
corresponding to the area under the normalized
pressure profile curve which can be obtained – Perform seismic free-field soil column analy-
from the integration of the pressure distribution sis and obtain the acceleration response spec-
over the height of the wall. The total area is trum at the base of the wall using a free-field
0.744H for a wall with height H. Keeping in analysis code such as SHAKE91 (Schnabel et
mind that the normalized shape of the pressure al. 1972) taking into account soil nonlinear-
distribution is similar to the response of SDOF ities with input motion specified at the ground
systems, the amplitudes of the seismic pressure surface or at the depth of foundation base mat,
can be obtained from the response spectrum Sa(free).
analysis of SDOF systems. Accordingly, the – Obtain total mass in the representative SDOF
total thrust applied to the wall is subsequently using Eq. 6.
obtained from the product of the total mass in the – Obtain the total seismic lateral force by mul-
equivalent SDOF system times the acceleration tiplying the mass of the representative SDOF
spectral value at the respective frequency of the by the spectral amplitude of the free-field
system. Having the total thrust, the frequency of response at the soil column frequency.
60 Analysis and Design Issues of Geotechnical Systems: Flexible Walls
Analysis and Design Issues of Geotechnical Systems: Flexible Walls, Fig. 7 Dynamic soil thrust acting on soil-
reinforced walls’ evaluation using different pseudostatic approaches (the symbols are defined in Table 1)
Analysis and Design Issues of Geotechnical Systems: which could be attributed to the intrinsic flexibil-
Flexible Walls, Table 1 Inertia and dynamic forces ity of soil-nailed wall system and possibly some
applied to the reinforced soil mass
level of conservatism in current design proce-
Dynamic
dures (Choukeir et al. (1997)). Accordingly, as
Design Inertial force force
method S/H Fi FD an extension of static limit equilibrium methods,
Seed and 0.50 0.5 mgH2 (3/8) mgH2 pseudostatic approaches in which dynamic earth
Mitchell pressure is computed based on conventional
(1981)
Mononobe-Okabe approach were developed.
Dhouib 0.30 + m/2 ð0:30þm=2ÞmgH 2 –
(1987)
4ð2Kþ3Þ Choukeir et al. (1997) presented a new
Se grestin 0.30 0.20 mgH2 l(3/8) pseudostatic method to analyze soil-nailed
and Bastick mgH2 slopes. Their proposed approach is derived as an
(1988) extension of the kinematical working-stress
K = 2.5 m, l = 0.6 design method developed by Juran et al. (1990)
for static loading of soil-nailed structures. The
dynamic-loading conditions. On the other hand, applicability of this working-stress design
unlike common gravity retaining structures, soil- method under static loading has been described
nailed walls have also shown remarkably well in details by Juran and Elias (1991) and is incor-
performance during many ground motions, porated in several design codes (FHWA 2003).
62 Analysis and Design Issues of Geotechnical Systems: Flexible Walls
Analysis and Design Issues of Geotechnical Systems: Flexible Walls, Fig. 8 Pseudostatic method suggested by
Choukeir et al. (1997), Hb = the projected height of the inclined wall
The main design assumptions of the kinematical mass of the wedge multiplied by the
analysis are as follows: pseudostatic horizontal acceleration ah = mg.
– Shear modulus of the soil is constant along the
– Failure occurs by a quasi-rigid body rotation depth.
of the active zone, which is limited by a – Shear strength parameters of the soil are con-
log-spiral failure surface. stant during earthquake loading.
– The locus of the maximum tension and shear – The soil is not saturated, and thus no effect of
forces at failure coincides with the failure sur- pore water pressure is present.
face developed in the soil mass. – The failure-reinforced wedge is assumed to be
– The shearing resistance of the soil, defined by subjected by its own weight, W, and the earth-
Coulomb’s failure criterion, is entirely mobi- quake loading from unreinforced soil mass
lized along the sliding surface. represented pseudostatically by dynamic soil
– The shearing resistance of stiff inclusions, thrust, FD, and its own inertial force, Fi
defined by Tresca’s failure criterion, is mobi- (Fig. 8)
lized in the direction of the sliding surface in
the soil. aps
K h ¼ n:m
,
,
n ¼
– The horizontal components of the inter-slice am
forces, Eh (see Fig. 8), are equal. 1
0:5 2 o
– The effect of a slope (or horizontal surcharge), ¼
,
r ¼ (10)
at the upper surface of the nailed soil mass, on 1r 2 on
the tension forces in the nails is linearly
Do check that all symbols in Eq. 10 have been
decreasing along the failure surface.
defined somewhere else.
18m
27m
b c
head (m)
during the harmonic base
excitation (Halabian
et al. 2010)
0.005
0
0 1 2 3 4 5 6 7 8 9 10
–0.005
Time (s)
path induced to soil mass during the earthquake wall crest decreases. Figure 11a indicates that
loading. After static equilibrium is achieved, the while the input frequency motion increases, the
full width of soil subgrade should be subjected to peak outward horizontal displacement of the wall
dynamic excitations. decreases. Conversely, as illustrated by Fig. 11b,
the peak horizontal acceleration response along
Harmonic Ground Excitation the wall facing is increased by an ascending trend
Despite having a single-frequency content and of input motion frequency. From the other results
short duration of the excitation, Halabian shown in Fig. 12, it is worth noting that increas-
et al. (2010) used a variable-amplitude harmonic ing input motion frequency made tensile forces to
ground motion to understand the effects of crucial less mobilize along nail bars due to the fact that
parameters on the performance of soil-nailed the critical state becomes more inaccessible since
structure under actual earthquake acceleration the input motion frequency gets higher than fun-
records. The accelerogram follows a trend damental frequency of the system.
which has both increasing and decaying peak
acceleration parts expressed as: Influence of Angle of Nail Inclination The
influence of nail inclination angle on dynamic
€ 1 2p response of the structure is shown in Fig. 13a.
U ðtÞ ¼ 1 cos t sin ð2pf tÞ (11)
2 T The peak outward horizontal displacement of
walls shows decreasing in value by increasing
in which T and f are the duration of net harmonic nails inclination angle while it varies in the
excitation and the frequency of excitation. range of 0–15
. Afterwards, increasing nail incli-
nation angle results in slight increase of wall
Influence of Harmonic Excitation deformations, especially when it ranges in
Frequency The effect of the excitation fre- 25–30
. As a result, it is shown that the increase
quency on the horizontal displacement time his- in nail inclination angle would have destabilizing
tories at the crest of the wall is presented in effect on the performance of structure during
Fig. 10. The displacement time histories indicate harmonic base excitations.
that the horizontal displacement at the crest of the When it comes to the peak horizontal accelera-
wall was influenced by the excitation frequency tion, as it is demonstrated in Fig. 13b, it can be
and increases monotonically with the time during concluded that increasing nail inclination angle
the harmonic base excitation. However, it can be would result in increase of peak horizontal acceler-
concluded that as the excitation frequency ation response of the wall. Figure 14 shows distri-
increases, the horizontal displacement at the bution of maximum tensile forces along nail bars.
Analysis and Design Issues of Geotechnical Systems: Flexible Walls 65
0.5
ω=2
ω=4
0.75
ω=6
ω=8
1
0 0.1 0.2 0.3 0.4 0.5
Peak Acceleration Response (m/s2)
0.75
1
0 10 20 30 40 50
Maximum Tensile Force (kN)
It is of interest to note that tensile forces along nail horizontal acceleration response of the wall
bars increase for the upper nail bars as their incli- would be in the range of 7.5–10 m. Therefore,
nation angle increases. there would be no considerable influence as the
nail length increases beyond the value of 12 m.
Influence of Nail Length The effect of nail The results of the maximum nails tensile forces
length on the peak outward horizontal displace- (Fig. 16) indicate that there would be no signifi-
ment of the wall is shown in Fig. 15a. It can be cant differences in the mobilized maximum ten-
noted that the peak outward horizontal displace- sile forces along nail bars (excluding the third
ment of the wall decreases as the nails length nail) as the nails length increases.
increases; while, the peak horizontal acceleration
along the wall facing has been greater for longer Influence of Soil Strength Properties The soil
nail bars (Fig. 15b). However, the major influ- strength properties, soil friction angle, f, and
ence of nail length on increasing the peak soil cohesion, c, can also influence on the
66 Analysis and Design Issues of Geotechnical Systems: Flexible Walls
b
0
Normalized Wall Depth
i=0
i=5
0.25 i=10
i=20
0.5
0.75
1
0 0.02 0.04 0.06 0.08
Peak Acceleration Response (m/s2)
i=0
0.75 i=5
i=10
i=20
1
0 20 40 60 80
Maximum Tensile force (kN)
wall performance such as horizontal deflections dynamic loads due to harmonic base excita-
along the wall facing and the maximum tions. As illustrated in Fig. 17a and b, peak
mobilized tensile forces along the nail bars dur- horizontal displacements along wall facing
ing an earthquake. Global performances of soil- decrease as the soil cohesion and internal fric-
nailed walls during earthquakes have shown tion angle increase. Figure 17c and d show that
that increasing in soil strength properties increasing soil strength properties would result
would improve soil shear strength so that in reduction of maximum tensile forces along
soil-nailed wall would better resist against nail bars.
Analysis and Design Issues of Geotechnical Systems: Flexible Walls 67
b 0
0.5
L=7.5 m
0.75 L=9.0 m
L=10.5 m
L=12.0 m
1
0 0.02 0.04 0.06 0.08
Peak Horizontal Acceleration Response (/g)
0.75
1
0 20 40 60 80
Maximum Tensile force (kN)
a 0
c
0
Normalized Wall Depth
0.5 0.5
C=5
C=5
C=10
0.75 0.75 C=10
C=15 C=15
C=20 C=20
1 1
0 0.002 0.004 0.006 0.008 0.01 0 20 40 60 80
Peak Horizontal Displacement (m) Maximum Tensile Force (kN)
b d
0 0
0.25 0.25
0.5 0.5
ϕ=28 ϕ=28
ϕ=32 ϕ=32
0.75 0.75
ϕ=36 ϕ=36
ϕ=40 ϕ=40
1 1
0 0.002 0.004 0.006 0.008 0.01 0 20 40 60 80
Peak Horizontal Displacement (m) Maximum Tensile force (kN)
Analysis and Design Issues of Geotechnical Systems: Normalized peak horizontal displacement (b) Normalized
Flexible Walls, Fig. 17 Normalized peak horizontal dis- peak horizontal displacement (c) Maximum tensile forces
placement and maximum tensile forces distribution along distribution along height of wall facing (d) Maximum
height of wall facing (Halabian et al. 2010) (a) tensile forces distribution
b 0
El-Centro
Kobe
0.25 Northridge
d/H
0.5
0.75
1
0 20 40 60 80 100 120 140
Maximum Tensile force (kN)
Analysis and Design Issues of Geotechnical Systems: Flexible Walls 69
d/H
height of wall (Halabian
et al. 2010) 0.6
0.8
1
0 20 40 60 80 100
Lateral Pressure behind Wall Facing (kPa)
Ostadan F, White WH (1997) Lateral seismic soil presented to compute the distribution of seismic
pressure-an updated approach. Bechtel Technical active earth pressure on a rigid retaining walls
Grant Report, Bechtel Corporation, San Francisco
Schnabel PB, Lysmer J, Seed HB (1972) SHAKE – a supporting cohesionless backfill in more realistic
computer program for earthquake response analysis manner by considering time and phase difference
of horizontally layered sites. Earthquake Engineering within the backfill.
Research Center, University of California, Berkeley,
Report no. EERC 72–12, Dec
Seed HB, Mitchell JK (1981) Earthquake resistant design
of reinforced earth walls. International study for the Types of Retaining Walls
reinforced earth company, Progress report, Berkeley
Segrestin P, Bastick MJ (1988) Seismic design of Earth retaining structures are structures that can
reinforced earth retaining walls. In: Proceedings of
the international geotechnical symposium on theory be used to support backfill along a slope or to
and practice of earth reinforcement, Fukuok Kyushu, support an excavation as illustrated in Fig. 1.
pp 577–582 Various types of retaining structures are
Veletsos A, Younan AH (1994) Dynamic soil pressure on adopted for different applications. Over the
rigid vertical walls. Earthq Eng Soil Dyn 23:275–301
Wood J (1973) Earthquake-induced soil pressures on time, the classical gravity rigid retaining walls
structures. Report EERL 73–05, California Institute evolved into reinforced concrete cantilever
of Technology, Pasadena walls (e.g., sheet piles), with or without buttresses
and counter forts (Fig. 2). These were then
followed by a variety of crib- and bin-type
walls. All these walls are externally stabilized
Analysis and Design Issues of walls or conventional gravity retaining walls.
Geotechnical Systems: Rigid Walls In order to stabilize the deep excavation projects
in metropolitan areas or even slopes where sequen-
Amir M. Halabian tial construction in comparison with other common
Department of Civil Engineering, Isfahan retaining walls is beneficial, the flexible retaining
University of Technology, Isfahan, Iran structures could be considered. During the sequen-
tial excavation, an in situ flexible retaining wall
system is constructed to provide stability and to
Synonyms minimize movements of the adjacent ground. Sol-
dier piles with shotcrete lagging are being used
Pseudo-dynamic method; Rigid retaining walls; extensively as an excavation support system, par-
Yielding walls ticularly in stiff soil conditions and where ground
water ingress into the excavated area is not prob-
lematic (Tomlinson 1995). Furthermore, since the
Introduction soldier piles are not contiguous, much fewer soldier
piles often need to be driven in comparison with
Earthquakes have caused transient and perma- sheet piles, thereby yielding significant savings in
nent deformations of retaining structures in past time and cost of installation and thus allowing
severe earthquakes followed by collapse of walls excavation to commence with a minimum of lead
in some cases. Therefore, knowledge of seismic time. These systems are usually being constrained
active earth pressure behind retaining walls is against certain types of movements by the presence
very important in the design of these structures of external bracing elements such as struts (Figs. 3
in seismically active regions. This article dis- and 4). In the case of basement walls or bridge
cusses commonly used pseudo-static approaches abutments, lateral movements at the top of the
such as the Mononobe-Okabe method, which retaining structure are restrained by the presence
gives the linear distribution of seismic earth pres- of the structure they support.
sure on rigid retaining walls in an approximate Soil reinforcement in various constructions is
way. A general pseudo-dynamic method is also also used to provide the stability for excavations
Analysis and Design Issues of Geotechnical Systems: Rigid Walls 71
Soldier Pile
Soldier Pile
As excavation proceeds, wood lagging or sheet- the failure zone. The anchors are stressed up to a
ing (concrete shotcrete) is placed to retain the soil chosen load when the grout has sufficiently
between the piles. Then, the anchors are installed cured. The process continues until the final exca-
at regular intervals and grouted in a zone beyond vation level is reached. Tiebacks eliminate
Analysis and Design Issues of Geotechnical Systems: Rigid Walls 73
Analysis and Design Issues of Geotechnical Systems: Rigid Walls, Fig. 8 Deformation mode shapes for yielding
rigid retaining walls
Steedman and Zeng 1990; Richards et al. 1999; Okabe method is obtained from the equilibrium
Choudhury and Singh 2006; Ghosh 2008). of the active or passive wedges (Fig. 9a and b).
In addition to the forces that exist under static
conditions of the failure wedge in a dry, cohe-
Mononobe-Okabe Approach sionless backfill, the wedge is also subjected to
horizontal and vertical pseudo-static forces
Okabe (1926) and Mononobe and Matsuo (1929) specified as the mass of the wedge multiplied
extended Coulomb’s theory and considered seis- by pseudo-static accelerations ah = khg and
mic forces by applying earthquake loads as av = kvg. The total active and passive thrusts
pseudo-static inertial forces to the Coulomb on a rigid wall retaining a dry, cohesionless
active or passive failure wedge. This pseudo- backfill can be presented in a form similar to
static method (known as the Mononobe-Okabe what is developed for the static conditions in
method) has been widely used in practical appli- the Coulomb’s theory:
cations (Mylonakis et al. 2007) and is
recommended by several building codes and
gb H 2
guidelines (e.g., EAU 1996; FHWA 1997; PAE ¼ ð1 kv Þ KAE (2)
CHBDC 1998; ASCE 4–98 2000; FEMA 2
369 2000; PIANC 2001; EN 1997 2002;
AASHTO 2006; IBC 2006) because of its sim- gb H 2
plicity in practical applications and reasonable PPE ¼ ð1 kv Þ KPE (3)
2
predictions of the actual dynamic pressures act-
ing on walls (Seed and Whitman 1970; Whitman in which gb is the backfill unit weight, H is the
1990; Ebeling et al. 1992; Veletsos and Younan height of the wall, and PAE and PPE are the
1994; Dakoulas and Gazetas 2008). dynamic active and passive earth pressures. KAE
The dynamic lateral earth pressure on rigid and KPE are the coefficient of dynamic active and
retaining structures proposed in the Mononobe- passive earth pressures given by:
cos2 ðjb j bÞ
KAE ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 (4)
ðjb þdÞ sin ðjb jiÞ
cos j cos2 b cos ðd þ j þ bÞ 1 þ sin cos ðdþbþjÞ cos ðiþbÞ
cos2 ðjb j þ bÞ
KPE ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 (5)
ðjb þdÞ sin ðjb jþiÞ
cos j cos2 b cos ðd þ j bÞ 1 sin cos ðdbþjÞ cos ðiþbÞ
76 Analysis and Design Issues of Geotechnical Systems: Rigid Walls
a b
i i
k VW kVW
khW khW
β β
W W
PPE φb R
φb
δ δ
R
PAE
Analysis and Design Issues of Geotechnical Systems: Mononobe-Okabe analysis. (b) Forces acting on passive
Rigid Walls, Fig. 9 Mononobe-Okabe approach’s fail- wedge in Mononobe-Okabe analysis
ure wedges. (a) Forces acting on active wedge in
jb is the internal friction angle of the backfill. i component, PA or PP, and the dynamic contri-
is the slope of the backfill with respect to the bution, DPAE or DPPE:
horizontal axis. d is the friction angle between
the inner face of the wall and the backfill. b is PAE ¼ PA þ DPAE , PPE ¼ PP þ DPPE (6)
the angle between the inner face of the wall and
the vertical axis. j ¼ tan1 1k
kh
v
in which kh and According to the Coulomb theory for linear back-
kv stand for the horizontal and vertical acceler- fill surfaces with no surcharge loading, the static
ations of the soil wedge in g unit. The wall is component acts at a point located H/3 above the
unstable when kh is higher than (1kv) tan j. height of the wall. On the other hand, the dynamic
During the earthquake, kh and kv are time vary- part can be taken to act at a point approximately
ing and chaos so that the seismic earth pressures 0.6H above the base of the wall (Seed and Whit-
in both active and passive cases are not constant. man 1970). Therefore, the total lateral earth pres-
The vertical ground excitation has a great influ- sure is applied at a height:
ence on dynamic stability of rigid retaining
structures during an earthquake. Despite time- PA ðH=3Þ þ DPAE ð0:6HÞ
h¼ (7)
varying nature of ground excitations accelera- PAE
tions, the vertical accelerations of earthquakes
usually assumed to be a fraction of their hori- above the base of the wall.
zontal accelerations.
Although the Mononobe-Okabe method sug-
gests that the total dynamic lateral earth pres- Effects of Water on Dynamic Lateral
sure on rigid retaining structures should apply at Pressures
a point H/3 above the base of the wall of
height, H, experimental results demonstrated Hydrodynamic Pressure
that it actually occurs at the higher point The retaining walls in practice are usually
under dynamic loading conditions (Fig. 10). designed with drainage systems to prevent
The total dynamic lateral earth pressure can be water from building up. In these cases, the
divided into the static active or passive dynamic lateral earth pressure can be
Analysis and Design Issues of Geotechnical Systems: Rigid Walls 77
ΔPAE
PΑΕ = PΑ+ΔPAE
PΑ
0.6H
h
H/3
Pwd
Pws
H
Pwd
Pws
determined based on soil dry unit weight using excitations, besides the hydrostatic pressure,
the Mononobe-Okabe theory. However, the pws, water in front of a retaining wall will
presence of water in retaining walls in water- exert dynamic pressures, pwd, on the face of the
front areas and in backfills behind retaining wall (Fig. 11). The hydrostatic pressure is calcu-
walls can play a significant role in determining lated by
seismic loads that act on the wall during
and after earthquakes. During any ground pws ðxÞ ¼ gw x (8)
78 Analysis and Design Issues of Geotechnical Systems: Rigid Walls
The application point of the hydrostatic thrust is Analysis and Design Issues of Geotechnical Systems:
a point H/3 above the base of the wall of Rigid Walls, Fig. 12 Effect of wall inclination on hydro-
height, H. dynamic pressure
Hydrodynamic pressure results from the seis- rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mic response of the water mass in front of the x x
x x
pwd ðx, aÞ ¼ Cm ðaÞ kh gw H 2 þ 2
wall and is usually estimated using H H H H
Westergaard’s theory (Westergaard 1931). (12)
Westergaard’s theory applies for a vertical rigid
wall retaining a very large (theoretically infinite) or approximately using Westergaard’s approach
extent of water basin with no backfill horizontally can be estimated by
excited by harmonic motion of its rigid base. The
amplitude of hydrodynamic pressure is then 7 pffiffiffiffiffiffiffiffiffi
pwd ðx, aÞ ¼ Cm ðaÞ kh gw H x=H (13)
determined using Westergaard’s theory as 8
7 pffiffiffiffiffiffiffiffiffi
pwd ðxÞ ¼ kh gw H x=H (10) in which Cm(a) is the inclination factor given by
8
a
The resultant hydrodynamic force is obtained as Cm ðaÞ ¼ 2:0 (14)
p
7
According to Westergaard’s theory, the hydrody- Pwd ðaÞ ¼ Cm ðaÞ kh gw H2 ¼ ð1:17 Cm ðaÞ kh Pws Þ
12
namic thrust can be taken to act at a point approx-
(15)
imately 0.4H above the base of the wall. The
excess hydrodynamic pressures could be applied Effect of Water Basin Length
toward or outward the wall depending on the base Westergaard’s theory was developed based on
excitation direction. Thus, the direction of the this assumption that the rigid wall is retaining a
excitation should be taken into account in deter- semi-infinite reservoir of water. For the case a
mining the total hydrodynamic pressure acting on finite water basin (Fig. 13), the hydrodynamic
the wall. pressure applied to the wall can be evaluated
using Werner and Sundquist’s (1943) suggestion
Effect of Wall Inclination by intruding the water basin length modification
The hydrodynamic pressure on the walls with factor, Cn, as
facing inclination (Fig. 12) could be estimated
taking the wall inclination using Zangar (1953) 7 pffiffiffiffiffiffiffiffiffi
pwd ðxÞ ¼ Cn kh gw H x=H (16)
and Chwang (1978): 8
Analysis and Design Issues of Geotechnical Systems: Rigid Walls 79
where Cn is defined in terms of ratio of water pressures resulting from the seismic response
basin length, L, over the wall height, H, as of the water in the backfill soils could be
estimated by Westergaard’s theory modifying
the effect of the soil permeability as
4 L=H
Cn ¼ 1:0
3 1 þ L=H (17)
pffiffiffiffiffiffiffiffiffi
Cn ¼ 1:0 L=H 2:7 7
pwd ðxÞ ¼ Ce kh gw H x=H (19)
8
The total hydrodynamic thrust considering the in which Ce is the correction factor that expresses
water basin length effect acts at a point located the portion the pore water which vibrates freely
0.4H above the base of the wall of height, H with within the soil:
amplitude
2p n gw H2
7 Ce ¼ 0:5 0:5 tan h log (20)
Pwd ¼ Cn kh gw H2 ¼ ð1:17 Cn kh Pws Þ (18) 7Ew KT
12
where n is the porosity, H is the water depth, Ew is
Dynamic Pressure for Saturated Backfill the bulk modulus of water (2 106 kPa), K is the
backfill permeability, and T is the predominant
Seismic response of rigid walls retaining saturated period of the excitation (Fig. 14).
backfills could be affected by the presence of water The total hydrodynamic thrust considering
changing the inertial forces within the backfill, the water within the backfill acts at a point
developing the hydrodynamic pressures with the located 0.4H above the base of the wall of
saturated soil and generating excess pore water height, H with amplitude
pressure resulting in the cyclic deformation of
the backfill soil (Matsuo and Ohara 1965). 7
Pwd ¼ Ce kh gw H2 ¼ ð1:17 Ce kh Pws Þ (21)
The soil permeability plays an important role 12
in developing hydrodynamic pressures within
the backfill soil. For soils with small perme- Earth Lateral Pressures on Walls
ability, the pore water moves within the soil in Retaining Saturated Backfills
restrained conditions during the ground shak-
ing; while if the permeability of the backfill is For walls retaining saturated soils with restrained
very high, the soil particles move easily pore water conditions, to take into account the
through the pore water, and the pore water presence of water within the backfill, the
may remain stationary. Hydrodynamic water Mononobe-Okabe approach is modified using
80 Analysis and Design Issues of Geotechnical Systems: Rigid Walls
Cen
n Free water
(1-Ce)n
H
Trapped water
V=1
1-n
Soil skeleton
A
ΔPAE
H ΔPw PAE
δ
Pw
0.6 H
0.4 H
H/3
Analysis and Design Issues of Geotechnical Systems: Rigid Walls, Fig. 16 Different components of acting thrust
on the wall
PA and DPAE with a common unit weight (e.g., 3 gdry
EAK 2002). In this case, the buoyant unit weight, DPAE ¼ kh ðgsat gw Þ H2 (31)
8 gsat -gw
gsat gw , and a modified seismic coefficient as
In other approach, by representing the excess pore
g water pressure in the backfill using the pore pres-
k ¼k (27)
gsat gw sure ratio, ru, the active earth pressure acting on wall
with cohesionless backfill can be computed using
For impermeable soils such as clayey sand,
clayey silts, and clayey gravels, Ce = 0, and pAE ¼ KAE gsub ð1 ru Þ ð1 kv Þ
therefore, should be used g ¼ gsat resulting
Pwd = 0 and
gsat kh
j ¼ tan 1 (32)
gsub ð1 ru Þð1 kv Þ
1 3
DPAE ¼ kh gsat H2 (28)
2 4 Steedman-Zeng Method
Hz
Qv mðzÞ ¼ gb dz (34)
g tan a
Qh
1 sin ða jb Þ
The seismic active earth pressure coefficient, KAE ¼
KAE, is defined as tan a cos ðd þ jb aÞ
kh 2pVs cos ða jb Þ
þ 2 m1
2PAE 2p tan a oH cos ðd þ jb aÞ
KAE ¼ (37)
gb H2 kv 2pVp sin ða jb Þ
2 m2
2p tan a oH cos ðd þ jb aÞ
Substituting for Fh and Fv, KAE is derived as (38)
Analysis and Design Issues of Geotechnical Systems: Rigid Walls 83
@PAE ðtÞ gz sin ða jb Þ kh gz cos ða jb Þ z
pAE ðtÞ ¼ ¼ þ sin o t
@z tan a cos ðd þ jb aÞ tana cos
ðd þ jb aÞ Vs
(40)
kv gz sin ða jb Þ z
sin o t
tan a cos ðd þ jb aÞ Vp
C.G.
mxG H
Pwd(t)
Pws, PA
hd
θ 0.4H
0.3H
Any procedure in the performance-based displacements of the wall, the importance of the
design of retaining walls should follow these wall, the effects of failure, and the cost of repair.
steps:
Choukeir M, Juran I, Hanna S (1997) Seismic design of Matsuzawa H, Ishibashi I, Kawamura M (1985) Dynamic
reinforced-earth and soil-nailed structures. Ground soil and water pressures of submerged soils. J Geotech
Improv 1:223–238 Eng ASCE 111(10):1161–1176
Chwang AT (1978) Hydrodynamic pressures on sloping Mononobe N, Matsuo H (1929) On the determination of
dams during earthquakes. Part 2. Exact theory. J Fluid earth pressures during earthquakes. In: Proceedings of
Mech 87:343–348 world engineering congress
Dakoulas P, Gazetas G (2008) Insight into seismic earth Mylonakis G, Kloukinas P, Papantonopoulos C (2007) An
and water pressures against caisson quay walls. alternative to the Mononobe-Okabe equations for seis-
Geophys J Roy Astron Soc 58(2):95–111 mic earth pressures. Soil Dyn Earthq Eng 27:957–969
EAK (2002) Empfehlungen des Arbeitsasschusses Okabe S (1926) General theory of earth pressure. Journal
Kustenschutzwerke. Die Kuste, H. 65 Japan Society of Civil Engineering, 12(1)
EAU (1996) Recommendations of the committee for PIANC (2001) Seismic design guidelines for port struc-
waterfront structures, harbours, and waterways, tures. A. A. Balkema, Tokyo (International Navigation
7th edn. Ernst & Sohn, Berlin Institute)
Ebeling RM, Morrison EE, Whitman RV, Liam Finn WD Richard R, Elms DG (1979) Seismic behavior of gravity
(1992) A manual for seismic design of waterfront retaining walls. J Geotech Eng, ASCE 105:449–464
retaining structures. US Army Corps of Engineers, Richards R, Huang C, Fishman KL (1999) Seismic earth
Technical report ITL-92-11 pressure on retaining structures. J Geotech Geoenviron
EN 1997 (2002) Eurocode 7 geotechnical design – Part 1: Eng, ASCE 125(9):771–778
General rules. CEN European committee for standard- Seed HB, Whitman RV (1970) Design of earth retaining
ization, Bruxelles structures for dynamic loads. In Proceedings, ASCE
EUROCODE 8 (EUROPEAN PRE-STANDARD) Specialty Conference on Lateral Stresses in the Ground
(1994) Design provisions for earthquake resistance of and Design of Earth Retaining Structures, pp 103–147
structures-Part 5: Foundations, retaining structures and Steedman RS, Zeng X (1990) The influence of phase on
geotechnical aspects. The Commission of the Euro- the calculation of pseudo-static earth pressure on a
pean Communities retaining wall. Geotechnique 40:103–112
FEMA 369 (2000) The 2000 NEHRP recommended pro- Tomlinson MJ (1995) Foundation design and construc-
visions for new buildings and other structures. Part 2: tion. Wiley, New York
Commentary. Federal Emergency Management Veletsos AS, Younan AH (1994) Dynamic soil pressures
Agency on rigid retaining walls. Earthq Eng Struct Dyn
FHWA (1997) Geotechnical engineering ciruclar #3. 23:275–301
Design guidance: geotechnical earthquake engineer- Werner PW, Sundquist KJ (1943) On hydrodynamic
ing for highways, design principles. FHWA-SA-97- earthquake effects. Trans Am Geophys Union
076, Federal Highway Administration, vol I, 30(5):636–657
U.S. Department of Transportation Westergaard H (1931) Water pressure on dams during
Ghosh P (2008) Seismic active earth pressure behind a earthquakes. In: Transportations of ASCE, Paper
non-vertical retaining wall using pseudo-dynamic no. 1835, pp 418–433
analysis. Can Geotech J 45(1):117–123 Whitman RV (1990) Seismic design and behavior of gravity
Horvath JS (2005) Integral-abutment bridges: geotechni- retaining walls. In: Proceedings conference on design
cal problems and solutions using geosynthetics and and performance of earth retaining structures, ASCE
ground improvement. In: Proceeding IAJB 2005: the geotechnical special publication, vol 25. pp 817–842
2005 FHWA conference on integral abutment and Whitman RV, Liao S (1985) Seismic design of retaining
jointless bridges, Baltimore, pp 281–291 walls, paper GL-85-1, U.S. Army engineer waterways
Huang C (2005) Seismic displacement of soil retaining experiment station, Vicksburg
walls situated on slope. J Geotech Geoenviron Eng Wood J (1973) Earthquake-induced soil pressures on
131(9):1108–1117 structures. Report EERL 73–05, California Institute
International Code Council (2006) IBC (International of Technology, Pasadena
Building Code). International Code Council, Country Wu Y, Prakash S (1996) On Seismic Displacement of Rigid
Club Hills, 664 pp retaining Walls. In: Prakash S (ed) ASCE Geotechnical
Japan Railway Technical Research Institute (JRTRI) special publication: Analysis and Design of Retaining
(1999) Design guidelines for railway structures- Structures Against Earthquakes, SP editor, pp 21–37
aseismic design. Maruzen Co. Ltd Wu Y, Prakash S (2001) Seismic displacements of rigid
Kramer S (1996) Geotechnical earthquake engineering. retaining walls. In: State-of-the-art paper number
Prentice-Hall, Upper Saddle River 705, proceedings fourth international conference on
Matsuo H, Ohara S (1965) Dynamic pore water recent advances in geotechnical earthquake engineer-
pressure acting on quay walls during earth- ing and soil dynamics, San Diego
quakes. In: Proceedings of the third world conference Zangar CN (1953) Hydrodynamic pressures on dams due
on earthquake engineering, New Zealand, vol 1, to horizontal earthquakes. Proc Soc Exp Stress Anal
pp 130–140 10:93–102
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Dynamic Procedures 87
For modern ductile structures or non-ductile record set that is scaled to several levels of inten-
buildings that are not first-mode dominated sity (typically by multiplying all natural
(e.g., plan-asymmetric or tall structures) or accelerograms by the same scale factor). At the
whenever the site of interest is subject to near- other end, scaling can be completely avoided
field motions, a strong set of records together (if possible, considering catalogue limitations)
with an improved IM (see earlier discussion) by using a separate set of natural records for
should be preferred. each level of IM (within some tolerance of
course). Extracting the P(EDP|IM) information
can be done in several ways. One approach is to
Analysis Strategies employ parametric regression, taking care to sep-
arately fit the non-collapsing IM-EDP points via a
Determination of fragility necessitates a wide- linear regression in logarithmic space. “Infinite
range assessment of structural response at multi- EDP” points indicative of collapse should be
ple levels of intensity. There are several ways to accounted for using logistic regression (see
organize the execution of nonlinear dynamic ana- Shome and Cornell 1999; Jalayer and Cornell
lyses, mainly differing in the manner of 2009). Alternatively, a nonparametric approach
postprocessing and in how they can employ can be used, employing, e.g., “running” 16/50/84
ground motion selection and scaling. The main percentiles to estimate the median and dispersion
candidates are three (Jalayer and Cornell 2009): of the EDP response at each level of the IM.
cloud analysis, stripe analysis, and incremental Stripe analysis is named after the characteris-
dynamic analysis (IDA). tic stripes of points aligned in distinct rows at
Of the three, cloud analysis is the least restric- different IM levels (Fig. 2). Like cloud analysis,
tive. Its name is derived from the characteristic it may involve any number of sets of records, and
cloud of results that appears in the IM versus EDP it also needs to employ at least some scaling to
plane, where each point corresponds to one anal- make sure that all runs at a given IM level actu-
ysis (see Fig. 1). One can employ any combina- ally perfectly correspond to the IM level
tion of scaling and record set selection. At one requested, without any tolerance. The importance
end, one can employ only scaling, using a fixed of this detail is that the distribution of EDP given
90 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Dynamic Procedures
IM is directly represented at each IM level by the Incremental dynamic analysis (IDA) is based
empirical distribution of the EDP results on scaling and scaling only (Vamvatsikos and
extracted from the pertinent analyses. In other Cornell 2002). Its focus is on individual records,
words, any statistical quantity of EDP given the each of which is scaled to several levels of inten-
IM (mean, 16/50/84 percentile, standard devia- sity, typically until collapse is reached. Although
tion, etc.) can be estimated directly from the closely related to stripe analysis, its main advan-
corresponding EDP values without any need for tage is that the IM levels of the different records
parametric or nonparametric regression, thus do not need to match. Instead, a fixed number of
considerably simplifying postprocessing. nonlinear dynamic analyses can be utilized for
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Dynamic Procedures 91
each record to achieve an accurate yet economi- Dolsek 2009; Vamvatsikos and Fragiadakis
cal representation of response from elasticity to 2010; Jalayer et al. 2011).
global dynamic instability. Then, interpolation of
the results pertaining to each record is employed
to generate continuous IDA curves (Fig. 3), as Determination of Fragility
well as stripes, and subsequently derive the
needed EDP|IM statistics (Vamvatsikos and Cor- A building-level fragility function is defined as a
nell 2004). probability-valued function of the IM that repre-
In practice, any of the above approaches can sents the probability of violating a limit state
be used to derive equivalent results. The distinc- given the IM level. Formally, it may be
tion may become important only in the sense of represented as P(C < D|IM), i.e., as the proba-
scaling versus selecting ground motion records bility of the seismic demand D exceeding the
that puts IDA at a potential disadvantage: It has to seismic capacity C given the IM. Two alternative
use the same set of ground motions for all inten- yet equivalent formats exist to define the C and
sity levels, needing a sufficient IM to provide D terms, namely, the IM basis and the EDP basis.
unbiased results. On the other hand, the represen- When a single EDP is used to define the limit
tation of the EDP|IM statistics via continuous state, this dichotomy is best understood by visu-
IDA curves offers a powerful visual tool for alizing the first exceedance of the limit state as a
understanding structural performance that can single point on each IDA curve, having unique
often tip the scales in its favor. In the end, the IM and EDP coordinates (Fig. 4).
choice lies with the analyst and his/her under- If we choose to characterize demand and
standing of the problem at hand. capacity by their IM counterparts, then P
As a final remark, it is important to remember (C < D|IM) simply becomes the cumulative dis-
that the aforementioned methods can only take tribution function (CDF) of the IM-valued capac-
into account the record-to-record variability. ity, IMc. Simply plotting the empirical CDF of
If other sources of uncertainty need to be incor- the IMc values, as, for example, in Fig. 5, pro-
porated, for example, model parameter variabil- vides us with a ready estimate of the fragility
ity, then more refined (and complex) approaches function. By estimating the appropriate statistics
should be adopted instead (Liel et al. 2009; (median, IMc50, and standard deviation of the log
92 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Dynamic Procedures
values, bIMc), one may also represent the results is deterministic, i.e., we presume to perfectly
via an analytic lognormal approximation, as know the value of EDP that causes violation of
shown in Fig. 5, and formally represented by the the limit state. Then, the capacity points of Fig. 4
following expression: align themselves into a single vertical line that
intersects all IDA curves at the limiting (single)
lnIM lnIMc50 EDPc value, meaning that all the dispersion in the
PðEDPc < EDPjIMÞ ¼ F :
bIMc fragility function (or the IMc values, equiva-
(1) lently) originates from the records themselves.
Inversely, if one wants to introduce the
If instead D and C are represented in EDP terms, (typically considerable) uncertainty in EDPc,
then they are both potentially random quantities and he/she had best go through the EDP route; other-
integration is needed: wise, it is not possible to obtain the inflated bIMc
dispersion of Eq. 1. Regardless of the route taken,
ð
þ1 though, the end result can always be represented
PðEDPc < EDPjIMÞ ¼ PðEDPc < EDPÞ in terms of the lognormal function parameters
0
IMc50 and bIMc of Eq. 1.
PðEDPjIMÞ dIM:
(2)
Summary
The first term of the integrant is essentially the
complementary cumulative distribution function A key ingredient of modern frameworks for
(CCDF) of EDPc, and the second is the probabil- performance-based earthquake engineering is
ity distribution function (PDF) of EDP given the the assessment of the distribution of structural
IM. Despite the apparent complexity of this demand given the seismic intensity. Expressed
approach, it is often the preferred method for as the probability of exceeding specific levels of
estimating fragility (although not necessarily for the structural response indicative of defined limit
representing it). This is best understood by con- states, this is also known as fragility. Recent
sidering the simpler case where the EDP capacity advances in computer technology have allowed
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Dynamic Procedures 93
the use of resource-intensive nonlinear dynamic intensity measure and probabilistic assessment proce-
analysis as the tool of choice for analytically dure. Proceedings of the 2nd US-Japan Workshop on
Performance-based Earthquake Engineering Method-
estimating such fragility functions. For practi- ology for RC Building Structures, Sapporo, Hokkaido A
cally implementing such an assessment, several Cornell CA, Krawinkler H (2000) Progress and challenges
strategies exist on how to select and/or modify in seismic performance assessment. PEER Center
ground motion records and apply them to a struc- News, http://peer.berkeley.edu/news/2000spring/
index.html. Accessed May 2014
tural model at several levels of intensity. Clouds, Dolsek M (2009) Incremental dynamic analysis with con-
stripes, and incremental dynamic analysis curves, sideration of modelling uncertainties. Earthq Eng
each characterized by the appearance of the Struct Dyn 38(6):805–825
corresponding results in the response versus Iervolino I, Cornell CA (2005) Record selection for
non-linear seismic analysis of structures. Earthq Spec-
intensity space, are among the most prominent tra 21(3):685–713
techniques. Despite their superficial differences, Jalayer F, Cornell CA (2009) Alternative non-linear
they may all achieve the same fidelity in their demand estimation methods for probability-based
results, practically differing only in the ease of seismic assessments. Earthq Eng Struct Dyn
38:951–972
postprocessing and interpretation. In all cases, Jalayer F, Elefante L, Iervolino I, Manfredi G (2011)
parameterization by the (typically scalar) inten- Knowledge-based performance assessment of existing
sity measure means that considerable attention RC buildings. J Earthq Eng 15(3):362–389
should be paid to its selection, as well as the Jalayer F, Beck JL, Zareian F (2012) Analyzing the suffi-
ciency of alternative scalar and vector intensity mea-
selection of the ground motions, to ensure the sures of ground shaking based on information theory.
unbiasedness of the estimates. J Eng Mech ASCE 138(3):307–316
Katsanos EI, Sextos AG, Manolis GD (2010) Selection of
earthquake ground motion records: a state-of-the-art
Cross-References review from a structural engineering perspective. Soil
Dyn Earthq Eng 30:157–169
▶ Analytic Fragility and Limit States [P(EDP| Liel AB, Haselton CB, Deierlein GG, Baker JW
(2009) Incorporating modeling uncertainties in the
IM)]: Nonlinear Static Procedures assessment of seismic collapse risk of buildings. Struct
▶ Assessment of Existing Structures Using Saf 31(2):197–211
Response History Analysis Lin T, Haselton CB, Baker JW (2013a) Conditional-
▶ Incremental Dynamic Analysis spectrum-based ground motion selection. Part I: Haz-
ard consistency for risk-based assessments. Earthq
▶ Nonlinear Dynamic Seismic Analysis Eng Struct Dyn 42(12):1847–1865
▶ Seismic Collapse Assessment Lin T, Haselton CB, Baker JW (2013b) Conditional-
▶ Seismic Fragility Analysis spectrum-based ground motion selection. Part II:
▶ Seismic Loss Assessment Intensity-based assessments and evaluation of alterna-
tive target spectra. Earthq Eng Struct Dyn
▶ Time History Seismic Analysis 42(12):1867–1884
Luco N, Bazzurro P (2007) Does amplitude scaling of
ground motion records result in biased nonlinear struc-
References tural drift responses? Earthq Eng Struct Dyn
36:1813–1835
Baker JW, Cornell CA (2006) Spectral shape, epsilon and Luco N, Cornell CA (2007) Structure-specific scalar
record selection. Earthq Eng Struct Dyn intensity measures for near-source and ordinary earth-
35(9):1077–1095 quake ground motions. Earthq Spectra 23(2):357–392
Bianchini M, Diotallevi PP, Baker JW (2009) Predictions Shome N, Cornell CA (1999) Probabilistic seismic
of inelastic structural response using an average of demand analysis of nonlinear structures. RMS pro-
spectral accelerations. In: Proceedings of the 10th gram, report no. RMS35, PhD thesis, Stanford Univer-
international conference on structural safety & reli- sity, CA
ability (ICOSSAR), Osaka Shome N, Cornell CA, Bazzurro P, Carballo JE
Bradley BA (2010) A generalized conditional intensity (1998) Earthquakes, records and nonlinear responses.
measure approach and holistic ground motion selec- Earthq Spectra 14(3):469–500
tion. Earthq Eng Struct Dyn 39(12):1321–1342 Tothong P, Cornell CA (2006) An empirical ground-
Cordova PP, Deierlein GG, Mehanny SS, Cornell CA motion attenuation relation for inelastic spectral dis-
(2000) Development of a two-parameter seismic placement. Bull Seismol Soc Am 96(6):2146–2164
94 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
Vamvatsikos D (2013) Derivation of new SAC/FEMA adopted logic tree approach for evaluating the
performance evaluation solutions with second-order effects of modelling uncertainties on the fragility
hazard approximation. Earthq Eng Struct Dyn
42:1171–1188 curves of masonry buildings. Alternatively, the
Vamvatsikos D, Cornell CA (2002) Incremental dynamic impact of the modelling uncertainties on the fra-
analysis. Earthq Eng Struct Dyn 31(3):491–514 gility curve can be simulated by utilizing the so-
Vamvatsikos D, Cornell CA (2004) Applied incremental called probabilistic SDOF model (Kosič et al.
dynamic analysis. Earthq Spectra 20(2):523–553
Vamvatsikos D, Cornell CA (2005) Developing efficient 2014). The advantage of pushover analyses in
scalar and vector intensity measures for IDA capacity comparison with nonlinear response history anal-
estimation by incorporating elastic spectral shape ysis lies in the former’s simplicity, since the
information. Earthquake Engineering and Structural results of pushover analysis are very intuitive.
Dynamics 34(13):1573–1600
Vamvatsikos D, Fragiadakis M (2010) Incremental This makes pushover-based methods attractive
dynamic analysis for estimating seismic performance for practical applications. On the other hand,
sensitivity and uncertainty. Earthq Eng Struct Dyn they are based on several assumptions. Thus, the
39(2):141–163 analyst should have good background knowledge
in order to be able to adequately interpret results
obtained by using pushover-based methods.
However, many studies have shown that, in the
Analytic Fragility and Limit States case of simple bridges and buildings up to around
[P(EDP|IM)]: Nonlinear Static eight stories high, the mean/median response can
Procedures be sufficiently accurately predicted by means of
conventional pushover analysis. In the case of
Matjaž Dolšek more complicated structures, the analyst should
Faculty of Civil and Geodetic Engineering, think about taking into account, at least approxi-
University of Ljubljana, Ljubljana, Slovenia mately, torsional effects and the so-called higher
mode effects (e.g., Reyes and Chopra 2011;
Kreslin and Fajfar 2012). Reyes and Chopra
Synonyms (2011) and Kreslin and Fajfar (2012) used linear
elastic analysis in estimating higher mode contri-
Fragility analysis; Fragility curve; Limit state; butions seismic demands. If the target displace-
Nonlinear static procedure; Pushover analysis; ment is obtained by nonlinear response history
SDOF model; Seismic demand; Target displace- analysis utilizing the single-degree-of-freedom
ment; Uncertainty (SDOF) model, then it is possible to consider
the different system failure modes obtained
from the pushover analyses (Brozovič and Dolšek
Introduction 2014). In this case, the failure-based SDOF
models are used in addition to the modal-based
Nonlinear static (pushover-based) procedures SDOF model associated with the first model.
were initially developed for the seismic perfor- It should be noted that the main objective of
mance assessment of structures (e.g., Saiidi and this entry is to introduce the use of conventional
Sozen 1981; Fajfar and Fischinger 1988; Paret pushover-based method for fragility analysis,
et al. 1996; Chopra and Goel 2002) and later on taking into account the randomness of ground
used also for the fragility analysis (e.g., motion and the epistemic (modeling) uncertainty.
Fragiadakis and Vamvatsikos 2010; Barbato Thus, not all issues associated with pushover-
et al. 2010; Dolšek 2012; Jalayer et al. 2011; based methods are precisely addressed.
Rota et al. 2014; Kosič et al. 2014). For example, The entry has been organized into three sub-
Jalayer et. al (2011) proposed the use of Bayesian entries. The theoretical background for the pre-
method for calculating fragility curve of the diction of the parameters of SDOF models is first
reinforced concrete frame. Rota et al. (2014) described. This subentry is subdivided into two
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures 95
which produce a vector U of displacements at any displacement equal to 1, and ut is the top displace-
individual floor level, at any instant in time. ment, which is a function of time. Based on such an
Despite simplifications, Eq. 2 is a system of assumption, it becomes clear that a multi-degree-
n-dependent nonhomogenous second-order dif- of-freedom (MDOF) model of a structure can be
ferential equations with nonlinear coefficients, transformed into a single-degree-of-freedom
where n stands for the number of stories of the (SDOF) model. This is achieved by substituting
building structure. Several methods exist for the all the kinematic quantities in the form of Eq. 4 and
solution of this equation at any instant in time, but pre-multiplying Eq. 2 by the vector CT:
there are also several reasons to further simplifi-
cation of the problem. CT M C u€t þ CT C C u_t þ CT K Cut ¼ CT M 1 u€g
For example, all nonlinear structural models (5)
are just an approximation of the realistic seismic
response of a structure. The majority of current Equation 5 can be rearranged into the following
structural models are capable of simulating sev- form:
eral of the failure modes that can be observed in
realistic structures. From this point of view, the mc u€t þ cc u_t þ kc ut ¼ m u€g (6)
question arises as to whether it is justified to treat
the whole system as dynamic, if the model used to by realizing that the results of the multiplication
represent a building structure is already a proxy of the matrices (Eq. 5) are scalars:
of reality. Furthermore, although the solution of
the equation of motion will provide exact results mc 5CT M C
from the theoretical point of view, the seismic cc 5CT C C (7)
loading is not known in advance. Thus, the selec- kc 5CT K C
tion of inappropriate ground motions could have
a greater impact on the results than the selected m 5CT M 1 (7a)
method of analysis. Finally, solving system of
equations of motion is extremely time demanding By introducing Eq. 6, it is clear that the system of
and, in the case of complex systems, also related n-dependent differential equation has been sim-
to poor convergence of the numerical results plified to just one equation of motion, where the
obtained. These are just some reasons that it is unknown quantities are the displacement, veloc-
often not practical to use response history analy- ity, and acceleration at the top of the building.
sis in order to simulate the seismic response of a However, it still does not have the form of the
structure. Therefore, Eq. 2 is often further conventional equation of motion, since m 6¼ mc.
simplified. In order to make Eq. 6 equivalent to the conven-
The solution of Eq. 2 would be significantly tional equation of motion, a so-called transforma-
easier if the motion of the structure could be tion factor, which relates the top displacement of
described with sufficient accuracy by reducing the structure to the displacement of the SDOF
the number of differential equations to just one, model, needs to be defined as follows:
e.g., in the case of building structures, by simu-
lating just the top (roof) displacement. From the
m CT M 1
theoretical point of view, such a simplification G¼ ¼ T (8)
mc C M C
can be achieved by assuming a displacement
vector U in the following form:
and the kinematic quantities of the SDOF model
as
U ¼ C ut (4)
Equation 6 can now be transformed into the con- It should be noted that the force-displacement rela-
ventional equation of motion of the SDOF model tionship of the SDOF model is, in general, nonlinear
by taking into account Eqs. 8 and 9: due to the nonlinear response of the structure. Thus, A
there is no exact answer as to which is the most
m u€ þ c u_ þ k u ¼ m u€g (10) appropriate shape of the normalized displacement
vector C or which are the most appropriate lateral
where forces F. If it is assumed that the structure oscillates
just in one given mode shape, then C5F. If it is
c ¼ G cc (11) further assumed that the mode shape F does not
change significantly with increasing damage to the
and structure and that the damping does not affect the
shape of the buildings’ oscillation, then it can be
k ¼ G kc (11a) shown, by analogy to the modal analysis, that the
shape of the equivalent static forces which causes
Determination of the Parameters of the SDOF lateral displacements that are equivalent to the
Model Using Pushover Analysis mode shape can be expressed as follows:
There are no exact instructions as to how to
determine the parameters of the SDOF model F ¼ MF (15)
since the mass m , the damping c , and the
stiffness k all depend on the assumed displace- Substituting Eq. 15 into Eq. 14 and realizing that
ment shape. In order to obtain a practical solution in such a case C5F, it can be shown that
to Eq. 10, the analyst has to determine the force-
displacement relationship of the SDOF model F ¼ FT MF (16)
taking into account certain hysteretic rules,
which are often assumed to be similar to those Multiplying and dividing Eq. 16 by m 5CT M 1
used for the structural components. However, the and taking into account Eq. 8 produces the fol-
displacement of the SDOF model can be simply lowing result:
determined according to Eq. 9. Some further
explanation is needed in order to be able to better FT M1
understand the background for the determination F ¼ (17)
G
of the forces of the SDOF model from the push-
over curve. In general, the product of the stiffness Eq. 18 can be further simplified by realizing that
k and u represents a force in the spring of the FT ¼ FT M:
SDOF model:
F T 1 Fb
F ¼ k u (12) F ¼ ¼ (18)
G G
which should be expressed in a different form by
where Fb is the so-called base shear, i.e., the sum
taking into account Eqs. 11a and 9:
of the lateral forces F. The force-displacement
relationship of the SDOF model is thus obtained
F ¼ k c u t (13)
simply by dividing the quantities at the structural
level by the so-called transformation factor.
Equation 13 can be further transformed by However, it should be emphasized that Eq. 18
substituting kc using Eq. 7 and with consideration may not be sufficiently accurate in the case
of Eqs. 3 and 4: when the pushover curve is based on a higher
mode shape. Equation 9 can be used, together
F ¼ C T F (14) with Eqs. 14 or 18, to determine the
98 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
force-displacement F u relationship of the whence it can be shown that the same damping
SDOF on the basis of the pushover curve. constants a and b of the SDOF model are equal to
It is often useful to know the period of the that assumed for the response history analysis of
SDOF model, which can be obtained by analyz- the entire structure. The damping constants a and
ing Eq. 10. It is well known that the radial b, or only one of these constants, if damping is
frequency of an oscillating body is proportional to the mass or stiffness matrix only,
are often determined by assuming a certain ratio
rffiffiffiffiffiffi
k 2p of critical damping. Describing the relationship
o ¼ ¼ (19) between the damping constants and the ratio of
m T
critical damping is, however, outside the scope of
whence it follows that this derivation.
Equation 10 can be solved by numerical inte-
rffiffiffiffiffiffi
m gration which is implemented in conventional
T ¼ 2p (20) software for response history analysis. However,
k
the solution depends, in general, on the ground
The stiffness of the SDOF model can be motion and assumptions adopted for the determi-
expressed by means of Eq. 12. The period of the nation of the parameters of the SDOF model, i.e.,
SDOF model thus equals the mass m , the damping model, the force-
displacement relationship F u , and the hys-
rffiffiffiffiffiffiffiffiffiffi teretic rules.
m u
T ¼ 2p (21) In the simplest and the most commonly used
F
approach, the parameters of the SDOF model are
In order to solve Eq. 10, it is necessary to define determined by performing pushover analysis tak-
the damping model c . The so-called Rayleigh ing into account invariant distribution of the lat-
damping model is often assumed at the level of eral forces. In this case, the procedure for the
the structure. In this case, the damping matrix is determination of the parameters of the SDOF
expressed as a linear combination of the mass and model can be decomposed into the following
stiffness matrices: steps:
pushover curve can be obtained, in the case of results of a pushover analysis are usually
the SDOF model, according to Eq. 14. In the presented in the form of a pushover curve,
case when the distribution of displacement which represents a relationship between the A
vector D is more or less constant for all levels base shear and top (roof) displacement of the
of damage, then the forces F can be deter- investigated structure (e.g., see Fig. 2). The push-
mined by Eq. 18. over curve is, in fact, a characteristic of the struc-
6. Assume hysteretic behavior of the SDOF ture. However, it depends on the type of pushover
model. analysis used.
7. Compute the mass m of the SDOF model The pushover analysis becomes an approxi-
according to Eq. 7. mate analysis method if it is used for the predic-
8. Compute the damping constants a and/or b tion of expected seismic performance of a
based on the assumed damping model and structure (e.g., Saiidi and Sozen 1981; Fajfar
the ratio of critical damping. and Fischinger 1988; Fajfar 2000). In this case,
it is important that the structural model implicitly
accounts for the cyclic deterioration. However,
Pushover Analysis and Limit States the results of pushover analysis are used for the
definition of the SDOF model, which represents a
Pushover analysis is a kind of nonlinear static link between the seismic action and the seismic
analysis which aims at determining the relation- demand.
ship between the lateral forces and the engineer- The seismic demand at the level of the SDOF
ing demand parameters (EDPs) for all model can be obtained by means of numerous
performance levels of a structure. The results of procedures. Once the displacement at the level
pushover analysis are very intuitive, which of the SDOF model has been obtained, it can be
makes it attractive for practical applications. It transformed to the selected EDP at the structural
is clear that the relationship between the forces level. Since several phenomena associated with
and the deformations are linearly elastic until the response of a structure during an earthquake
damage occurs to the first element. Beyond this are neglected in pushover-based methods, it is
level, an increment of lateral forces causes a clear that the results of such simplified proce-
greater increment in the deformations in compar- dure are approximate also from the theoretical
ison with that observed in the case of linear elas- point of view. Due to the approximate nature of
tic analysis, since the stiffness of the structure is pushover-based procedures, many types of push-
reduced due to the damage which has occurred to over analysis have been developed in recent
some of the elements. When the deformations decades. Some of them are briefly described
reach a certain level, the maximum strength of below.
the structure is reached. From here on, the
strength of the structure decreases due to the Types of Pushover Analysis
softening of the material (e.g., see Fig. 2) or due Conventional pushover analysis (e.g., Saiidi and
to second-order effects, if they are considered in Sozen 1981; Fajfar and Fischinger 1988) is based
the analysis. In less general case, when the soft- on an invariant distribution of the lateral forces.
ening of material and second-order effects is not The shape of the lateral forces is therefore inde-
taken into account, the pushover curve becomes a pendent of deformations, whereas the shape of
straight line once a plastic mechanism forms. the deformations is dependent on the level of
The strength and deformation capacity, the structural damage. This is the most common
available ductility, and the system failure mode type of pushover analysis, which is recommended
represent the global results of the pushover anal- by some guidelines and building codes (e.g.,
ysis, which should always be checked in the case FEMA 356, Eurocode 8). It is usually prescribed
when seismic performance assessment of struc- that pushover analysis should be performed for at
ture is based on a nonlinear analysis method. The least two vertical distributions of the lateral loads
100 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
in order to assess the range of EDPs that might however, did not provide significantly improved
occur during actual dynamic response. Eurocode prediction in comparison to the force-based algo-
8 suggests the use of the so-called uniform and rithms that are used for adaptive pushover
modal load patterns, which are, respectively, pro- analysis.
portional to the story masses and the lateral forces Conventional or adaptive pushover analysis
from the elastic analysis. According to FEMA provides a single pushover curve, which can be
356, two patterns have to be selected from two used in conjunction with the SDOF model to
groups. The first load pattern should be consistent obtain the expected values of EDPs. For some
with the modal pattern, whereas a second pattern types of buildings and EDPs, such a simulation
can be based on a uniform load pattern or on an of the seismic response of a structure provides
adaptive load distribution which accounts for the sufficiently accurate results. However, the ana-
redistribution of the lateral loads, taking into lyst has to be careful when interpreting the results
account the properties of the yielded structure. of the seismic performance assessment of a struc-
Due to the simplicity of conventional pushover ture on the basis of one pushover curve, since it is
analysis, even more than two pushover analyses not possible to assess all types of EDPs on the
can be easily performed. For example, an itera- basis of single pushover analysis. In general, the
tive pushover-based procedure (IPP) (Celarec dynamic response of a structure is far more com-
and Dolšek 2013) involving model adaptation plex, especially if it suffers significant damage. In
was recently introduced, which aims at improv- general, several system failure modes can be
ing the capability of simulations based on simpli- observed due to the randomness of ground
fied nonlinear models. It was shown that the IPP motions. Additionally, in the case of taller build-
can be used to approximately simulate the shear ings, several modes can significantly contribute
failure of columns (Celarec and Dolšek 2013; to the results. On the other hand, mode shapes
Kosič et al. 2014), although simplified nonlinear depend on the scale of inflicted damage, which
models are not capable of direct simulation of gradually increases during a strong earthquake.
such effects. Thus, the maximum values of EDPs do not
Adaptive pushover analysis is an alternative to appear at the same instant in time. Furthermore,
conventional pushover analysis. In force-based the impact of mode shapes varies with respect to
adaptive pushover analysis, the load vectors are the EDPs. For example, the higher mode effects
gradually updated due to the damage occurring to can be significant for the prediction of story drifts
the structure, which affects the mode shapes. One and for the internal forces in the upper parts of
of the first procedures for the adaptive lateral load taller buildings. Much effort has been expended
for pushover analysis was proposed by Bracci in order to try to approximately account for the
et al. (1997). Many different procedures for so-called effect of higher modes. One of the first
force-based adaptive pushover analysis followed. attempts in this direction has been made by Paret
Most approaches involve response spectrum et al. (1996), who performed several pushover
analysis which is performed for the different analyses using a lateral load pattern based on
modal characters of the structure which are asso- different elastic mode shapes. A theoretical
ciated with selected stages of the pushover anal- basis for the modal pushover analysis procedure
ysis. However, Aydinoğlu (2003) has argued that for estimating the seismic demand for buildings
the load patterns should be based on an inelastic was introduced by Chopra and Goel (2002). The
rather than an elastic response spectrum. It was method involves several pushover analysis and
soon realized that force-based adaptive pushover corresponding modal-based SDOF models,
analysis offers only a relatively minor advantage whereas the overall response is obtained by com-
in comparison with conventional pushover anal- bining the results of all considered modes.
ysis. As a result of such an observation, a Recently, a general pushover analysis was intro-
displacement-based adaptive pushover procedure duced (Sucuoğlu and G€unay 2011). It is based on
was proposed (Antoniou and Pinho 2004), which, several pushover analyses, which are performed
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures 101
on the basis of the lateral forces obtained as a Each country can prescribe return periods of the
combination of the modal forces. It was also design seismic action associated with these limit
shown that the so-called higher system failure states. By default, it is assumed that appropriate A
modes can significantly contribute to the levels of protection are achieved when the
response of taller buildings. Brozovič and Dolšek structural performance for the DL, SD, and NC
(2014) have shown that system failure modes limit states is checked for seismic actions
observed due to the randomness of ground corresponding to return periods of 225, 475, and
motion can be sufficiently accounted for by the 2475 years, respectively. The prestandard for the
envelope-based pushover analysis procedure, seismic rehabilitation of buildings FEMA
which is based on several pushover analyses and 356 (2000), which was issued a few years before
the use of failure-based SDOF models. The the Eurocode 8-3, is based on similar metrics for
failure-based SDOF models utilize the displace- checking the target building performance. The
ment vectors corresponding to the system failure basic structural performance levels of a building
modes that are observed from pushover analyses are immediate occupancy (IO), life safety (LS),
in the case of second or higher modes. This sig- and collapse prevention (CP). FEMA 356 states
nificantly improves the prediction of the intensity that building performance is a combination of the
causing a designated limit state, if the structure is performance of both the structural and the
sensitive to the effects of higher modes. nonstructural components. The description of
performance levels (i.e., limit states), which are
Definition of Limit States discrete damage states selected from among the
As already mentioned above, pushover analysis is infinite spectrum of possible damage states that
the simplest nonlinear analysis method, which buildings could experience during an earthquake,
has been included in several guidelines and stan- consists of estimates rather than precise predic-
dards for seismic performance assessment and tions (FEMA 2000).
the retrofitting of buildings. The results of push- According to the brief overview of limit
over analysis can be used, in conjunction with the states provided in some standards, it is clear
target displacement for a given seismic action, that there are many different definitions of
to check whether the structural performance sat- them. In general, each limit state is associated
isfies the defined performance objectives. For this with a certain degree of damage. However, there
purpose, building codes provide limit states at the are some restrictions regarding the definition of
component or structural level. Eurocode 8-3 limit states on the basis of pushover analysis. For
(CEN 2005) utilizes the following limit states: example, the pushover analysis is not able to
damage limitation (DL), significant damage explicitly simulate cycling deterioration. Thus,
(SD), and near collapse (NC). These limit states in the case of pushover analysis, limit states
are explicitly defined at the component level, cannot be dependent on the level of cyclic dete-
whereas definitions at the structural level are rioration. Since damage models are often
only descriptive. For example, it is prescribed defined by story drifts or the rotations in plastic
that the near-collapse limit state at the structural hinges, such a limitation of pushover analysis,
level is attained when a structure is heavily dam- associated with an inability to simulate cycling
aged, with some residual lateral strength and stiff- deterioration, is not critical, provided that the
ness, but is still capable of sustaining vertical nonlinear model used for pushover analysis
loads and aftershocks of moderate intensity. How- implicitly accounts for cyclic deterioration of
ever, moderate permanent drifts are present, and it the structural components. In the case of
is not likely, from the economic point of view, that nonstructural components, it is common to
repair of the building could be justified. Note that define a limit state by the level of story acceler-
the NC limit state is often defined at the global ation, which is not explicitly simulated by push-
level, i.e., when the strength in the post-capping over analysis. A simplified model for story
range of pushover curve decreases for 20 %. acceleration is therefore also required.
102 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
Although there are no critical restrictions for demand parameter. In the case of an IM-based
the definition of limit states in the case when definition of a limit state, which is common for
seismic demand is obtained by the pushover- the collapse limit state, the fragility function can
based method, it should be realized that the accu- be defined as follows:
racy of pushover-based methods is limited. The
analyst should therefore be careful when ln im ln mimLS
PðLSjIM ¼ imÞ ¼ F (26)
checking the limit states, especially in the case sln imLS
when target performance is defined at the com-
ponent level. It is thus recommended that the where PðLSjIM ¼ imÞ is the probability of
limit states should be defined at the global and exceeding the limit state LS if the intensity mea-
structural levels, taking into account the global sure assumes a value equal to im, mimLS is the
parameters obtained from the pushover curve. median limit-state intensity, and sln imLS is the
Even in this case, it is recommended that the corresponding standard deviation of the natural
structural damage associated with such defined logarithms. The fragility function is therefore
limit states should be checked, since it is possible defined by two parameters, the median value
that a flexible structure has not suffered signifi- medp or mimLS and the corresponding standard
cant damage even though the corresponding point deviation sln edp or sln imLS. Many different pro-
on the pushover curve is beyond the post-capping cedures for the estimation of these parameters
point. Such performance can be observed in based on pushover analysis have been devel-
the case of flexible frames, where the P-D effect oped. It should also be noted that many different
has a quite significant impact on structural procedures for fragility analysis exist. Some of
performance. them can also be used on the basis of seismic
performance assessment utilizing pushover
analysis (e.g., Jalayer et al. 2011; Rota
Pushover-Based Seismic Fragility et al. 2014).
Analysis The accuracy of determination of the fragility
parameters (m, sln) depends on the method for
The result of fragility analysis is the fragility seismic performance assessment of a structure
function, which represents the probability and the level of the simulations, which accounts
that an engineering demand parameter (EDP) for the effect of uncertainty.
will exceed a certain limit-state value edp
given an intensity IM ¼ im . In the simplest Deterministic Approach for Assessing the
case, it is assumed that the seismic fragility Parameters of the Fragility Function
function is based on a lognormal distribution In the simplest case, the effects of uncertainty can
function: be assumed by the default values of sln. Conse-
quently, the deterministic seismic performance of
PðEDP > edpjIM ¼ imÞ ¼ a structure can be used. This enables the use of
analytical models for the determination of the
ln edp ln medp (25)
1F target displacement. Parametric studies have
sln edp
shown that the sln imLS depends on the limit
state and the type of seismic intensity measure.
ln edpln medp
where F sln edp represents the standard If the intensities causing collapse are assessed by
normal distribution function, medp is the median the spectral acceleration corresponding to the first
value of the EDP given the intensity im, and vibration period, the sln imLS is in the interval
sln edp is the standard deviation of the natural between 0.3 and 0.5 (Lazar and Dolšek 2014).
logarithms of the EDP given that IM ¼ im. The The sln imLS increases with respect to the period of
above definition is suitable when the limit state is structure if it is assessed on the basis of peak
based on a threshold value of an engineering ground acceleration. It varies between 0.5 and
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures 103
0.75. The value of sln imLS can additionally all the variables in Eqs. 27 and 28 is described in
increase if epistemic uncertainties are considered the corresponding references (CEN 2004; FEMA
in addition to the record-to-record randomness. 2000). A
In Europe, the most commonly used approach
for the determination of expected seismic Consideration of Ground-Motion
response based on pushover analysis is the N2 Randomness in Estimation of Parameters of
method (e.g., Fajfar 2000), which has been Fragility Function
implemented in Eurocode 8 (CEN 2004), The first level of simulations involves consider-
whereas the so-called coefficient method is used ation of the effect of ground-motion randomness.
in the United States (e.g., FEMA 356 (2000)). In general, the target displacement according to
According to Eurocode 8 and FEMA 356, the this approach is based on nonlinear dynamic anal-
target displacement (dt and dt), respectively, can ysis of the SDOF model. The analyst should
be determined as follows: therefore select an appropriate set of ground
motions and define the SDOF model, which
mðqu , T , T C Þ T 2 could be used for nonlinear dynamic analysis as
dt ¼ G
Se ðT Þ 2 ; (27)
qu 4p previously described in this entry. The results of
such analyses are samples of EDP values given
T 2e the different values of the intensities or the sam-
dt ¼ C0
C1 ðR, T S , T e ÞC2 C3
Sa ðT e Þ ; (28) ple of limit-state intensities, which could be used
4p2
to estimate the parameters of the fragility func-
where the parameters G and C0 have the same tion. Some alternatives exist which can be used in
meaning, i.e., they are the so-called transforma- order to avoid the use of nonlinear dynamic anal-
tion or modification factors which relate the spec- ysis at the level of the SDOF model. Vamvatsikos
tral displacement of an equivalent SDOF model and Cornell (2006) developed the SPO2IDA soft-
to the roof displacement of the MDOF model. For ware tool, which is capable of recreating the
the experienced reader, it is also trivial that the seismic behavior of a single-degree-of-freedom
last parts of Eqs. 27 and 28 represent the elastic (SDOF) model with a quadri-linear force-
spectral displacement Sde, whereas the middle displacement relationship. They performed a
part of Eqs. 27 and 28 corresponds to the parametric study by suitably varying the five
so-called inelastic displacement ratio C (e.g., parameters of the force-displacement relation-
Miranda 2000), which depends on many parame- ship, which are negative and nonnegative hard-
ters, such as the ground motion type, the hysteretic ening, the residual plateau, ductility at the
behavior, the force-displacement relationship, beginning of strength degradation, and ductility
the masses, damping, and others. For reasons of at collapse. Based on the results of the parametric
simplicity, formulas for the determination of the study, regression analysis was used to define each
inelastic displacement ratio have been introduced segment of the approximate 16th, 50th, and 84th
based on the results of parametric studies. For fractile IDA curves. Although SPO2IDA is a
example, in Eurocode 8, the simplified useful software tool, it is not easy to extend its
R m T relationship proposed by Fajfar applicability, since a new regression analysis is
(2000) is used to define the inelastic displacement required if the seismic response parameters are to
ratio, whereas in FEMA 356 (2000), two addi- be computed for additional ground motion
tional parameters are used to define the inelastic records. More recently, Peruš et al. (2013) intro-
displacement ratio, i.e., the parameter C2, which duced a web-based methodology for the predic-
represents the effect of a pinched hysteretic shape, tion of approximate IDA curves, which consists
stiffness degradation, and strength deterioration on of two independent processes. The result of the
maximum displacement, and the parameter C3, first process is a response database of the single-
which represents the increased displacement due degree-of-freedom model, whereas the second
to dynamic P-D effects. Note that the meaning of process involves the prediction of approximate
104 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
IDA curves from the response database by using should be generated in such a way that the corre-
n-dimensional linear interpolation. Such an lation structure between the random variables is
approach makes possible user-friendly prediction simulated in the best possible manner. This rep-
of the seismic response parameters with high resents the second step in the process of the
accuracy. In order to demonstrate the capabilities determination of the sample of random variables.
of the proposed methodology, a web application The problem can be successfully solved by the
for the prediction of the approximate 16th, 50th, stochastic optimization method called simulated
and 84th fractile response of a reinforced concrete annealing. More details are available elsewhere
structure was developed (http://ice4risk.slo- (e.g., Dolšek 2012).
projekt.info/WIDA/). Note that web-based meth- However, all the referenced procedures are
odology for the prediction of the approximate IDA based on simulations which are performed at the
curves estimates an error of the predicted IDA level of a MDOF model. Recently, an approxi-
curves of a SDOF model. mate procedure utilizing a deterministic MDOF
model and uncertainty analysis at the level of a
Consideration of Ground-Motion SDOF model was proposed (Kosič et al. 2014).
Randomness and Epistemic Uncertainty in It can be classified as being somewhere in
Estimation of Parameters of Fragility between the aforementioned procedures for the
Function determination of fragility parameters on the basis
In a general case, fragility parameters should be of pushover methods, both in terms of computa-
based on simulations which take into account the tional time and in terms of accuracy. Such an
effects of ground-motion randomness and the approach is computationally less demanding,
other knowledge-based sources of uncertainty, since all the simulations are performed at the
such as physical uncertainties, modeling uncer- level of the so-called probabilistic SDOF model.
tainties, human errors, uncertainties associated It can therefore be attractive for the fragility
with the knowledge level about the structure, analysis of more extensive building stock.
and others. In order to account for knowledge-
based uncertainties, the simulations can be Example: Pushover-Based Seismic Fragility
performed at the level of a MDOF model (e.g., Analysis of the RC Frame Building With
Fragiadakis and Vamvatsikos 2010; Dolšek Consideration of Ground-Motion
2012) or a SDOF model (Kosič et al. 2014). Randomness and Modeling Uncertainty
Many different simulation methods can be used A simple variant of a pushover-based fragility
to perform pushover analysis with consideration analysis is demonstrated by means of an illustra-
of knowledge-based uncertainties (e.g., Barbato tive example. The procedure involves conven-
et al. 2010; Jalayer et al. 2011). It is convenient tional pushover analyses and the nonlinear
and straightforward to utilize Monte Carlo simu- dynamic analyses of corresponding SDOF
lation with Latin hypercube sampling (e.g., models. The effects of ground-motion and
Fragiadakis and Vamvatsikos 2010; Dolšek modeling uncertainties are accounted for, respec-
2012). The LHS technique uses stratification of tively, by the set of ground motions and the set of
the probability distribution function of the ran- structural models which are generated by Monte
dom variables Xi and consequently requires sig- Carlo with LHS. The step-by-step procedure for
nificantly fewer simulations in comparison with the determination of the fragility curve is as
the ordinary type of Monte Carlo simulation. In follows:
general, two steps are needed to determine the
sample of random variables, which are directly 1. Preparation of data regarding the structure
applied in the structural model. First, each ran- and the seismic hazard at the location of the
dom variable is sampled by inverse method using structure. It is important to obtain as accurate
equidistant points between sample probabilities. as possible data regarding the seismic hazard
If the random variables are correlated, the sample in order to reduce the effect of uncertainties.
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures 105
Information regarding the seismic hazard Different types of pushover analysis can be
and the soil type are important for the selec- performed. Conventional pushover analysis,
tion of appropriate ground motions. which is based on the invariant lateral load, is A
2. Definition of limit states. In general, limit the simplest and is the most often applied.
states can be defined at the component or The results of the pushover analysis are push-
structural level. As briefly discussed above, over curves and the seismic demand on the
there is no critical restriction regarding the structure given the top displacement. In gen-
definition of limit states if the seismic eral, more than one pushover analysis per
demand is based on a pushover analysis structural model is required.
procedure. 7. Determination of the SDOF models. This
3. Identification of epistemic uncertainties. step requires the transformation of top dis-
Uncertainties are often divided into aleatoric placement and base shear to the displacement
and epistemic, although there is no need to (Eq. 9) and force (Eq. 14 or Eq. 18) of the
make such a distinction. In this example, the SDOF model. The force-displacement rela-
aleatoric uncertainties are accounted for by tionship of the SDOF model should be multi-
the set of ground motions (step 5), whereas linear. Negative post-capping stiffness can
the effects of the epistemic uncertainties are also be applied in order to estimate the col-
incorporated by the set of structural models. lapse capacity. It is usually assumed that the
The most important sources of epistemic hysteretic behavior is the same as that pre-
uncertainty should be modeled by appropri- scribed for the components of the structural
ate random variables. models. The mass of the SDOF model should
4. Determination of the set of structural models be determined according to Eq. 7a, whereas
utilizing Monte Carlo with LHS. The LHS the damping can be proportional to mass
technique uses stratification of the probabil- and/or stiffness (Eq. 24).
ity distribution function of the random vari- 8. Response history analyses for all SDOF
ables and consequently requires fewer models and ground motions. Since, in the
simulations in comparison with a crude case of SDOF models, the response history
Monte Carlo simulation. However, any analysis is not computationally demanding, it
appropriate LHS technique can be used in is recommended that the records should be
order to generate the sample of the uncertain scaled and that the relationship between the
input parameters. It has been previously ground-motion intensity and the displace-
shown (e.g., Dolšek 2012) that the effect of ment of the SDOF model should be com-
epistemic uncertainties can be simulated puted for several levels of damage, e.g.,
with sufficient accuracy by a number of from elastic behavior to collapse.
structural models, which is at least twice the 9. Assessment of seismic demand at the level of
number of random variables. It is worth structural model. The displacement obtained
emphasizing that the nonlinear structural from the response history analysis of the
model should implicitly account for cyclic SDOF model should be transformed to the
deterioration. top displacement using Eqs. 8 and 9. Any
5. Selection of an appropriate set of ground other engineering demand parameter at the
motions. Many different procedures exist level of structure can be obtained from the
for the selection of an appropriate set of results of the pushover analysis on the basis
ground motions. Ground motions are usually of the known top displacement. However, if
selected to match the target spectrum, which more than one pushover analysis is
is, in the simplest case, the elastic accelera- performed per structural model, in order to
tion spectrum prescribed by a building code. account for the effect of higher failure
6. Calculation of pushover curves for all the modes, then the total demand can be obtained
structural models simulated in step 4. by enveloping the results of all pushover
106 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures, Fig. 1 A view of the investigated
four-story building and the reinforcement in selected columns and beams
analyses (e.g., Brozovič and Dolšek 2014) or ductility class high (behavior factor = 5). Con-
by combining the results together according crete of quality C25/30 and B500 Tempcore
to an appropriate combinational rule (Chopra reinforcing steel were used to construct the
and Goel 2002). The results of this step are building. Minimum and maximum mean con-
samples of the engineering demand parame- crete strengths of 32 MPa and 56 MPa were
ters given the intensities or samples of limit- measured. The modulus of elasticity of the
state intensities. concrete varied from 28.5 to 35.3 GPa. The
10. Calculation of the fragility parameters and mean yield strength exceeded the characteris-
the fragility curve. Fragility curves can be tic yield strength by between 10 % and 20 %,
directly estimated on the basis of the sample depending on the diameter of the reinforcing
values. In the case when the fragility curves bars. However, the average value of the yield
are based on an assumed lognormal distribu- strength of the reinforcing steel amounted to
tion, the method of moments, the maximum about 580 MPa.
likelihood method, the method of counted Step 2. Fragility analysis was performed for the
percentiles, or any other suitable method can limit states of damage limitation (DL), signif-
be used to calculate the median engineering icant damage (SD), and near collapse (NC), as
demand parameters or the limit-state intensi- defined in Eurocode (CEN 2005). It was
ties and the corresponding dispersions. assumed that the DL limit state is attained at
the structural level if the longitudinal rein-
The application of the above-described step-by- forcement in all the columns in any story starts
step procedure for the determination of the fragil- to yield. It was further assumed that the SD
ity curves is demonstrated by means of a four-story and NC limit states are reached, at the struc-
reinforced concrete frame building (Fig. 1), which tural level, when the moment exceeds the
was previously analyzed (Dolšek 2012). Just a maximum moment in the case of the first col-
brief overview of the procedure is presented. umn or the rotation exceeds the ultimate rota-
tion in the case of first column, respectively.
Step 1. The structure was designed for a peak However, additional criteria for the occur-
ground acceleration of 0.3 g, soil type B, and rence of the SD and NC limit states were
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures 107
defined. It was assumed that the top displace- affect the estimated median seismic response
ment corresponding to the NC limit state parameters and the corresponding dispersion.
should not be less than that associated with The results of this step were 20 structural A
80 % of the maximum base shear measured in models based on the sample of random vari-
the post-capping range of the pushover curve. ables. For comparison, an additional “deter-
Similarly, it was assumed that the top dis- ministic” structural model was also created by
placement corresponding to the SD limit assuming median values of the random
state should be not less than the 75 % of the variables.
NC limit-state top displacement. Step 5. A set of 14 ground motions (Ngm = 14)
Step 3. Since full knowledge about the building were selected from European Strong Motion
structure was available, only the following Database to match the Eurocode-based accel-
modeling uncertainties were considered in eration spectrum. All the selected records
the fragility analysis: mass, strength of the were recorded on stiff soil and had peak
concrete and the reinforcing steel, effective ground acceleration greater than 0.1g. The
slab width, damping, and the model for deter- mean acceleration spectrum of the selected
mining the initial stiffness and ultimate rota- ground motions matched reasonably well
tion in the plastic hinges of the beams and with the target Eurocode-based spectrum. It
columns. A normal or lognormal distribution should be noted that ground motions selected
was assumed for the majority of the nine ran- in this way often overestimate the seismic
dom variables (Nvar = 9). The statistical char- response with respect to that based on
acteristics of the input random variables were hazard-consistent ground motions.
taken from literature as reported elsewhere Step 6. The lateral load pattern for conventional
(Dolšek 2012). All the considered input ran- pushover analyses was based on the first vibra-
dom variables were assumed to be tion mode (Eq. 15). For simplicity, the struc-
uncorrelated with respect to the story masses, ture was analyzed in the Y direction only (see
which were assumed to be perfectly correlated Fig. 1). The pushover curves for all structural
and modeled by one random variable. The models from Step 4 are presented in Fig. 2.
highest value of the coefficient of variation An extensive scatter can be observed in the
was adopted for the prediction of the ultimate case of the deformation capacity of the real-
rotations in the beams (0.6) and columns (0.4). ized structural models, whereas the difference
Step 4. Realization of the random variables was in the strength is less. However, no significant
based on the LHS technique. The size of the differences can be observed between the
sample was assumed to be equal to 20, which deterministic model and the so-called median
is slightly larger than twice the number of all pushover curve. The limit-state top displace-
the random variables considered in the fragil- ment and the corresponding base share are
ity analysis. The correlation matrix of the highlighted for the DL, SD, and NC limit
sample was almost identical to the target cor- states. The high coefficient of variation of the
relation matrix. Note that this is only a neces- limit-state top displacement, which increases
sary condition for the prediction of the seismic gradually with respect to the severity of the
response parameters with the required accu- limit state, is partly the consequence of the
racy, whereas the sufficient condition is high coefficient of variation of the input
related to the number of simulations, which parameters and partly the consequence of the
is, in the case of the example, relatively low formation of different system failure modes,
(Nsim = 20). However, the results of which were observed from pushover analyses.
performed studies have shown that, in the Step 7. The pushover curves were idealized by
case when Nsim is larger than twice the number means of a trilinear force-displacement rela-
of random variables (Nvar), an increase in the tionship, also taking into account negative
number of simulations does not significantly post-capping stiffness. Transformation of the
108 Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures
idealized pushover curve to the parameters of shown in Fig. 3a are presented for the deter-
an equivalent SDOF model was performed by ministic SDOF model, taking into account
dividing the base shear and top displacement just ground-motion randomness. The vari-
of the idealized pushover curve by a transfor- ability can be observed only for the limit-
mation factor G (Eq. 8). The mass of the SDOF state intensities, while in the adjacent figure
model and the force-displacement relation- (Fig. 3b), variability can be observed also for
ships of the SDOF model were obtained the limit-state displacements. The latter vari-
according to Eq. 7a and Eqs. 9 and 18, respec- ability is the consequence of the effect of
tively. Note that the number of the SDOF epistemic uncertainty. The sample of limit-
models was the same as the number of push- state intensities presented in Fig. 3b was
over curves. All the parameters of the SDOF used to estimate the parameters of the fragil-
models differed from one another due to the ity curves. The median values and the
epistemic uncertainties. However, the peak- corresponding dispersion of natural loga-
oriented hysteretic model and damping pro- rithms were estimated according to maximum
portional to the tangent stiffness were likelihood method. The fragility curves were
assumed to be invariant. then obtained using Eq. 26. The results are
Step 8. The response history analyses were presented in Fig. 4.
performed for each SDOF model by scaling
the ground-motion intensity in order to obtain A flatter fragility curve can be observed in the
the complete relationship between the peak case of the DL limit state if the effects of episte-
ground acceleration and the displacement of mic uncertainty are considered in the fragility
the SDOF model. The results are presented in analysis. This is in agreement with expectations
Fig. 3, where the highlighted points indicate and can be modeled by using the mean estimate
the limit-state intensities which were used for approach, in which it is assumed that the effects
the calculation of the fragility parameters. of epistemic uncertainties on the fragility curve
Steps 9 and 10. The limit-state top displacements can be modeled by simply inflating the disper-
highlighted in Fig. 2 were transformed into sion. However, in the case of the NC limit state, it
the displacement of the corresponding SDOF is clear that incorporating the epistemic uncer-
model, which was then used to determine the tainties in the determination of the fragility curve
limit-state intensities. The limit-state dis- causes a shift of the fragility curve to the left,
placements and the limit-state intensities which means that the effects of epistemic
Analytic Fragility and Limit States [P(EDP|IM)]: Nonlinear Static Procedures 109
Analytic Fragility and Limit States [P(EDP|IM)]: SDOF model and (b) the SDOF models based on the LHS
Nonlinear Static Procedures, Fig. 3 The limit-state technique. The median limit-state points for the DL, SD,
peak ground accelerations and displacements associated and NC limit states are also presented
with the set of structural models and (a) the deterministic
step-by-step procedure and demonstrated by means Dolšek M (2012) Simplified method for seismic risk
of an example of a four-story building. It has been assessment of buildings with consideration of aleatory
and epistemic uncertainty. Struct Infrastruct Eng
shown that pushover-based methods can be suc- 8(10):939–953
cessfully used for fragility analysis. However, the Fajfar P (2000) A nonlinear analysis method for
analyst should always be aware of the approximate performance-based seismic design. Earthq Spectra
nature of pushover-based methods. Such an aware- 16(3):573–592
Fajfar P, Fischinger M (1988) N2 – a method for
ness will allow him to decide to which type of non-linear seismic analysis of regular buildings. Pro-
structures and for what purpose individual ceedings of ninth world conference on earthquake
pushover-based methods can be applied. engineering, vol V, Tokyo-Kyoto
FEMA (2000) Prestandard and commentary for the seis-
mic rehabilitation of buildings, vol 356, FEMA. Federal
Emergency Management Agency, Washington, DC
Cross-References Fragiadakis M, Vamvatsikos D (2010) Fast performance
uncertainty estimation via pushover and approximate
▶ Reinforced Concrete Structures in Earthquake- IDA. Earthq Eng Struct Dyn 39(6):683–703
Jalayer F, Elefante L, Iervolino I, Manfredi G (2011)
Resistant Construction Knowledge-based performance assessment of existing
▶ Seismic Fragility Analysis RC buildings. J Earthq Eng 15(3):362–389
▶ Seismic Reliability Assessment, Alternative Kosič M, Fajfar P, Dolšek M (2014) Approximate seismic
Methods for risk assessment of building structures with explicit
consideration of uncertainties. Earthq Eng Struct Dyn
▶ Uncertainty Theories: Overview 43(10):1483–1502
Kreslin M, Fajfar P (2012) The extended N2 method
considering higher mode effects in both plan and ele-
References vation. Bull Earthq Eng 10(2):695–715
Lazar N, Dolšek M (2014) Incorporating intensity bounds
Antoniou S, Pinho R (2004) Development and verification for assessing the seismic safety of structures: does it
of a displacement-based adaptive pushover procedure. matter? Earthq Eng Struct Dyn 43(5):717–738
J Earthq Eng 8(5):643–661 Miranda E (2000) Inelastic displacement ratios for struc-
Aydinoğlu MN (2003) An incremental response spectrum tures on firm sites. J Struct Eng 126(10):1150–1159
analysis procedure based on inelastic spectral dis- Paret TF, Sasaki KK, Eilbeck DH, Freeman SA
placements for multi-mode seismic performance eval- (1996) Approximate inelastic procedures to identify
uation. Bull Earthq Eng 1(1):3–36 failure mechanisms from higher mode effects. In: Pro-
Barbato M, Gu Q, Conte JP (2010) Probabilistic push-over ceedings of 11th world conference on earthquake engi-
analysis of structural and soil-structure system. J Struct neering (P.No. 966), Acapulco
Eng 136(11):1330–1341 Peruš I, Klinc R, Dolenc M, Dolšek M (2013) A web-based
Bracci JM, Kunnath SK, Reinhorn AM (1997) Seismic methodology for the prediction of approximate IDA
performance and retrofit evaluation of reinforced con- curves. Earthq Eng Struct Dyn 42(1):43–60
crete structures. J Struct Eng 123(1):3–10 Reyes JC, Chopra AK (2011) Three-dimensional
Brozovič M, Dolšek M (2014) Envelope-based pushover modal pushover analysis of buildings subjected to
analysis procedure for the approximate seismic response two components of ground motions, including its eval-
analysis of buildings. Earthq Eng Struct Dyn 43(1):77–96 uation for tall buildings. Earthq Eng Struct Dyn
Celarec D, Dolšek M (2013) Practice-oriented probabilis- 40:789–806
tic seismic performance assessment of infilled frames Rota M, Penna A, Magenes G (2014) A framework for the
with consideration of shear failure of columns. Earthq seismic assessment of existing masonry buildings
Eng Struct Dyn 42(9):1339–1360 accounting for different sources of uncertainty. Earthq
CEN (2004) Eurocode 8: design of structures for earth- Eng Struct Dyn 43(7):1045–1066
quake resistance – part 1: general rules, seismic actions Saiidi M, Sozen MA (1981) Simple nonlinear seismic
and rules for buildings, EN 1998–1. European Com- analysis of R/C structures. J Struct Div ASCE
mittee for Standardisation, Brussels 107(5):937–953
CEN (2005) Eurocode 8: design of structures for earth- Sucuoğlu H, G€ unay MS (2011) Generalized force vectors
quake resistance – part 3: assessment and retrofitting of for multi-mode pushover analysis. Earthq Eng Struct
buildings, EN 1998–3. European Committee for Dyn 40(1):55–74
Standardisation, Brussels Vamvatsikos D, Cornell CA (2006) Direct estimation of
Chopra AK, Goel RK (2002) A modal pushover analysis the seismic demand and capacity of oscillators with
procedure for estimating seismic demands for build- multi-linear static pushovers through IDA. Earthq Eng
ings. Earthq Eng Struct Dyn 31(3):561–582 Struct Dyn 35(9):1097–1117
Ancient Monuments Under Seismic Actions: Modeling and Analysis 111
Ancient Monuments a b
Under Seismic Actions: b b
Modeling and Analysis,
Fig. 2 Rocking of a rigid A
block on a rigid base
without sliding: (a) positive
angle of rotation; (b)
negative angle of rotation
CM CM
.. ϕ ..
mxg −ϕ mxg
h
h
mg θ θ mg
r r
O' O
ϕ O O' −ϕ
.. ..
xg xg
is a complicated problem that has attracted the about corner O of the base (Fig. 2a), which cor-
interest of many researchers in the last century. responds to positive angles of rotation
The basic scope of these investigations has been (clockwise), or about corner O0 of the base
the estimation of the required intensity of the (Fig. 2b), which corresponds to negative angles
ground motion to overturn slender bodies during of rotation (counterclockwise). The distinction
strong earthquakes. The first attempts towards between positive and negative angles of rotation
this aim were made by Milne (1885), Milne and is essential, because a different equation governs
Omori (1893) and Kirkpatrick (1927). However, the motion in each case.
it was Housner (1963) who first tackled the prob- Under a horizontal base excitation x€g ðtÞ, the
lem systematically, deriving the equations of equation of motion can easily be derived by
motion that govern the rocking response of applying Newton’s second law for the moments
a rigid, freestanding block and examining the about the pole of rotation (point O for positive
overturning risk under simple ground pulses. rotations, point O0 for negative rotations) and
taking under consideration that, apart from the
Equation of Motion weight, a horizontal inertial force (d’ Alembert
Let us consider the rocking block shown in Fig. 2. force) acts at CM, equal to m€ xg ðtÞ. Then, the
The block and the base are considered rigid, and the equation of motion can be written in the form:
coefficient of friction is assumed to be large enough
to prevent sliding of the block. In Fig. 2, a rectan- I O ’€ mgr sinðy ’Þ ¼ m€
xg r cosðy ’Þ
gular block is shown, with dimensions: width of the (1)
base = b and height of the center of mass
(CM) from the base = h. However, the equations where ’ is the rocking angle, m is the mass of
of motion and the basic features of the response that the block, r is the distance of CM from the pole of
are presented in the following are also valid for any rotation (point O or O0 ), y is the slenderness angle
symmetric block in the vertical direction with such defined by y = tan1b/(2h), IO is the mass
dimensions and arbitrary shape. moment of inertia about O or O0 , and g is the
Assuming that the block has been set into acceleration of gravity. In this equation, when-
rocking motion, the rotation can either happen ever a double sign appears, the upper one
114 Ancient Monuments Under Seismic Actions: Modeling and Analysis
corresponds to ’ > 0 (clockwise rotation) and equilibrium assuming that |y ’| << 1, thus
the lower one to ’ < 0 (counterclockwise cos(y ’) 1 and sin(y ’) (y ’).
rotation). Then, Eq. 1 becomes
Introducing the parameter p defined by
I O ’€ mgr’ ¼ mgry mr€
xg (7)
mgr
p ¼
2
(2)
IO It is noted that, as evident from Eqs. 6 and 7,
the rocking block corresponds to a system with
and using the sgn() function, Eq. 1 can be written negative stiffness. Also, for slender blocks with
in the following general form: slenderness angle y less than about 20
, the line-
arized Eqs. 6 and 7 do not differ significantly,
I ’€ þ p2 sin ½ysgnð’Þ ’ ¼ p2 cos ½ysgnð’Þ ’ since tan y sin y y. Analyses of such blocks
(3) showed that the linearized equations can predict
with acceptable accuracy for engineering pur-
The parameter p is a characteristic frequency poses the response and the overturning risk of
of the system. Actually it is equal to the pendu- the block.
lum frequency of the block when hung about the
corner of the base. Impact with the Ground
As mentioned above, Eqs. 1 and 3 are valid for During rocking, the pole of rotation alternates
any rigid block with an axis of symmetry about from point O to O0 , or vice versa. The alteration
the vertical axis through CM, with IO being the of the pole of rotation takes place when the block
mass moment of inertia about the corner of the hits the ground and is accompanied by energy
base. In the special case of a homogeneous rect- dissipation due to the change in the velocity of
angular block, IO is given by the following the center of mass which is schematically shown
relation: in Fig. 3. Let us assume that the angular velocity
h i before impact is ’_ 1 and after impact is ’_ 2 . Then,
m b2 þ ð2hÞ2 one can write
4mr2
IO ¼ ¼ , (4)
3 3 ’_ 2 ¼ e
’_ 1 (8)
and the characteristic frequency p becomes
where e is a coefficient of restitution. If we
rffiffiffiffiffi assume that the impact is fully plastic, i.e., that
3g
p¼ (5) there is no rebound of the block on the base, the
4r value of e can be calculated applying the principle
of conservation of the angular momentum about
For small angles of rotation (’ << 1), Eq. 1
point O0 before and after impact (Housner 1963),
can be linearized about the equilibrium position
which gives
assuming that cos ’ 1 and sin ’ ’ and
neglecting the second order term containing the
mb2
product ’
x€g ðtÞ. Taking under consideration that e¼1 (9)
2I O
rsin y = b/2 and rcos y = h, the following line-
arized equation can be derived:
Equation 9 implies that the coefficient of res-
b titution e and the dissipated energy depend solely
I O ’€ mgh’ ¼ mg mh€
xg (6) on the geometry of the block. However, experi-
2
mental investigation (e.g., Priestley et al. 1978;
For large angles of rotation, when the block is Aslam et al. 1980) showed that the actual value of
close to the verge of overturning, the linearization e might be significantly different than the theo-
should be performed about the point of unstable retical value of Housner, depending on the
Ancient Monuments Under Seismic Actions: Modeling and Analysis 115
Ancient Monuments a b b b
Under Seismic Actions:
Modeling and Analysis,
Fig. 3 Change in the A
velocity of the CM during
impact from positive to
negative angles of rotation:
(a) immediately before
impact, when the pole of
rotation is point O; (b) CM
immediately after impact, CM
v2
when the pole of rotation is
point O0 v1
θ θ
r r
O' O O' O
materials of the block and the base. For this equilibrium position with positive angles of
reason, in many analyses e is considered an inde- rotation. When ’ becomes zero, the block hits
pendent parameter of the problem rather than the base and the rotation continues around point
been calculated from Eq. 9. O0 with the motion been described by the equa-
tion describing negative rotations. According to
Free Vibrations Eq. 8, the initial angular velocity of this motion
In case that the block is rocking freely is reduced compared with the velocity of
ð€
xg ðtÞ ¼ 0Þ, the linearized equation of motion the block when it hits the base. Due to the dif-
(7) becomes ferent equations of motion describing each
regime of the response, the overall behavior is
’€ p2 ’ ¼ p2 y (10) nonlinear.
If there was not any dissipation of
where p is given by Eq. 2. Note that p is not the energy during impact, the time elapsed from
eigenfrequency of the system, since, due to the the angle of maximum tilt (initially equal to
negative stiffness, rocking blocks do not possess ’0) up to the point when the block reaches the
eigenfrequencies in the classical sense. The solu- equilibrium position (’ = 0) would correspond
tion of Eq. 10 is to one fourth of the period T of free vibrations.
Therefore, setting ’ = 0 for t = T/4 in
’ðtÞ ¼ A coshðptÞ þ B sinhðptÞ y (11) Eq. 12, the period of free vibrations can be
calculated:
where the coefficients A and B are determined
from the initial conditions. For example, for an
4 1
initial tilt of the block, ’0 < y, the initial condi- T ¼ cosh1 (13)
p 1 ’0 =y
tions are ’(t = 0) = ’0 and ’_ (t = 0) = 0, and
Eq. 11 becomes
However, in real structures dissipation of
’ðtÞ ¼ y ðy ’0 ÞcoshðptÞ (12) energy does occur at each impact. For this reason,
the block will attain a reduced maximum tilt
Equation 12 describes the motion of the block after each impact, and the period of free vibra-
around point O as it comes back to the tions will continuously decrease. Using Eq. 8,
116 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Ancient Monuments 25
Under Seismic Actions:
Modeling and Analysis,
Fig. 4 Dependence of the 20
period of free vibrations on
the normalized initial tilt,
15
’0/y
pT
10
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
j0 / q
0.4
0.4
0.2
0.2
0.0
0 1 2 3 4 5 6 7 8
Number of impacts, n
Housner (1963) calculated that the amplitude ’n a critical value, x€g cr. At that time, the overturning
of the free vibrations after the nth impact is moment about O or O0 due to the d’ Alembert
inertial force, MA = m x€g cr h, becomes equal
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
’n ’
2 with the restoring moment due to the weight of
¼1 1 en 1 1 0
(14) the block, ME = mgb/2; therefore
y y
The plot of the dimensionless product pT ver- x€g cr ¼ ð tan yÞg (15)
sus the normalized initial tilt angle, ’0/y (Eq. 13),
is given in Fig. 4, while in Fig. 5 the decrease of For slender blocks (y < 20
), the critical base
the amplitude with the number of impacts is acceleration can be approximated by
plotted for e = 0.7 and various values of the
normalized initial tilt, ’0/y.
x€g cr ¼ yg (16)
Initiation of Rocking Under Earthquake
Excitation If the peak acceleration
of the base motion,
Under a seismic ground excitation x€g , rocking pga, is smaller than x€g cr , the earthquake is not
starts when the base acceleration reaches strong enough to initiate rocking of the block.
Ancient Monuments Under Seismic Actions: Modeling and Analysis 117
Ancient Monuments
Under Seismic Actions:
Modeling and Analysis,
Fig. 6 (a) El Centro A
(1940) earthquake. (b)
Rocking response of an
orthogonal block of
dimensions b = 0.50 m,
2 h = 1.5 m for the El
Centro earthquake
amplified to several values
of pga (for e = 0.85)
As evident from Eq. 15, the slenderer the block, 0.80 g the block overturns in the opposite direc-
the smaller is the required base acceleration to set tion (negative rotations). However, if the base
it into rocking motion. motion is amplified even more to pga = 0.90 g,
the response is stable again and overturning does
Nonlinearity and Sensitivity of the Response not occur (gray line).
As mentioned above, the rocking response is Apart from the nonlinearity, another charac-
highly nonlinear. This is illustrated in Fig. 6b, in teristic of the response is its sensitivity to even
which the time history of the angle of rotation of trivial changes of the parameters. This sensitiv-
an orthogonal block with dimensions ity has been proven by the non-repeatability
b = 0.50 m, 2 h = 1.50 m is shown for the El of the same experiment (Yim and Chopra
Centro (1940) earthquake (Fig. 6a) amplified to 1984). In Fig. 7, the sensitivity of the response
four different values of the peak ground acceler- of the abovementioned block to the value of the
ation: pga = 0.60 g, 0.70 g, 0.80 g, and 0.90 g. In coefficient of restitution e is shown. In this plot,
all cases, the coefficient of restitution was set to the response of the block is shown for the El
e = 0.85, which corresponds to Housner’s theo- Centro record amplified to pga = 0.50 g and
retical value according to Eq. 9. It is seen that the for three values of the coefficient of restitution:
response of the block is stable for pga = 0.60 g e = 0.85 (Housner’s value), e = 0.87, and
(blue line), while the block overturns in the direc- e = 0.88.
tion of positive rotations for pga = 0.70 g (green It is seen that the response for e = 0.87 (green
line). Increasing the base excitation to pga = line) is very similar with the one for e = 0.85
118 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Ancient Monuments
Under Seismic Actions:
Modeling and Analysis,
Fig. 7 Rocking response
of an orthogonal block of
dimensions b = 0.50 m,
2 h = 1.5 m for the El
Centro earthquake
amplified to pga = 0.50 g
for various values of the
coefficient of restitution e
(blue line), except for an additional small rocking g
’max g tan y
response of the block around t = 5 s, which does ¼ f pT p , , tan y, e (17)
pga pga
not occur for e = 0.85. However, if we slightly
increase the coefficient of restitution to e = 0.88, Assuming that e is known, Eq. 17 implies that:
intense rocking occurs after t = 4 s with signifi-
cantly larger amplitude than the amplitude in the • For a given base excitation (given pga and
time interval 2.0 < t < 3.5 s when all the Tp), the response depends on the slenderness
rocking response takes place for e = 0.85. It is y and the characteristic frequency p. The lat-
interesting to notice that this intense rocking for ter is a function of the size of the block. The
e = 0.88 occurs after the strong motion of the value of p decreases inversely with the size of
ground excitation (see Fig. 6a). the block (e.g., see Eq. 5 for an orthogonal
It must also be noted that although, in general, block), measured with the distance r. There-
a decrease in the value of e leads to smaller fore, for the same slenderness there is an
rocking amplitude, due to the larger dissipation important size effect on the response. Actu-
of energy during impact, it is also possible that ally, among two blocks with the same slen-
a smaller coefficient of restitution produces larger derness y but different size, the smaller one
rocking response (Aslam et al. 1980). This coun- will experience more intense rocking than the
terintuitive phenomenon is attributed to the larger one. This is shown in Fig. 8, in which
nonlinearity of the response. the response of two blocks with tan y = 0.5
but different size (b = 0.50 m for the left
Main Features of the Rocking Response block and b = 1.5 m for the right) is shown
Except for the abovementioned nonlinearity and for the same impulse base excitation. It is
the sensitivity of the response, an important fea- seen that the small block overturns, while
ture of the dynamic behavior is its dependence on the large one does not.
the dimensionless quantity pTp with Tp being the • For a given block (given y and p), the rocking
predominant period of the ground motion. Actu- response and the overturning risk greatly
ally, for harmonic excitation (which can be depend on the predominant period of the
extrapolated to pulse-like ground motions) the base excitation. In general (for details see
normalized response can be expressed solely as Zhang and Makris 2001; Dimitrakopoulos
a function of four dimensionless terms (Zhang and DeJong 2012), the required normalized
and Makris 2001; Dimitrakopoulos and DeJong amplitude of the base acceleration, pga/(gtan
2012): y), to cause overturning decreases as the
Ancient Monuments Under Seismic Actions: Modeling and Analysis 119
Ancient Monuments
Under Seismic Actions:
Modeling and Analysis,
Fig. 8 Rocking response A
of two similar orthogonal
blocks of the same
slenderness (tan y = 0.5)
and different size (base
width b = 0.50 m in the left
block and b = 1.50 m in
the right) for the same
impulse base excitation
Difficulties and Uncertainties of the Dynamic Ancient Monuments Under Seismic Actions: Model-
ing and Analysis, Fig. 9 Photo of a single standing and
Analysis
a fallen column of the Olympieion of Athens, Greece,
As mentioned in the Introduction, columns of showing the multi-drum construction of ancient columns
ancient monuments are made of individual struc-
tural elements, called drums, which are put one drums of the columns were aiming at centering
on top of the other without mortar or any other the stones during construction and, practically, do
connecting material (Fig. 9). The wooden dowels not have any effect on the seismic response.
that were usually placed at the joints between the In few cases steel connections (dowels) are
120 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Ancient Monuments
Under Seismic Actions:
Modeling and Analysis,
Fig. 10 The four rocking
“modes” of vibration of
a two-block assembly
(Psycharis 1990)
provided at the joints, which restrict, up to their governed by the motion of the stones, which can
yielding, sliding but do not affect, in general, rock and slide individually or in groups. In case of
rocking. cylindrical drums, wobbling also occurs during
Due to their spinal construction, ancient col- rocking. The dynamic analysis of such systems is
umns respond to strong earthquakes with intense a difficult task that can only be treated numeri-
rocking and sliding of the drums. In addition, due cally, since there are many different “modes” in
to the cylindrical shape of the drums, wobbling which the structure can vibrate. For example, for
also occurs during rocking. Usually, rocking dom- a system of two blocks, there are four “modes” of
inates the response of multi-drum columns, which, rocking vibrations, shown in Fig. 10 (Psycharis
for this reason, is characterized by the strong 1990). In case of multi-drum columns, the
nonlinearity and the sensitivity discussed in the corresponding modes increase exponentially
previous section. In addition, the analysis is with the number of the drums and, for typical
extremely difficult and complicated due to the columns with many drums, can be tens of
many “modes” of response in which multi-block thousands.
systems can respond, as it will be discussed in the It is noted that the term “modes” of vibration is
ensuing. used here to denote the different patterns of the
Furthermore, there is a number of other uncer- response. These “modes” of response must not be
tainties associated with the seismic analysis of confused with the “eigenmodes” of continuous
ancient monuments: the existing damage, which systems. Rocking structures do not possess natu-
might be crucial to the stability of the monument ral modes in the classical sense.
to future earthquakes but is difficult to implement Under an earthquake excitation, the response
in the numerical models; the difficulties in of the column continuously alternates from one
representing accurately the real geometry; and “mode” to another (Fig. 11). Each “mode” is
the “vague” properties that must be assigned to governed by a different set of equations of
the joints in the numerical models. All these motion, while criteria must be defined for the
issues will be discussed in the following; how- transition between “modes.” It is evident, there-
ever, taking into consideration the sensitivity of fore, that sophisticated numerical codes are
the rocking response to even small changes in the needed for the dynamic analysis of ancient mon-
parameters, one must have in mind that, no matter uments, able to take under consideration the slid-
how accurate an analysis is, there is always an ing of the drums, the opening of the joints
inherent uncertainty in the results. (rocking), and even complete separation and
recontact of the drums.
Analysis of the Dynamic Response In general, numerical models for the analysis
The dynamic response of multi-drum columns, of masonry structures may be classified into two
and, in general, of stacks of rigid blocks, is major conceptual classes: (i) equivalent
Ancient Monuments Under Seismic Actions: Modeling and Analysis 121
Ancient Monuments
Under Seismic Actions:
Modeling and Analysis,
Fig. 11 Response of two A
columns of Olympieion of
Athens at two different time
instances during intense
ground shaking. The
geometry of the two
columns is slightly
different (the left has
14 drums and the right 15)
leading to different
“modes” of vibration
(numerical results obtained
with 3DEC software)
continuum models, in which the influence of The method calculates the displacements and
the joints between blocks is introduced by the forces in the individual blocks and applies
means of special constitutive relations, and compatibility laws to detect new contacts
(ii) discontinuous models, in which the joints between the bodies. DEM employs an explicit
are represented explicitly, leading to an idealiza- algorithm for the solution of the equations of
tion of the structure model as a block assemblage. motion of the blocks, taking into account large
The finite element method is the preferred numer- displacements and rotations (Itasca 1998;
ical tool for models of the first type, while either Papantonopoulos et al. 2002).
finite element implementations with joint ele- Comparison of experimental (shaking table)
ments or discrete element methods are capable data concerning the seismic response of
of handling the second type. a marble drummed replica of a column of the
For stone masonry structures, particularly Parthenon in 1:3 scale with the numerical results
those composed of multi-drum columns and produced by 3DEC (Papantonopoulos et al. 2002)
architraves as ancient monuments, block models showed satisfactory agreement in the maximum
are an obvious option. Deformation and failure of displacements during the seismic motion and in
these structures is mainly governed by the rela- the residual deformation of the column (Fig. 12).
tive movements between blocks. The blocks can Exact agreement could not be obtained, due to the
be assumed rigid without significant loss of sensitivity of the dynamic response to even trivial
precision. changes in the parameters (Mouzakis et al. 2002;
One method that has been proven to be very Dasiou et al. 2009a). For example, repetition of
efficient in dealing with such systems is the dis- the same experiment led to different results in
tinct (or discrete) element method (DEM) intro- many cases. An example is shown in Fig. 13, in
duced by Cundall in the 1970s in the context of which the recorded response during two similar
rock mechanics and later extended to three- experiments (EQ14 and EQ15) is depicted, show-
dimensional problems (Cundall 1988). The code ing significantly different residual deformation.
3DEC (Itasca 1998), which was used for most of The differences between these “similar” experi-
the analyses presented herein, is based on DEM. ments were (i) trivial differences in the initial
This method provides the means to apply the geometry of the column due to slightly different
conceptual model of a masonry structure as (not visible) relocation of the drums at their initial
a system of blocks. The system deformation is position and (ii) small difference (less than 2 %)
concentrated at the joints, where frictional sliding in the shaking table motion. It is interesting
or complete separation may take place. to note that the numerical analyses also
122 Ancient Monuments Under Seismic Actions: Modeling and Analysis
a 200 b 100
Experiment Experiment
Absolute Displacement (mm)
0 0
−100 −50
−200 −100
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (sec) Time (sec)
Ancient Monuments Under Seismic Actions: Model- a 1:3 replica of a column of the Parthenon under seismic
ing and Analysis, Fig. 12 Comparison of experimental excitation: (a) x-direction; (b) y-direction
(shaking table) and numerical (3DEC) results concerning (Papantonopoulos et al. 2002)
the horizontal displacement at the top of the capital of
a 100 b 100
EQ14 EQ14
Relative Displacement (mm)
Experiment Experiment
50 3DEC 50 3DEC
0 0
−50 −50
−100 −100
4 8 12 16 20 4 8 12 16 20
Time (sec) Time (sec)
Ancient Monuments Under Seismic Actions: Model- Parthenon under seismic excitation for the same experi-
ing and Analysis, Fig. 13 Horizontal displacement at ment repeated twice (Papantonopoulos et al. 2002)
the top of the capital of a 1:3 replica of a column of the
showed different response, caused by the the following). In this sense, experimental data of
abovementioned small difference in the base the seismic response of multi-drum columns are
excitation (the geometry was exactly the same extremely valuable since they can be used for the
in the numerical models). These comparisons, calibration of the numerical models.
and additional ones concerning more compli-
cated structural systems comprising of three col- Modeling Aspects
umns connected with architraves (Dasiou Regardless of the numerical method that will be
et al. 2009b), showed that DEM, and specifically used for the analysis of the seismic response,
3DEC, can reliably predict the seismic response assigning values to the parameters of the numer-
of classical monuments. ical model is not a straightforward procedure.
It must be emphasized that the application of Due to the sensitivity of the response, the param-
sophisticated numerical models for the prediction eters of the numerical model and the more or less
of the seismic response of ancient columns and accurate geometrical representation of the struc-
colonnades requires the knowledge of the value ture can affect the results significantly. For this
of several parameters which, in general, are not reason, experimental data are generally needed in
known a priori (this issue is discussed in detail in order to calibrate these parameters, especially the
Ancient Monuments Under Seismic Actions: Modeling and Analysis 123
ones concerning the properties at the joints. In the mk, where mk ms. This change in friction
following, several critical aspects related to the depends not only on the normal stresses on the
sensitivity of the response to several parameters drum interface during sliding but also on the A
are discussed. number of reversals and the amount of cumula-
tive slip. A considerable number of laboratory
Joint Parameters test results on surfaces of limestone and marble
In the distinct element method, the interaction specimens of different roughness can be found in
forces between two blocks are applied at a set of literature. Although the test procedures differ, the
contact points located at the vertex-to-face and evidence is that, with increasing normal stress sn,
the edge-to-edge intersections. These forces the static friction coefficient increases, while the
depend on the relative displacement between the difference (ms – mk) decreases. This means that,
blocks according to the constitutive model other things being equal, it is easier for drums to
adopted for the joints. In the normal direction, slide in the upper parts of a multi-drum column
the joint behavior is governed by the normal than near its pedestal and also that, once sliding
stiffness coefficient, kn, which relates the contact begins, the drop of the static coefficient ms to the
stress with the normal contact displacement. No residual value of mk may lead to larger and more
tensile strength is considered, so this spring ele- rapid displacement in the upper parts than at
ment is only active in compression. In the shear lower levels of the shaft.
direction, an elastoplastic stress-displacement The proper assessment of the joint parameters is
law is usually assumed. The elastic range is char- essential for the analysis, since they affect the
acterized by the shear stiffness, ks, while the shear response significantly. An example of the effect
strength is governed by the Coulomb friction of the value of kn and ks on the response of a
coefficient, m, with no cohesive strength six-drum column with a two-piece capital is illus-
component. trated in Fig. 14 for three different combinations of
The appropriate values to be assigned to the the normal and shear coefficients (Toumbakari and
stiffness coefficients kn and ks depend on the Psycharis 2010). It is evident that the values of kn
material of the drums. Since typical values for and ks affect both the amplitude of the response
various materials are not provided in the litera- and the residual deformation of the column.
ture, the assignment of these values is not easy. If Significant might also be the effect of the
experimental data are available, the coefficients coefficient of friction. An example is shown in
can be calibrated against these data. However, Fig. 15 for the column of Olympieion of Athens
such data are very few and concern specific mate- (the existing dowels at the joints were not con-
rials only (usually marble). It is noted that loose sidered in this analysis) under a strong earthquake
contacts at the joints due to deteriorations of the motion. Three values of the coefficient of friction
contact surfaces might also affect the results. were examined: m = 0.55, 0.75, and 1.05. It is
One way to overcome this difficulty is to cal- noted that the typical value of m for marble is
ibrate kn and ks so that the natural period of the about 0.70–0.75; thus, the value m = 0.55 is con-
column for small amplitude oscillations matches sidered rather low, while the value m = 1.05 is
the one determined by ambient vibration mea- unrealistically high. It is seen that the value of the
surements. It is noted that for low-amplitude coefficient of friction affected mainly the
vibrations multi-drum columns behave like con- response and the residual dislocation of the
tinuous systems and do possess natural modes. drums at the upper part of the column, where
This approach has the advantage that the effect of the developed accelerations were large enough
existing imperfections at the joints can be to initiate sliding.
included in the joint stiffness.
Concerning the coefficient of friction, it varies Damping
from a static value ms at the initiation of sliding to The dissipation of energy due to sliding is auto-
a lower value, which degrades to the kinetic value matically taken under consideration with the
124 Ancient Monuments Under Seismic Actions: Modeling and Analysis
20 Case #4
Displacement [cm]
Case #5
Case #6 Case kn ks
10 [Pa/m] [Pa/m]
4 5x109 1x109
0 5 0.5x109 0.1x109
9
6 1x10 0.2x109
−10
0 10 20 30
Time [s]
Ancient Monuments Under Seismic Actions: Model- column for three different sets of the joint stiffness
ing and Analysis, Fig. 14 Time histories of the dis- (Toumbakari and Psycharis 2010)
placement at the top of the capital of a freestanding
Capital
Drum #13
Drum #10
Base
Capital
Drum #13
Drum #10
Base
Ancient Monuments Under Seismic Actions: Modeling and Analysis, Fig. 15 Time histories of the displacement
at various positions along the height of a freestanding column for three different values of the coefficient of friction
elastoplastic model of the shear stiffness. dissipation of energy due to the impacts between
However, since elastic properties are assumed drums during rocking.
at the joints, extra damping must be introduced Shaking table experiments on multi-drum col-
to the numerical model to account for the umns showed very low attenuation. These results
Ancient Monuments Under Seismic Actions: Modeling and Analysis 125
a 200 b 200
Absolute Displacement (mm)
0 0
−100 −100
−200 −200
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (sec) Time (sec)
Ancient Monuments Under Seismic Actions: Modeling and Analysis, Fig. 16 Comparison of numerical results
with experimental data for (a) zero damping and (b) 0.5 % stiffness-proportional damping (Papantonopoulos et al. 2002)
lead to the conclusion that, to be conservative, the modeling is practically impossible, due to the
numerical simulations should be performed with complicated geometry of the architectural details
very low or even zero damping, at least during the of the columns and the random shape of existing
strong shaking. This is illustrated in Fig. 16, in damage. In some cases, even the gross geometry
which the numerical results for zero damping and (number and dimensions of drums) might not be
0.5 % stiffness-proportional damping are com- easy to implement due to lack of information,
pared with experimental data (Papantonopoulos since, in most monuments, the height of each
et al. 2002). It is seen that the introduction of even drum and sometimes even the number of the
a small value of damping decreased unrealisti- drums of the columns are not constant (see
cally the amplitude during the strong shaking. Fig. 9). This happens because drums were made
However, it is necessary to introduce some of marble of superior quality and the height of
damping towards the tail of the response in each drum depended on the available marble
order to attenuate the free vibrations and be able pieces, as ancient builders were trying to avoid
to calculate the residual deformation of the any unnecessary loss of material.
column. In addition, numerical models containing
In general, whenever damping is used it is large number of blocks might require extremely
preferable to use the stiffness-proportional com- high computational time; thus, it might be desir-
ponent of Rayleigh damping. However, explicit able to simplify the model as much as possible.
time-stepping algorithms, such as the one in An example of the effect of the number of
3DEC, require rather small time steps if drums considered in the analysis on the response
stiffness-proportional damping is used. Thus, in of a column with a statue on top is given in Fig. 17
order to avoid the increase in the computational (Ambraseys and Psycharis 2011). In this case, the
effort, only the mass proportional component of investigation concerned the risk of toppling of the
Rayleigh damping can be adopted, with a low statue. The column consisted of 15 drums, but
value though. analyses were also performed for columns
of 5 drums, 30 drums, and 60 drums (see top of
Required Accuracy of the Geometric Fig. 17). In Fig. 17, the uplift of one corner of the
Representation statue is shown. It is seen that the response varied
One question that arises in conducting numerical in each case while the statue overturned only in
analyses of ancient monuments is how accurately the case of the column with 15 drums.
the model must represent the actual geometry. As It is evident therefore that the geometry must
a general rule, the geometry should be be implemented as accurately as possible in order
implemented as accurately as possible, due to to reach correct conclusions. However, if only
the sensitivity of the response. However, exact gross results are sought, it might be adequate to
126 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Ancient Monuments Under Seismic Actions: Model- 15-drum column, (c) 30-drum column, and (d) 60-drum
ing and Analysis, Fig. 17 Time histories of the vertical column. Overturning of the statue occurs only in case (b)
displacement of corner 1 of the base of the statue (blue (Ambraseys and Psycharis 2011)
line) and the capital (red line): (a) 5-drum column, (b)
apply very simple geometries, which can be ana- original columns. It is seen that the very simple
lyzed much easier. An example is given in single-block representation could predict with
Fig. 18, in which the required amplitude of a har- acceptable accuracy the overturning risk of the
monic excitation to cause overturning versus the columns and thus it could be used as a first-order
period of the excitation (stability threshold) is approximation within a decision making
shown for (a) the column of the temple of Zeus procedure.
at Nemea, Greece, and (b) the column of the
temple of Apollo at Bassae, Greece (Psycharis Representation of Existing Damage
et al. 2000). The results obtained for the actual In their current condition, ruins of ancient struc-
geometry of the multi-block columns are com- tures present many different types of damage.
pared with the ones for the equivalent single- Most common are missing pieces (cutoffs) that
block columns, i.e., the monolithic columns reduce the horizontal sections in contact, founda-
with the same overall dimensions with the tion problems resulting in tilting of the columns,
Ancient Monuments Under Seismic Actions: Modeling and Analysis 127
Ancient Monuments
Under Seismic Actions:
Modeling and Analysis,
Fig. 18 Comparison of the A
stability threshold for the
overturning of the multi-
drum column and the
equivalent single-block
column: (a) column of the
temple of Zeus; (b) column
the temple of Apollo
(Psycharis et al. 2000)
dislocated drums from previous earthquakes, It is evident therefore that the damage that can
and cracks in the structural elements that in be observed today in the monuments must be
some cases split the block in two. Such imper- implemented in the numerical models. It should
fections affect significantly the stability of be mentioned, however, that this is not an easy
the columns which, thus, are much more vulner- task for several reasons: (i) because the existing
able to earthquake excitations compared with damage has not always been mapped in the
the original intact structures. Therefore, the required detail, (ii) because the shape of missing
abovementioned impressive stability of ancient pieces is irregular and very difficult to be
monuments against earthquakes might not be modeled accurately, and (iii) because certain
a valid assumption any more, if significant dam- types of damage, as cracks in the stone blocks,
age is present. are either unknown or impossible to take under
An example of the effect of existing imperfec- consideration.
tions on the stability of ancient columns is shown
in Fig. 19 for the column of the Parthenon in Modeling of Clamps and Dowels
Athens (Psycharis et al. 2003). The maximum In ancient monuments, iron clamps and dowels
permanent displacement of the column is plotted exist between the beams of architraves and the
versus the pga of the ground motion, and it is seen stones of walls. In columns, iron dowels were
that the presence of the imperfections shown in rarely put at the joints, connecting adjacent
the left drawing of Fig. 19 leads to larger dis- drums. In such cases, these connectors must be
placements and significantly earlier collapse. included in the numerical models because they
128 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Collapse
1000
displacement (mm)
800
600 no
imperfections
400
200 with
imperfections
0
0 0.5 1 1.5 2
pga (g)
Ancient Monuments Under Seismic Actions: Model- Aigion, Greece (1995), earthquake amplified to several
ing and Analysis, Fig. 19 Maximum permanent dis- values of pga without and with the imperfections shown in
placements of a column of the Parthenon under the the left diagram (Psycharis et al. 2003)
0.00
−0.50
0 4 8 12 16 20
Time(sec)
0.40
Column 7.5 - Capital
Kalamata earthquake - PGV= 80cm/sec
Displacement in E-W direction
0.20 A
Displacement (m)
0.00
−0.20
−0.40
0 5 10 15 20
N
W Time (sec)
E 0.40
S Column 7.5 - Capital
Kalamata earthquake - PGV= 80cm/sec
Displacement in N-S direction
0.20
Displacement (m)
0.00
−0.20
−0.40
0 5 10 15 20
Time (sec)
Ancient Monuments Under Seismic Actions: Model- direction – top diagram) and the out of plane (N–S
ing and Analysis, Fig. 21 Olympieion of Athens: dis- direction – bottom diagram)
placement at the top of the E column in the in plane (E–W
It must also be noted that, for large rocking directions. For cylindrical blocks, the pole of
angles, it is possible that a dowel disengages from rotation continuously changes its position, mov-
the upper drum. In that case, the dowel might not ing along the perimeter of the base of the drum
be able to reinsert in the mortise during the (wobbling). It is interesting to note that it was
reversed motion due to the wobbling and the observed during shaking table experiments for
sliding of the drums, blocking thus the proper purely plane excitation that even small distur-
sitting of the upper drum. In most cases, however, bances in the direction normal to the plane of
shear dowels are inserted several centimeters in rocking could cause significant amplification of
the drums, and, thus, it is not probable that the response in that direction (Papantonopoulos
rocking can cause their disengagement, even for et al. 2002).
strong ground motions. In general, comparisons between 2D and 3D
analyses show that the 2D approach underesti-
2D Versus 3D Analysis mates the response, predicting greater stability.
For freestanding columns, and despite the sym- For systems involving many columns in line
metry of the column about the vertical axis, there or in corner (colonnades), 3D analyses are evi-
is a significant difference in the response between dently necessary. In such cases, it has been
two-dimensional and three-dimensional analysis. observed that the end column of each row suffers
Two-dimensional analysis is unable to capture all significant out-of-plane deformation which, for
the aspects of the real response, mainly the rota- strong earthquake motions, might be significantly
tion of the drums around the vertical axis due to larger than the in-plane one. An example is
the simultaneous rocking in two normal shown in Fig. 21 for a set of two columns of the
130 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Ancient Monuments Under Seismic Actions: Model- columns under harmonic excitations: (a) columns of the
ing and Analysis, Fig. 22 Stability threshold for the temple of Apollo; (b) columns of the temple of Zeus
collapse of the freestanding column and the set of two (Psycharis et al. 2000)
Olympieion of Athens connected with an archi- period of excitation, significantly larger acceler-
trave. Under a three-dimensional earthquake ation is needed to overturn the larger columns of
excitation, the displacement at the top of the Zeus compared with the smaller columns of
east column was much larger in the out-of-plane Apollo.
direction (bottom diagram) than in the in-plane Another interesting observation is that the sta-
one (top diagram). bility threshold of each monument was similar for
the freestanding column and the set of two col-
Size Effect umns. This means that restoration of fallen archi-
Similarly to the single rocking block (see section traves does not necessarily lead to enhanced
“Main Features of the Rocking Response”), the stability of the monument against future earth-
size of ancient monuments affects their dynamic quakes. Figure 22 shows that such restoration of
response and their vulnerability to earthquakes, the architraves might be favorable or unfavorable
with larger structures being more stable than depending on the characteristics of the structure:
smaller ones. This is shown in Fig. 22, in which in the case of Apollo, it was generally unfavor-
the minimum required acceleration amplitude of able; in the case of Zeus, it was generally
a harmonic excitation of varying period to cause favorable.
collapse (stability threshold) is shown for two It should be mentioned that the above obser-
cases: (a) the columns of the temple of Apollo vation concerns the in-plane collapse of the col-
at Bassae, Greece, of height 5.95 m and (b) the umns (2D analyses). However, shaking table tests
columns of the temple of Zeus at Nemea, Greece, on sets of columns connected with architraves in
of height 10.33 m (Psycharis et al. 2000). Results line or in corner have shown that the architrave
are given for the freestanding column and the set beams are the most vulnerable parts of monu-
of two columns. It is seen that, for the same ments, as they are the first pieces that fall down
Ancient Monuments Under Seismic Actions: Modeling and Analysis 131
(in the out-of-plane direction). The collapse of arbitrarily. The identification of active or poten-
the architraves endangers the stability of the tially active faults, including blind faults, the
whole monument, since it is possible that they long-term seismic history of the region and reli- A
hit the columns during their fall. able magnitudes as well as regionally representa-
tive data on strong ground motions are important
parameters that must be considered in this choice.
Selection of Ground Motions Among the important characteristics of ground
motion, time histories to be used in 3D numerical
The earthquake response of ancient monuments analyses are their maximum velocity and associ-
is dominated by the rocking of the drums of the ated period as well as the directivity effects which
columns. As mentioned in section “Main Fea- will be present depending on the proximity of the
tures of the Rocking Response,” rocking is monuments to a potentially active fault.
greatly affected by the predominant period of From the geological point of view, three inde-
the excitation. It is evident, therefore, that the pendent source parameters are required to
vulnerability of a monument depends on the fre- describe an earthquake: the fault dimension, the
quency content of the ground motion, with long- static moment M0, and either the mean stress or
period earthquakes being much more dangerous energy released by faulting or strain energy
than high-frequency ones. This is shown in change. For sites close to the causative fault, the
Fig. 22, in which the stability threshold decreased ground motions will show pronounced directivity
exponentially as the period of excitation effects in their ground velocity and displace-
increased in all cases. Previous analyses ments. These will appear as long-period ampli-
(Psycharis et al. 2000) have shown that tude pulses linearly related to the seismic
low-frequency earthquakes force the structure to moment. Additionally, the effect of the source
respond with intensive rocking, whereas high- mechanism on directivity will be that ground
frequency ones produce significant sliding of accelerations due to strike-slip faulting will be
the drums, especially at the upper part of the larger than those due to normal faulting, while
structure. ground velocities and displacements will be
It is evident that the choice of the earthquakes larger for thrust faulting compared with normal
that will be used in the analyses is very important, earthquakes.
as the dynamic response of multi-drum columns It is evident, therefore, that the choice of
and colonnades and the danger of collapse are which time histories to include and which to
sensitive to the energy and frequency content of exclude in order to constrain ground motions is
the time history of the input ground motion. Apart an important decision. There is a balance to be
from the abovementioned strong effect of the struck between being not restrictive enough in the
predominant period of the ground motion, the time histories used, leading to unreliable results
time sequence of the various phases in the record and hence predictions due to errors and uncer-
might also be significant. In this sense, it is essen- tainties, and being too restrictive, which leads to
tial to constrain the selection of the base excita- a too small set of time histories and hence
tions to what one may call suitable surrogate nonconclusive results.
ground acceleration time histories that could rep-
licate as closely as possible the time histories of
past and anticipated earthquakes. Summary
Records must be chosen from past earthquakes
which are associated specifically with the tecton- In this chapter the main characteristics of the
ics and the seismicity of each region or with response of ancient monuments to earthquakes
earthquakes from other regions of similar tecton- and the difficulties and uncertainties encountered
ics. They may not be always available or in suf- in the analysis, mainly in what regards the values
ficient numbers, but they should not be selected assigned to the parameters of the numerical
132 Ancient Monuments Under Seismic Actions: Modeling and Analysis
Earthquake archeology can be considered as earthquake activity over millennial time spans,
a synonym for archeoseismology. But earthquake archeoseismology can indeed extend the archive
archeology can also be seen to serve objectives of earthquakes beyond written sources, thus
proper to archeology, i.e., reconstructing human becoming a legitimate and complementary
history, in particular attempting to better under- source of seismic-hazard information.
stand the true impact of earthquakes on human After a short historical note, a summary of the
history. The term earthquake archeology can be different types of archeological evidence for
traced to the Japanese term jishin kōkogaku, ancient earthquakes, commonly used in
referring to a research field developed in the archeoseismology, is given. Subsequently, the
mid-1980s in Japan primarily through the initia- strengths, challenges, and pitfalls of
tive of Sangawa Akira, a geomorphologist at the archeoseismology are discussed. Some new
Geological Survey of Japan, focusing on sedi- developments in archeoseismology will be intro-
ment deformation features within archeological duced. In conclusion, some issues and perspec-
contexts (Barnes 2010). tives in archeoseismology are presented.
Since the book Archeoseismology (Stiros and
Jones 1996), a series of special issues of journals
has reflected the evolution of the burgeoning A Historical Note
discipline over the last two decades towards an
ever increasing multidisciplinary discipline In the first volume of the Palace of Minos,
(McGuire et al. 2000; Galadini et al. 2006a; published in 1921 (Evans 1921), Sir Arthur
Caputo and Pavlides 2008; Sintubin et al. 2010; Evans did not mention earthquakes as a possible
Silva et al. 2011). cause for the destructions observed during exca-
In the current entry, archeoseismology is con- vation works at the Bronze Age, Minoan site of
sidered as a discipline belonging to the broad Knossos (Crete, Greece). In the second volume of
research realm of earthquake sciences, reflected the Palace of Minos, published in 1928 (Evans
by a continuum of overlapping and complemen- 1928), though, tectonic earthquakes became the
tary research fields, each focusing a particular primary destructive agent, not only leaving
source of earthquake data, applying appropriate a clear marker horizons in the archeological stra-
methods and techniques, and targeting a specific tigraphy but also causing cultural change as
time window. In this respect, archeoseismology evidenced by discontinuities in ceramic style
bridges the gap between instrumental and and architecture. So, what happened in those
historical seismology on the one side and 7 years that completely changed Evans’ thinking?
paleoseismology and earthquake geology on On 20 April 1922, during the excavation of the
the other. Archeoseismology focuses on cultural “House of the Sacrificed Oxen” and the “House
material data spanning the last few millennia. It of the Fallen Blocks” (Fig. 1) at Knossos, Evans
shares, however, a common goal with the other experienced an earthquake, leading him to
disciplines, i.e., a better understanding of the believe that earthquakes of tectonic nature
earthquake history within a region in an attempt (so not related to the volcanic activity of the
to assess the seismic hazard and mitigate the Thera/Santorini volcano) may very well have
seismic risk. Most valuable contribution of caused damage to the Minoan buildings during
archeoseismology to seismic hazard assessment the Bronze Age. But only after experiencing the
is situated in earthquake-prone regions with dramatic earthquake that hit the Eastern Mediter-
a long and lasting cultural heritage. Seismic- ranean on 26 June 1926 and caused severe dam-
hazard practitioners are confronted with the prob- age in the region of Heraklion, Sir Arthur Evans
lem that the instrumental record is too short (only got convinced that earthquakes are the primary
spanning somewhat over a century) and the his- destructive agent responsible for the main stages
torical record too incomplete or even inexistent. of destruction observed at Knossos. He devel-
By having the potential of determining oped a seismic archeological stratigraphy,
Archeoseismology 135
Ancient Earthquakes
Archeoseismology, Fig. 2 The wall of the Crusader 2 m. The fortress has been built astride the Dead Sea
Fortress of Vadum Iacob (Ateret, Israel) (cf. Marco transform plate boundary between the African and Ara-
et al. 1997) is displaced left laterally over more than bian plates # Sintubin
partly displaced due to coseismic surface ruptur- The most spectacular cases of such faulted
ing on an active fault. As already mentioned, it is relics can be found astride strike-slip faults, in
definitively not a coincidence that archeological which case there can be no doubt of the tectonic
sites are astride active faults. nature of the displacement. Notorious examples
As cultural piercing features, these faulted are the Crusader Fortress of Vadum Iacob
relics not only serve to identify active faults, (Israel) (Fig. 2) (cf. Marco et al. 1997) and the
but they are also used to determine the type of Al Harif Roman aqueduct (Syria) (cf. Sbeinati
faulting (normal, reverse, strike slip), the amount et al. 2010), both located astride the Dead Sea
of coseismic slip related to a single earthquake, as fault, a left-lateral transform plate boundary
well as the cumulative fault slip. All these data between the African and Arabian plates. By
eventually allow to derive time-averaged fault- the way, as long, linear structures, aqueducts
slip rates over time spans of centuries to are ideal cultural piercing features because
millennia, very comparable to paleoseis- chances are they cross more than one fault.
mological work. The fault-slip rates obtained Also in extensional settings, giving rise to the
from an archeological context can subsequently typical seismic landscapes of the Mediterranean
be compared to long-term slip rates from (from Turkey to Spain), faulted relics are often
paleoseismological work, enabling the evaluation encountered in archeological sites (Fig. 3)
of potential slip deficits and thus increased seis- (e.g., Similox-Tohon et al. 2006). In these
mic hazards. cases of dip-slip, normal-fault movements,
138 Archeoseismology
Archeoseismology, Fig. 3 The mosaic floor of the activity (terminus post quem) can be inferred based on
Roman Neon Library at Sagalassos (SW Turkey) has the archeological evidence (cf. Similox-Tohon et al. 2006)
been cut by a normal fault, of which a post-363 AD # Sintubin
Archeoseismology, Fig. 4 In situ broken vessels, that has been attributed to an earthquake in the thirteenth
evidencing a collapse on a “living surface,” as part of century BC (cf. Jusseret and Sintubin 2012) # J. Driessen
a destruction horizon at the Minoan site of Sissi (Crete)
Archeoseismology, Fig. 5 Recycling of building material in repair works of a wall at Susita (Golan Heights)
(cf. Sintubin 2011), possibly after a destructive earthquake # Sintubin
archeological sites, again giving rise to the Finally, other archeological evidences, such
danger of amalgamating regionally distinct as repairs, recycling of building materials
earthquakes. (Fig. 5), complete or partial abandonment, and
Other issues concerning destruction horizons architectural and/or cultural changes, can further
are preservation and disturbance. Little is known contribute to the identification of ancient
about the way ancient societies coped with the earthquakes.
aftermath of a major earthquake. Earthquake
debris may have been cleared from streets and Structural Damage Due to Ground Shaking
buildings and disposed at particular dumpsites, so and Ground Motion
that the earthquake destruction horizon is no lon- A third type of earthquake evidence in the
ger preserved in the archeological record. Valu- archeological record are typical strain struc-
ables and/or victims may have been recovered tures in the building fabric that are primarily
from the debris, while material may have been caused by coseismic ground shaking and ground
reused in rebuilding, leaving behind a highly dis- motion. These earthquake-related damage
turbed earthquake destruction horizon. It is there- features are most conspicuous on monumental
fore fair to conclude that the visibility of buildings and constructions, such as temples,
earthquake destruction in the archeological stra- fountains, theaters, basilicas, pavements, col-
tigraphy is rather dependent on social factors than umns, statues, aqueducts, etc.
on physical parameters (cf. Jusseret and Sintubin A first systematic inventory of possible
2012). earthquake-related damage typologies has been
Archeoseismology 141
Archeoseismology, Fig. 7 A dropped keystone in the Roman Baths at Sagalassos (SW Turkey), interpreted as
a potential earthquake archeological effect (cf. Similox-Tohon et al. 2006) # Sintubin
irrefutably distinguish between damage caused integration of principles and practices of a very
by earthquakes and that caused by other natural wide range of sciences, from history, anthropol-
agents or human intervention, the main issue ogy, archeology and sociology, over geology,
archeoseismology is confronted with is the ques- geomorphology, geophysics, and seismology to
tion how earthquake evidence in destruction hori- architecture and structural engineering. Arguably
zons and/or disturbed buildings can be the principle difficulty in archeoseismology is
meaningfully translated into physical earthquake the lack of a shared protocol and of a rigorous
parameters, such as intensity, magnitude, dis- and transparent standardized methodology
tance to epicenter, date of earthquake, ground (cf. Sintubin and Stewart 2008 and references
acceleration, etc. Ultimately, because the limita- therein).
tions of the archeoseismological record are all too Through the years, different, primarily quali-
obvious when it comes to claiming its potential tative, archeoseismological schemes have been
role in seismic-hazard studies, it all boils down to proposed, consisting of points of interest (Karcz
the question what the true added value is of and Kafri 1978; Rapp 1986; Nikonov 1988;
archeoseismology. Stiros 1996), key research questions (Guidoboni
1996), or flow charts (Galadini et al. 2006b) that
In Search of a Shared Protocol and ought to be considered during excavation works
Standardized Methodology at an archeological site by collaborative teams
Archeoseismology’s greatest challenge – and its of seismologists, geologists, archeologists, etc.
foremost attraction – remains to date the (see Sintubin and Stewart 2008 for synthetic
Archeoseismology 143
Besides this added value of quantitative indeed shed light on the societal factors defin-
archeoseismology with respect to monumental ing the resilience or vulnerability of past socie-
architecture, also the value of rubble architecture ties. Eventually, by highlighting how our A
and associated destruction horizons get again ancestors coped with earthquake disasters,
particular interest in a more quantitative, inte- archeoseismology could find new aspirations
grated territorial approach that starts from the in establishing local earthquake cultures in
specific seismotectonic context and the empirical earthquake-prone regions (cf. Sintubin et al.
ground-motion relationships of potential earth- 2008), possibly providing a substantial contri-
quake sources and focuses on well-documented, bution to the mitigation of the earthquake
high-visibility archeological contexts, charac- risk, by improving earthquake risk literacy
terized by very rapid ceramic change (~100- and awareness.
year time window) (cf. Jusseret and Sintubin
2012).
Summary
Guidoboni E (1996) Archaeology and historical seismol- Rucker JD, Niemi TN (2010) Historical earthquake
ogy: the need for collaboration in the Mediterranean catalogues and archaeological data: achieving synthe-
area. In: Stiros S, Jones RE (eds) Archaeoseismology. sis without circular reasoning. In: Sintubin M,
Institute of Geology & Mineral Exploration and Brit- Stewart IS, Niemi TM, Altunel E (eds) Ancient earth- A
ish School of Athens, Athens, pp 7–13 quakes, Geological Society of America special paper,
Hinzen K-G (2005) The use of engineering seismological 471. Geological Society of America, Boulder,
models to interpret archaeoseismological findings in pp 97–106
Tolbiacum, Germany: A case study. Bull Seismol Soc Sbeinati MR, Meghraoui M, Suleyman G, Gomez F,
Am 95:521–539 Grootes P, Nadeau MJ, Al Najjar H, Al-Ghazzi
Hinzen K-G, Fleischer C, Reamer SK, Schreiber S, R (2010) Timing of earthquake ruptures at the Al
Sch€utte S, Yerli B (2011) Quantitative methods in Harif Roman aqueduct (Dead Sea fault, Syria) from
archaeoseismology. Quat Int 242:31–41 archaeoseismology and paleoseismology. In:
Jackson J (2006) Fatal attraction: living with earthquakes, Sintubin M, Stewart IS, Niemi TM, Altunel E (eds)
the growth of villages into megacities, and earthquake Ancient earthquakes, Geological Society of America
vulnerability in the modern world. Philos Trans R Soc special paper, 471. Geological Society of America,
364:1911–1925 Boulder, pp 243–267
Jusseret S, Sintubin M (2012) All that rubble leads to Schaeffer CFA (1948) Stratigraphie Comparée et
trouble: reassessing the seismological value of archae- Chronologie de l’Asie Occidentale. Oxford University
ological destruction layers in Minoan Crete and Press, London
beyond. Seismol Res Lett 83(4):736–742 Silva PG, Sintubin M, Reicherter K (2011) Earthquake
Jusseret S, Sintubin M (2013) The origins of an old myth: archaeology and paleoseismology. Quat Int
Sir Arthur Evans, Claude Schaeffer and the seismic 242(1):1–253
destruction of Late Bronze Age Eastern Mediterranean Similox-Tohon D, Sintubin M, Muchez P,
civilizations. Seismol Res Lett 84(1):94–100 Vanhaverbeke H, Verhaert G, Waelkens M (2005)
Karcz I, Kafr, U (1978) Evaluation of supposed Identification of a historical morphogenic earthquake
archaeoseismic damage in Israel. J Archaeol Sci through trenching at ancient Sagalassos (SW Turkey).
5:237–253 J Geodyn 40:279–293
Marco S, Agnon A, Ellenblum R, Eidelman A, Basson U, Similox-Tohon D, Sintubin M, Muchez P, Verhaert G,
Boas A (1997) 817-year-old walls offset sinistrally 2.1 Vanneste K, Fernandez M, Vandycke S,
m by the Dead Sea transform, Israel. J Geodyn Vanhaverbeke H, Waelkens M (2006) The identifica-
24:11–20 tion of an active fault by a multidisciplinary study at
McGuire WG, Griffiths DR, Hancock PL, Stewart IS (eds) the archaeological site of Sagalassos (SW Turkey).
(2000) The archaeology of geological catastrophes, Tectonophysics 420:371–387. [+Erratum in
Special publications, 171. Geological Society, London Tectonophysics, 435 (2007), 55–62]
Michetti AM, Hancock PL (1997) Paleoseismology: Sintubin M (2011) Archaeoseismology: past, present and
understanding past earthquakes using quaternary geol- future. Quat Int 242:4–10
ogy. J Geodyn 24(1–4):3–10 Sintubin M, Stewart IS (2008) A logical methodology for
Michetti AM, Esposito E, Guerrieri L, Porfido S, Serva L, archaeoseismology: a proof of concept at the archaeo-
Tatevossian R, Vittori E, Audemard F, Azuma T, logical site of Sagalassos, southwest Turkey. Bull
Clague J, Comerci V, G€ urpinar A, McCalpin J, Seismol Soc Am 98(5):2209–2230
Mohammadioun B, Mörner NA, Ota Y, Roghozin Sintubin M, Stewart IS, Niemi TM, Altunel E (2008)
E (2007) Intensity scale ESI 2007. Memorie descrittive Earthquake archaeology – just a good story? Seismol
della Carta Geologica d’Italia 74:41 Res Lett 79(6):767–768
Nikonov AA (1988) On the methodology of Sintubin M, Stewart IS, Niemi TM, Altunel E (eds)
archeoseismic research into historical monuments. (2010) Ancient earthquakes, Geological Society of
In: Marinos PG, Koukis GC (eds) Engineering Geol- America special paper, 471. Geological Society of
ogy of Ancient Works, Balkema, Rotterdam, America, Boulder
pp 1315–1320 Stiros SC (1996) Identification of earthquakes from
Nur A, Cline EH (2000) Poseidon’s horses: plate tectonics archaeological data: methodology, criteria and limita-
and earthquake storms in the late bronze age aegean tions. In: Stiros S, Jones RE (eds) Archaeoseismology.
and eastern mediterranean. J Archaeol Sci 27:43–63 Institute of Geology & Mineral Exploration and Brit-
Rapp GJ (1986) Assessing archaeological evidence for ish School of Athens, Athens, pp 129–152
seismic catastrophies. Geoarchaeology 1:365–379 Stiros SC, Jones RE (1996) Archaeoseismology. Institute
Rodrίguez-Pascua MA, Pérez-López R, Giner-Robles JL, of Geology & Mineral Exploration and British School
Silva PG, Garduño-Monroy VH, Reicherter K (2011) of Athens, Athens
A comprehensive classification of Earthquake Archae- Tuttle MP, Schweig ES (1995) Archeological and pedo-
ological Effects (EAE) in archaeoseismology: appli- logical evidence for large prehistoric earthquakes in
cation to ancient remains of Roman and Mesoamerican the New Madrid seismic zone, central United States.
cultures. Quat Int 242:20–30 Geology 23(3):253–256
148 Artificial Offshore Seismic Sources
General Considerations
Artificial Offshore Seismic Sources
The goal of marine seismic reflection surveys is to
Wolfgang Rabbel and Dennis Wilken image geological structure down to a predefined
Institute of Geosciences, University of Kiel, target depth at highest possible resolution. There-
Kiel, Germany fore, the pressure pulses generated by a marine
seismic source are usually designed to have a
sharp onset and short duration (some to some
10 ms; exception marine vibrator source)
Synonyms
corresponding to a broadest possible frequency
spectrum. The short duration, respective large
Marine seismic sources; Seismic energy sources
bandwidth, is necessary to enable distinguishing
of reflections (echoes) arriving at the receivers
subsequently from interfaces at increasing depths.
Introduction Seismic wave propagation is affected by absorption
that leads to a notable continuous decrease of
Artificial Offshore Seismic Sources are instru- amplitudes starting from the high-frequency end
ments that are used for generating short- of the spectrum. Therefore, penetration depth
duration pressure pulses in the marine water increases with decreasing signal frequency. While
column, usually with constant repetition rate. it propagates, the wave front widens and signal
They provide the acoustic energy for marine amplitudes decrease accordingly (“geometrical
seismic surveys. Marine seismic surveys are spreading”). This effect needs to be compensated
performed in order to image geological struc- for by increasing the initial strength of the pressure
ture (layers and faults) beneath the seafloor and pulse in order to keep reflected signals of deep
to characterize rocks and rock strength in terms horizons detectable. Finally, in order to obtain a
of seismic wave velocity or dynamic elastic reliable depth section, the measured profiles have
constants. Therefore, short pressure pulses are to be sampled spatially with a sufficiently high
applied to the water column that travel as density realized by repeating the measurements
acoustic waves down to the sea bottom where while the vessel is moving. This implies that the
they are converted into seismic body and inter- marine seismic source needs to be reloadable
face waves (elastic waves). These seismic within a certain time interval depending on the
waves penetrate deeper into the underground velocity of the ship. Typical repetition intervals
and are partially reflected at geological inter- are between 0.5 up to some 10 s. The target range
faces and/or refracted back to the surface relevant for seismic exploration – for example in
(Fig. 1, right). This returning part of the earthquake engineering – may extend from near-
wave field is recorded either at the seafloor seafloor layering down to several kilometers depth,
with seismographs or in the water column for example, in order to detect and trace deep
with hydroacoustic receivers. In order to reaching fault systems. Correspondingly, different
allow for profiling both seismic source and sorts of seismic sources are in use, each accounting
receivers are towed behind a vessel carrying for different target depth or infrastructure related
out repeated measurements while moving for- situations (Table 1).
ward. The technique of marine seismic has
been developed and improved continuously Types of Marine Seismic Sources and
since the late 1940s. This entry provides an Their Key Principles
overview of the purpose and principles marine
seismic sources. More detailed information can Airguns and GI-guns
be found, for example, in Jones (1999) and Airguns or G-(Generator)-Guns generate an
Meunier (2011). acoustic impulse by releasing a high pressure air
Artificial Offshore Seismic Sources 149
GI-Gun
streamer/hydrophone array
layers
sedimentary
Boomer
Artificial Offshore Seismic Sources, Fig. 1 Right: or propagation velocity change; wave propagation paths
Principle of marine seismic data acquisition; pulse-like indicated by rays (black lines). Left: examples of seismic
seismic waves travel through geological layers and are sources for near-surface prospecting; from top to bottom:
reflected back to the surface at interfaces where density Airgun (GI-gun), Sparker, Boomer
bubble underwater. The radiation pattern of the bubble oscillation, GI-(Generator Injector)-Guns
airgun source corresponds to an acoustic dipole are very common that inject a second air pulse to
because an image-source mirrored at the water collapse the initial bubble before it starts to oscil-
surface has to be considered in order to account late. G-Guns and GI-Guns can be mounted in
for the free-surface condition. It causes a arrays to enhance the signal amplitude and, when
so-called ghost reflection following the primary fired with time offsets, to modulate the shape of
pulse with negative polarity and a time delay the signal.
depending on source depth. While moving to The generated frequency content mainly
the sea surface, the generated bubble depends on the volume of the airgun chamber,
oscillates around the equilibrium of air- and hydro- the assigned pressure, and the tow depth.
static water pressure with a natural period T.
T depends on the energy of the injected air E, the Water Guns
confining hydrostatic pressure P of the water col- A water gun is a pneumatic seismic source oper-
umn, and the fluid density d according to the ated by compressed air, too. The air is used to
Rayleigh-Willis formula T 1.14d1/2 P5/6E1/3. drive a piston pressing water from a chamber
This oscillation generates an undesired lengthen- inside the gun into the water column. The seismic
ing of the seismic signal. In order to overcome the impulse is generated by the implosion of a cavity
150 Artificial Offshore Seismic Sources
Artificial Offshore Seismic Sources, Table 1 Technical data and properties of typical marine seismic sources. The
given value of structural resolution corresponds to a quarter wavelength at 1,600 m/s wave speed
Energy per pulse (J) / Resolution
pressure Main Typical at main
Source Chamber (dB reference 1 mPat Spectral frequency exploration frequency
type volume (l) 1 m) width (Hz) (Hz) depths (m) Power supply unit
Airgun 0.2–32 ~10 kJ/170–230 dB 10–200 Some 2–40 Compressor
100 m to
some km
GI Gun 0.4–4 ~10 kJ/170–230 dB 50–200 Some 2–8 Compressor
100 m to
some km
Water 0.2–2 ~30 kJ 50–1,000 Some 0.4–8 Compressor
Gun 100 m to
some km
Sparker 300–2,400 J 100–3 k ~1,500 Upto 0.25 Capacitor bank
~200 m
Boomer 300–700 J 500–4 k ~3,000 Upto 0.15 Capacitor bank
~100 m
Pinger ~100 J 2 k–150 k Upto ~10 m 0.2–0.03 Electricalamplifier
Vibro Some 10 kW times 6–100 Some 4–100 Electrical current
source sweep length (s) 100 m to
some km
created behind the expelled water. It is free of pulse to the source. The discharge is led through a
bubble oscillations and contains higher frequen- flat coil inside the source body. From below an
cies than typically generated by airguns. aluminum plate is pressed against the coil by
springs. The magnetic field of the coil induces
Sparker eddy currents inside this plate. These currents
The sparker source generates an acoustic produce magnetic fields that repel the plate from
impulse by a high-voltage spark discharge the coil (Lenz’s law). The spring then moves the
between two electrodes in the water. The spark plate back, leaving a collapsing cavitational vol-
consists of a high-pressure plasma whose expan- ume that generates the acoustic impulse.
sion and collapse generates the acoustic peak.
The high voltage needed to generate such a Pinger
plasma is stored in a bank of capacitors, which Sediment echo sounders (SES) use piezoelectric
is reloaded for each shot. Sparkers can be used at “pingers” as electroacoustic converters.
different energy levels, depending on the capac- Some materials (especially crystalline) tend to
itor bank. Commonly sparkers are used in arrays show electromechanical properties which means
to enhance the signal amplitude and to maintain that the material is stretched when voltage is
the signal shape which usually changes because applied. Closely packed piezoelectric crystals
the electrodes tend to tip-off during usage. are very effective electroacoustic converters and
Sparkers can be applied in brackish water and can be used as acoustic sources.
seawater. These pingers are applied as single point
sources or – even more often – in arrays which
Boomer enable, when used with time offsets, a modula-
Boomers are also based on combining electro- tion of the radiation pattern of the source. Alter-
magnetic and mechanical principles. Similar to native developments are based on the parametric
the sparker case, electric charges are assembled nonlinearity effect to enhance the acoustic effi-
in a bank of capacitors releasing a high-voltage ciency and the stability of the signal.
Assessment and Strengthening of Partitions in Buildings 151
frames. In addition, efforts have been made to for strengthening existing partition walls are
utilize infill walls as a means of producing eco- presented.
nomic designs by reducing the sizes of the mem-
bers of the bounding frames.
A large number of researchers have studied Behavior of Partitions
the behavior of infilled frames. It is evident
from their studies that infill walls can provide The modeling of the behavior of infilled frames
both an economic and practical means for the under lateral loading (and mainly earthquake-
lateral stability of framed structures and a viable induced loads) is a complex issue because these
alternative for retrofitting existing structures to structures exhibit a highly nonlinear response due
resist seismic, wind, and blast loads. Despite this, to the interaction between the masonry infill
there is reluctance from the engineering commu- panel and the surrounding frame. This results to
nity to use this structural system widely and treat several modes of failure, each of which has a
infill walls as structural elements. different failure load, and hence a different ulti-
Therefore, although the subject of infilled mate capacity and overall behavior.
frames has been studied for more than 60 years, At moderate loading levels, the infill of a
there are still no definitive answers either about non-integral infilled frame separates from the sur-
their behavior and interaction with the bounding rounding frame and the infill acts as a diagonal
frame or about the estimation of their stiffness strut (Fig. 1). As the racking load is increased,
and strength. Some problems that make it diffi- failure occurs eventually in either the frame or
cult for practicing engineers to use infill walls as the infill. The usual mode of frame failure results
structural elements are the inherent nonlinearity, from tension in the windward column or from
the high degree of variability, and the inherent shearing in the column or beams or plastic hinging
degradability of infill walls. The fact that infill in columns or beams; however, if the frame
walls exhibit completely different in-plane and strength is sufficient enough to prevent its failure
out-of-plane behavior makes the problem even by one of these modes, the increasing racking load
more difficult to tackle. This reveals the difficul- eventually produces failure of the infill. In the
ties associated with this problem and the need for most common situations, the in-plane lateral load
more research to answer these long-standing applied at one of the top corners is resisted by a
questions. truss formed by the loaded column and the infill
Nevertheless, the presence of infill walls has along the diagonal connecting the loaded corner
shown that in most cases they have beneficial and the opposite bottom corner. The state of stress
effects on the behavior of structures during earth- in the infill gives rise to a principal compressive
quakes and have contributed in the prevention of stress along the diagonal and a principal tensile
collapse of many structures. It is therefore for this stress in the perpendicular direction. If the infill is
reason that efforts have been made to use made of concrete, successive failures, initially by
engineered infills to retrofit existing seismic- cracking along the compression diagonal and then
deficient buildings, since they present an eco- by crushing near one of the loaded corners or by
nomically viable solution. crushing alone, will lead to collapse; if the infill is
In this entry, the in-plane behavior and in-plane made of brick masonry, an alternative possibility
failure modes of infilled frames are presented. of shearing failure along the mortar planes may
Then, macro- and micromodels for infill walls arise (Fig. 1).
are presented followed by the results of experi- Based on both experimental and analytical
mental investigations for the replacement of results during the last five decades, different
masonry infill walls with reinforced concrete failure modes of masonry-infilled frames
infills, as an economically viable method for were proposed that can be classified into five
retrofitting existing structures. Finally, methods distinct modes given below:
Assessment and Strengthening of Partitions in Buildings 153
Assessment and Strengthening of Partitions in compression (DK) modes. (b) Sliding shear (SS), frame
Buildings, Fig. 1 Failure modes of non-integral infilled failure (FF), and diagonal cracking (DK) modes (Asteris
frames: (a) Corner crushing (CC) and diagonal et al. 2011 with permission from ASCE)
1. The corner crushing (CC) mode, which repre- 5. The frame failure (FF) mode, which is seen in
sents crushing of the infill in at least one of its the form of plastic hinges developing in the
loaded corners, as shown in Fig. 1a. This mode columns or the beam-column connections, as
is usually associated with infilled frames shown in Fig. 1b. This mode is associated with
consisting of a weak masonry infill panel a weak frame or a frame with weak joints and
surrounded by a frame with weak joints and strong members infilled with a rather strong
strong members (Mehrabi and Shing 1997; infill.
El-Dakhakhni 2002; Ghosh and Amde 2002;
El-Dakhakhni et al. 2003). Ghosh and Amde (2002), based on the finite
2. The diagonal compression (DC) mode, which element method and including interface elements
represents crushing of the infill within its cen- at the frame-infill interface, confirmed the order of
tral region, as shown in Fig. 1a. This mode is occurrence of the above five distinct failure modes.
associated with a relatively slender infill, Of the five modes, only the CC and SS modes are of
where failure results from out-of-plane buck- practical importance (Comité Euro-International
ling of the infill. du Béton CEB 1996), since most infills are not
3. The sliding shear (SS) mode, which represents slender (El-Dakhakhni et al. 2003) and therefore
horizontal sliding failure through bed joints of the second mode (DC) is not favored. The fourth
a masonry infill, as shown in Fig. 1b. This mode (DK) should not be considered as a failure
mode is associated with weak mortar joints mode, due to the post-cracking capacity of the infill
in the infill and a strong frame. to carry additional load. The fifth mode (FF) relates
4. The diagonal cracking (DK) mode, which to the failure of the frame, and it is particularly
is seen in the form of a crack across the important when examining existing structures,
compressed diagonal of the infill panel and which in many cases exhibit frame weakness. It
often takes place with simultaneous initia- should be noted that these failure modes are only
tion of the SS mode, as shown in Fig. 1b. seen/applicable to the case of infill walls without
This mode is associated with a weak frame openings on the diagonal of the infill panel.
or a frame with weak joints and strong Kappos and Ellul (2000), based on an analyti-
members infilled with a rather strong infill cal study of the seismic performance of masonry-
(Mehrabi and Shing 1997; El-Dakhakhni infilled RC-framed structures, found that taking
2002). into account the infill in the analysis resulted in
154 Assessment and Strengthening of Partitions in Buildings
Assessment and
Strengthening of
Partitions in Buildings,
Fig. 3 Six-strut model for
masonry infill panel in
frame structures
(Chrysostomou 1991)
More complex macromodels were hence pro- They postulated that this resulted in a system of
posed, still typically based on a number of diag- significant strength, stiffness, and ductility that
onal struts. behaves much like the modern eccentrically
Thiruvengadam proposed the use of a braced frame systems. Such behavior could, in
multiple-strut model to simulate the effect of the part, be responsible for the observed good perfor-
infill panel. This particular model consists of a mance of these buildings in the 1906 San
moment-resisting frame with a large number of Francisco earthquake (Hamburger and Meyer
pin-jointed diagonal and vertical struts. Initially, 2006).
a perfect frame-infill bond condition is assumed, The main advantage of the multiple-strut
and the lateral stiffness of the infill, by its shear models, in spite of the increase in complexity, is
deformation, is modeled by a set of pin-ended the ability to represent the actions in the frame
diagonal struts running in both directions. These more accurately. Syrmakezis and Vratsanou
diagonals represent the shear and axial stiffness (1986) employed five parallel struts in each diag-
of the masonry infill. Similarly, the vertical stiff- onal direction. It was stressed how different con-
ness contribution is accounted for by providing tact lengths play a significant effect on the
vertical struts. The objective of the aforemen- bending moment distribution in the frame
tioned study was a realistic evaluation of the members.
natural frequencies and modes of vibration, pur- Chrysostomou (1991, Chrysostomou
poses for which the nonlinear phenomena do not et al. 2002) aimed at obtaining the response of
play an important role. This model has been infilled frames under earthquake loading by tak-
adopted by many researchers to investigate the ing into account both stiffness and strength deg-
effect of infill on the behavior of infilled frames radation of infills. Each infill panel was modeled
and has been also included in FEMA-356, due to by six compression-only inclined struts (Fig. 3).
the great number of struts, as a method for model- Three parallel struts were used in each diagonal
ing the special case of infilled frames with open- direction and the off-diagonal ones were posi-
ings. Similarly, Hamburger and Chakradeo tioned at critical locations along the frame mem-
proposed a multiple-strut configuration that can bers. These locations are specified by a parameter
account for the openings also, but the evaluation a, which represents a fraction of the length or
of the characteristics of the struts is rather com- height of a panel and is associated with the posi-
plicated. They showed that for panels of typical tion of the formation of a plastic hinge in a beam
configuration, the formation of these struts pro- or a column. At any point during the analysis of
tects the beam-to-column connections, which the nonlinear response, only three of the six struts
have limited capacity in withstanding significant are active, and the struts are switched to the
flexural demand, with plastic hinges forming opposite direction whenever their compressive
instead within the mid-span region of the beam. force reduced to zero.
Assessment and Strengthening of Partitions in Buildings 157
In order to conduct nonlinear analysis, the one diagonal and two off-diagonal struts, on mak-
force-displacement relationships corresponding ing use of the orthotropic behavior of the masonry
to the equivalent strut model must be adequately wall as well as on some experimental observa- A
defined. The modeling of hysteretic behavior tions and analytical simplifications, in order to
increases not only the computational complexity simplify the nonlinear modeling of these
but also the uncertainties of the problem. structures.
In Chrysostomou’s model, the hysteretic Crisafulli investigated the influence of differ-
behavior of the six struts is defined by a hysteretic ent multiple-strut models on the structural
model, which consists of two equations. The first response of infilled reinforced concrete frames,
equation defines the strength envelope of a struc- focusing on the stiffness of the structure and the
tural element and the second defines its hysteretic actions induced in the surrounding frame.
behavior. The shape of the envelope and the Numerical results, obtained from the single-,
hysteretic loops is controlled by six parameters, two-, and three-strut models, were compared
all of which have physical meaning and can be with those corresponding to a refined finite ele-
obtained from experimental data. The advantage ment. The lateral stiffness of the structure was
of this strut configuration over the single diagonal similar in all the cases considered, with lower
strut is that it allows the modeling of the interac- values for two- and three-strut models. It must
tion between the infill and the surrounding frame be noted that, for the multiple-strut models, the
and it takes into account both strength and stiff- stiffness may significantly change depending on
ness degradation of the infill, which is vital for the separation between struts. The single-strut
determining the response of infilled frames model underestimates the bending moment
subjected to earthquake load. because the lateral forces are primarily resisted
Saneinejad and Hobbs developed a method by a truss mechanism. On the other hand, the
based on the equivalent diagonal strut approach two-strut model leads to higher values than
for the analysis and design of steel or concrete those corresponding to the finite element model.
frames with concrete or masonry infill walls A better approximation is obtained from the
subjected to in-plane forces. The proposed ana- three-strut model, although some differences
lytical model assumes that the contribution of the arise at the ends of both columns. Although the
infill panel to the response of the infilled frame single-strut model constitutes a sufficient tool for
can be modeled by replacing the panel by a sys- the prediction of the overall response and the
tem of two diagonal masonry compression struts. triple-strut model is superior in precision,
The method takes into account the elastoplastic Crisafulli adopted the double-strut model
behavior of infilled frames considering the lim- approach, accurate enough and less complicated
ited ductility of infill materials. Various compared to the other models.
governing factors such as the infill aspect ratio, More recently, Crisafulli and Carr (2007)
the shear stresses at the infill-frame interface, and proposed a new macromodel in order to repre-
relative beam and column strengths are sent, in a rational but simple way, the effect of
accounted for in this development. This model masonry infill panels. The model is implemented
has been adopted by Madan et al. (1997) for static as a four-node panel element which is connected
monotonic loading, as well as quasi-static cyclic to the frame at the beam-column joints. Inter-
loading. This model for masonry infill panels was nally, the panel element accounts separately for
implemented in IDARC 2D Version 4.0 (Vales the compression and shear behavior of the
et al. 1996), a computer-based analytical tool for masonry panel using two parallel struts and a
the inelastic analysis and damage evaluation of shear spring in each direction. This configura-
buildings and their components under combined tion allows an adequate consideration of the
dynamic, static, and quasi-static loading. lateral stiffness of the panel and of the strength
El-Dakhakhni (2000, 2002) and El-Dakhakhni of masonry panel, particularly when a shear fail-
et al. (2001) suggested replacing the infill wall by ure along mortar joints or diagonal tension
158 Assessment and Strengthening of Partitions in Buildings
failure is expected. Furthermore, the model is capturing the force redistribution and cracking
easy to apply in the analysis of large infilled processes of brittle materials.
frame structures. The main limitation of the To this end, constitutive relationships must be
model results from its simplicity, since the defined for the use of finite element models of
panel is connected to the beam-column joints infilled frames. The use of 3D solid, instead of
of the frame, being thus not able to properly linear, elements in constitutive models requires a
predict the bending moment and shear forces in considerably higher level of model sophistica-
the surrounding frame. A detailed presentation tion. Models of concrete behavior are based
as well as references to the work on either on regression analyses of experimental
macromodels of the various researchers men- data (empirical models) or on continuum
tioned above is given by Asteris et al. (2011). mechanics theories, which should also be verified
against experimental data. Many such models
Micromodeling have been proposed, but the application of FE
The ability to model accurately the behavior of a packages in practical structural analysis has
structural system, including the infilled RC shown that the majority of constitutive relation-
frame, depends on experimental work. A review ships are case dependent, since the solutions
of the available published experimental data obtained are realistic only for specific types of
describing the effect of infill walls on the overall problems. The application of these packages to a
structural response of RC frames finds significant different set of problems requires modification,
scatter in the obtained results, due to the large sometimes significant, of the constitutive rela-
number of uncertainties involved in the various tionships. The situation is better for the reinforce-
investigations carried out to date. This data can ment. However, complications arise with the
describe qualitatively the effect of infill walls, but introduction of bond-slip laws, which results in
it is not yet possible to quantify this effect exper- large discrepancies in predicted behavior.
imentally. Since the formulation and validation Regarding the modeling of the infill panels,
of mathematical (numerical or analytical) models the great number of influencing factors, such as
is based on available experimental data, the valid- dimension and anisotropy of the bricks, joint
ity of the yield predictions of current software width and arrangement of bed and head joints,
packages is currently questionable. material properties of both brick and mortar, and
Due to the difficulties, limitations, and high quality of workmanship, make the simulation of
costs associated with the testing of infilled RC plain brick masonry extremely difficult. The level
frames, resort is frequently made to the use of of complexity of the analytical model depends on
nonlinear finite element analysis (NLFEA). The whether masonry is considered as one-, two-, or
use of the FE method can provide a more detailed three-phase material, where the three phases are
description of the effect of infill walls on the comprised of the units, mortar, and unit-mortar
response of infilled frames. At the same time, it interface.
allows the investigation to be extended to struc- Another level of complexity is added with the
tural forms more complex than the simple struc- modeling of the infill-frame interface, which
tural elements that are usually studied determines the boundary conditions of the infill
experimentally (i.e., scaled models of one- or and its interaction with the bounding frame.
two-story infilled frames with column height to Springs and interface elements have been used
beam length equal to values from 0.75 to 1.25). by various researchers to represent the interaction
To qualify for accurately capturing the behavior between deformable structures such as these,
of a concrete structure, a FEA package must where interfacial separation and sliding may
predict the structural response within an error of arise.
up to 20 %. Such a package should also employ Infill walls have also been represented by
appropriate constitutive models and adopt boundary elements. These are combined with
nonlinear numerical procedures capable of FE, which can represent the bounding frame.
Assessment and Strengthening of Partitions in Buildings 159
Reinforced Concrete Infills in Reinforced other test campaigns, the thickness of a preexisting
Concrete Frames thin wall was increased by 100–150 % without any
direct connection of the new wall with the
The construction of new walls is the most effec- bounding frame. The failure mode of all specimens
tive and economic method for retrofitting was in shear (including sliding at the interface).
multistory reinforced concrete (RC) buildings, It is interesting to note that for epoxy-grouted
especially those with pilotis (soft ground story). dowels, the force resistance of the infilled frame
Their structural and economic effectiveness was on average 87 % of the integral one, while for
increases when selected bays of an existing RC the mechanical connections it was 80 % on aver-
frame are fully infilled with integral RC walls age. For the increased thickness of an existing thin
replacing masonry ones. Such a method is infill wall, the force resistance was on average
appealing since the intervention is concentrated 92 % of the monolithic specimen, while the dis-
in only certain bays of the frame and reduces the placement at failure was on average 13 % smaller
disturbance to the inhabitants as compared to the than for the integral specimen. For the epoxy-
case of using concrete jackets on the columns. grouted dowels and for the mechanical connection,
Most of the experimental research work the ultimate deformation was on average 55 % and
performed in the last decades has focused on 115 % larger than in the integral specimen, respec-
other frequently used types of retrofitting, in par- tively. The results show that although a deform-
ticular on fiber-reinforced polymers (FRP) and able connection gives a somewhat reduced
concrete jackets. Research on the use of RC infill strength with respect to the monolithic case, the
walls has mainly targeted what is feasible: testing ultimate deformation of the retrofitted structure is
of one- to two-story specimens. However, data is considerably increased.
lacking for taller full-scale specimens that reflect Among the specimens tested in Turkey
real-life applications, due to the practical difficul- (Chrysostomou et al. 2013, 2014), some were
ties associated with the high forces needed for the single-story one-bay 1:2 and 1:3 scale, with RC
tests. Regarding code provisions, Eurocode 8 – infill thickness 25 % and 33 % of the width of the
Part 3 fully covers retrofitting with FRP or con- frame members. Others were two-story one-bay
crete jackets, while it does not address the scaled at 1:3, with infill wall thickness 33 % and
retrofitting of RC frames with the addition of 40 % of the width of the members of the bounding
new walls created by infilling selected bays. frame. The RC infill was in most cases fully
Experimental research on reinforced concrete connected on the perimeter with dowels, in
frames converted into walls by infilling with RC some cases there was a gap between the infill
has been carried out almost exclusively in Japan and the columns, while in some other cases
and Turkey. The experiments in Japan there was no connection other than simple bear-
(Chrysostomou et al. 2013, 2014) were ing. In one case, the rebars of the infill were
performed on a total of 27 1:3–1:4 scale single- welded to those of the members of the frame,
story one-bay RC-infilled frames with RC infill instead of using dowels, and monolithic speci-
walls with a thickness of 26–60 % (on the average mens (not exactly similar to the infilled ones)
43 %) of the width of the frame members. The were included for comparison purposes. Finally,
test results were compared in most cases with there was a two-story three-bay specimen scaled
monolithically cast specimens of the same geo- at 1:3, with the middle bay infilled with a wall
metric characteristics (in which the frame and the with 63 % thickness of the width of the frame
infill wall were cast at the same time and inte- members. The connection was made with epoxy-
grally connected). The connection of the RC infill grouted dowels and the failure mode was pre-
to the bounding frame was done by means of dominantly flexural. In all other cases, the
epoxy-grouted dowels (17 specimens) or through single-story walls failed in shear, while the
mechanical devices, such as shear keys and two-story walls failed in a combination of flexure
dowels without epoxy (6 specimens). In four and shear sliding at the base.
Assessment and Strengthening of Partitions in Buildings 161
Assessment and Strengthening of Partitions in (Chrysostomou et al. 2014 with kind permission from
Buildings, Fig. 6 (a) Dowels and starter bars. (b) Springer Science and Business Media)
Dowels, starter bars, and web reinforcement
4. The “weak” south frame behaved equally have proposed methods for improving the per-
well as the “strong” north frame. formance of these elements.
5. The slip displacements at the horizontal One of the methods is to use shotcreting on the A
interfaces of the ground-floor walls were of faces of masonry infills which can increase the
the order of 0.8 mm, which is very close to stiffness and the lateral load capacity of the infilled
the full engagement of the starter bars but not frame and reduce the lateral drift at the ultimate
of the dowels. load. Calvi and Bolognini (2001) have performed
6. The slip displacements between the wall and full-scale tests in which they have placed a 10 mm
the bounding columns of the ground floor plaster on both sides of a masonry infill wall cov-
were of the order of 0.4 mm. ering either reinforcement (Ø5 mm or Ø6 mm) or
7. The behavior of the wall was mainly flexural; wire meshes and have studied the behavior of the
yielding took place at both the ground-floor strengthened infilled frame. Based on their obser-
and the first-floor walls. vations from the experiments, the introduction of
8. The distribution of strains along the some reinforcement in the mortar layers, with a
bounding columns of the walls shows that geometrical percentage lower than 1 %, will
the ones for the ground floor are much larger almost double the acceleration levels for the occu-
than those of the first floor, while those of the pational and damage limit states.
second and third floors are negligible. Another technique is the use of fiber-reinforced
9. The two connection arrangements used polymer (FRP) reinforcement to enhance the
behaved satisfactorily. overall response of such systems. Several experi-
10. Higher mode effects appeared in the mental studies have been performed which dem-
response of the structure. onstrate that significant improvement of strength
11. Some vertical cracks appeared at the connec- and energy absorption capacity can be achieved if
tion of the beams to both the exterior and the adequate anchoring is provided.
wall columns. Altin et al. (2008) investigated experimentally
12. A horizontal crack appeared at the ground the behavior of strengthened masonry-infilled
beam of the walls, which was the main cause reinforced concrete (RC) frames using diagonal
for the loss of strength of the south frame. carbon fiber-reinforced polymer (CFRP) strips
under cyclic loads. Test results indicated that,
It was demonstrated that this is a viable CFRP strips significantly increased the lateral
method for retrofitting and it can be used to strength and stiffness of perforated clay brick-
strengthen existing ductility and strength- infilled nonductile RC frames. Specimens receiv-
deficient structures. The recorded global and ing symmetrical strengthening showed higher lat-
local behavior of the structure provides data for eral strength and stiffness, compared to the ones
the development of numerical models, to facili- at which CFRP strips of the same width were
tate the proposal of design guidelines for such a applied to one of the interior or exterior surface
retrofitting method. of the infill wall. In the latter case, similar lateral
strength and stiffness was obtained, irrespective
of the side of placement of the strips.
Strengthening of Partitions Lunn and Rizkalla (2009) used glass fiber-
reinforced polymer (GFRP) systems for increasing
In general, infill walls and in particular the the out-of-plane resistance of infill masonry walls
masonry infill panels exhibit a small plastic to loading. They have concluded that GFRP
region on the stress-strain curve due to signifi- strengthening of infill masonry walls is effective
cant decrease in stiffness, strength, and energy in increasing the out-of-plane load-carrying capac-
absorption capacity. In order to enhance the ity when proper anchorage of the FRP laminate is
behavior of infill partitions and remove provided. They also note that the latter has a signif-
their inherent deficiencies, various researchers icant effect on the failure mode of the assemblage.
164 Assessment and Strengthening of Partitions in Buildings
Papanicolaou et al. (2007) have used the have contributed in the prevention of collapse
textile-reinforced mortar (TRM), which is a new of many structures. It is therefore for this reason
structural material, for testing its capability for that efforts have been made to use engineered
increasing the load-carrying capacity and infills to retrofit existing seismic-deficient build-
deformability of unreinforced masonry walls ings, since they present an economically viable
subjected to cyclic in-plane loading. TRMs com- solution. Several equations have been intro-
prise fabric meshes made of long woven, knitted, duced to define the stiffness of infilled frames
or even unwoven fiber rovings in at least two and their capacities, which are related to a num-
directions. The density of rovings in each direc- ber of failure modes. These were considered in
tion can be controlled independently, thus affect- macro- and micro mathematical models that
ing the mechanical characteristics of the textile have been proposed to simulate their behavior
and the degree of penetration of the mortar matrix and their interaction with the bounding frame.
through the mesh openings. The former are simpler (especially the single-
Based on the experimental results, the authors strut models), while the latter require a higher
stated that, in terms of strength, TRM jackets are at level of modeling sophistication. The construc-
least 65–70 % as effective as FRP jackets with tion of new walls is the most effective and eco-
identical fiber configurations. In terms of nomic method for retrofitting multistory
deformability, of crucial importance in seismic reinforced concrete (RC) buildings, especially
retrofitting of unreinforced masonry walls, TRM those with pilotis (soft story). Their structural
jacketing for shear walls is about 15–30 % more and economic effectiveness increases when
effective than FRP. They therefore conclude that selected bays of an existing RC frame are fully
TRM jacketing is an extremely promising solution infilled with integral RC walls replacing
for strengthening and seismic retrofitting of masonry ones. Recent experimental investiga-
unreinforced masonry walls subjected to in-plane tions make possible the quantification of this
loading. methodology and the proposal of code equations
A more recent development is the use of for the design of such systems. Several methods
engineered cementitious composites (ECC) exist for improving the properties of infilled
which are a special class of fiber-reinforced panels avoiding brittle failure and increasing
cement-based composite materials, typically their load-carrying capacity as well as their
reinforced with polyvinyl alcohol fibers. energy absorption during an earthquake.
Dehghani et al. (2013) have tested in diagonal
compression a number of specimens with differ-
ent ECC-strengthening configuration, and they Cross-References
evaluated their in-plane deformation and strength
properties, including the post-peak softening ▶ Numerical Modeling of Masonry Infilled
behavior. They state that the proposed technique Reinforced Concrete Frame Buildings
can effectively increase the shear capacity of ▶ Retrofitting and Strengthening of Structures:
masonry panels (1.5–2.8 times), improve their Basic Principles of Structural Interventions
deformability, enhance their energy absorption ▶ Seismic Fragility Analysis
capacity (35 times), and prevent the brittle ▶ Seismic Strengthening Strategies for Existing
failure mode. (Code-Deficient) Ordinary Structures
▶ Seismic Vulnerability Assessment: Reinforced
Concrete Structures
Summary ▶ Strengthening Techniques: Code-Deficient
R/C Buildings
The presence of infill walls has shown that in ▶ Strengthening Techniques: Code-Deficient
most cases they have beneficial effects on the Steel Buildings
behavior of structures during earthquakes and ▶ Time History Seismic Analysis
Assessment of Existing Structures Using Inelastic Static Analysis 165
Data
ag
m2
T
TC TD
Force/moment
Sa
T = TC
μ=1
ag 3 2
Displacement/ 6 T = TD
chord rotation
Sd
ADRS format
Analysis
c Pushover d Equivalent SDOF model
D
P1
V D* F*
P2
m*
k*
V D
D*
F*
Dt* Sd
V
Assessment of Existing Structures Using Inelastic Static Analysis, Fig. 1 Version of the inelastic pushover-based
analysis, included in EC8/3
seismic intensity levels. Besides other superior on the global behavior of the structure. They
features, they can provide more realistic informa- can identify the critical regions in the structure.
tion about the force demand on brittle elements This information is particularly useful when the
and the information on the influence of the existing structures have been evaluated and the
strength degradation of individual components strengthening techniques have been selected.
168 Assessment of Existing Structures Using Inelastic Static Analysis
backbone curve
a idealization of the b c
backbone curve
A
Force/moment
Force/moment
Force/moment
Displacement/chord rotation Displacement/chord rotation Displacement/chord rotation
Ductile response Semi-ductile response Brittle response
Assessment of Existing Structures Using Inelastic Static Analysis, Fig. 2 Idealized numerical models
are approximated based on the type of the evaluation (e.g., Mergos and Kappos 2008;
response. Typical examples are presented in Fischinger et al. 2012). Similar observations can
Fig. 2. be applied to other mechanisms, which reduce the
Contrary to the structures designed according ductility capacity of the structural components
to modern codes, where the brittle types of failure and the whole structure. Therefore, it is feasible
are in general avoided, in structures which were to explore and compare different available
designed before the modern principles of seismic options before the numerical model, and its prop-
engineering were established, different types of erties are defined. All uncertainties related to the
brittle or semi-ductile failure of their structural material properties should also be properly
components may be expected. Thus for the reli- explored.
able estimation of their response, the appropriate In the pushover analysis, the MDOF model of
numerical models should be employed. Their the structure is subjected to lateral forces that are
properties can be defined based on different pro- intended to simulate the inertial forces expected
cedures, defined in the (pre)standards, guidelines, in the building during an earthquake. They are
and literature (e.g., CEN 2005, FEMA-356, usually distributed according to a selected
ATC-40, CEB-FIP 2003). It is, however, worth (mostly invariant) pattern. It significantly influ-
noting that the available models which describe ences the pushover curve and further determines
quite complicated mechanisms that reduce the the relative magnitudes of the shear forces, the
ductility capacity of the structural components moments, and deformations within the structure;
have been less frequently investigated and are therefore, it should be carefully selected,
often less reliable. The properties of these models depending on the properties of the analyzed struc-
are often defined using empirical or semiempiri- ture and the expected response. The distribution
cal procedures. The results of these procedures of the lateral inertial forces will in general vary
can be considerably different. continuously during the earthquake response,
For example, the value of the shear strength of depending on the gradual formation of the plastic
RC components can strongly depend on the hinges in the structure. To take into account these
method used to estimate it. The differences changes, in many codes and guidelines, it is
between different methods depend on many recommended to perform an analysis, taking
parameters, e.g., the amount of the flexural and into consideration at least two lateral force pat-
shear reinforcement, the shear span ratio, the terns. Typically the distribution of forces propor-
axial force, etc. Moreover, the numerical models tional to the fundamental mode of vibration in the
that can take into account the complex interaction elastic range and the uniform distribution are
between shear and flexural response in the recommended. Then the most adverse results
nonlinear range are still under investigation and are taken into consideration. Other solutions
170 Assessment of Existing Structures Using Inelastic Static Analysis
Assessment of Existing
Structures Using
Inelastic Static Analysis,
Fig. 3 In some structures
(e.g., with “soft” story), the
deflection line can Plastic hinges
considerably change after
yielding Plastic hinges
Displacement Displacement
shape before shape after
yielding yielding
recommended in the codes are presented in sec- pattern is used or the adaptive methods are
tion “Different Applications of the Inelastic applied, the analysis procedure is more complex
Pushover-Based Analysis, Adopted in the (Pre) and more demanding.
Standards and Guidelines.” Similar observations to those presented above
The approach described in the previous para- can be applied, e.g., to the RC dual and wall
graph is in general appropriate for many struc- buildings, with the reinforcement that does not
tures, which are designed according to modern meet the requirements of current codes (e.g.,
standards. However, in some older existing struc- where plastic hinges can form at the elevations
tures, this solution is not always suitable. above the foundations).
For example, in medium-rise frame buildings, The accuracy of different lateral load patterns
which are designed according to capacity-design in structures where the higher modes have an
principles (where the columns’ flexural strength important influence on the response is discussed
exceeds the beam flexural strength), the triangu- in sections “Some Issues that Influence the Accu-
lar and uniform distribution are typically applied. racy of the Inelastic (Single-Mode) Pushover-
In existing medium-rise buildings, this is also an Based Methods” and “Application of the Inelastic
appropriate solution, but only if the beam-sway Static Methods to the Analysis of Bridges”.
mechanism is the most likely to be developed. In
other words, the structure does not include a
so-called soft. In opposite cases, the uniform dis- Different Applications of the Inelastic
tribution may not be able to frame the range of Pushover-Based Analysis, Adopted in
actions that may occur during the actual dynamic the (Pre)Standards and Guidelines
response. Thus a rather complex choice of the
lateral force pattern may be needed. The response The static pushover analysis has no rigorous the-
of such structures in the elastic range will be oretical foundation. It is based on the assumption
approximately linear (see Fig. 3). In the nonlinear that the response of the structure (MDOF system)
range, the displacement pattern can considerably can be estimated using the results of the analysis
change after the formation of the plastic hinges, of an equivalent SDOF oscillator (see Fig. 1).
as is illustrated in Fig. 3. In such cases, some (pre) This means that it is assumed that the response
standards (e.g., FEMA-356, ATC-40) recom- is governed by one invariant mode of vibration.
mend the use of the adaptive load pattern, which In general this is incorrect. However, the assump-
can take into account changes of the displace- tion is approximately fulfilled in many (regular)
ment response shape corresponding to the seis- structures, where the influence of the higher
mic intensity applied. An alternative solution is modes is negligible and the deflection shape is
the adaptive pushover-based methods (one of almost invariable. Thus the seismic response of
which is presented in section “Some Alternatives these MDOF systems is quite accurately esti-
to Single-Mode Pushover-Based Methods”). mated based on the analysis of an equivalent
Note, however, that if either the adaptive load SDOF model. It is not the intention of this article
Assessment of Existing Structures Using Inelastic Static Analysis 171
T*
Seismic demand for
effective damping βeq
Dt* Sd Dt * Sd
to explain the theoretical background of the for- structures, the displacements are corrected for
mulation of the equivalent SDOF system, since it short-period structures. This approach is applied
can be found elsewhere (e.g., Krawinkler and in, e.g., Eurocode 8/3 (EC8/3) and implicitly in
Seneviratna 1998). Here, different applications FEMA-356 (see sections “Eurocode 8/3 (CEN
of this basic concept, which are typically used 2005)” and “FEMA-356”). In FEMA-356, the
for the assessment of the existing structures, are maximum seismic displacements, estimated
overviewed. based on the analysis of SDOF system, are addi-
In most of the modern (pre)standards and tionally corrected to take into account different
guidelines for the assessment of existing struc- issues which are not included in the pushover-
tures, the simplest form of the pushover-based based analysis such as strength degradation, P-D
analysis is included. All methods that are effect, etc.
included in these documents are based on the If the secant stiffness is used to define the
same basic concept presented above. However, properties of the equivalent SDOF oscillator, typ-
these methods are not the same. The procedures ically the overdamped acceleration spectra are
which are used to define the properties of the used (see Fig. 4d). Actually the capacity spectrum
equivalent SDOF model and those which are method approach is followed. It is explained later
used to define the seismic demand of this in section “ATC-40” where the method included
model are different. In general two different into the ATC-40 is presented.
approaches are used. The properties of the In the NZSEE 2006, different inelastic
equivalent SDOF system are defined either methods are included. The one which is concep-
based on the equivalent pre-yielding stiffness tually different from those in other standards is
or based on the equivalent secant stiffness (see described in this article (see section “FEMA-
Fig. 4a, b). 440”). Contrary to the previously mentioned pro-
If the first approach is employed, the maximum cedures, the maximum seismic displacement at
response of the SDOF oscillator is typically the reference point is assumed at the beginning of
defined based on the 5 % damped acceleration the analysis taking into account the estimated
spectra proposed in the codes. The target displace- capacity of the structure. It is then used to esti-
ment of the equivalent SDOF system is obtained mate the properties of an equivalent SDOF
using the equal displacement rule approximation model. The dynamic properties of the SDOF
(see Fig. 4c). Since this approximation is only model are defined based on the secant stiffness
suitable for the medium- and long-period corresponding to the assumed maximum
172 Assessment of Existing Structures Using Inelastic Static Analysis
Assessment of Existing a b V
Structures Using
F*
Inelastic Static Analysis, Vt
Fig. 5 (a) Idealization of Fy*
the capacity curve in Vy αKe
EC8/2, (b) idealization of
0.6 Vy
the pushover curve in
FEMA 356
Ke
dy* dm* d* δt D
relationship between dt* and det*in EC8/3. For buildings, the influence of the higher modes
long- and medium-period structures, C1 is 1, and should be properly taken into account. The higher
for short-period structures, it is mode effects may be determined by doing higher
mode pushover analyses.
ð R 1Þ
T s The application of the capacity spectrum tech-
1þ
Te nique means that both the structural capacity
C1 ¼ (9) curves and the demand response spectra are plotted
R
in the spectral acceleration versus the spectral dis-
R is the ratio of elastic demand to the calcu- placement domain and compared. Therefore in the
lated strength capacity (a similar meaning to qu in next step, the pushover curve is converted to the
EC8/3). capacity spectrum curve using the modal shape
The C2 modification factor represents the vectors, participation factors, and modal masses
effect of a pinched hysteretic shape, stiffness obtained from a modal analysis of the structure.
degradation, and strength deterioration on the The capacity spectrum curve represents the rela-
maximum displacement response. tionship between accelerations Sa and displace-
The C3 modification factor represents increased ments Sd of the equivalent SDOF oscillator. Then
displacements due to dynamic P-D effects. the standard elastic acceleration spectrum
(corresponding to 5 % damping) is converted to
ATC-40 the ADRS format, where the spectral accelerations
While several similarities can be found between are presented as a function of the corresponding
the EC8/3 and FEMA-356, the simplified spectral displacements (see Fig 4d). In this way,
nonlinear procedure in ATC-40 is rather differ- the capacity curve and the seismic demand can be
ent. It is based on the equivalent linearization. plotted on the same axes and compared.
This is a version of the capacity spectrum The capacity spectrum method of equivalent
method, which was first introduced by Freeman linearization assumes that the equivalent
et al. (1975). The basic assumption is that the damping of the system is proportional to the
maximum displacement of the nonlinear SDOF area enclosed by the capacity curve. The equiva-
system can be estimated from the maximum dis- lent period, Teq, is assumed to be the secant
placement of a linear elastic SDOF system that period at which the seismic ground motion
has the period and damping ratios that are larger demands, reduced by the equivalent damping,
than those of the initial values for the nonlinear intersect the capacity curve (FEMA-440). Since
system. The elastic SDOF system that is used to the equivalent period and damping are both
estimate the maximum inelastic displacements of a function of the displacement, the solution to
the nonlinear system is usually referred to as the determine the maximum inelastic displacement
equivalent or the substitute system. The period (i.e., performance point) is iterative.
and damping of the equivalent system are The equivalent period Teq and effective vis-
referred to as the equivalent period and equiva- cous damping beq (which is used to reduce the
lent damping ratio, respectively. seismic demand) are defined as
As in the previous two methods, the pushover
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
analysis is performed first. In general, ATC-40 m
T eq ¼ T0 (10)
recommends the distribution of the lateral forces 1 þ am a
proportional to the fundamental mode pattern. In
structures where the response considerably 2
ðm 1Þ
ð1 aÞ
beq ¼ 0, 05 þ k (11)
changes after yielding (e.g., in structures with p
m
ð1 þ am aÞ
soft stories – see the discussion in section “Push-
over Analysis: Numerical Models and the Lateral where
Load Pattern”), the adaptive load pattern is T0 is the initial period of vibration of the
required. In high-rise buildings or irregular nonlinear system, a is the post-yield stiffness
Assessment of Existing Structures Using Inelastic Static Analysis 175
ratio, and k is an adjustment factor to approxi- Some Observations and Proposed Improvements
mately account for the changes in the hysteretic of the Procedure, Included in FEMA-356
behavior in reinforced concrete structures. It was observed that the characteristic periods A
The adjustment factor k depends on the hys- which are used to differentiate the response of
teretic behavior of the system. Three equivalent short- versus medium- and long-period structures
damping levels are defined. Type A corresponds were found to be shorter than those observed
to structures with reasonably full hysteretic from nonlinear response history analyses. This
loops, similar to the elastic-perfectly plastic can result in an underestimation of the inelastic
oscillator. Type C corresponds to structures with deformations for the periods between the charac-
severely degraded loops, and type B denotes the teristic period and the periods that are approxi-
hysteretic behavior between types A and C. The mately 1.5 times the characteristic period.
value of k and the corresponding equivalent The equal displacement rule approximation
damping is the largest for systems with was found to lead to a relatively good approxi-
a hysteretic behavior of type A. Their values mation of the maximum inelastic deformations
decrease for degrading systems B and C. The for systems with elastic-perfectly plastic behav-
existing buildings are in general categorized as ior and periods longer than 1 s. This is not always
structures of type C or B, depending on the shak- applicable to soft soil sites and near-fault records.
ing duration and hysteretic behavior of the struc- The limiting values of coefficient C1, which
tural components. defines the ratio of elastic and inelastic deforma-
Based on the comparison of the capacity curve tions, can lead to a large underestimation of the
and seismic demand, the target displacement is displacements in short-period structures. Even if
defined. This displacement is then converted to this limitation was not taken into account, the
roof displacement and other aspects of the magnification of inelastic displacement demands
response are defined. with a decreasing lateral strength for short-period
The ATC-40 method is also well accepted due structures was found to be larger than that
to the clear and useful visualization of the proce- suggested by FEMA-356. The corrections of
dure. Note, however, that the procedure, defined coefficient C1 were proposed for short-period
in EC8/3, can also be presented in a similar man- structures.
ner; nevertheless, it is essentially different. The It was observed that in many cases the cyclic
capacity curve as well as the seismic demand, degradation does not increase the maximum dis-
defined in EC8/3, can also be converted to the placements. Thus the use of the related coeffi-
ADRS format, plotted on the same axes and com- cient C2 is recommended only for structures with
pared, as is presented in Fajfar (1999) and illus- significant stiffness and/or strength degradation.
trated in Fig. 4c. Coefficient C2 is corrected.
It was found that coefficient C3, which is
FEMA-440 intended to represent an increase of displacements
FEMA-440 is the document where the previously due to the dynamic P-D effects, cannot adequately
described procedures (FEMA-356 and ATC-40) take into account the possibility of dynamic insta-
are evaluated. Improvements of both methods are bility. It was proposed that this coefficient be elim-
recommended. Since the document is extended inated. Instead, it was proposed to use the NRHA
with many important observations, particularly for all structures where the strength is below the
for the existing structures, only some of the con- limit proposed in FEMA-440.
clusions related to the methods included in
FEMA-356 and ATC-40 are provided in this sec- Some Observations and Proposed Improvements
tion. Some of them are also presented in section in the Procedure Included, in ATC-40
“Some Issues that Influence the Accuracy of the The accuracy of the estimated maximum dis-
Inelastic (Single-Mode) Pushover-Based placement response in long-period structures
Methods.” depends on the hysteretic behavior type.
176 Assessment of Existing Structures Using Inelastic Static Analysis
In structures with hysteretic behavior type A and pushover analysis is performed taking into con-
periods longer than about 0.7 s, the ATC-40 pro- sideration the same distribution of lateral load as
cedure underestimates the maximum displace- in FEMA-356).
ments. In structures with a hysteretic behavior Since the displacement-based method (DBM)
type B and periods longer than about 0.6 s, is different from those presented in the previous
small underestimations or overestimations of the sections, it is described in more detail in the next
maximum displacements were observed. This paragraphs. The basis of the method was devel-
depends on the level of lateral strength and on oped by Priestley (1995) and further evaluated in,
the site class. For structures with hysteretic e.g., Priestley et al. (2007). The method places
behavior type C, the ATC-40 procedure leads to a direct emphasis on establishing the ultimate
an overestimation of the lateral displacements displacement capacity of the lateral force-
regardless of the period of the structure. resisting elements. Contrary to the methods
In short-period structures (with periods shorter presented in the previous sections, where the
than noted in the previous paragraph), seismic displacements were calculated using
a significant overestimation of the maximum dis- acceleration spectra, in the DBM, the displace-
placements were observed. The overestimation ment spectra are considered.
increases with decreasing strength. In the first step, the flexural and shear strengths
It was also found that the ATC-40 assumption, of the critical sections of the structural compo-
where it is supposed that the inelastic deforma- nents, assuming that no strength degradation
tion demands of structures with hysteretic type occurs due to the cyclic lateral loading in the
B will be larger than those in structures with post-elastic range, are estimated. In the second
hysteretic type A, does not agree with the results step, the post-elastic deformation mechanism of
of the NRHA. According to the NRHA, these the structure and the probable horizontal seismic
deformations were approximately the same and base shear capacity (Vprob) are determined. It is
in some cases even slightly larger in structures recommended that the post-elastic mechanism is
with hysteretic behavior type A. The provisions defined using the (a) simple lateral mechanism
of ATC-40 do not address the potential dynamic analysis (presented in the guidelines),
instability that can arise in systems with in-cycle (b) inelastic response history analysis, and
strength degradation and P-D effects. (c) nonlinear pushover analysis. When the
The suggested improvements of the ATC-40 nonlinear pushover analysis is used to estimate
procedure are presented in chapter 6 of FEMA- the nonlinear response, it is performed in the
440. They are focused on improved estimates of same manner as in FEMA-356 (the same distri-
the equivalent period and equivalent damping. It bution of the lateral load is taken into account).
is concluded that generally the optimal effective In the third step, the plastic rotation capacities
period is less than the secant period and the opti- of the structural members are determined. It is
mal effective damping is also less than that spec- then eventually corrected if the shear failure
ified in ATC-40. occurs before the limits to the flexural plastic
More details about the accuracy, advantages, rotation capacity are reached. The story inelastic
and drawbacks of both methods can be found in drift capacity is estimated from the plastic rota-
Appendix A of FEMA-440. tion capacities.
Considering the critical storey drift and the
NZSEE 2006 previously defined post-elastic deformation
The NZSEE 2006 addresses different methods mechanisms, the overall displacement capacity
of analysis, referred to as (a) force-based method, Usc and the ductility capacity m are determined
(b) displacement-based method, (c) consolidated at the effective height heff of the substitute
force-displacement-based method, and (d) structure (an equivalent SDOF model of the
method using a nonlinear pushover analysis (the structure). The effective secant stiffness keff at
Assessment of Existing Structures Using Inelastic Static Analysis 177
the maximum displacement Usc is determined as Some Issues that Influence the Accuracy
keff = Vprob/Usc. Then the corresponding effec- of the Inelastic (Single-Mode) Pushover-
tive period of vibration Teff is calculated as Based Methods A
sffiffiffiffiffiffiffi
M In this section, some issues that can influence the
T eff ¼ 2p (12) accuracy of the pushover-based methods are
keff
presented: the idealization of the capacity curve,
where M is the effective mass of the substitute the influence of the higher modes, and the influ-
structure (an equivalent SDOF model of the ence of the in-plan torsion. In the last two sub-
structure). Alternatively Teff can be estimated sections, the basic observations from FEMA-440
directly using the Rayleigh-Ritz equation. about the strength degradation and the soil struc-
Based on the evaluated ductility capacity, the ture interaction are cited.
equivalent viscous damping xeff of the structure is
defined. It is recommended to use the method
suggested in Pekcan et al. (1999). In general the Idealization of the Capacity Curve
equivalent viscous damping is determined in In some methods presented in section “Different
a similar manner as in ATC-40, summing the Applications of the Inelastic Pushover-Based
viscous and hysteretic damping. Analysis, Adopted in the (Pre)Standards and
The structural performance factor Sp is calcu- Guidelines,” the actual capacity curve is ideal-
lated according to NZS 1170.5 (2004) and taking ized either by elastic-perfectly plastic or
into consideration the detailing used in the struc- by a bilinear relationship (see Figs. 4 and 5).
ture. It is used to reduce the seismic design This approximation is one of the key issues,
actions on a structure. The displacement demand since it defines the equivalent stiffness of the
is defined using the overdamped elastic displace- equivalent SDOF model. This stiffness further
ment spectrum. It is defined reducing the dis- influences the period of the SDOF oscillator and
placement spectrum d(T), derived from the 5 % the estimated value of the target displacement.
damped elastic acceleration spectrum. The dis- Thus the approximation should be carefully
placement spectrum d(T) is reduced using the performed.
correction factor Kx: The approximation of the capacity curve
is usually performed equating the area bounded
rffiffiffiffiffiffiffiffiffiffiffi
7 (energy) with the actual and idealized curve (see
Kx ¼ (13) Fig. 5). At this stage, the target displacement
2þx
should be assumed, because the compared areas
where x = xeff. depend on the maximum displacement. Thus if
Thus the spectral displacement demand of the the assumed target displacement differs consid-
analyzed structure at height heff is defined as erably from the value obtained at the end of the
analysis of the SDOF system, iterations are
Usd ¼ Sp ð%NBSÞt d T eff K x (14) strongly recommended. Otherwise the target
displacement can be poorly estimated.
where (%NBS)t is the target percentage of the In structures that do not exhibit considerable
new building standard. The %NBS is essentially strain hardening in the nonlinear range, the
the assessed structural performance of the build- elastic-perfectly plastic idealization is accurate.
ing compared with the requirements for a new When the strain hardening is considerable, e.g., in
building. some types of bridges, bilinear idealization is
Then the displacement capacity Usc and the more appropriate (see section “Application of
demand Usd are compared. If Usc/Usd 1, retrofit the Inelastic Static Methods to the Analysis of
is unnecessary to achieve (%NBS)t and vice versa. Bridges” for more details).
178 Assessment of Existing Structures Using Inelastic Static Analysis
7.1
5 × 4.5 = 22.5
6.0 3.8 4.6
25.5
21.5
3.0
7 9 × 5.4 = 48.6 7 6.0 3.8 4.6 7.1
62.6 21.5
Assessment of Existing Structures Using Inelastic Static Analysis, Fig. 6 An example of the structure, irregular in
plan and elevation
The Influence of Higher Modes shears, and overturning moments can deviate
It was mentioned before that the basic variants of significantly from the NRHA.
the methods described in section “Different In buildings the importance of a higher-mode
Applications of the Inelastic Pushover-Based effect increases with the amount of inelasticity
Analysis, Adopted in the (Pre)Standards and (note that in some bridges quite the opposite trend
Guidelines” can be classified as single-mode was observed – see section “Application of the
nonadaptive methods since they assume that the Inelastic Static Methods to the Analysis of Brid-
response is controlled by a single predominant ges” for more details). Typical examples are, e.g.,
mode which remains unchanged after yielding RC shear walls, where this phenomenon has been
occurs. This assumption limits their application. observed years ago. An explanation can be found
For example, when they are used for the analysis elsewhere (e.g., Rejec et al. 2012).
of existing structures which are irregular in ele- Higher-mode contributions become more sig-
vation and/or in plan (see an example in Fig. 6), nificant for structures with fundamental periods
certain extensions are needed in order to take into that fall into the constant-velocity part of the
account the important influence of higher modes response spectrum. Forces which are developed
and the in-plan torsion. In high-rise buildings, the due to the important influence of higher modes
influence of the higher modes can be important can considerably influence the failure mecha-
regardless of their irregularity. In such structures, nism. Thus the pushover analysis cannot always
multimode pushover methods (see section identify this mechanism.
“Some Alternatives to Single-Mode Pushover- In FEMA-440, a single first-mode distribution
Based Methods” for more details) or NRHA can was found sufficient for the estimation of dis-
be used as an alternative to single-mode placement and other quantities that were not sig-
pushover-based methods. nificantly affected by higher modes. The adaptive
The influence of the higher modes has been the load distribution was sometimes better and some-
subject of different studies. The one presented in times worse than the first-mode distribution. The
FEMA-440 outlines the most important observa- uniform distribution is recommended in most of
tions. Some of them are provided in the following the codes as the second choice in order to frame
paragraphs. the response quantities. However, in FEMA-440
The nonlinear static pushover procedures it was found that it often did not fulfill this role.
appeared to be reliable for the design and The uniform distribution was found to be the
evaluation of low-rise buildings. In relatively worst with regard to all the monitored response
tall frame buildings, where the higher-mode quantities. Thus it was not recommended as
response is significant, interstory drifts, story a stand-alone option.
Assessment of Existing Structures Using Inelastic Static Analysis 179
All the codes addressed in this entry recognize stiffness in a force-deformation relationship for
that the single-mode pushover-based methods are a structural model using nonlinear static proce-
less efficient when the higher modes are impor- dures. The performance implications depend on A
tant. Thus they require these methods to be com- the type of strength degradation (cyclic or
bined with the results of the linear dynamic in-cycle strength degradation). For structures
procedures or the multimode pushover analysis that are affected by component strength losses,
is required. Specific solutions can be found in the including P-D effects, occurring in the same cycle
particular standard. as yielding (in-cycle strength degradation), the
negative post-elastic slope can lead to the
The Influence of In-Plan Torsion, dynamic instability of the structural model
Two-Dimensional (2D) Versus Three- (FEMA-440 2005). For this reason, it is
Dimensional (3D) Analysis suggested that the pushover-based methods are
The in-plan torsion can importantly influence the only used if the strength of the structures is above
response of the existing structures which are irreg- a certain prescribed limit. Otherwise the use of
ular in plan. In such structures, the 3D analysis is the NRHA is recommended.
generally needed. The considerable influence of
in-plan torsion can be expected in torsionally flex- The Soil-Structure Interaction
ible structures, where the first mode of vibration is There is a perception among many in the practic-
torsional. Substantial torsional effects can also be ing engineering community that short, stiff build-
obtained in one direction of structures where the ings do not respond to seismic shaking as
second mode of vibration is torsional. adversely as might be predicted analytically.
It was observed that the inelastic static There are several reasons why short-period struc-
methods can significantly underestimate the dis- tures may not respond as conventional analysis
placements on the stiff/strong side of such build- procedures predict. Among these are (a) radiation
ings (see, e.g., Kreslin and Fajfar 2012). Thus the and material damping in supporting soils,
displacements on the stiff/strong side of the struc- (b) structures with basements that experience
ture should be increased. reduced levels of shaking, (c) incoherent input
Most of the (pre)codes recognized this phenom- to buildings with relatively large plan dimen-
enon. Thus they require an increase in the seismic sions, and (d) inaccuracies in modeling, including
displacements, defined by inelastic static methods, dumping of masses, neglecting the foundation’s
due to the torsional effects. All of them require flexibility, and some elements that contribute to
a 3D analysis of the structures where the in-plan the strength (FEMA-440 2006). In FEMA-440
torsion is important. According to all codes, the procedures, it is proposed that soil-structure
accidental eccentricity should be taken into interaction is incorporated into the nonlinear
account in all analyses regardless of their torsional static analyses.
flexibility. The specific requirements related to an
increase of displacements due to the torsional
effects differ from standard to standard. Some Alternatives to Single-Mode
In general, the 3D pushover-based analysis Pushover-Based Methods
and torsional effects are topics that require addi-
tional investigations. Thus, they are the subjects Multimode procedures are considered as an alter-
of numerous researches, reported in the literature. native approach for the analysis of structures
An overview of these researches is recently pro- where the single-mode methods are less accurate.
vided in Bhatt (2011). There are many multimode pushover methods
described in the literature. In this section, two of
The Strength and Stiffness Degradation them are briefly presented. The multimode push-
The strength degradation, including P-D effects, over analysis (MPA) (Chopra et al. 2004) is
can lead to an apparent negative post-elastic selected as being representative of the
180 Assessment of Existing Structures Using Inelastic Static Analysis
nonadaptive multimode pushover-based method, which means that it can take into account
methods, and incremental response spectrum the influence of the higher modes as well as their
analysis (IRSA) (Aydinoğlu 2003) is selected as changes depending on the seismic intensity. The
the representative of the adaptive multimode contributions of the different modes are consid-
pushover-based methods. Both methods were ered in an incremental pushover analysis.
addressed in FEMA-440. The MPA was evalu- When the structure enters the nonlinear range,
ated as an alternative to the single-mode its dynamic properties are changed each time
pushover-based methods, and the IRSA was rec- a new plastic hinge is developed. In regular struc-
ognized as the potential improvement of the tures, these changes are typically quite small. In
inelastic analysis techniques that can be used to irregular structures, the mode shapes as well as
reliably address the MDOF effects. their contributions to the overall response can
significantly change. IRSA can take into account
The MPA Method these changes. More importantly, IRSA is also
The MPA method was developed by Chopra and capable of taking into account the effect of modal
Goel (e.g., Chopra and Goel 2002). The analysis coupling to the formation of the plastic hinges.
is performed in a similar manner as that presented Different studies (e.g., Kappos et al. 2012)
in section “Eurocode 8/3 (CEN 2005).” However, have confirmed that the method is quite reliable;
the number of the analyses depends on the num- however, it also has certain limitations, and thus
ber of important modes, identified in the it is not universal and cannot replace the NRHA
initial – elastic – state. The pushover analysis is in all cases, for example, in certain types of
performed separately for each important mode. bridges.
The lateral load is proportional to the shape of the
vibration mode. Based on each pushover analy-
sis, the contribution of the related mode of vibra- Application of the Inelastic Static
tion to the seismic displacements is defined. Methods to the Analysis of Bridges
These contributions are then combined using the
appropriate combination rule (e.g., SRSS). The With regard to their dimensions and structural
method supposes that the modes of vibrations are systems, and in general with regard to their seis-
invariant. Thus it can be classified as the mic response, bridges are quite different from
nonadaptive method. This limits its applicability. buildings. Therefore the application of different
When considerable changes of the mode shapes pushover-based methods, which were originally
can be observed in the nonlinear range developed for buildings, is not straightforward,
(an example is described in section “Pushover particularly when the bridges are analyzed in the
Analysis: Numerical Models and the Lateral transverse direction. For the analysis of bridges in
Load Pattern”), the method is in general less the transverse direction, the pushover methods in
reliable. general should be applied in a slightly different
It was examined in FEMA-440 as an alterna- manner. The analysis differs mainly regarding:
tive to single-mode pushover-based methods. (1) the choice of the reference point where the
Taking into account a study of five buildings of displacements are registered, (2) the distribution
very different properties, it was found to be more of the lateral load, and (3) the idealization of the
accurate than the single-mode pushover methods, capacity curve. These issues are briefly presented
but not completely reliable. Similar observations in the following subsections. In the last subsec-
were obtained based on other studies, e.g., studies tion, the applicability of the pushover methods
of bridges, presented in Kappos et al. (2012). for the analysis of bridges is discussed.
c Displacements of V213P
20 P1 P3
Displacement [cm]
P2 MD
15
10
0
0 50 100 150 200
Station [m]
Assessment of Existing Structures Using Inelastic Static Analysis, Fig. 8 Pushover curves and displacement
envelopes, obtained in a highly irregular viaduct, based on different reference points
to the fundamental mode of the bridge in the overestimated maximum displacement (see
elastic range and (b) uniform distribution. Note Fig. 11) can be obtained. Namely, in bridges
however that the shape of the load pattern, pro- with pinned abutments, the capacity curve can
portional to the fundamental mode, depends on exhibit a considerable strain hardening slope,
the type of the supports above the abutments as it which should be properly taken into account.
is presented in Figs. 9 and 10. In regular bridges, This is illustrated in Fig. 11.
pinned at the abutments, the parabolic distribu-
tion (see Fig. 9c) is also feasible. Applicability of Pushover-Based Methods for
the Analysis of Bridges
The Idealization of the Capacity Curve The single-mode methods can accurately predict
The idealization of the capacity curve can signif- the response of regular bridges where the influ-
icantly influence the stiffness of the equivalent ence of higher modes is not important. This is the
SDOF model and the estimated value of the max- case where the effective mass of the predominant
imum seismic displacement. When this stiffness mode exceeds 80 % of the total mass. When the
is not adequately estimated, the actual and esti- methods are nonadaptive, they can be accurately
mated maximum displacement can be signifi- used if the mode shapes do not significantly
cantly different. In some methods, such as that change based on the seismic intensity.
included in the EC8/3, the capacity curve is In bridges where the superstructure is consid-
approximated using the elastic perfectly plastic erably stiffer than the supporting elements
idealization. However, in viaducts which are (piers), the influence of the higher modes is in
pinned at the abutments, this idealization can be general negligible. Typical representatives are
inappropriate, since an underestimated equiva- short- and medium-span bridges (e.g., the length
lent stiffness of the SDOF system and an of a bridge is less than 500 m), which are not
Assessment of Existing Structures Using Inelastic Static Analysis 183
L L
X
m1 ... mj mn m1 mj mn
Fj = pj mj Fj ...
A
F1
Fj Fn F1 Fj Fn Fj = pj mj Fj
Φ1 Φj Φn Φ1 Φj Φn
c parabolic
Assessment of Existing Structures Using Inelastic
4 4
Φj = − x2 + x Static Analysis, Fig. 10 Distributions of the lateral
2 L
L load, appropriate for bridges with roller supports at the
abutments
Assessment of Existing Structures Using Inelastic
Static Analysis, Fig. 9 Distributions of the lateral load,
appropriate for bridges that are pinned at the abutments
significant influence of the higher modes to the
deflection line of the superstructure.
supported by very stiff (very short) piers (e.g., in In many bridges, the accuracy of the single-
single-column piers, the height of the columns mode methods increases with the seismic inten-
exceeds 10 m). sities (nonlinearity). Columns yield, their stiff-
Considerable changes of the mode shapes can ness is reduced, and thus the superstructure has
be expected (and were observed) first of all in short a more important role for the overall response.
bridges, where the displacements of the super- However, this is not the rule. In bridges where the
structure above the abutments are not restrained. torsional stiffness decreases when the columns
In such bridges, the predominant mode usually yield, the single-mode methods cannot accurately
changes considerably, when the damage of the predict the response at high intensity levels, due
side columns reduces their stiffness to such an to the emphasized torsional rotations.
extent that the torsional stiffness of the bridge In bridges with roller supports above the abut-
becomes lower than its translational stiffness. ments, considerable changes of mode shapes can
The accuracy of the single-mode pushover- occur, particularly when the side spans are rela-
based methods depends on: (a) the ratio of the tively long. When they are supported by short stiff
superstructure stiffness and the stiffness of the columns in the central part of the superstructure,
bents (the length of the bridge and number and considerable torsional effects can be obtained.
the location of the short columns along the bridge); In long bridges, the influence of the higher
(b) the relative strength of the columns, compared modes in the majority of cases does not depend
to the seismic intensity; and (c) the boundary con- on the stiffness of the columns and their strength.
ditions at the abutments (mostly in short bridges). In such bridges, the superstructure of the standard
When the superstructure is stiff compared to types becomes quite flexible, and consequently
the supporting columns (piers), it has the higher modes become important, regardless
a predominant role defining the response of the of the stiffness and the strength of the columns.
bridge. In such cases, the response is typically When the response of the bridge is consider-
influenced by one predominant mode. However, ably influenced by the higher modes or the modes
if the bridge is supported by short stiff columns, of vibration are changing considerably depending
they govern the response and cause the on the seismic intensity, the multimode pushover
184 Assessment of Existing Structures Using Inelastic Static Analysis
Assessment of Existing a Bridges pinned at the abutments b Bridges with rollers at the abutments
Structures Using
Inelastic Static Analysis, Bilinear idealization Elastic-perfectly plastic idealization
Fig. 11 Idealization of the
pushover curve
Force
Force
Maximum
displacement Maximum
displacement
Displacement Displacement
For intensity-based assessment (which is the The main goal of the latest generation of
default option in most codes and standards for performance-based earthquake engineering
seismic assessment, including ASCE 41, EC8-3, guidelines (such as FEMA P-58) is to develop
and the NZSEE Recommendations, reviewed in tools to convert engineering performance objec-
Section “Published Codes, Standards and Guide- tives into metrics that are more meaningful to
lines”), one or more performance objectives will other stakeholders, such as building owners,
be checked, each comprising a return period at building tenants, and insurance companies. Per-
which the probabilistic seismic hazard is evalu- formance may be quantified in terms of direct
ated and the desired structural and nonstructural monetary costs (for repair or replacement),
performance at this hazard level. In ASCE 41, for human impact (chance of casualties or fatalities),
example, the “basic safety objective” (BSO) is and indirect costs (time to repair, reoccupy, or
considered to be satisfying a life safety perfor- return to function) – snappily summarized by the
mance level under a 475-year return period 3 Ds: “dollars, deaths, and downtime.” Perfor-
ground motion and a collapse prevention perfor- mance objectives expressed in this format of
mance level under a 2,475-year return period. course place a higher burden on the analyst – par-
Nonstructural performance objectives are also ticularly for areas outside the United States, where
given, including a life safety objective to ensure the default fragility functions given in PACT for
that heavy equipment or façades do not fall and various nonstructural components may not apply.
endanger life or impede egress. The recently developed REDiTM Rating System
Performance objectives both for assessment (Resilience-based Earthquake Design Initiative;
and (if required) retrofit design for existing struc- Almufti and Willford 2013) bases its platinum,
tures are often allowed to be less onerous than in gold, and silver performance resilience objectives
the design of new structures. This is justified partly on numerical estimates of post-earthquake
primarily on the basis of pragmatism: incorporat- reoccupancy and functional recovery times and
ing seismic-resistant detailing into new buildings provides a useful framework for discussion of
usually comes at a relatively small premium, enhanced performance objectives with clients
whereas the cost of retrofit (including both direct and other stakeholders.
and indirect costs) may be prohibitive. This is
especially true of historical structures, where
extensive strengthening measures may not be Modeling Considerations
acceptable from a cultural heritage point of
view (of course, this cultural heritage is exactly As with any engineering or scientific numerical
what seismic strengthening is trying to preserve modeled exercise, the outputs from NRHA are
in the long term). Less onerous performance only as reliable as the specific modeled tech-
objectives are also sometimes justified on the niques used and the numerical inputs. Although
basis of a lower remaining design life than the NRHA is perceived as giving the most accurate
nominal value adopted for existing structures estimates of building seismic performance, it has
(and therefore lower exposure period to seismic the disadvantage that model checking (of both
hazard). The BSO in ASCE 41, for example, is inputs and outputs) may be hampered by com-
considered to represent a potentially higher risk plexity and potentially by computational power
than expected for properly designed and (e.g., the time taken to envelope results of large
constructed new buildings. In jurisdictions where structural analysis models over multiple ground
seismic assessment and retrofit are not obligatory, motions can discourage the analyst from
building owners may decide on performance checking all response quantities thoroughly).
objectives for themselves, based on a cost-benefit Analysts may be overconfident of the results of
analysis of the cost of strengthening and the risk NRHA and feel that detailed modeled obviates
reduction in terms of future repair costs and poten- the need to apply engineering principles to the
tial loss of life following an earthquake. assessment of structures.
Assessment of Existing Structures Using Response History Analysis 189
The following subsections provide a summary been reached among researchers and codes of
of some modeled decisions and interpretations practice on the most appropriate methods. Cate-
that face an analyst employing NRHA on an gories of ground motions applied in seismic A
assessment project. Many of these aspects are assessment are:
covered in more detail in other chapters in this
entry. The discussion is divided into aspects relat- • Recorded by accelerometers in real earth-
ing to the seismic input (ground acceleration his- quakes, applied to the structural model
tories), structural modeled, and treatment of unmodified
outputs from the analysis. • Recorded in real earthquakes multiplied by
a linear scaling factor to achieve compatibility
Seismic Input with the peak ground acceleration or other
From the point of view of the structure, seismic scalar measure of seismic intensity (e.g., spec-
action is felt through acceleration of the ground at tral acceleration at the fundamental period of
the base – in both horizontal and vertical the structure)
directions – and therefore NRHA requires appro- • Recorded in real earthquakes modified to
priate ground acceleration histories to apply. In achieve spectrum compatibility across
most cases, the seismic hazard is defined in the a range of structural periods (spectral
form of a response spectrum, and therefore com- matching)
patibility with a target spectrum is one of the most • Generated from a purely mathematical algo-
important considerations in definition of ground rithm, such as modified white noise
acceleration histories, but nonstationary charac- • Derived from a physics-based simulation of
teristics such as duration can also be important. the seismic wave generation and propagation
Results can be very sensitive to the ground process.
motions applied, and the ground motions to be
used for a given application depend not only on NEHRP Consultants Joint Venture (2011)
the seismic hazard and expected seismic shaking provides a discussion of the various consider-
at the site but also on the objectives of the ations for development of ground motions for a
analysis. project.
For the assessment of mean (expected) Most literature on ground motion develop-
response only, current codes allow as few as ment considers only a single component of hori-
three ground motions to be assessed, although if zontal input and does not discuss the proper
fewer than seven are used, the maximum of each treatment of horizontal variability and the
response quantity across all the analyses should application of multiple components of ground
be used. These requirements date back to the level motion to a 3D analysis model. Ground motions
of computing power available in the 1980s and recorded in real earthquakes are typically not
1990s, and nowadays, seven is generally consid- axisymmetric – i.e., they do not shake each hor-
ered to be a sensible minimum, from which the izontal direction equally strongly. In fact, in some
arithmetic mean response can be taken (NEHRP cases, structures with a certain period may be
Consultants Joint Venture 2011). For estimates of shaken strongly in one direction, whereas struc-
the statistical distribution of structural response tures with a different period may be shaken more
quantities, many more ground motions will be strongly in a different direction. Figure 1 shows
required – NEHRP Consultants Joint Venture an example of such a case, where the 1.3-s struc-
(2011) suggests 30 records, although this depends tural response is strongest at 79
from the x-axis
on the variability of structural response and the and the 3.0-s structural response is strongest at
ground motions affecting the site. 161
. Grant (2011) developed the software,
There is a substantial body of literature on the RspMatchBi, to spectrally match recorded
development of ground motions appropriate for ground motions to retain the axes of strong shak-
the NRHA of structures, and no consensus has yet ing of the original motion while matching the
190 Assessment of Existing Structures Using Response History Analysis
a b
Assessment of Existing Structures Using Response major and minor axes of response for PEER Record
History Analysis, Fig. 1 Response orbits (displacement 184 from Imperial Valley event; (a) 1.3-s period response
of single degree of freedom oscillators) and definition of and (b) 3.0-s period response (From Grant et al. 2011)
spectra of the record to target spectra. NEHRP input seismic demand is lower than the elastic
Consultants Joint Venture (2011) provides some response of the structure, nonlinear deformation
discussion on the application of multicomponent is not expected, and a linear model would suffice.
ground motions to a structural model. On the other hand, when a degree of nonlinearity
is expected in the structure’s seismic response,
Structural Modeling explicitly modeled this nonlinearity may be nec-
The most appropriate type of response history essary, allowing a more accurate estimation of
analysis model (or analysis model in general) is the overall structural response (e.g., peak dis-
very application specific. If a simple and conser- placements or storey drifts), but also a better
vative model is adequate to demonstrate accept- understanding of how seismic demand is plasti-
able seismic performance, then there is little cally redistributed among parallel load paths
benefit in introducing extra detail and complex- following first yield.
ity. In practice, modeled requirements are differ-
ent for every project, and it is not possible to give 2D Versus 3D
a recipe that can be followed for every applica- Regular structures in plan, for which torsional
tion. The following are some aspects of structural response is expected to be minimal, are often
analysis that are important for the use of NRHA analyzed with separate 2D models for each prin-
in both the assessment of existing structures and cipal axis of the structure. In engineering practice
in new design. (in contrast to academic research), projects on
which NRHA is considered worthwhile are sel-
Linear Versus Nonlinear dom particularly regular, although there are of
Although the focus of this entry is on nonlinear course many exceptions. 3D analysis does not
analysis, the extra effort involved to introduce introduce significant complexity in pre- and
material and geometric nonlinearity is not always post-processing when using modern analysis
justified – particularly when target performance packages, with the exception that the ground
objectives restrict the response to the elastic or motion input is more complicated.
near-elastic range. This also applies at
a component level – for example, force- Level of Detail
controlled (brittle) elements are usually modeled A given structure can be analyzed in different
linearly, given that their nonlinear response is levels of detail: from an equivalent single degree
generally unacceptable. Of course, even if of freedom representation through to a detailed
heavily nonlinear response is acceptable, if finite element model. Often, several different
Assessment of Existing Structures Using Response History Analysis 191
Assessment of Existing Structures Using Response History Analysis, Fig. 2 Different levels of analysis detail for
modeling an unreinforced masonry house
levels of analysis will be considered for one of the wall piers. Figure 2b shows masonry walls
project: for example, simplified models at an modeled in 2D shell elements; masonry material
initial stage and for subsequent model checking, models may be of a smeared crack type, in which
a main structural model (e.g., frame members the development of cracks in the walls is modeled
modeled as 2D beam elements) for overall per- explicitly, or a phenomenological hysteresis
formance assessment, and detailed local models model in which backbone and cyclic hysteretic
to calibrate aspects of the main model (e.g., rules, based on experimental behavior, are used.
modeled force transfer within a RC joint region Finally, Fig. 2c shows a model in which individ-
using 3D solid elements for concrete, 2D beam ual bricks are modeled with 3D solid elements.
elements for reinforcing bars, and bond and fric- In this case, the interaction between bricks is
tion between them defined with a mathematical described with a contact surface formulation,
model). which allows for opening and closing of cracks
Three different levels of detail are shown in between the bricks and the transfer of compres-
Fig. 2, for the NRHA of an unreinforced masonry sion and traction across the interfaces. If cali-
(URM) house. Figure 2a shows an equivalent brated correctly, this model can explicitly track
frame model, in which masonry pier flexural the development of failure mechanisms in the
and shear responses are modeled with nonlinear URM wall up until the point of collapse.
spring or link elements. For this model, the mass Another example of a multilevel approach, as
is lumped at an equivalent roof height, and a rigid applied to the new design of an offshore platform,
diaphragm connects the mass to the tops of each may be found in Gibson et al. (2012).
192 Assessment of Existing Structures Using Response History Analysis
Assessment of Existing
Structures Using Monotonic backbone
Response History Cyclic response
Analysis,
Fig. 4 Component
force–deformation
hysteretic behavior,
generated using the Ibarra
Force
et al. (2005) hysteresis
model
Displacement
adopted is able to accurately describe available the capacity of the bar in question is reduced by
experimental data for the level of strain expected a linear ratio of the provided to required devel-
in the analysis and the cyclic conditions expected. opment length and also significantly reduces
plastic rotation limits by up to a factor of 10.
Modeling Deficiencies in Structural Design
NRHA is most commonly adopted for the assess- Model Checking and Post-Processing
ment of existing structures when the structural The amount of post-processing required for
detailing is deficient with respect to modern a given NRHA model is in some ways inversely
code requirements, and therefore adequate seis- related to the amount of rationalization and sim-
mic performance cannot be demonstrated by sim- plification that has gone into the modeled effort.
pler analysis methods. Typical examples of For example, if URM walls are modeled with
detailing deficiencies are inadequate develop- a discrete element approach (e.g.,
ment lengths and splice lengths for reinforcing Fig. 2c) – wherein all bricks and their interaction
bars, transverse reinforcement in beam–column are explicitly modeled – then crack widths can be
joints and plastic hinge regions, pre-Northridge evaluated explicitly, and the “no collapse” per-
steel beam–column connections, presence of formance requirement can be checked by obser-
unreinforced masonry infills in frame buildings, vation of the state of the model at the end of the
inadequate shear strength and/or failure to satisfy analysis. In frame analyses – with 2D beam ele-
a capacity design hierarchy in beams and col- ments used for beams and columns – if mono-
umns, and unreinforced masonry buildings in tonic and cyclic degradation are properly
general. Although it may be desirable to address accounted for, collapse response can also be
any of these deficiencies through retrofit when observed directly. More commonly, adequate
they are present, there are many cases when performance of deformation-controlled elements
acceptable behavior can be demonstrated. Gener- is evaluated on the basis of observed peak plastic
ally, detailing deficiencies such as splice lengths deformations (e.g., peak plastic hinge rotations in
and beam–column joint details are addressed by beams and columns, peak plastic axial deforma-
limiting both the strength and plastic deformation tion in braces, peak drifts in RC or URM walls)
capacity of the affected members. For example, compared with allowable limits from published
when reinforcing bars are anchored with inade- guidelines (e.g., those considered in section
quate development length, ASCE 41 requires that “Published Codes, Standards and Guidelines”)
Assessment of Existing Structures Using Response History Analysis 195
or experiments from the literature carried out earthquake may indicate some cracking that
specifically for the project. Performance of closes up following the earthquake and does not
force-controlled elements is evaluated on the require structural repair. Residual deformations A
basis of observed forces from the analysis com- are sensitive to backbone characteristics (such as
pared with capacities evaluated according to post-yield stiffness, line BC in Fig. 3), as well as
guidelines or engineering first principles. cyclic hysteretic characteristics (such as
As noted in section “Seismic Input”, analyses unloading stiffness and pinching). Furthermore,
are typically repeated for multiple earthquake estimating residual deformations in NRHA can
ground motions, and potentially for multiple require a significantly longer analysis, allowing
application orientations of each ground motion, time after the earthquake ground motion for the
to account for the fact that record-to-record structure to settle. Residual deformations are not
uncertainty is relatively large compared to other explicitly required to be checked in most
sources of uncertainty. Appropriate processing of published guidelines, although approximate
structural analysis results will depend on the con- limits are given in ASCE 41. FEMA P-58
siderations discussed in that section. When the discussed the estimation of residual interstory
peak response from multiple analyses is consid- drifts in an appendix. On the basis that estimation
ered (say if fewer than seven ground motions are of residual response is very sensitive to modeled
considered as input), this does not introduce sig- and ground motion characteristics, it suggests
nificant difficulty – every analysis carried out using predefined relationships between residual
must demonstrate acceptable response. Consider- and peak drifts and to not attempt to evaluate
ation of mean response, however, introduces sev- them in the main NRHA model.
eral difficulties. For purely numerical quantities Nonstructural components are seldom
(e.g., plastic hinge rotation), typically codes are included in a NRHA model, and therefore their
interpreted to refer to arithmetic mean, whereas assessment is carried out in the post-processing
the geometric mean would be more consistent phase. Some so-called nonstructural components,
with the common assumption of log-normally such as URM infill panels in frame buildings,
distributed structural response. For qualitative should be included in the structural model,
responses, such as “collapse” versus “no col- when their inclusion is expected to affect struc-
lapse,” or “uplift at foundation” versus “no tural response. Nonstructural components are
uplift,” means cannot be evaluated. In these typically classified as either displacement/drift-
cases, the median is a more appropriate measure sensitive (such as partitions or façade panels) or
of central tendency of analysis results. Many acceleration-sensitive (such as electrical equip-
engineers consider more than just statistics ment), and the performance of some equipment
when interpreting analysis results – even if four may be rated on the basis of a limiting peak drift
out of seven analyses demonstrate no collapse, or acceleration level. For more important equip-
most engineers would be interested in assessing ment, response can be evaluated with a response
how collapse occurs in the other three and, if history analysis using floor acceleration histories
economically possible, prevent these modes of recorded in the main NRHA model or a response
collapse from occurring. NEHRP Consultants spectrum analysis using secondary response
Joint Venture (2011) provides further discussion spectra, also from the main model. Finally, sta-
of how analysis and performance objectives tistical performance of nonstructural components
affect ground motion development and analyses is often expressed in the form of a fragility func-
carried out. tion, which gives the probability of reaching dif-
Residual deformations can be more important ferent damage levels for a given level of
than peak transient deformations for evaluating acceleration or drift. Fragility functions for com-
performance criteria relating to post-earthquake mon types of equipment and other nonstructural
operability or possibility of occupancy. For components (at least for US practice) are given in
example, peak deformations during the FEMA P-58.
196 Assessment of Existing Structures Using Response History Analysis
Published Codes, Standards, and note below about the updated version ASCE
Guidelines 41/SEI 41-13). Some deformation-controlled
actions – such as flexure in columns of moment
This section provides a review of available inter- frames – may be inhibited for new structures
national guidelines on the seismic assessment and using capacity design procedures, but for existing
retrofit design of existing structures. Regulatory structures, it would be excessively conservative
requirements for demonstrating seismic perfor- to condemn a noncompliant structure if adequate
mance vary, but in real projects, it is helpful to seismic performance can be demonstrated.
demonstrate compliance with documents such as ASCE 41 tabulates nonlinear deformation
the following to establish credibility – although acceptance criteria for all types of structural com-
this should establish a minimum duty of care and ponents; in fact, these criteria are routinely
should not restrict the engineer from applying applied also for the performance-based seismic
state-of-the-art methods and models when they design of new structures and are referred to in
contradict the published guidelines. guidelines for analysis and design of high-rise
buildings (e.g., ATC 2010). Allowable plastic
ASCE/SEI 41-06: Seismic Rehabilitation of deformations are tabulated in terms of relevant
Existing Buildings characteristics such as performance objective,
ASCE/SEI 41-06 (ASCE 2007) – herein ASCE member slenderness (for steel braces), presence
41 – is perhaps the most complete published of high shear and axial load and compliance of
guideline available for assessment and retrofit transverse reinforcement with code requirements
design for existing buildings, including extensive (for concrete beams and columns), and other non-
provisions on NRHA. Acceptable nonlinear seis- compliances with respect to modern codes.
mic performance is evaluated on a component- A disadvantage of this tabular format is that some-
by-component basis (e.g., plastic rotations in times small changes in assumptions (e.g., a change
beam and column plastic hinges), and all struc- from “noncompliant” transverse reinforcement to
tural elements must satisfy these checks for com- “compliant” transverse reinforcement) can make
pliance. Different limits are quantified for a large step change to the allowable deformation,
different performance objectives: Immediate whereas the real effect modeled is likely to be
Occupancy, Life Safety, and Collapse Preven- continuous.
tion. Overall measures of structural performance Nonlinear behavior of deformation-controlled
such as interstory drifts are not explicitly actions is modeled following the generalized
checked, except to evaluate possibility of impact force–displacement or moment–rotation back-
with adjacent buildings and in checking elements bone illustrated in Fig. 3. Parameters
that span between stories such as reinforced con- (deformations and forces) to describe these
crete (RC) shear walls and nonstructural curves are tabulated for every component in
components. ASCE 41. Note that ASCE 41 does not provide
ASCE 41 distinguishes between deformation- guidance on modeled the cyclic response of com-
controlled and force-controlled actions, as ponents following the backbone in Fig. 3; there-
discussed in section “Material and Section Prop- fore, modeled of hysteretic phenomena such as
erties”. Deformation-controlled actions are stiffness and strength degradation and pinching is
modeled with their expected (mean) strength not foreseen (see section “Hysteretic Response of
and nonlinear response quantities, while force- Components”). ATC-72-1 (ATC 2010) notes that
controlled actions are restricted to their lower- the force–displacement backbone represents
bound (mean minus one standard deviation) a “cyclic envelope curve” (i.e., represents cycli-
strength. The nonlinear response of force- cally degraded response) rather than the force
controlled actions is not usually modeled, since envelope that one would get from a monotonic
exceeding their linear capacity would result in test on the component. Furthermore, restricting
a sudden shedding of their load (although see deformations of primary structural components
Assessment of Existing Structures Using Response History Analysis 197
to within point C on Fig. 3 ensures that the region (limit states and associated return periods of seis-
in which cyclic degradation dominates is not mic action) that must be satisfied, either before or
reached. after strengthening. Performance objectives con- A
The next edition of ASCE 41 – ASCE/SEI sidered in the standard are defined as Near Col-
41-13 (ASCE 2014) – has just been published lapse, Significant Damage and Damage
and expands the remit of ASCE 41 to include Limitation – inconsistently with the performance
the multitiered seismic assessment approach pre- objectives in EC8-1 for new structures – but the
viously found in ASCE/SEI 31-03 (ASCE 2003). actual objective(s) to be checked is left to the
The nonlinear response history analysis proce- National Annex.
dures have been expanded somewhat, reflecting Similarly to ASCE 41, EC8-3 distinguishes
significantly increased application of this method between ductile and brittle actions. For nonlinear
and the availability of research efforts such as analyses, capacities are evaluated on the basis of
Deierlein et al. 2010; NEHRP Consultants Joint deformation for ductile actions and strength for
Venture 2010, since the publication of ASCE/SEI brittle ones. Column flexure is considered a
41-06. Nonlinear behavior of force-controlled ductile action, although interestingly “strong
actions is now permitted to allow explicit column–weak beam” behavior is enforced for
modeled of post-failure redistribution following steel moment frames only (not reinforced concrete).
brittle failure. Finally, and importantly for this Acceptance criteria for nonlinear actions are
entry, the unreinforced masonry (URM) wall pro- contained in informative Annexes A, B, and C for
visions have been updated; major changes are RC, steel, and masonry, respectively. No guid-
that bed-joint sliding is now considered a ance is given for the assessment of timber struc-
deformation-controlled action, while the pier tures. Generally, for elements classified as
rocking mechanism is now limited to lower primary components of the seismic force
axial load ratios and for piers at least 600 in resisting system, allowable deformations are
thickness. explicitly noted as mean minus one standard
deviation estimates, and for secondary compo-
Eurocode 8 Design of Structures for nents, they are the mean values.
Earthquake Resistance Part 3: Assessment For RC beams and columns, allowable total or
and Retrofitting of Buildings plastic rotations are expressed in terms of regres-
Eurocode 8 Part 3 (CEN 2005) – herein sion relationships on experimental data, includ-
EC8-3 – deals with both the seismic assessment ing such aspects as section depth, shear and axial
and retrofitting of existing buildings. It is inter- force ratios, longitudinal and transverse rein-
esting to note that other Eurocodes do not cover forcement ratio, and material properties. Modifi-
non-seismic assessment of existing structures, cations are also provided for lack of compliance
but it was recognized by the code developers with modern requirements, such as lap splices in
that many older structures in Europe did not plastic hinge zones and the use of smooth
take into account seismic resistance in their reinforcing bars. Alternative expressions are
design (whereas presumably gravity and wind also provided for curvatures rather than rotations.
have been taken into account – at least by com- Annex A is also notable for a ductility-dependent
mon design practice if not explicit calculation). In shear strength calculation and specific rules for
common with the other standards considered concrete, steel, and fiber-reinforced polymer
here, EC8-3 leaves the development and details (FRP) jacketing of seismically deficient ele-
of a national program for seismic risk mitigation ments, including estimates of the strength and
to the member countries, including such critical deformation capacities provided by these inter-
aspects as what triggers the need for assessment ventions that could be used in NRHA.
of an existing building, timescales over which Acceptability criteria for steel elements,
assessment and retrofitting must take place contained in Annex B, are more comparable to
(if required), and the performance objectives the equivalents in ASCE 41. They are expressed
198 Assessment of Existing Structures Using Response History Analysis
The NZSEE Recommendations contain use- distributions of other demand parameters associ-
ful discussion on the modeled of URM struc- ated with damage to structural and nonstructural
tures; they note the importance of including the components (e.g., peak floor acceleration and A
distribution of mass in a masonry building and peak storey drift ratio). A building performance
discuss the use of finite element modeled with model is developed, which takes into account all
2D or 3D elements. However, the guidance is the nonstructural components in the building and
more targeted towards linear analysis; they rec- relates them to specific demand parameters (e.g.,
ommend using a ductility factor of 1.0 and an partitions in a floor are related to the storey drift
equivalent viscous damping value of 15 % to in that floor). Fragility functions and conse-
account for energy dissipation due to cracking, quence functions are provided for typical
sliding, and rocking, which do not seem to be nonstructural components present in US build-
appropriate for nonlinear analysis. Nevertheless, ings and incorporated in the software, PACT.
deformation limits are given as 1 % for elements A Monte Carlo analysis is carried out to sample
failing in a rocking mode and 0.5 % for all other each of these fragility functions to assess the
elements. Unique among the documents consid- expected damage and losses under the design
ered here, the Recommendations also give earthquake ground motion.
a rotation limit for spandrels of 0.5 % (based on FEMA P-58 evaluates economic and human
the clear span). losses as continuous variables, rather than
checking for compliance with discrete limit states
FEMA P-58: Seismic Performance Assessment (such as “Life Safety” and “Collapse Prevention”
of Buildings used in ASCE 41). It is left to the building owner,
The ATC-58 and ATC-58-1 projects developed engineer, or future code developer to define
the so-called next-generation performance-based acceptable performance for a given project. Eco-
seismic design guidelines following on from nomic losses can be compared with the costs of
“first-generation” projects such as ASCE 41. seismic retrofit or with other insurance or invest-
The outcome of these projects was FEMA P-58, ment costs.
published in 2012 (ATC 2012). This project takes Downtime calculations in FEMA P-58 are
one of the main original objectives of limited to estimates of repair time resulting
performance-based design – communicability of from earthquake damage, which does not account
seismic performance to non-engineers – to the for delays for required inspections, mobilization
next level by providing guidance on converting of engineering and contractors resources, financ-
results from analytic models into estimates of ing, permitting, and utility disruption. It also only
casualties, repair costs, and repair time. To focuses on the time to achieve full recovery of the
achieve this in a meaningful way, the framework facility, but does not consider intermediate goals
explicitly incorporates variability and uncertainty such as the time to reoccupy or return to func-
in both structural and nonstructural response. The tionality, which are of more direct concern to
probabilistic loss calculations are implemented in building owners. The downtime calculation was
a computer program, PACT (Performance therefore improved as part of the development of
Assessment Calculation Tool). the REDiTM Rating System (Resilience-based
FEMA P-58 considers the three basic assess- Earthquake Design Initiative) guidelines
ment types discussed in section “Performance (Almufti and Willford 2013), taking into account
Objectives and Assessment Methodology.” For input from contractors, cost estimators, bankers,
each of these assessment types, a similar building department officials, and researchers.
workflow is followed. Structural performance is The REDiTM Rating System uses this extended
assessed with NRHA (or other methods) and is FEMA P-58 loss assessment approach to demon-
quantified both in terms of collapse fragility strate that the REDiTM resilience objectives are
(probability distribution of the ground motion satisfied – primarily for new buildings, but also
level required to cause collapse) and probability applicable for existing structures.
200 Assessment of Existing Structures Using Response History Analysis
Gibson R, Wilcock TR, Grant DN, Gration DA (2012) design. NIST GCR 10-917-9. National Institute of
Performance based seismic design of an offshore plat- Standards and Technology, Gaithersburg
form in the Caspian Sea. Struct Eng 90:14–22 NEHRP Consultants Joint Venture (2011) Selecting and
Grant DN (2011) Response spectral matching of two hor- scaling earthquake ground motions for performing A
izontal ground motion components. Journal of Struc- response-history analyses. NIST GCR 11-917-15.
tural Engineering 137: 289–297 National Institute of Standards and Technology,
Grant DN, Padilla D, Greening PD (2011) Orientation Gaithersburg
dependence of earthquake ground motion and structural NEHRP Consultants Joint Venture (2012) Soil-structure
response. In: Dolšek M (ed) Protection of built environ- interaction for building structures. NIST GCR
ment against earthquakes. Springer, Dordrecht, 352 pp 12-917-21. National Institute of Standards and Tech-
Ibarra LF, Medina RA, Krawinkler H (2005) Hysteretic nology, Gaithersburg
models that incorporate strength and stiffness deterio- NZSEE (2006) Assessment and improvement of the
ration. Earthq Eng Struct Dyn 34:1489–1511 structural performance of buildings in earthquakes.
NEHRP Consultants Joint Venture (2010) Applicability of New Zealand Society for Earthquake Engineering,
nonlinear multiple-degree-of-freedom modeling for Wellington
B
Base-Isolated Systems,
Reliability-Based
Characterization of,
Fig. 1 Schematic
representation of a base-
isolated finite element
building model
Super-Structure
z
y Ground
Motion
x
Base
Platform
Isolator
Bearing
stochastic model for the excitation. The reliabil- three-dimensional linear elastic system, while
ity assessment and the structural response of the base is assumed to be rigid in plane and it is
base-isolated buildings are discussed in sections modeled using 3 degrees of freedom. Let xs(t) be
“Reliability Measures” and “System Response,” the nth dimensional vector of relative displace-
respectively. In section “Application Problem,” ments of the superstructure with respect to
the proposed characterization is illustrated the base and Ms, Cs, Ks be the corresponding
by means of a based-isolated finite element build- mass, damping, and stiffness matrices. Also, let
ing model. The entry closes with some final xb(t) be the vector of base displacements with
remarks. three components and Gs be the matrix of earth-
quake influence coefficients of dimension n 3,
that is, the matrix that couples the excitation
Structural Model components of the vector x€g ðtÞ to the degrees of
freedom of the superstructure. The equation of
Finite element building models with a relatively motion of the superstructure is expressed in the
large number of degrees of freedom are consid- form
ered for modeling the superstructure, i.e., the
structure above the isolation system. For illustra- Ms x€s ðtÞ þ Cs x_ s ðtÞ þ Ks xs ðtÞ ¼ Ms Gs x€b ðtÞ þ x€g ðtÞ
tion purposes a schematic representation of (1)
a base-isolated finite element building model
under ground motion is shown in Fig. 1. In where x€b ðtÞ is the vector of base accelerations
general, base-isolated buildings are designed relative to the ground. To complete the formula-
such that the superstructure remains elastic. tion of the combined model, the equation of
Hence, the superstructure is modeled as a motion for the base platform is written as
Base-Isolated Systems, Reliability-Based Characterization of 205
Base-Isolated Systems,
Reliability-Based
Characterization of,
Dr
Fig. 2 Schematic
representation of a rubber Di
bearing B
Steel Plate
Rubber Layer
(thickness: tr)
GTs Ms Gs þ Mb x€b ðtÞ þ x€g ðtÞ þ GTs Ms x€s ðtÞ matrix of linear elastic isolation components,
þ Cb x_ b ðtÞ þ Kb xb ðtÞ þ f is ðtÞ ¼ 0 and fis(t) is the vector containing the nonlinear
isolation elements forces.
(2)
Rewriting the previous equations, the
where Mb is the mass matrix of the rigid base, Cb combined equation of motion of the base-
is the resultant damping matrix of viscous isola- isolated structural system can be formulated in
tion components, Kb is the resultant stiffness the form
Ms Ms Gs x€s ðtÞ Cs O x_ s ðtÞ
þ þ
GTs Ms Mb þ GTs Ms Gs x€b ðtÞ O Cb x_ b ðtÞ
(3)
Ks O xs ð t Þ Ms G s 0
¼ €
x g ð t Þ
O Kb xb ð t Þ Mb þ GTs Ms Gs f is ðtÞ
The above formulation can be extended to other rubber being vulcanized to the steel plates. Rub-
cases as well, for example, the consideration of ber bearing systems are able, in principle, to
nonlinear models for the superstructure. support the superstructure vertically, to provide
the horizontal flexibility together with the restor-
ing force, and to supply the required hysteretic
Isolation Model damping. Figure 2 shows a schematic represen-
tation of a rubber bearing, where Dr represents
Rubber bearings are considered for modeling the the external diameter of the isolator, Di indicates
isolation system. Such devices have been used the internal diameter, and Hr = trnr is the total
over many years in a number of seismically iso- height of rubber in the device, where tr is the layer
lated buildings worldwide (Kelly 1986; Su thickness and nr is the number of rubber layers.
et al. 1990; Makris and Chang 1998). They An analytical model that simulates measured
require minimal initial cost and maintenance restoring forces under bidirectional loadings is
compared to other passive, semi-active, and used in the present formulation. The model is
active energy absorption devices. A rubber bear- based on a series of experimental tests conducted
ing consists of layers of rubber and steel, with the for real-size rubber bearings (Yamamoto
206 Base-Isolated Systems, Reliability-Based Characterization of
Base-Isolated Systems, Reliability-Based Characterization of, Fig. 3 Schematic representation of the decompo-
sition of the restoring force
et al. 2009; Minewaki et al. 2009). The loading of the base displacement vector xb(t); u_ ðtÞ is the
tests of seven full-scale isolators were carried out velocity vector; ^u_ ðtÞ and ^dðtÞ are the unit direc-
using the Caltrans Seismic Response Modifica- tional vectors of u_ ðtÞ and d(t), respectively; and
tion Device Test Facility at the University of || || indicates the Euclidean norm. The parame-
California, San Diego. The specimens used in ters a = 0.2Hr and b = 0.7 are positive con-
the tests were made with high damping rubber stants that relate to the yield displacement and
compounds. In particular, horizontal bidirec- smoothness of yielding, respectively. Once the
tional loading tests for isolators with a diameter vector d(t) has been derived, the restoring force
of 0.7 m and 1.3 m were conducted. On the basis f(t) on the isolator (in the x and y direction)
of the test results, the model assumes that the is expressed in terms of the unit directional
restoring force on the rubber bearing is composed vector ^uðtÞ and the vector d(t) as
of a force directed to the origin of the isolator and f ðtÞ ¼ ^uðtÞ f e ðtÞ dðtÞf s ðtÞ , where fe(t) is the
another force approximately opposite to the nonlinear elastic component and fs(t) is the
direction of the movement of the isolator. Such elastoplastic component. The previous decom-
decomposition of the restoring force is shown position of the restoring force to elastic and
schematically in Fig. 3. dissipative components simplifies the restoring
According to the model, the direction of the force characteristics and results in a simple and
movement of the isolator d(t) is defined in terms accurate model. These two forces have different
of the components in the x and y direction of the appearance mechanisms and different influenc-
base displacement vector xb(t) by means of the ing factors. For example, the dependency of
nonlinear differential equation these forces on factors such as temperature, ver-
tical pressure, and maximum experienced shear
1 h i
strain would differ. Therefore, the previous
d_ ðtÞ ¼ k u_ ðtÞ k ^u_ ðtÞ k dðtÞkb ^dðtÞ ,
a (4) decomposition has the potential for more
uð0Þ ¼ 0, dð0Þ ¼ 0 general modeling of other dependencies. Based
on experimental results, the components of the
where the components of the vector u(t) corre- restoring force on the rubber bearing can be
spond to the components in the x and y direction approximated as
(
0:35g
ðtÞA if 0 gðtÞ 1:8
f e ðt Þ ¼ (5)
0:35gðtÞ þ 0:2ðgðtÞ 1:8Þ2 A if gðtÞ 1:8
Base-Isolated Systems, Reliability-Based Characterization of 207
a2
t
ea3 ðtn Þ
t
hðt, M, r Þ ¼ a1 (7)
tn
where
e a 2 0:2lnð0:05Þ a2
a1 ¼ , a2 ¼ , a3 ¼
0:2 1 þ 0:2ðlnð0:2Þ 1Þ 0:2
(8)
Base-Isolated Systems, Reliability-Based Character-
ization of, Fig. 4 Comparison of analytical and experi-
mental hysteresis loops for a medium-size rubber bearing. On the other hand, the total spectrum of the
Dr = 0.8 m, Hr = 0.16 m, Di = 0.15 m motion at a site S( f, M, r) is expressed as the
208 Base-Isolated Systems, Reliability-Based Characterization of
product of the contribution from the earthquake Base-Isolated Systems, Reliability-Based Character-
source E( f, M), path P( f, r), site G( f), and type of ization of, Table 1 Parameters for the stochastic ground
acceleration model
motion I( f), i.e., S( f, M, r) = E( f, M) P( f, r)
G( f ) I( f ). Numerical Numerical
Parameter value Parameter value
The source component is given by
rs (gm/cc) 2.8 bs (km/s) 3.5
E( f, M) = C M0(M) Sa( f, M), where C is pffiffiffi
V 1= 2 RF 0.55
a constant, M0(M) = 101.5M + 10.7 is the seismic
F 2.0 R0 (km) 1.0
moment, and the factor Sa is the displacement
source spectrum given by
Base-Isolated Systems, Reliability-Based Characterization of, Fig. 5 Envelope function, ground motion spec-
trum, and a sample ground motion
ð
seismological parameters M and r; and Ih is a set
PF ð yÞ ¼ pðyÞd ðyÞ (11)
of indices. The controllable system parameters gðy, yÞ0
are taken as the external diameter and the total
height of rubber of the isolators, that is, yT = where p(y) is the multidimensional probability
< Dr, Hr >. The probability that performance density function that characterizes the parameters
conditions are satisfied within a particular refer- involved in the excitation model. For systems
ence period T is used as a reliability measure. under stochastic excitation, the probability inte-
In this context a failure event F is defined gral involves in general a large number of uncer-
in terms of a performance function g(y,y) as tain parameters. Thus, the reliability estimation
F(y,y) = g(y,y) 0, where constitutes a high-dimensional problem. This
problem is solved by applying an advanced sim-
jhi ðt, y, yÞj ulation technique called subset simulation
gðy, yÞ ¼ 1 maxi Ih maxt ½0, T
hi (Au and Beck 2001). In the approach, the failure
(10) probabilities are expressed as a product of condi-
tional probabilities of some chosen intermediate
and h*i , i In are the corresponding acceptable failure events, the evaluation of which only
response levels. In particular, the probability of requires simulation of more frequent events.
failure events associated with the base drift and Therefore, a rare event simulation problem is
superstructure absolute acceleration are consid- converted into a sequence of more frequent
ered in the present formulation. The probability event simulation problems. Validation calcula-
of failure PF(y) evaluated at some particular tions have shown that subset simulation can be
value of the isolation system parameters y can applied efficiently to first excursion response
be expressed in terms of the probability integral problems for a wide range of dynamical systems.
210 Base-Isolated Systems, Reliability-Based Characterization of
Details of this simulation procedure from the a proper integration scheme such as the
theoretical and numerical viewpoint can be Crank-Nicolson method (Burden and Faires
found in Au and Beck (2001), Ching 2011).
et al. (2005), and Zuev et al. (2011). (v) The iteration starting with solving Eq. 3
needs to be repeated until the variation of
the quantities dj(t + Dt), j = 1, . . ., nI is
sufficiently small between two consecutive
System Response iterations (Jensen and Sepulveda 2011).
Base-Isolated Systems,
Reliability-Based
Characterization of,
Fig. 6 Building model
with base isolation system
B
10−2 10−2
10−3 10−3
10−4 10−4
PFailure
PFailure
10−5 10−5
10−6 10−6
10−7 10−7
10−8 10−8
0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.15 0.16 0.17 0.18 0.19 0.20 0.21 0.22 0.23 0.24 0.25
DR [m] HR [m]
Base-Isolated Systems, Reliability-Based Characterization of, Fig. 7 Probability of the failure event related to the
base drift as a function of the isolation system parameters
10−2 10−2
−3
10 10−3
10−4 10−4
PFailure
PFailure
10−5 10−5
10−6 10−6
−7
10 10−7
10−8 10−8
0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 0.15 0.16 0.17 0.18 0.19 0.20 0.21 0.22 0.23 0.24 0.25
DR [m] HR [m]
Base-Isolated Systems, Reliability-Based Characterization of, Fig. 8 Probability of the failure event related to the
superstructure absolute acceleration as a function of the isolation system parameters
increases. In this case the isolation system characteristics on the reliability of such systems.
becomes more flexible. The results of the example problem indicate that
The effects of the isolation system parameters the minimization of the superstructure absolute
on the superstructure absolute acceleration are acceleration and the minimization of the base
shown in Fig. 8. In this figure the probability of displacement are conflicting goals. For example,
failure in terms of the absolute acceleration of the provision of additional stiffness to the isola-
the superstructure is presented. The probability tion system decreases the superstructure reliabil-
of failure increases as the rubber diameter ity. Contrarily, the flexibility of the isolators has
increases. In contrast, as the height of the rubber a positive impact on the reliability of the super-
increases, the probability of failure is decreased. structure. However, this flexibility may induce
Clearly these effects are opposite to the effects of undesirable effects on the base isolation system.
the isolation system parameters on the base Therefore, the performance of the superstructure
response. For example, the introduction of addi- as well as the response of the base isolation sys-
tional stiffness in the isolation system makes the tem should be considered simultaneously for
structural response ineffective. In other words, a proper characterization of base-isolated sys-
this additional stiffness which controls the base tems. It is noted that additional considerations
displacement makes the superstructure more vul- such as the evaluation of the superstructure reli-
nerable. This in turn affects the primary objec- ability in terms of interstory drifts can also be
tive of the isolation system, namely, the taken into account. This could lead to important
reduction of the superstructure response. The results and conclusions. Similarly, the explicit
previous results and observations show that consideration of near-field ground motions may
the protection of the superstructure and the have important implications for flexible struc-
control of the base displacement are conflicting tures such as base-isolated systems. For these
objectives. systems near-field ground motions may lead to
excessive base deformations and superstructure
deformations with important implications for the
Summary reliability of the combined structural system
(isolation system and superstructure). In sum-
A reliability-based characterization of base- mary, the proposed reliability-based characteri-
isolated systems under stochastic excitation has zation of base-isolated systems under stochastic
been proposed. The characterization allows to excitation provides a valuable basic tool for the
evaluate the effect of important design isolator analyst and designer.
Base-Isolated Systems, Reliability-Based Characterization of 213
This led to the time-domain formulation in Yuen where it is justified to assume that the modal
and Katafygiotis (2001a), a spectral density for- (not physical) force has a constant power spectral
mulation in Yuen and Katafygiotis (2001b) and density in the vicinity of the resonance band
later an FFT formulation in Yuen and dominated by the mode.
Katafygiotis (2003). As mentioned before, the Following a Bayesian approach in the fre- B
time-domain formulation is not preferred for its quency domain, use is made of the (scaled) FFT
assumption on the prediction error and modal of the measured acceleration response at a
force. The spectral density formulation involves number of dofs on the structure, denoted by
n o
statistical averaging and may not make full use of ^x€j Rn : j ¼ 0, . . . , N 1 and abbreviated as
the information contained in the data. The Bayes- n o
ian formulation based on FFT is therefore the ^x€j , where n is the number of measured dofs.
n o
preferred choice. The above formulations are The (scaled) FFT of ^x€j is the complex-valued
only computationally feasible for a small number sequence fFk Cn : k ¼ 0, . . . , N 1g where
of measured degrees of freedom (dofs). This
partly explains why the formulations did not rffiffiffiffiffiffiffi N 1
2Dt X ^
flourish and there were no applications for a Fk ¼ x€j expð2pijk=N Þ
N j¼0 (1)
number of years after the formulations first
appeared. Based on the FFT formulation, fast k ¼ 0, . . . , N 1
but equivalent formulations that allow
mathematical analysis and feasible implementa- and i2 ¼ 1 and Dt is the sampling interval. For a
tion in general were developed in Au (2011) (see given k, the FFT Fk corresponds to frequency
also Zhang and Au 2013) for well-separated
modes, Au (2011, 2012a) for multiple (possibly
f k ¼ k=NDt k ¼ 0, . . . , int½N=2 (2)
close) modes, and Au and Zhang (2012a) and
Zhang et al. (2015) for well-separated modes
but multiple setups. Application to field or exper- where int[N/2] is the integer part of N/2 and is the
imental data can be found in Au et al. (2012a, b), index corresponding to the Nyquist frequency.
Au and To (2012), Au and Zhang (2012b), and The FFT Fk has been scaled so that
Fan et al. (2012). See also Yuen and Kuok Fk Fk Cnn gives the one-sided sample spectral
(2010a, b) for application of the Bayesian spec- density (periodogram) matrix. Summing the i-th
tral density approach. Asymptotic uncertainty diagonal element of Fk Fk over k ¼ 0, . . . , N 1
laws that provide insights on the achievable and multiplying by the frequency resolution
limits operational modal analysis have been Df ¼ 1=NDt gives twice the mean square value
developed in Au (2014a, b). Recent overviews of the i-th measured dof (Parseval equality).
can be found in Yuen and Kuok (2011) and Au For the purpose of identifying the modes of
et al. (2013). interest, it is sufficient and preferable to make use
of the FFT data only on a selected frequency band
around the mode(s). This is because in reality, the
Context data contains a variety of dynamic (colored)
activities over its sampling spectrum (up to the
The primary objective of ambient modal identifi- Nyquist frequency), most of which are irrelevant
cation is to identify the natural frequencies, to identifying the mode(s) or are difficult to
damping ratios, and mode shapes of a constructed model.
structure using (output) vibration response mea- Suppose there are m modes within the selected
surement at a number of measured degrees frequency band and let fFk g denote the collection
of freedom (dofs) under “ambient conditions.” of FFT data within the band. It is modeled as a
The latter is a notion relative to the mode of sum of contributions from the structural vibration
interest. It refers specifically to the situation (the target to be measured) and a prediction error
216 Bayesian Operational Modal Analysis
(noise) that accounts for the difference between u, denoted by pðujfFk gÞ. Using Bayes’ theorem,
the theoretical and measured response: it is given by
X
m
Fk ¼ Fi €ik þ «k (3) pðujfFk gÞ ¼ pðfFk gÞ1 pðfFk gjuÞpðuÞ (6)
i¼1
where Fi Rn (i ¼ 1, . . . , m) is the mode shape of The first term on the RHS does not depend on u.
the i-th mode in the selected frequency band; As far as the identification of u is concerned, it
€ik C is the FFT at frequency fk of the i-th can be ignored. The middle-term pðfFk gjuÞ is
modal acceleration response; and «k Cn is the called the “likelihood function.” It is derived
FFT of the prediction error, which may arise due based on modeling assumptions that relate the
to measure (channel) noise and modeling error response of the theoretical model for a given u
(e.g., unaccounted dynamics). Assuming classi- to the FFT data. The middle-term p(u) is the
cally damped modes, the time-domain counter- “prior distribution” that reflects one’s knowledge
part of ik satisfies the uncoupled equation of about u in the absence of data. In modal identifi-
motion: cation problems with sufficient data, it can be
taken as a constant because it is slowly varying
€i ðtÞ þ 2zi oi _ i ðtÞ þ o2i i ðtÞ ¼ pi ðtÞ (4) compared to the likelihood function. As a result,
it is sufficient for modal identification problems
where oi ¼ 2pfi ; fi, zi, and pi(t) are, respectively, to use
the natural frequency (in Hz), damping ratio, and
modal force. pðujfFk gÞ / pðfFk gjuÞ (7)
In the above context, the set of modal param-
eters to be identified from the FFT data fFk g , Deriving the likelihood function requires deriv-
denoted by u, consists of the parameters: ing the joint PDF of fFk g for a given u. This PDF
is generally complicated, but it turns out to admit
(1) Natural frequencies f1 , . . . , fm Rþ an asymptotic closed form when the number of
(2) Damping ratios z1 , . . . , zm ð0, 1Þ data points N is large (Brillinger 1981). In partic-
(3) Mode shapes F1 , . . . , Fm Rn ular, as N ! 1, fFk g at different k’s are asymp-
(4) PSD matrix of modal forces S Cnn totically independent. For a given k, the real and
(assumed constant within the band) imaginary parts of Fk are jointly Gaussian with
(5) PSD of prediction error Se Rþ (assumed zero mean. Thus,
constant within the band) 1=2
pðujfFk gÞ / pðfFk gjuÞ ¼ ð2pÞnNf P detCk
Each mode shape Fi ¼ ½F1i , F2i , . . . , Fni T is (
k
)
1 X ReFk
T
subjected to a unit norm constraint, i.e., 1 ReFk
exp Ck
2 k ImFk ImFk
X
n
jjFi jj2 ¼ FTi Fi ¼ F2ji ¼ 1 i ¼ 1, . . . , m (8)
j¼1
" #
1 FðReHk ÞFT FðImHk ÞFT Se in the presence of the data. Mathematically, the
Ck ¼ þ I2n posterior covariance matrix is equal to the inverse
2 FðImHk ÞFT FðReHk ÞF T 2
of the Hessian of the NLLF at the MPV.
(9) The above results can be reasoned as follows.
where F ¼ ½F1 , . . . , Fm Rnm is the mode
Given the data, let the MPV of the modal param- B
eters be denoted by the vector u.^ Approximating
shape matrix and Hk Cmm is the modal transfer
the NLLF by a second-order Taylor expansion
matrix whose (i, j) entry is given by ^
about u,
Hk ði, jÞ ¼ Sij hik hjk (10)
LðuÞ L u ^ þ ∇L u ^ uu ^
hik ¼
b2ik 1 þ ð2zi bik Þi
1 1
T
^ ∇2 L u
^ uu ^
(11) þ uu (15)
2
bik ¼ f i =f k (12) where ∇L u ^ and ∇2 L u ^ denote, respectively,
the gradient (row) vector and Hessian matrix of
and Sij is the (i, j) entry of the PSD matrix of ^ Since the NLLF is minimized at u,
the NLLF at u. ^
modal force, S Cmm . Note that Hk and S are
∇L u ^ ¼ 0 (a zero vector) and ∇2 L u ^ are
Hermitian (i.e., equal to its conjugate transpose);
Ck is real symmetric. positive definite matrix. Applying this on Eq. 15
and then substituting into Eq. 14 gives
Posterior Most Probable Value and
T
Covariance Matrix ^ 1 uu
pðujfFk gÞ / exp L u ^ ∇2 L u ^ uu
^
2
For convenience in analysis or computation, the 1 T
/ exp u u ^ ∇2 L u ^ uu ^
posterior PDF is written in terms of the NLLF 2
1 T 1
(negative log-likelihood function): / exp u u ^ C ^ uu ^
2
T
1X 1 X ReFk 1 ReFk
(16)
LðuÞ ¼ lndetCk þ Ck
2 k 2 k ImFk ImFk where
(13)
1
^ ¼ ∇2 L u
C ^ (17)
so that
is a positive definite matrix. Eq. 16 is just the
pðujfFk gÞ / expðLðuÞÞ (14) variable part of a joint Gaussian PDF with zero
mean and covariance matrix equal to C. ^ The
Note that the constant nNf ln(2p) has been omit- proportionality constant of the equation is
ted from the NLLF.
1=2
^
ð2pÞny =2 det C (ny is the number of
With sufficient data often encountered in prac-
tice, the posterior PDF has a single peak in the modal parameters), which can be obtained by
parameter space of u, which is the most probable integration or by probability reasoning. Thus,
value (MPV) of the modal parameters given the
1=2
data. Equivalently, the NLLF has a unique mini- ^
pðujfFk gÞ ¼ ð2pÞny =2 det C
mum at the MPV. Approximating the NLLF by a
1
T 1
(18)
second-order Taylor expansion about the MPV ^ ^ ^
exp u u C uu
leads to a Gaussian approximation of the posterior 2
PDF. The resulting Gaussian distribution has a
mean at the MPV, and its covariance matrix, called Computational Problems
the “posterior covariance matrix,” reflects the The MPV of modal parameters cannot be
remaining uncertainties of the modal parameters obtained analytically because the relationship
218 Bayesian Operational Modal Analysis
X
n E0k ¼ aHk aT þ Se Im0 (29)
jjFi jj ¼
2
FTi Fi ¼ F2ji ¼ 1 i ¼ 1, . . . , m
j¼1
is an m0 -by- m0 Hermitian matrix. The signifi-
(25) cance of Eq. 28 is that the mode shape basis B has
been segregated out in the last term, which is
In this case the most probable mode shape cannot
quadratic in nature. The most probable basis
be obtained directly by solving a standard eigen-
should therefore minimize the quadratic form
value problem. The mathematical structure of the
under orthonormal constraints. This does not
optimization problem is more complex because
lead to a standard eigenvalue problem, but pro-
the mode shapes are not necessarily orthogonal to
cedures have been developed that allow the most
each other. However, it is possible to reduce the
probable basis to be determined efficiently by
complexity by representing the mode shape via a
Newton iteration. This resolves the potentially
set of orthonormal basis and noting that the
high-dimensional optimization with respect to
dimension of the subspace spanned by such
the mode shape. Based on this, a strategy has
basis does not exceed the number of modes. The
been developed for determining the MPV of dif-
mode shape matrix F ¼ ½F1 , . . . , Fm Rnm is
ferent groups of parameters, iterating until con-
represented as
vergence (Au 2012a).
F ¼ B0 a (26)
Modal Signal-to-Noise Ratio
where
The term “signal-to-noise ratio” (s/n ratio) is
0 often used to describe the quality of data to indi-
B0 ¼ B01 , . . . , B0m0 Rnm (27)
cate the contribution of the signal targeted to be
measured compared to the contribution of
contains in its columns a set of orthonormal “noise,” which inevitably arises from sensor,
“mode shape basis” B0i Rn : i ¼ 1, . . . , m0 cables transmitting the signal (especially for ana-
spanning the “mode shape subspace;” log signals), digitizing hardware, etc. The actual
0
a Rm m contains in its columns the coordinates quantitative definition differs depending on the
of each mode shape with respect to the mode particular application, however. In ambient
shape basis; m0 minðn, mÞ is the dimension of modal identification, the s/n ratio that affects the
the mode shape subspace, which can be deter- identification results is related to the spectral
mined with the help of a singular-value spectrum. contribution of the modal response compared to
The MPVs of B0 and a need to be determined in the spectral contribution of the noise in the vicin-
the identification process. ity of the resonance frequency band. In the case of
Based on Eq. 26, it can be shown that the well-separated modes, a definition of the s/n ratio
NLLF can be rewritten as that is found to govern fundamentally the
220 Bayesian Operational Modal Analysis
characteristics of the identification results is the modal s/n ratio in Eq. 30 always increases with
PSD of the theoretical acceleration response at the number of measured dofs.
resonance (SDk ¼ S=4z2 ) divided by the PSD of
the prediction error (Se): High s/n Asymptotics
It turns out that the Bayesian modal identification
S results in terms of the MPV and posterior covari-
g¼ (30)
4Se z2 ance matrix reduce to intuitive forms when the
modal s/n ratio is large (theoretically infinity).
Governing Factors This is referred to as “high s/n asymptotics.”
The modal s/n ratio in Eq. 30 depends on the Asymptotically for large modal s/n ratio, the
following factors: terms in the NLLF can be separated into two
terms: one quadratic term depending only on the
(1) The prediction error PSD Se. This includes mode shape (and the prediction error PSD in a
contributions from the channel noise as well multiplicative manner) and the other term
as modeling errors, e.g., due to unaccounted depending only on the remaining parameters
modes and colored nature of modal force. (natural frequency, damping ratio, modal force
(2) The damping ratio z. The smaller the damping, PSD, prediction error PSD). The mode shapes are
the higher the modal s/n ratio, as a result of therefore asymptotically uncoupled from the
higher modal response PSD at resonance. remaining parameters. This has important impli-
(3) The modal force PSD S. This in turn depends cations on the behavior of the MPV and the pos-
on the intensity of environmental excitation terior covariance matrix. The former provides a
and (less trivially) on the measured dofs. sound basis for setting initial guess and designing
Specifically, let j denote the “full” mode fast algorithms for determining the MPV of modal
shape of the structure containing all parameters. The latter leads to remarkably simple
(possibly an infinite number of) dofs, and it closed-form formulas for the posterior variance of
is scaled to have unit norm, i.e., jT j ¼ 1 . the modal parameters. These are collectively
Then it can be reasoned that known as “uncertainty laws,” which govern the
! achievable limits of operational modal analysis.
X
S ¼ Sp xi2
(31) The asymptotic behavior of the most probable
i mode shape is discussed in sections Single Mode
and Multiple Modes for a single mode and mul-
where the sum is over the measured dofs, and
tiple modes in the selected frequency band,
respectively. The posterior covariance matrix is
jT SF j discussed in the context of uncertainty laws in
Sp ¼ 2 (32)
jT Mj section Uncertainty Laws.
is the modal force PSD consistent with the scaling Single Mode
of the global mode shape j; M is the mass matrix For a single mode in the selected band, asymp-
of the structure. Note that in reality, the full mode totically the most probable mode shape F is
shape j and the mass matrix M can hardly be equal to the eigenvector with the largest eigen-
identified and so is Sp. Instead, the measured value of the following matrix:
mode shape F and the modal force PSD
X
S consistent with the scaling FT F ¼ 1 is identi- A0 ¼ ReFk ReFTk þ ImFk ImFTk (33)
fied. Eq. 31 shows that the modal force PSD k
S always increases with the number of measured
dofs, although the rate depends on the mode This matrix depends only on the data, and so the
shape of the dof incrementally added to the most probable mode shape can be calculated
existing set of measured dofs. As a result, the explicitly without optimization.
Bayesian Operational Modal Analysis 221
The above result can be reasoned as follows. Minimizing the NLLF is equivalent to maximiz-
Asymptotically the matrix A in the NLLF Eq. 21 ing the sum on the rightmost. Since it contains no
becomes A0 in Eq. 33, and so the NLLF becomes cross terms between Bi0 and Bj0 (i 6¼ j), maximiz-
ing it gives the most probable basis as the
LðuÞ
nN f ln2 þ ðn 1ÞN f lnSe eigenvectors of A0 corresponding to the first m0 B
X
þ lnðSDk þ Se Þ þ S1 largest eigenvalues.
e d
k
S1
e F A0 F
T
(34)
Quantification of Identification
In the above expression, only the last term Uncertainty
depends on F, which is a quadratic form with a
constant coefficient matrix A0. Minimizing it with Mode Shape Uncertainty
respect to F and subjected to the norm constraint For scalar-valued modal parameters such as the
natural frequency and damping ratio, the poste-
FT F ¼ 1 gives the most probable mode shape as
rior uncertainty can be conveniently described in
the eigenvector of A0 with the largest eigenvalue.
terms of the posterior standard deviation or in a
nondimensional manner the posterior coefficient
Multiple Modes
of variation (c.o.v.), equal to the posterior stan-
When there is more than one mode in the selected
dard deviation divided by the MPV. The posterior
band, asymptotically the most probable mode shape
standard deviation is equal to the square root of
basis is given by the eigenvectors of A0 in Eq. 33
the posterior variance, which can be obtained
corresponding to the first m0 largest eigenvalues.
from the corresponding diagonal entry of the
Note that this does not give directly the most prob-
posterior covariance matrix.
able mode shapes because the MPV of the coordi-
The above quantification cannot be applied to
nates with respect to the basis, i.e., a in Eq. 26, is not
the mode shape because it is vector valued and is
known but its determination is still nontrivial.
subject to the norm constraint. One way to quan-
The above can be reasoned as follows.
tify the uncertainty of the mode shape is through
Asymptotically, the term Im0 Se E0k in the
the Expected modal assurance criterion (MAC),
quadratic term of the NLLF in Eq. 28 becomes
defined analogously as in the deterministic case.
the identity matrix Im0 and so
According to the posterior distribution, which
is approximated by a Gaussian distribution,
LðuÞ
nN f ln2 þ ðn m0 ÞN f lnSe
X given the measured data, the mode shape is a
þ S1
e dþ lnjdetE0k j Gaussian vector with mean vector equal to the
k most probable mode shape and a covariance
X
S1 Fk B0 B0 Fk
T
(35) matrix CF Rnn . The latter is given by
e
k the corresponding n n partition of the full
covariance matrix (containing all modal parame-
This expression depends on the mode shape ters). The Expected MAC is defined as the
basis B0 only through the sum in the last term, expected value of the cosine of the hyper angle
which is quadratic in nature. Substituting between the random mode shape vector and the
B0 ¼ B01 , . . . , B0m0 , the sum can be rewritten as most probable mode shape vector. This value
ranges between 0 and 1; the higher the value,
!
X X
m0 X the smaller the mode shape uncertainty. An
Fk B0 B0 Fk B0i Fk Fk B0i
T T
¼ exact closed-form expression for the Expected
k i¼1 k MAC has not been developed, but an asymptotic
X
m 0
expression is available, which is adequate for
B0i A0 B0i
T
¼ (36) practical purposes. It can be shown that
i¼1
asymptotically
222 Bayesian Operational Modal Analysis
The Bayesian and frequentist measure of iden- the data duration, and T is the natural period
tification uncertainty has different roles. The of the target mode. Note that N f ! 1 and
Bayesian measure reflects fundamentally the z ! 0 imply that N c ! 1 as well.
remaining uncertainty associated with the param- (3) The selected frequency band is assumed to be
eters for a given identification model and a given f ð1 kzÞ, where k is called the “bandwidth B
set of data. It does not reflect what the identifica- factor.” The bandwidth factor is a trade-off
tion results would be if the experiment is between identification information (the
repeated, because modeling error can exist and larger, the better) and modeling error risk
conditions of the next experiment can change. It (the smaller, the better). Typically it ranges
does not describe ensemble variability. On the between 3 and 10.
other hand, the frequentist measure is an aggre- (4) The FFT data fFk g in the selected band is
gate effect of identification uncertainty, modeling assumed to indeed result from a process
error, and system changes over the repeated obeying the identification model, i.e., no
experiments. modeling error. Note that prediction error is
still present because it is part of the identifi-
cation model.
Uncertainty Laws
The selected frequency bandwidth is assumed
Under the Bayesian framework, the posterior to be proportional to the damping ratio because
uncertainty of the modal parameters can be the width of the resonance band around the natu-
calculated in terms of the posterior covariance ral frequency does, e.g., the half power band is
for a given set of measured data. This process f ð1 zÞ, corresponding to k ¼ 1. An asymptoti-
is computational and implicit in nature. It cally small damping ratio implies that the modal
does not yield much insight about how the s/n ratio g ¼ S=4Se z2 in Eq. 30 is asymptotically
identification uncertainty depends on test high. It is therefore implied in the uncertainty
configurations. laws that the modal s/n ratio is high.
In the case of a single mode in the selected
band, it has been possible to derive closed-form Main Results
analytical expressions for the posterior covari- The asymptotic expressions of the posterior coef-
ance matrix under asymptotic conditions, ficient of variation of the modal parameters are
namely, small damping and long data duration, summarized in Table 1. In terms of correlation,
which are typically met in applications. The except for that between the damping ratio z and
expressions are collectively referred as “uncer- the modal force PSD S, the correlation between
tainty laws” (Au 2014a, b). They provide insights any pair of the modal parameters among {f, z, S,
on the scientific nature of the ambient modal Se, F} is asymptotically small, at most O(z).
identification problem and its fundamental limits The correlation between z and S is O k1=2 .
on identification uncertainty. The uncertainty
laws can also be used for drafting specifications Implications
for ambient vibration tests. The following implications can be drawn from
the uncertainty laws:
Context
The uncertainty laws have been derived under the (1) For small damping ratio, the posterior uncer-
following context: tainties of the natural frequency and mode
shape are significantly smaller than those of
(1) The damping ratio is small, z ! 0. the damping ratio and the modal force PSD.
(2) The spectral information for identifying the The former are O(z1/2) and the latter are
mode is large, in the sense that
O z1=2 .
N f ¼ 2kzN c ! 1 where N c ¼ T d =T , Td is
224 Bayesian Operational Modal Analysis
1
Modal force PSD S Squared posterior c.o.v. 2
BS ðkÞ ¼ 1 k2 ð tan 1 kÞ tan 1 k þ k2kþ1
d2S ¼ Nf B1S ðkÞ
Prediction error PSD Se Squared posterior c.o.v. –
d2Se ¼ ðn11 ÞNf
Mode shape F Covariance matrix BF ðkÞ ¼ tan 1 k
CF ¼ Nc BnzF ðkÞ In FFT
Expected hyper angle (squared)
d2F ¼ Nðcn1 Þnz
tan 1 k
(2) The damping ratio has the greatest uncer- with probability logic and modeling assumptions.
tainty in terms of posterior c.o.v. and is there- This is especially relevant for ambient modal
fore expected to govern the accuracy identification where the input excitation is not
requirement in ambient vibration tests. measured and where significant uncertainty can
(3) The mode shape is the only modal property exist in the identified modal parameters even in
whose leading order uncertainty is affected the presence of data. A frequency domain
by the modal s/n ratio. Its posterior uncer- approach is preferred over a time-domain
tainty is typically small. This implies that in approach because it significantly relaxes the
reality the quality of the mode shape is likely broadband assumption on the modal force and
to be governed by other sources such as sen- prediction error. By making inference using the
sor alignment error. FFT on a selected frequency band around the
(4) When the modal s/n ratio is high, reducing mode of interest, the modal force and prediction
the channel noise level (e.g., with better error only need to have a flat spectrum within the
equipment) or increasing the measured dofs band. The likelihood function can be formulated
has little or no effect on reducing the poste- from first principles using random-vibration the-
rior c.o.v.s of modal parameters. It is because ory and signal processing asymptotics. The
these factors influence only the modal s/n resulting form, however, does not allow the pos-
ratio, but the latter has no role in the terior statistics of the modal parameters to be
uncertainty laws. computed easily because its dependence on the
(5) Assuming a damping ratio of 1 % and a band- modal parameters is complicated and there are
width factor of 6, the minimum data duration typically a moderate to large number of parame-
to achieve a posterior of c.o.v. of 30 % in the ters. Using linear algebra techniques, fast equiv-
damping ratio is 300 natural periods. alent formulations have been developed that
allow efficient algorithms to be developed. The
Bayesian identification results are governed char-
Summary acteristically by the modal signal-to-noise ratio.
The posterior covariance matrix also admits
A Bayesian approach for modal identification remarkably simple closed form when the s/n
provides a fundamental means for processing ratio is high, leading to a set of uncertainty laws
the information contained in the data to make that govern the limit of ambient modal
inference on the modal parameters consistent identification.
Bayesian Statistics: Applications to Earthquake Engineering 225
of the authors and do not necessarily reflect the multinomial distribution with parameters
views of the National Science Foundation. Y ¼ yj , j ¼ 1, . . . , nc , a random vector whose
coordinates yJ capture earthquake
Xnc probabilities in
each cell j, such that y ¼ 1 . We call
Xnc j¼1 j
Bayesian Framework in Earthquake NZ ¼ Z the total number of newly B
j¼1 j
Engineering recorded earthquakes. The new records are used
in the Bayesian framework to update earthquake
Seismic Hazard Analysis probabilities pj.
Numerical example. For illustration, a hypo-
Earthquake Probabilities thetical example is considered. It is assumed that
A seismic activity matrix is defined as a at a site only three earthquakes j = 1, 2, 3 are
two-dimensional histogram which gives the likely to occur with known probabilities
probability of occurrence pJ of earthquakes char- p1 = 0.2, p2 = 0.3, p3 = 0.5. The seismic activ-
acterized by moment magnitude m and source-to- ity matrix, shown in Fig. 3a, has nc = 3 cells. If
site distance r, denoted by (m, r)J. Each point in an earthquake of type j = 1 occurs at that site
this matrix with coordinates (m, r)J is referred to and it is used for the update of the earthquake
as a cell j = 1,. . .,nX
C, where nc is the total number probabilities p1, p2, p3, then NZ = 1 and
nc
of cells. Note that p ¼ 1. Seismic activity
j¼1 j z ¼ ½ z1 z2 z3 ¼ ½ 1 0 0 , which is a
matrices can be calculated for each zip code in sample of Z.
the United States by using the earthquake proba- The unknown parameters fy1 , . . . , ync g are
bility tool available on the US Geological Survey assumed to have a Dirichlet distribution with
website (USGS 2009). Figure 1 shows the seis- parameters fa1 , . . . , anc g, that is,
mic activity matrix for Los Angeles.
A Bayesian approach is used to update earth- nc
a 1
quake probabilities in light of new data. For this, f y1 , . . . , yn c / ∏ yj j , 8yj 0: (2)
j¼1
we seek a probabilistic parametric model with
parameters, the earthquake probabilities. Let ZJ Moreover, we impose that
be a random variable which counts the number
aj
of new earthquakes recorded in each cell j. yj ¼ ¼ pj ; (3)
The vector Z ¼ Zj , j ¼ 1, . . . , nc has a a0
Bayesian Statistics:
Applications to
Earthquake Engineering,
Fig. 1 Seismic activity
matrix for Los Angeles 0.01
0.005
pj
0
220
180
140
5.5
100 6.0
r [km] 6.5
60 7.0 m
20 8.5
9
228 Bayesian Statistics: Applications to Earthquake Engineering
a b
p(θ1 , θ2 , 1 − θ1 − θ2 |z)
10 10
f(θ1, θ2, 1−θ1−θ2)
5 0 5 0
0.5 0.5
0 0
1 1
θ2
0.5 θ2 0.5
1 1
0 θ1 0
θ1
Bayesian Statistics: Applications to Earthquake Engineering, Fig. 2 (a) Prior density f ðy1 , y2 , 1 y1 y2 Þ and
(b) posterior density pðy1 , y2 , 1 y1 y2 jzÞ
a 0.5 b 0.5
0.4 0.4
B
0.3 0.3
E[qj|z]
pj
0.2 0.2
0.1 0.1
0 0
1 2 3 1 2 3
Cell j Cell j
Bayesian Statistics: Applications to Earthquake Engineering, Fig. 3 Two-dimensional seismic activity matrix:
(a) original earthquake probabilities pj, j = 1, 2, 3, and (b) updated earthquake probabilities yj jz , j = 1, 2, 3
a b
x 10−6 x 10−7
1.6 14 ϕ1,j (n)
ϕ2,j (n)
1.4 12 ϕ3,j (n)
1.2 10
1 8
gj(n|Θ)
6
0.8
4
0.6
2
0.4
0
0.2
−2
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
n [rad/s] n [rad/s]
Bayesian Statistics: Applications to Earthquake Engineering, Fig. 4 (a) Power spectral density gj(n) and (b)
linear components ’k,j, k = 1, 2, 3, for (m, r)J = (8,200 km)
algorithmic form. For the Bayesian approach, a functions which depend on (m, r)J. In the Bayes-
parametric model for gj(n) is sought, and it is ian context, the vector Y ¼ f yk k ¼ 1, 2, 3 g is
written as seen as a random vector with mean ½Y ¼ ½111.
The range C of Y is defined such that it ensures
the existence of the spectral density, i.e.,
gj ðv : YÞ ¼ y1 ’1, j þ y2 ’2, j þ y3 ’3, j (7)
gj ðnÞ 0, 8j ¼ 1, . . . , nc , 8n 0, and 8y C.
A uniform prior f(Y) on C is assumed for the
using the singular value decomposition, where vector of unknown parameters Y. Figure 4 shows
f yk ¼ 1, k ¼ 1, 2, 3 g are some global param- the spectral density gj(n) for (m, r)J = (8, 200 km)
eters and ’k,j, k = 1, 2, 3, are deterministic and its components ’k,j, k = 1, 2, 3.
230 Bayesian Statistics: Applications to Earthquake Engineering
Bayesian Statistics: a 8 b
Prior Prior
Applications to Posterior 2 Posterior
Earthquake Engineering,
2 0.5
0 0
0.6 0.8 1 1.2 1.4 0.5 1 1.5
θ1 θ2
c
0.6
0.4
0.3
Prior
Posterior
0.2
0 0.5 1 1.5 2
θ3
Let x ¼ xj, i , i ¼ 1, . . . , nj , j ¼ 1, . . . , nc be a matrix Cj(Y) of the ground motion process in
Xnc
set of n ¼ j¼1 j
n ground motion records avail- cell j. The components of the covariance matrix
can be calculated from the Fourier transform of
able at a site, where nj is the number of earth-
the spectral density function gj(n; Y) (Grigoriu
quakes with (m, r)J and nc is the number of cells in
2012, Section 3.6).
the seismic activity matrix in Fig. 1. It is assumed
The posterior probability density function
that xJ,I are the discrete versions of the records’
p(Y | x) follows directly from Eq. 1.
strong motion parts, i.e., the part of the earth-
Numerical example. For numerical results,
quake where most of its energy is concentrated
n = 5,000 samples of zero-mean, stationary,
(Trifunac 1971). Under the assumption that xJ,I
Gaussian processes with spectral densities gj(n),
are samples of a zero-mean, stationary, Gaussian
j = 1,. . .,nc, were simulated in cells j according
process (Zerva 2009), Chap. 7; Rezaeian and
to the multinomial distribution given by the seis-
Kiureghian 2010), the logarithmic form of the
mic activity matrix of Los Angeles (Fig. 1). The
likelihood function is written as
marginal prior and posterior densities of the
n unknown parameters yi, i = 1, 2, 3 are shown in
lnðlðxjYÞÞ / ln Cj ðYÞ Fig. 5.
2
The resulting posterior 95 % confidence inter-
1X nc Xnj
xT Cj ðYÞ1 xj, i ; (8) val for the spectral density gj(n; Y) shown in
2 i¼1 j¼1 j, i
Fig. 6, i.e., the interval [glw up
j (n), gj (n)] such that
n h io
up
where ln denotes the natural logarithm and ℙ gj ðn; YÞ glw j ðnÞ, gj ðnÞ ¼ 0:95 , is also
|Cj (Y)| is the determinant of the covariance calculated based on the density p(Y | x).
Bayesian Statistics: Applications to Earthquake Engineering 231
x 10−6 D yi , y0i , xi ¼ 2o0 z0 y0i þ o20 yi þ xi : (10)
95%C.I.
2
SBM
Two intensity measures for the ground motion
xi(t) are defined:
1.5 B
1. Peak ground acceleration (PGA)
gj(ν⏐Θ)
1
xi ¼ max jxi ðtÞj (11)
0tt
0.5
2. Pseudo-spectral acceleration (PSa)
0
0 5 10 15 20 25 30 xi ðo; zÞ ¼ max jyi ðt; o, BÞj; (12)
ν [rad/s] 0tt
Bayesian Statistics: Applications to Earthquake where yi(t; o, z) is the solution of Eq. 10.
Engineering, Fig. 6 Ninety-five percent confidence
interval (CI) for the spectral density gj(n; Y), j = 1, 2, 3
For illustration purposes, we present the meth-
odology for calculating fragility curves for the
Seismic Fragility Analysis single-degree-of-freedom system in Eq. 10. In
Seismic fragility of a structural or nonstructural this context, we define the seismic fragility as
system is the probability that the system’s
response exceeds a critical value under seismic Pf ðx; ycr Þ ¼ ℙ max jyi ðt; o0 , B0 Þj > ycr jxi ¼ x ;
ground motions of specified intensities, e.g., peak 0tt
a x 10–5 b x 10–4
p(μP GA, σP GA|data)
c −3 d −3
−3.5 −3.5
μP GA
μP GA
−4 −4
−4.5 −4.5
0.2 0.4 0.6 0.8 1 1.2 1.4 0.2 0.4 0.6 0.8 1 1.2 1.4
σP GA σP GA
a 1 b 1
0.8 0.8
B
0.6 0.6
Pf
Pf
0.4 0.4
Bayesian Statistics: Applications to Earthquake Engineering, Fig. 8 Fragility curves (mean) Pf (PGA) and 95 %
confidence intervals (CIs) for (a) n =10 and (b) n = 50 records
a x 10−3
b
p(mP Sa, P Sa|data)
3
p(mP Sa, P Sa|data)
0.04
2
0.02
1
0 0
−4 0.5 −4 0.5
0.4 0.4
0.3 0.3
0.2 0.2
mP Sa −4.5 0.1 sP Sa mP Sa −4.5 0.1 sP Sa
c d
−3.8 −3.8
−3.9 −3.9
−4 −4
mP Sa
mP Sa
−4.1 −4.1
−4.2 −4.2
−4.3 −4.3
−4.4 −4.4
−4.5 −4.5
0.1 0.2 0.3 0.4 0.5 0.1 0.2 0.3 0.4 0.5
sP Sa sP Sa
a 1 b 1
0.8 0.8
0.6 0.6
Pf
Pf
0.4 0.4
Bayesian Statistics: Applications to Earthquake Engineering, Fig. 10 Fragility curves (mean) Pf(PSa) and 95 %
confidence intervals (CIs) for (a) n = 10 and (b) n = 50 records
nD = number of possible damage states, nD 1 before it can enter damage state 2. Damage
{1, 2, . . .}. state 1 is defined as “minor cracked joints and
X = uncertain excitation, e.g., peak zero-period tile.” Damage state 2 is defined as “cracked joints
acceleration at the base of the asset in ques- and tile.” y1 = 0.0020, b1 = 0.70, y2 = 0.0050,
tion. Here excitation is called demand param- and b2 = 0.40. The repair for D = 1 is described B
eter (DP), using the terminology of FEMA as “carefully remove cracked tile and grout at
P-58 (Applied Technology Council 2012). cracked joints, install new ceramic tile and
FEMA P-58 builds upon work coordinated re-grout joints for 10 % of full 100 foot length
by the Pacific Earthquake Engineering of wall. Existing wall board will remain in place.”
Research (PEER) Center and others. PEER Repair for D = 2 is “install ceramic tile and grout
researchers use the term engineering demand all joints for full 100 foot length of wall. Note:
parameter (EDP) to mean the same thing. Usu- gypsum wall board will also be removed and
ally X fℜ 0g but it does not have to replaced which means the removal of ceramic
be. Note that X fℜ 0g means that X is a tile will be part of the gypsum wall board
real, nonnegative number. removal.”
x = a particular value of X, i.e., with no
uncertainty. The Lognormal Distribution
Fd(x) = a fragility function for damage state The lognormal is ubiquitous in probabilistic seis-
d evaluated at x. mic hazard analysis (PSHA) and probabilistic
F(s) = standard normal cumulative distribution seismic risk analysis (PSRA). To understand it,
function (often called the Gaussian) evaluated consider first the normal (not the lognormal) dis-
at s, e.g., normsdist(s) in Excel. tribution. If a quantity X is normally distributed
ln(s) = natural logarithm of s. with mean m and standard deviation s, it can take
yd = median capacity of the asset to resist dam- on any scalar value in –1 < X < 1. Its cumula-
age state d measured in terms of X. Usually tive distribution function (CDF) can be expressed
yd fℜ > 0g but it could in principle have a as follows:
vector value and perhaps negative values. The
x m
subscript d appears because a component can
P½ X x ¼ F (2)
sometimes have nD > 1. s
bd = logarithmic standard deviation of the uncer-
tain capacity of the asset to resist damage Note that:
state d.
m {ℜ}, meaning that m is any real scalar value
For example, see the PACT fragility database s {ℜ > 0}, meaning that s is any positive
at https://www.atcouncil.org/files/FEMAP-58- scalar value.
3_2_ProvidedFragilityData.zip (Applied Tech-
nology Council 2012). See the tab PERFOR- One can also find the value x associated with a
MANCE DATA, the line marked specified nonexceedance probability, p.
C3011.002c. It employs the lognormal form to
propose two fragility functions for Wall Partition, x ¼ m þ s F1 ðpÞ (3)
Type: Gypsum + Ceramic Tile, Full Height,
Fixed Below, Slip Track Above w/ returns Now consider to the lognormal distribution. If a
(friction connection). The demand parameter is variable is lognormally distributed, that means its
“Story Drift Ratio,” meaning the time-maximum natural logarithm is normally distributed, which
absolute value of the peak transient drift ratio for means it must take on a positive real value and the
the story at which partition occurs. For that com- probability of it being zero or negative is zero.
ponent, nD = 2, which occur sequentially, mean- One can write the CDF several different, equiva-
ing that a component must enter damage state lent, ways:
238 Beginner’s Guide to Fragility, Vulnerability, and Risk
lnx lny and never confuse a mathematical simplification
P½ X x ¼ F or model with reality. Ideally one’s model
b
approximates reality, but the model is not the
lnðx=yÞ
¼F (4) thing itself.
b
lnx mlnX
¼F Multiple Fragility Functions
slnX Consider situations where a component, asset, or
person can experience multiple possible damage
where mlnX denotes the mean value of ln(X),
states. This entry considers only discrete damage
which is the same as the natural logarithm of the
states, meaning that one can number the damage
median, ln(y). It is sometimes desirable to calcu-
states D = 1, 2, etc., but not 1.5. If a component is
late y and b in terms of m and s. Here are the
damaged, one can number damage states in at
conversion equations. (The reader can learn more
least one of three ways. First, let D = 0 denote
about probability distributions from various
the undamaged state. If the component or asset is
sources, such as the free online source NIST/
damaged, let the damage state be denoted by D
SEMATECH (2013) or Ang and Tang (1975).)
{1, 2, . . . nD}. The three kinds of fragility func-
Let v denote the coefficient of variation of X.
tions dealt with here are:
Then
1. Sequential damage states. A component must
s
v¼ (5) enter damage state 1 before it can enter dam-
m age state 2, and it must enter 2 before 3, and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi so on.
b¼ lnð1 þ v2 Þ (6) 2. Simultaneous damage states. A damaged
component can be in more than one damage
m
y ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi (7) state at the same time. Order does not
1 þ v2 matter.
3. Mutually exclusive and collectively exhaus-
Note well: there is nothing fundamental about tive (MECE) damage states. A damaged com-
the lognormal distribution that makes it ideal ponent can be in one and only one damage
or exact for fragility functions, ground motion, state. Order does not matter.
and so on. It is commonly used for several
reasons: There are other ways to express fragility,
such as with a vector that combines numbers
1. Simplicity. It only requires two parameter of various types. For example, one might want
values to completely define. to talk about a scalar quantity Q of a compo-
2. Precedent. The lognormal has been used for nent that is in a particular sequential damage
fragility functions for decades. state D, such as the fraction of a reinforced
3. Information-theory reasons. Given a median concrete shear wall that has cracks of at least
and logarithmic standard deviation of some 3/8 in. width, which would be indicative of the
uncertain positively valued quantity like com- quantity of steel that has to be replaced. But
ponent capacity, the lognormal is the most the present discussion is limited to the three
uncertain distribution, that is, it assumes the kinds noted above.
least amount of information.
4. Sometimes it fits the data well. Sequential Damage States
In sequential damage states, the damage states are
But the lognormal may fit data badly, some- ordered (D is therefore an ordinal number, sub-
times worse than other competing parametric and ject to certain mathematical rules of ordinal num-
nonparametric forms. Beware oversimplification, bers), so one can talk about lower damage states
Beginner’s Guide to Fragility, Vulnerability, and Risk 239
and higher ones. The probability of reaching or equal to the probability of reaching or exceeding
exceeding a lower damage state is greater than or a higher damage state.
The first line is the probability that the com- Simultaneous Damage States
ponent is undamaged. The next is the proba- With simultaneous damage states, one can eval-
bility that the component is damaged, but not uate the probability that a component is in each
in damage state nD, called the maximum dam- damage state independently of the others.
age state. The last line is the probability that Because order does not matter, damage states
the component is damaged in the maximum D are nominal numbers, like the numbers on
damage state. football jerseys without any order, although
D = 0 is reserved for the undamaged state.
P D ¼ d i & D 6¼ d j jX ¼ x ¼ P½D 1, D ¼ di jX ¼ x ∏ 1 P D ¼ d j jD 1, X ¼ x
j
¼ P½D 1jX ¼ x P½D ¼ d i jD 1, X ¼ x ∏ 1 P D ¼ dj jD 1, X ¼ x
j
¼ P ½ D 1j X ¼ x P ½ D ¼ d i j D 1 ∏ 1 P D ¼ d j j D 1
j
(12)
240 Beginner’s Guide to Fragility, Vulnerability, and Risk
Consider now the probability that the component that the component is in two damage states
is in exactly two damage states. Let D1 denote the D1 = di and D2 = dj but not any other damage
first of two nonzero damage states that the com- state dk (k 6¼ i, k 6¼ j, i 6¼ j, i 6¼ 0, j 6¼ 0, and k 6¼ 0)
ponent is in and D2 the second. Let P[D1 = di & given X = x. It is given by
D2 = dj & D 6¼ dk | X = x] denote the probability
P D 1 ¼ d i & D 2 ¼ d j & D ¼ d k jX ¼ x
¼ P½D 1jX ¼ x P½D ¼ di jD 1 P D ¼ d j jD 1 ∏ ð1 P½D ¼ dk jD 1Þ
k
(13)
One could repeat for three damage states i, j, and MECE Damage States
k by repeating the pattern. It is the product of the Remember that MECE means that, if the compo-
probabilities that the component is in each dam- nent is damaged (denoted by D 1), it is in one
age state i, j, and k and the probabilities that it is and only one nonzero damage state. One can
not in each remaining damage state l, m, n, etc. evaluate it by
P[D 1 | X = x] = probability that the com- function, the analyst should define failure in
ponent is damaged in some way, which one eval- unambiguous terms that do not require the exer-
uates with a single fragility function. If the cise of judgment, i.e., where two people observ-
fragility function is taken as a lognormal CDF, ing the same specimen would reach the same
see Eq. 10. Note that the fragility function does conclusion as to whether a specimen has failed
not have to be a lognormal CDF, but that is or not. Beware damage scales that do not meet
common. this test. Second, define the excitation to which
P[D = d | D 1] = probability that, if dam- specimens are subjected (maximum base accel-
aged, it is damaged in damage state d (and not any eration, peak transient drift ratio, etc.) in similarly
other value of D). Since under MECE damage unambiguous terms. Third, select specimens
states a component can only be in one damage without bias with respect to failure or nonfailure.
state, That is, one cannot use data about specimens that
were observed because they were damaged or
! because the damage was prominent or interesting
X
ND
P½ D ¼ d j D 1 ¼1 (15) in some way. (Failure data gathered by reconnais-
d¼1 sance surveys tend to be biased in this way.)
Fourth, ensure that specimens were subjected to
Creating Fragility Functions multiple levels of excitation.
function to the data as follows. This kind of data values of y and b that have the highest likelihood
is referred to here as type A data, A for actual of producing the observed data. At any level of
failure excitation. Before using the following excitation ri, there is a probability of any individ-
math, the analyst must clearly define “failure” ual specimen failing that is given by the lognor-
and should know both the means of observing mal CDF. Let pi denote this probability: B
specimen excitation and failure, as well as a
clear definition of the component or other asset lnðr i =yÞ
pi ¼ F (18)
category in question: b
ni = number of specimens, ni 2. Assume that the failure of any two different spec-
i = index to specimens, i {1, 2, . . . ni}. imens is independent conditioned on excitation.
ri = excitation at which specimen i failed. In that case, if one were to estimate the number of
failed specimens in bin i, it would be proper to
1X ni
take that number as a random variable with a
y¼ lnðr i Þ (16)
ni i¼1 binomial distribution. Let Fi denote that random
variable. The following equation gives the prob-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ability that one will observe fi failures among ni
1 X ni
b¼ ðlnðr i =yÞÞ2 (17) specimens with the per-occurrence failure prob-
ni 1 i¼1 ability pi:
ni ! f
P½ Fi ¼ f i ¼ p i
Bounding-Failure Excitation f i !ðni f i Þ! i
Suppose one possesses observations where at
ð1 pi Þni f i (19)
least one specimen did not fail, at least one spec-
imen did fail, and one knows the peak excitation
to which each specimen was subjected, but not This is the binomial distribution. One finds the y
the actual excitation at which each specimen and b values that maximize the likelihood of
failed. These data are referred to here as observing all the data {n1, f1, n2, f2, . . ..} given
bounding, or type B, data. Specimens are grouped excitations {r1, r2, . . .}. That likelihood is given
by the maximum level of excitation to which each by the product of the probabilities in Eq. 19,
specimen was subjected. Assume the fragility multiplied over all the bins. That is, find y and b
function is reasonably like a lognormal cumula- that maximize L(y, b) in:
tive distribution function and find the parameter mi
values y (median) and b (logarithmic standard Lðy, bÞ ¼ ∏ P½Fi ¼ f i (20)
i¼1
deviation) as follows:
One can explicitly maximize L(y, b), but it is
mi = numbers of levels of excitation among the easier to use Excel or similar MATLAB or other
data, referred to here as bins, mi 2. software. Excel’s solver is straightforward.
i = bin index, i {1, 2, . . . mi}. It may be easier to remember a more approx-
ri = maximum excitation to which specimens in imate approach of minimizing the weighted
bin i were subjected. squares of the difference between the observed
ni = number of specimens in bin i, ni {1, data and the idealized fragility function. That is,
2, . . .}. find y and b that minimize the squared error term
fi = number of specimens in bin i that failed, fi e2(y, b) in:
{0, 1, . . . ni}.
X
mi
2
f
One proper way to estimate y and b is by the e ðy, bÞ ¼
2
ni pi i (21)
ni
maximum likelihood method, i.e., by finding the i¼1
242 Beginner’s Guide to Fragility, Vulnerability, and Risk
The difference between the fragility functions data. If any of the following is true, use a mini-
derived by these two different methods generally mum value of bu = 0.25; otherwise, use
appears to be small compared with the scatter bu = 0.10.
about the regression lines, so arguments about
what is more proper tend to be academic com- 1. Data are available for five or fewer specimens.
pared with the choice of which one is easier. 2. In an actual building, a component can be
There is at least one more reasonable approach, installed in a number of different configura-
called logistic regression, but again this approach tions, but all specimens were tested with the
tends to produce roughly the same values of y and same configuration.
b, with differences that are small compared with 3. All specimens were subjected to the same
data scatter. loading protocol.
4. Actual behavior of the component is expected
Other Data Conditions to be dependent on two or more demand
There are cases where none of the specimens parameters (e.g., simultaneous drift in two
failed (type C, or capable, data), where failure is orthogonal directions), but specimens were
derived by structural analysis (type D data), loaded using only one demand parameter.
where expert opinion is used (type E), or where
Bayesian updating is used to update an existing In the case of type B data, increasing b using
fragility function with new evidence (type U). Eq. 22 introduces a bias in long-term failure prob-
For such situations, see Porter et al. (2007). ability and can cause the fragility function not to
pass through the actual failure data well. If the data
Dealing with Under-Representative Specimens generally lie at excitation levels below the derived
If the specimens used to create the fragility func- median capacity y, which is common, then
tion are very few in number or unrepresentative increasing b without adjusting y will cause the
of the broader class whose fragility is desired, or fragility function to move up (to higher probabil-
if the excitation to which they were subjected was ities) relative to the data. To increase b while still
unlike real-world earthquake shaking, one can ensuring that the derived fragility function passes
reflect added uncertainty associated with through the data, one can adjust y as follows:
unrepresentative conditions by increasing the fra-
gility function’s logarithmic standard deviation. X
Ni
ri ni
The FEMA P-58 guidelines, for example, suggest
i¼1
always increasing the logarithmic standard devi- r¼ (23)
X
Ni
ation of a fragility function that is derived from ni
test data or observations, as follows: i¼1
1.00 org/files/FEMAP-58-3_2_ProvidedFragilityData.
zip. “Component” means a building component,
like an RSMeans assembly such as glass curtain
0.75 walls. The component fragility functions in
Failure probability
undesirable outcome, called loss here, in terms of scales, some used by civil engineers and others
repair costs, life-safety impacts, and loss of func- used by public health professionals. Before using
tionality (dollars, deaths, and downtime) or in a terms such as minor injury, one should be sure
terms of environmental degradation, quality of its meaning is entirely clear and meaningful to the
life, historical value, and other measures. When intended user of the vulnerability information.
loss is depicted as a function of environmental Civil engineers sometimes use casualty scales
excitation, the function can be called a vulnera- that are ambiguous or not useful to public health
bility function. A seismic vulnerability function professionals.
relates uncertain loss to a measure of seismic Downtime is commonly measured in terms of
excitation, such as spectral acceleration response days or fractions of a year during which the asset
at some damping ratio and period. A seismic cannot be used for its intended purposes. Some-
vulnerability function usually applies to a partic- times it measures the time from the environmen-
ular asset class. tal shock (the earthquake, in the case of a seismic
Vulnerability is not fragility. Vulnerability vulnerability function) to the time when all
measures loss; fragility measures probability. repairs are completed, which includes both the
Vulnerability functions are referred to many time required to perform the repairs and a previ-
ways: damage functions, loss functions, vulnera- ous period during which damage is assessed,
bility curves, and probably others. repairs are designed, financing is arranged,
When a vulnerability function measures repair repairs are bid out, and the repair contractor
cost, it is commonly normalized by replacement mobilizes to the site.
cost new (RCN), a term which here means the Many vulnerability functions are expressed
cost of a similar new property having the nearest with conditional probability distributions that
equivalent utility as the property being appraised, give a probability that loss will not exceed some
as of a specific date (American Society of specified value given the excitation, for a partic-
Appraisers 2013). RCN excludes land value and ular asset class. The distribution is often assigned
usually refers to part or all of the fixed compo- a parametric form such as lognormal or beta, in
nents of a building or other assets (structural, which case the parameters of the distribution are
architectural, mechanical, electrical, and plumb- all required, some or all of them conditioned on
ing components) or to its contents. Repair cost excitation. Figure 2 presents a sample probabilis-
divided by RCN is referred to here as damage tic vulnerability function.
factor (DF). Some authors call it damage ratio,
fractional loss, or other terms. The expected Brief Summary of Vulnerability Derivation
value of DF conditioned on excitation is com- Methods
monly called mean damage factor (MDF). Some- While this entry does not provide guidance on
times it is assumed that if DF exceeds some how to derive vulnerability functions, it is worth-
threshold value such as 0.6, the property is not while to understand the different approaches to
worth repairing and is a total loss, so repair-cost doing so and some of the relevant history. There
vulnerability functions can jump abruptly from are three distinct approaches – sometimes
0.6 (or other threshold value) to 1.0 with increas- referred to as empirical, analytical, and expert
ing excitation. In principle DF can exceed 1.0 opinion – and some hybrids that combine aspects
because it can cost more to repair a building of two or more approaches. They are briefly
than to replace it. explained here.
When a vulnerability function measures life- An empirical vulnerability function is one
safety impacts, it commonly measures the frac- derived by regression analysis of observations
tion of indoor occupants who become casualties of pairs of excitation and loss for specimens of
(that is, they are killed or experience a nonfatal an asset class. Professional users of vulnerability
injury to some specified degree) as a function of functions such as insurance loss estimators tend
excitation. There are a variety of human-injury to prefer empirical vulnerability functions over
Beginner’s Guide to Fragility, Vulnerability, and Risk 245
results.
0.50 Expert opinion can quickly provide vulnera-
bility functions where empirical data are missing
and the analyst lacks the resources for an analyt-
E[Y |Sa =s] ical model. Briefly, one convenes a group of
0.25
experts familiar with the performance of the
asset class of interest and with a structured inter-
0.00
view process elicits their judgment of the perfor-
0 0.5 1 1.5 2 mance of the class at each of many levels of
Sa(1.0 sec, 5%), g excitation. ATC-13 (1985) represents one of the
earliest and most thorough expert-opinion
Beginner’s Guide to Fragility, Vulnerability, and models of the seismic vulnerability of buildings
Risk, Fig. 2 Sample vulnerability function that relates and other asset classes. Jaiswal et al. (2012) offers
damage factor to 1-s spectral acceleration response for
some asset class. The solid line depicts the expected a more recent example.
value of damage factor, and the dotted curve depicts the
probability density function of damage factor at 1 g of
excitation Hazard
other approaches, because they are known to be Probabilistic Seismic Hazard Analysis
based on real-world observations. One of the best Seismic hazard refers here to an uncertain rela-
examples of an empirical vulnerability function tionship between the level of some measure of
is the one derived by Wesson et al. (2004), who seismic excitation and the frequency or probabil-
performed regression analysis of losses to single- ity of a particular location experiencing at least
family housing from ground motion in the 1994 that level of excitation. It is not the measure of
Northridge earthquake. Among the challenges to excitation, the occurrence of an earthquake, nor
deriving empirical vulnerability functions are the the probability or frequency of excitation. These
scarcity of nearby ground motion observations, terms are not to be used interchangeably.
the difficulty getting reliable loss information, Seismic hazard is quantified many ways. One
and relatively few observations at high levels of is through a hazard curve, commonly depicted on
ground motion. an x-y chart where the x-axis measures shaking
Analytical methods are useful when empirical intensity at a site and the y-axis measures either
data are lacking and one has the time and other exceedance probability in a specified period of
resources to create and analyze a model. An ana- time or exceedance rate in events per unit time.
lytical approach generally involves defining one See Fig. 3 for an example Cornell (1968) applied
or more specimens to represent the class, creating the theorem of total probability to create a hazard
and analyzing a structural model to estimate curve. What follows here is a summary of current
structural response as a function of ground procedures to perform probabilistic seismic haz-
motion, estimating component-level damage ard analysis (PSHA), but is conceptually identi-
given structural response, and then estimating cal to Cornell’s work.
loss given component damage. The interested Engineers sometimes refer to the quantity by
reader is referred to FEMA P-58 (ATC 2012) which intensity is measured as the intensity mea-
for the current state of the practice for individual sure. Earth scientists call it ground motion. The
buildings and to Porter et al. (2009a, b; 2010, present work uses the term intensity measure (IM).
246 Beginner’s Guide to Fragility, Vulnerability, and Risk
lows. Let:
0.1
H = uncertain severity of ground motion at the
building site, e.g., Sa(T, 5 %).
0.01 h = a particular value of H.
G(h) = frequency (events per unit time) with
which H > h, i.e., the number of events per
0.001 unit time in which at least once during the
event, H > h.
nE = number of earthquake rupture forecast
0.0001 models to consider, nE {1, 2, . . .}. For
0.0 0.5 1.0 1.5 2.0
Sa(1.0 sec, 5%), g
example, in UCERF2, there were 480 discrete
combinations of fault model, rupture model,
Beginner’s Guide to Fragility, Vulnerability, and magnitude-area relationship, and several
Risk, Fig. 3 Sample hazard curve. The solid line shows other modeling choices that together are
the mean rate at which various levels of 1 s spectral represented by a logic tree that begins with
acceleration response are exceeded. The dotted line
shows that 0.5 g is exceeded on average about once in fault model, branches to rupture model, then
600 years to magnitude-area relationship, etc. For
UCERF2, nE = 480.
E = the “correct” earthquake rupture forecast,
Some authors distinguish between the intensity which is uncertain. E {1, 2, . . . nE}.
measure type (IMT), such as 5 % damped spectral e = an index to a particular earthquake rupture
acceleration response at 0.2 s period, and the inten- forecast, e {1, 2, . . . nE}.
sity measure level (IML), a particular value of the P[E = e] = Bayesian probability assigned to
IM such as 0.4 g. In any case, IMT must be earthquake rupture forecast e. In UCERF2,
completely specified. If using damped elastic this would be the product of the conditional
spectral acceleration response, one states the probabilities (weights) of the individual
period, damping ratio, and whether one is referring branches in the logic tree that represents
to the geometric mean of two orthogonal direc- UCERF2. Conditional probability means that
tions or the maximum direction or other direc- each branch’s probability can (though only
tional references. sometimes do) depend on which choices
To estimate seismic hazard, one applies the came before it.
theorem of total probability to combine the nA = number of attenuation relationships (also
uncertain shaking at the site caused by a particu- called ground motion prediction equations) to
lar fault rupture and the occurrence frequency or be employed, nA {1, 2, . . .}
probability of that rupture. Earth scientists create A = the “correct” ground-motion-prediction
models called earthquake rupture forecasts that equation, i.e., the one that expresses the true
specify the locations and rates at which various state of nature, which is uncertain.
fault produce earthquakes of various sizes, e.g., a = an index to ground-motion prediction
the Uniform California Earthquake Rupture equations.
Forecast version 2 (UCERF2, Field et al. 2007). P[A = a] = Bayesian probability assigned to
The uncertain shaking given a fault rupture is ground-motion-prediction equation a, i.e., the
quantified using a relationship variously called probability that ground-motion-prediction
an attenuation relationship or a ground-motion equation a actually reflects the true state of
prediction equation, such as the next-generation nature and is therefore the correct ground-
attenuation (NGA) relationships presented in the motion-prediction equation.
Beginner’s Guide to Fragility, Vulnerability, and Risk 247
nF(e) = number of fault sections in earthquake P[H > h | m,r,v,a] = probability that H > h given
rupture forecast e, nF(e) {1, 2, . . .}. earthquake magnitude m at distance r on soil
f = an index to faults section, f {1, 2, . . . with soil type v using ground-motion predic-
nF(e)}. tion equation a. Note that, given a known site
m0 = minimum magnitude to consider. location, fault segment f, and location along the B
Dm = an increment of magnitude, say, Dm = 0.1. fault segment o, distance r from the site to the
Mmax(f|E = e) = maximum magnitude of fault is known. Note that ground-motion pre-
which fault f is deemed capable under earth- diction equations use a variety of distance mea-
quake rupture forecast e, Mmax(f|E = e) sures. As of this writing, most ground-motion
{m0+ ½ Dm, m0 + 3/2 Dm, . . .}. prediction equations give an equation for the
no(f,m) = number of locations within fault section mean of the natural logarithm of H given m, r,
f that can generate a rupture of magnitude m, and v, for the logarithmic standard deviation of
no(f,m) {1, 2, . . .}. H (i.e., the standard deviation of the natural
o = an index to locations on fault section f, in o logarithm of H), and assume a lognormal dis-
{1, 2, . . . no(f,m)}. tribution of H conditioned on m, r, and v. Let
G(f,m,o|E = e) = frequency (events per unit time) mlnH and slnH denote the mean of ln(H) and the
with which fault section f ruptures at location o standard deviation of ln(H), respectively, given
producing earthquake of magnitude m ½ m, r, and v, assuming ground-motion-
Dm given earthquake rupture forecast e. prediction equation a. Under these conditions,
VS = uncertain site soil, potentially measured by
NEHRP site soil classification or shearwave lnðhÞ mlnH
P½H > hjm, r, v, a ¼ 1 F
velocity in the top 30 m of soil (Vs30) or slnH
something else. To find Vs30, see the site (25)
data application at www.opensha.org/apps
for U.S. locations or Google the USGS’s One can now estimate the hazard curve by apply-
Global Vs30 map server for other locations ing the theorem of total probability. Suppose one
v = a particular value of V. always knew soil conditions V with certainty. Then
X
nE X f ðeÞXnoX
nA nX ðf , mÞ
G ð hÞ ¼ P½H > hjm, r, v, a Gðm, f, ojE ¼ eÞ P½A ¼ a P½E ¼ e (26)
e¼1 a¼1 f ¼1 m o¼1
where the summation over m means that one events per unit time) not probability (chance that
considers each m {m0 + ½ Dm, m0 + 3/2 it occurs one or more times in a given time
Dm. . . Mmax(f|E = e)}. period).
The two can be related using a concept
Hazard Rate Versus Probability called a Poisson process. From the Wikipedia’s
Seismic hazard is often expressed in terms of article on the Poisson Process: “In probability
exceedance probability, rather than in terms of theory, a Poisson process is a stochastic pro-
exceedance rate. The distinction is this: exceed- cess which counts the number of events and the
ance probability is the probability that shaking of time that these events occur in a given time
H > h will occur at least once in a given period of interval. The time between each pair of con-
time. That means that it is the probability that it secutive events has an exponential distribution
H > h will occur exactly once, plus the probabil- with parameter G [the parameter is the occur-
ity that it will occur exactly twice, etc. For some rence rate per unit time] and each of these
calculations, the analyst wants rate (number of inter-arrival times is . . . independent of other
248 Beginner’s Guide to Fragility, Vulnerability, and Risk
inter-arrival times [meaning that the time Some Commonly Used Measures of Ground
between the second and third occurrence is Motion
independent of the time between the first and Peak ground acceleration (PGA). This is the
second occurrence – knowing one tells you maximum value of acceleration of a particular
nothing about the other, so a Poisson process point on the ground at any time during an earth-
is called memoryless]. The process is named quake. Often PGA is estimated as the geometric
after the French mathematician Siméon-Denis mean (the square root of the product) of the
Poisson and is a good model of radioactive maximum values of PGA parallel to each of two
decay, telephone calls, and requests for a par- orthogonal horizontal axes. PGA is sometimes
ticular document on a web server, among many called zero-period acceleration (ZPA), meaning
other phenomena.” the spectral acceleration response of a single-
Modeling earthquakes as Poisson arrivals is degree-of-freedom elastic oscillator with
convenient in part because in a Poisson process, zero, or near-zero, period. Before using a PGA
arrival rate and occurrence probability have this value, be sure you know whether it refers to
relationship: geometric mean, maximum-direction value, or
something else.
lnð1 PÞ Peak ground velocity (PGV). This is like
G¼ (27)
t PGA, except maximum velocity of a point on
the ground rather than acceleration.
where Peak ground displacement (PGD). This is
like PGA, except maximum displacement rela-
G = occurrence rate, i.e., events per unit time, tive to a fixed datum.
P = probability that at least one event will occur Spectral acceleration response (Sa(T,z)). This
in time t. usually refers to damped elastic spectral acceler-
ation response at some specified index period
So if a hazard curve were represented as P(h), such as T = 0.3 s, 1.0 s, 3.0 s, etc., and specified
the probability that H > h at least once in a damping ratio such as z = 5 %, at a particular
particular period of time t, one could use Eq. 27 point on the ground. To be precise, Sa(T,z) is the
to estimate the occurrence rate G(h), the average absolute value of the maximum acceleration rel-
number of times that H h per unit time. One can ative to a fixed datum of a damped elastic single-
also rearrange Eq. 27 to give probability as a degree-of-freedom harmonic oscillator with
function of rate: period T and damping ratio z when subjected to
a particular one-degree-of-freedom ground-
motion time history at its base. In practice, it
P ¼ 1 eGt (28)
typically refers to the geometric mean of spectral
acceleration response parallel to each of two
So with the occurrence rate G(h) of earthquakes orthogonal horizontal axes. It is so often measured
causing H > h, one can calculate the probability for z = 5 % that damping ratio is often not men-
that at least one earthquake with H > h will occur tioned. Before using Sa(T,z), be sure you know T,
in a given time t. z, and whether it is a geometric mean value or the
maximum-direction value or something else.
Measures of Seismic Excitation Spectral displacement response (Sd(T,z)).
These are many common measures of seismic This is like Sa(T,z) but for relative spectral dis-
excitation. Some measure ground motion and placement (displacement of the oscillator relative
some measure structural response or excitation to its base, not relative to a fixed datum) rather
to which the components of a building, bridge, than absolute acceleration of the oscillator.
or other structural systems are subjected. This Pseudoacceleration response (PSA(T,z)).
section introduces common ones. PSA(T,z) is defined as Sd(T,z) o2, where o is
Beginner’s Guide to Fragility, Vulnerability, and Risk 249
angular frequency, 2 p/T. Some authors prefer to measures, especially MMI. One reason is that
use PSA(T,z) rather than Sa(T,z), but for values of MMI observations can be made by people
z less than about 20 %, the two measures are exposed to shaking or who make post-earthquake
virtually identical. observations, whereas instrumental measures
Modified Mercalli Intensity (MMI) and require an instrument. B
European Macroseismic Scale (EMS). These Ground motion to intensity conversion equa-
are macroseismic intensity measures, meaning tions (GMICE). These estimate macroseismic
that they measure seismic excitation over a intensity as a function of instrumental measures
large area, not at a particular point on the ground. of ground motion. There are several leading
They are measured with an integer scale in GMICEs. When selecting among them, try to
Roman numerals from I to XII. They measure match the region, magnitude range, and distance
whether and how people in a region such as a range closest to the conditions where the GMICE
neighborhood or city felt and reacted to earth- will be applied. More data for conditions like the
quake motion (did they run outside?) and what ones in question are generally better than less
they observed to happen to the ground, buildings, data, all other things being equal. When consid-
and contents around them, such as plates rattling ering building response, GMICEs that convert
and weak masonry being damaged. They are sub- from Sa(T,z) to macroseismic intensity are gen-
jective and generally easier for nontechnical erally better than those that use PGA or PGV,
audiences to understand than instrumental mea- which do not reflect anything building-specific.
sures. On the MMI scale, building damage begins Two recent GMICEs for the United States are as
around MMI VI and it is rare for an earthquake to follows:
produce shaking of MMI X. EMS is similar to As of this writing, Worden et al.’s (2012)
MMI, but building-damage observations are relationships in Eqs. 29 and 30 seem to be the
related to common European building types. best choice for estimating MMI from ground
A version of EMS defined in 1998 is often motion and vice versa for California earthquakes.
referred to as EMS-98. For detail see Table 1, The reason is they employ a very large dataset of
Wood and Neumann (1931) and European California (ground motion, MMI) observations.
Seismic Commission Working Group— The dataset includes 2092 PGA-MMI observa-
Macroseismic Scales (1998). tions and 2074 PGV-MMI observations from
Japan Meteorological Agency (JMA) seismic 1207 California earthquakes M = 3.0–7.3, MMI
intensity scale. This is like MMI, but with a 0–7 2.0–8.6, R = 4–500 km. It includes no observa-
scale. It has both a macroseismic sense (observed tions from continental interior. It includes regres-
effects of people and objects) and an instrumental sions for Sa(0.3 s, 5 %), Sa(1.0 s, 5 %), Sa(3.0 s,
sense (in terms of ranges of PGA). See http://en. 5 %), PGA, and PGV that operate in both direc-
wikipedia.org/wiki/Japan_Meteorological_Agency_ tions, meaning that one can rearrange the rela-
seismic_intensity_scale for details. tionships to estimate instrumental measures in
Instrumental intensity measure (IMM). This terms of MMI, as well as MMI in terms of instru-
is a positively valued measure of intensity mental measures. The reason that the relation-
that can take on fractional values, e.g., 6.4. This ships are bidirectional is that Worden
is an estimate of MMI using functions of instru- et al. (2012) used a total least-squares data model-
mental ground-motion measures such as PGA ing technique in which observational errors on
and PGV. both dependent and independent variables are
taken into account. Equation 30 includes the
Conversion Between Instrumental and option to account for the apparent effects of mag-
Macroseismic Intensity nitude M and distance R. The columns for resid-
It is often desirable to convert between ual standard deviations show a modest reduction
instrumental ground-motion measures such as in uncertainty when accounting for M and
PGA or PGV and macroseismic intensity R (Table 2).
250 Beginner’s Guide to Fragility, Vulnerability, and Risk
Beginner’s Guide to Fragility, Vulnerability, and Risk, Table 1 MMI and EMS-98 macroseismic intensity scales
(abridged)
MMI Brief description EMS-98 Brief description
I. Instrumental Generally not felt by people unless I. Not felt Not felt by anyone
in favorable conditions
II. Weak Felt only by a couple people that are II. Scarcely felt Vibration is felt only by individual
sensitive, especially on the upper people at rest in houses, especially
floors of buildings. Delicately on upper floors of buildings
suspended objects (including
chandeliers) may swing slightly
III. Slight Felt quite noticeably by people III. Weak The vibration is weak and is felt
indoors, especially on the upper indoors by a few people. People at
floors of buildings. Standing rest feel swaying or light trembling.
automobiles may rock slightly. Noticeable shaking of many objects
Vibration similar to the passing of a
truck. Indoor objects may shake
IV. Moderate Felt indoors by many people, IV. Largely The earthquake is felt indoors by
outdoors by few. Some awakened. observed many people and outdoors by few.
Dishes, windows, and doors A few people are awakened. The
disturbed, and walls make cracking level of vibration is possibly
sounds. Chandeliers and indoor frightening. Windows, doors, and
objects shake noticeably. Like a dishes rattle. Hanging objects
heavy truck striking building. swing. No damage to buildings
Standing automobiles rock. Dishes
and windows rattle
V. Rather strong Felt inside by most or all and V. Strong Felt indoors by most, outdoors by
outside. Dishes and windows may many. Many sleeping people awake.
break. Vibrations like a train A few run outdoors. China and
passing close. Possible slight glasses clatter. Top-heavy objects
damage to buildings. Liquids may topple. Doors and windows swing
spill out of glasses or open
containers. None to a few people are
frightened and run outdoors
VI. Strong Felt by everyone; many frightened VI. Slightly Felt by everyone indoors and by
and run outdoors, walk unsteadily. damaging many outdoors. Many people in
Windows, dishes, glassware buildings are frightened and run
broken; books fall off shelves; some outdoors. Objects on walls fall.
heavy furniture moved or Slight damage to buildings; for
overturned; a few instances of fallen example, fine cracks in plaster and
plaster. Damage slight to moderate small pieces of plaster fall
to poorly designed buildings; all
others receive none to slight damage
VII. Very strong Difficult to stand. Furniture broken. VII. Damaging Most people are frightened and run
Damage light in buildings of good outdoors. Furniture is shifted and
design and construction; slight to many objects fall from shelves.
moderate in ordinarily built Many buildings suffer slight to
structures; considerable damage in moderate damage. Cracks in walls;
poorly built or badly designed partial collapse of chimneys
structures; some chimneys broken
or heavily damaged. Noticed by
people driving automobiles
(continued)
Beginner’s Guide to Fragility, Vulnerability, and Risk 251
Beginner’s Guide to Fragility, Vulnerability, and Risk, Table 2 Parameter values for Worden et al. (2012) GMICE
for California
Eq. 29 Eq. 30
Y c1 c2 c3 c4 c5 c6 c7 t1 t2 sMMI slog10Y sMMI slog10Y
PGA 1.78 1.55 1.60 3.7 0.91 1.02 0.17 1.57 4.22 0.73 0.39 0.66 0.35
PGV 3.78 1.47 2.89 3.16 0.90 0.00 0.18 0.53 4.56 0.65 0.40 0.63 0.38
PSA 1.26 1.69 4.15 4.14 1.05 0.60 0.00 2.21 4.99 0.84 0.46 0.82 0.44
(0.3 s)
PSA 2.50 1.51 0.20 2.90 2.27 0.49 0.29 1.65 4.98 0.80 0.51 0.75 0.47
(1.0 s)
PSA 3.81 1.17 1.99 3.01 1.91 0.57 0.21 0.99 4.96 0.95 0.69 0.89 0.64
(3.0 s)
252 Beginner’s Guide to Fragility, Vulnerability, and Risk
Beginner’s Guide to Fragility, Vulnerability, and Risk, Table 3 Parameter values for Atkinson and Kaka (2007)
GMICE for the United States
sMMI
Y c1 c2 c3 c4 c5 c6 c7 Eq. 31 Eq. 32
PGA 4.37 1.32 3.54 3.03 0.47 0.19 0.26 0.80 0.76
PGV 2.65 1.39 1.91 4.09 1.96 0.02 0.98 1.01 0.89
PSA(0.3 s) 2.40 1.36 1.83 3.56 0.11 0.20 0.64 0.88 0.79
PSA(1.0 s) 3.23 1.18 0.57 2.95 1.92 0.39 0.04 0.84 0.73
PSA(2.0 s) 3.72 1.29 1.99 3.00 2.24 0.33 0.31 0.86 0.72
interchangeable – it is inappropriate to simply Eq. 33, except with the subscript 0 on y and e.
rearrange terms of a GMICE to produce an It is practical to calculate for a given location,
IGMCE. The reason is that both GMICE and intensity measure type, and intensity measure
IGMCE are derived by regression analysis. level, the contribution of each fault segment,
Given (x,y) data, a least-squares regression of magnitude, rupture location, and value of e0 to B
y as a function of x will generally produce a the frequency with which the site is expected
different curve than a least-squares regression of to experience ground motion of at least the
x as a function of y. specified intensity measure level. In fact, Eq. 26
shows that the site hazard is summed from
Hazard Deaggregation such values. (For simplicity that equation
When evaluating the risk to an asset, it is often omits mention of e, but the extension is modest.)
desirable to perform nonlinear dynamic structural See Baker and Cornell (2005) for more
analyses at one or more intensity measure levels. information.
To do so, one needs a suite of ground-motion time Rather than leading the reader through the
histories scaled to the desired intensity. The math, suffice it to say that there are online tools
ground-motion time histories should be consis- to do that hazard deaggregation, and an example
tent with the seismic environment. That is, they is given here. The USGS offers a website that
should reflect the earthquake magnitudes m and does interactive hazard deaggregation for the
distances r that would likely cause that level of United States. As of this writing, the URL
excitation in that particular place. The reason is includes the year associated with the hazard
that magnitude and distance affect the duration model, so it will change over time. The most
and frequency content of the ground-motion time recent tool at this writing is https://geohazards.
history, which in turn affects structural response. usgs.gov/deaggint/2008/. When that site
See McGuire (1995) or Bazzurro and Cornell becomes obsolete, the reader should be able to
(1999) for more information. find the current one by googling “interactive haz-
There is another term (commonly denoted ard deaggregation USGS.”
by e) that also matters. It relates to how the Consider an imaginary 12-story building in
spectral acceleration response at a specified San Diego, California at 1126 Pacific Hwy, San
period in a particular ground-motion time history Diego CA, whose geographic coordinates are
differs from its expected value, given magnitude 32.7166 N -117.1713 E (North America has neg-
and distance. Let y denote the natural logarithm ative east longitude). Suppose its small-
of the intensity measure level, e.g., the natural amplitude fundamental period of vibration is
logarithm of the spectral acceleration response at 1.0 s, its Vs30 is 325 m/s, and its depth to bedrock
the building’s estimated small-amplitude funda- (defined as having a shearwave velocity of
mental period of vibration. Let m and s denote the 2,500 m/s) is 1.0 km. One wishes to select several
expected value and standard deviation of the nat- ground-motion time histories with geometric
ural logarithm of the intensity measure level, mean Sa(1.0 s, 5 %) equal to that of the motion
respectively, calculated from a ground-motion with 10 % exceedance probability in 50 years.
prediction equation. The e term is a normalized The input data look like Fig. 4. The results look
value of y, as follows: like Fig. 5, which shows that 10%/50-year motion
at this site tends to result from earthquakes with
ym Mw 6.6 at 1.8 km distance and a value of
e¼ (33)
s e0 = 1.22. One can then draw sample ground-
motion time histories with approximately these
When calculating the motion y0 that has a speci- values of magnitude, distance, and e0 from a
fied exceedance probability p0, one labels the e database such as PEER’s strong motion database,
from a specific source and this particular value of currently located at http://peer.berkeley.edu/
motion y0 as e0. The equation is the same as peer_ground_motion_database.
254 Beginner’s Guide to Fragility, Vulnerability, and Risk
Beginner’s Guide to Fragility, Vulnerability, and Risk, Fig. 4 USGS interactive hazard deaggregation website
Convenient Sources of Hazard Data same location gridded uniform seismic hazard
For California sites, see www.OpenSHA.org/ data such as the values of Sa(1.0 s, 5 %) with
apps for a very powerful hazard curve calculator. 2 % exceedance probability in 50 years (the
See http://earthquake.usgs.gov/hazards/products/ so-called MCEG map). For that map, see the
for several sets of gridded hazard data from the data file labeled “Gridded Hazard Map 1 Hz
National Seismic Hazard Mapping Program. For (1.0 s) 2 % in 50 Years” at http://earthquake.
example, for the 2008 NSHMP hazard curves for usgs.gov/hazards/products/conterminous/2008/
Sa(1.0 s, 5 %), see http://earthquake.usgs.gov/ data/. For elsewhere in the world, see www.
hazards/products/conterminous/2008/data/ and globalquakemodel.org.
look for the data file labeled “Hazard curve data The risk analyst typically also needs a conve-
1 Hz (1.0 s).” If you use NHSMP’s hazard curves nient and authoritative source of soil information.
(as opposed to uniform seismic hazard maps), the As of this writing, the USGS offers a global map
data will reflect hazard on site class B, so adjust of estimated average shearwave velocity in the
for other site classes. A good way to do that is by upper 30 m of soil (Vs30 in ASCE 7-10 notation).
multiplying by the site coefficient Fa (for the See http://earthquake.usgs.gov/hazards/apps/
3.33 Hz or 5 Hz curves) or Fv (for the 1 Hz vs30/ or Google “Global Vs30 Map Server” for
curves) from ASCE 7-10 Tables 11.4-1 and this useful resource. Find your latitude and lon-
11.4-2, respectively. You will also find at the gitude using Google Earth, then look up your
Beginner’s Guide to Fragility, Vulnerability, and Risk 255
Beginner’s Guide to Fragility, Vulnerability, and Risk, Fig. 5 Sample output of the USGS’ interactive hazard
deaggregation website
location at the USGS site by centering a grid possible values of that undesirable outcome – it
around it, adding and subtracting 0.01 degrees occurs or it does not occur – one can apply the
of latitude and longitude for a Vs30 map about theorem of total probability, combining fragility
2km 2km grid centered at your location. See and hazard, to estimate the mean frequency with
Wald and Allen (2007) for its technical basis. which it occurs or the probability that it will occur
In the United States, the site data app at www. in a specified period of time. If the undesirable
OpenSHA.org/apps is more powerful and easier outcome is measured in terms of loss, then one
to use. It offers the Wald-Allen Vs30 along with can apply the theorem of total probability, com-
other sources, such as the Wills and Clahan bining vulnerability and hazard, to estimate the
(2006) map of Vs30 in California. mean annualized loss or the probability that at
least a specified degree of loss will occur in a
specified period of time. This entry deals with
Risk for a Single Asset risk only for a single asset.
where G(s) = mean annual frequency of shaking Probability of Failure During a Specified
exceeding intensity s. One can also use integra- Period of Time
tion by parts and show that If one assumes that hazard and fragility are
memoryless and do not vary over time, then fail-
ð
1
dFðsÞ ure is called a Poisson process, and the probabil-
l¼ GðsÞds (35) ity that failure will occur at least once in time t is
ds
s¼0 given by
X
n ð
1
l ¼ Fi1 Gi1 ð1 expðmi Dsi ÞÞ dGðsÞ
EAL ¼ V yðsÞ ds (39)
i¼1 ds
! 0
DFi 1 1
Gi1 expðmi Dsi Þ Dsi þ
Dsi mi mi where V refers to the replacement value of the
X
n asset and y(s) is the expected value of loss given
¼ ðFi1 ai DFi bi Þ shaking s as a fraction of V. Equation 39 can only
i¼1 rarely be evaluated in closed form. More com-
(37) monly, y(s) and G(s) are available at discrete
values of s. If one has n + 1 values of s, at
where which both y(s) and G(s) are available, and
these are denoted by si, yi, and Gi: i = 0, 1,
Dsi ¼ si si1 DFi ¼ Fi Fi1 2, . . . n, respectively, then EAL in Eq. 39 can be
mi ¼ lnðGi =Gi1 Þ=Dsi for i ¼ 1, 2, . . . n replaced by
Beginner’s Guide to Fragility, Vulnerability, and Risk 257
X
n 1
EAL ¼ V yi1 Gi1 ð1 expðmi Dsi ÞÞ
i¼1
!
Dyi 0.1
Exceedance rate, yr −1
1 1
Gi1 expðmi Dsi Þ Dsi þ B
Dsi mi mi
X
n
0.01
¼V ðyi1 ai Dyi bi Þ
i¼1
1/250-year loss
(40)
0.001
where
ð
1
where
dGðsÞ
RðxÞ ¼ ð1 P½X xjS ¼ sÞ ds
ds
s¼0 y(s) = median vulnerability function, i.e., the
(41) value of the damage factor with 50 % exceed-
ance probability when the asset is exposed to
where excitation s,
v(s) = coefficient of variation of vulnerability,
X = uncertain degree of loss to an asset, such as i.e., the coefficient of variation of the damage
the uncertain damage factor, factor of the asset exposed to excitation s,
x = a particular value of X, b(s) = logarithmic standard deviation of the
s = a particular value of the excitation, such as vulnerability function, i.e., the standard devi-
the shaking intensity in terms of the 5 % ation of the natural logarithm of the damage
damped spectral acceleration response at factor when the asset is exposed to
some index period of vibration, excitation s.
258 Beginner’s Guide to Fragility, Vulnerability, and Risk
If one has the mean vulnerability function y(s) of loss at 0.90, when conditioned on shaking with
and coefficient of variation of loss as a function 10 % exceedance probability in 50 years.
of shaking v(s), use Eqs. 6 and 7 to evaluate y(s) For example, assume that loss is lognormally
and b(s). distributed conditioned on shaking s, with
Suppose the analyst has y(s), v(s), and G(s) at a median y(s) and logarithmic standard deviation
number n of discrete values of s, denoted here by b(s) as described near Eq. 42, which are related to
si, where i is an index i {1, 2, . . . n}. One can the mean vulnerability function y(s) and coeffi-
numerically integrate Eq. 41 by cient of variation v(s) as in Eqs. 6 and 7. Under
the assumption of Poisson arrives of earthquakes,
X
n shaking with 10 % exceedance probability in
Rð x Þ ¼ pi1 ðxÞGi1 ð1 expðmi Dsi ÞÞ 50 years is the shaking with exceedance rate
i¼1
G(sPML) = 0.00211 per year, so PML can be
!
Dpi ðxÞ 1 1 estimated as a fraction of value exposed by
Gi1 expðmi Dsi Þ Dsi þ
Dsi mi mi
PML ¼ yðsPML Þ expð1:28 bðsPML ÞÞ (46)
X
n
¼ ðpi1 ðxÞ ai Dpi ðxÞ bi Þ
i¼1
where sPML = G1(0.00211 year1), that is, the
(43) hazard curve (events per year) inverted at
0.00211.
where
acceleration response. The entry introduced earthquake rupture forecast, version 2 (UCERF 2).
probabilistic seismic hazard analysis, distin- USGS open file report 2007–1437
Jaiswal KS, Aspinall WP, Perkins D, Wald D, Porter KA
guished probability from exceedance rate, sum- (2012) Use of expert judgment elicitation to estimate
marized some leading measures of excitation and seismic vulnerability of selected building types. Proc
conversion between them, explained hazard 15th World Conference on Earthquake Engineering, B
deaggregation, and offered some convenient September 2012, Lisbon, Portugal
Johnson GS, Sheppard RE, Quilici MD, Eder SJ,
sources of hazard data. Risk relates the degree Scawthorn CR (1999) Seismic reliability assessment
of loss to its rate of exceedance, or it can express of critical facilities: A handbook, supporting documen-
long-term average loss. This entry presented tation, and model code provisions, MCEER-99-0008.
methods for estimating and depicting risk for a Multidisciplinary Center for Earthquake Engineering
Research, Buffalo, NY, 384 pp
single asset. Kaestli P, Faeh D (2006) Rapid estimation of
macroseismic effects and ShakeMaps using
macroseismic data. In: Proceedings of First European
conference on earthquake engineering and seismol-
References ogy, Geneva
Kaka SI, Atkinson GM (2004) Relationships between
American Society of Appraisers (2013) Definitions of value instrumental ground-motion parameters and Modified
relating to MTS assets, Machinery & Technical Special- Mercalli intensity in Eastern North America. Bull
ties Committee of the American Society of Appraisers. Seismol Soc Am 94(5):1728–1736
Reston VA, http://www.appraisers.org/Disciplines/ McGuire RK (1995) Probabilistic seismic hazard analysis
Machinery-Technical-Specialties/mts-appraiser-resources/ and design earthquakes: closing the loop. Bull Seismol
DefinitionsOfValue. Accessed 18 May 2015 Soc Am 85(5):1275–1284
Ang AHS, Tang WH (1975) Probability concepts in engi- Murphy JR, O’Brien LJ (1977) The correlation of peak
neering planning and design, 1 – basic principles. ground acceleration amplitude with seismic intensity
Wiley, New York, 409 pp and other physical parameters. Bull Seismol Soc Am
Applied Technology Council (ATC) (1985) ATC-13, 67:877–915
Earthquake Damage Evaluation Data for California. National Institute of Building Sciences and Federal Emer-
Redwood City, CA, 492 pp gency Management Agency (NIBS and FEMA) (2009)
Applied Technology Council (ATC) (2012) Seismic per- Multi-hazard loss estimation methodology, Earth-
formance assessment of buildings volume 1 – method- quake Model, HAZUS ®MH MR4 Technical
ology FEMA P-58-1. Federal Emergency Manual. Federal Emergency Management Agency,
Management Agency, Washington, DC. http://goo.gl/ Washington, DC
QN8AQz NIST/SEMATECH (2013) e-Handbook of statistical
Atkinson GM, Kaka SI (2007) Relationships between felt methods. http://www.itl.nist.gov/div898/handbook/.
intensity and instrumental ground motion. Bull Accessed 3 Mar 2015
Seismol Soc Am 97:497–510 Porter KA (2009a) Cracking an open safe: HAZUS
Baker JW, Cornell CA (2005) A vector‐valued ground vulnerability functions in terms of structure-
motion intensity measure consisting of spectral accel- independent spectral acceleration. Earthq Spectra
eration and epsilon. Earthq Eng Struct Dyn 25(2):361–378
34(10):1193–1217 Porter KA (2009b) Cracking an open safe: more HAZUS
Bazzurro P, Cornell A (1999) Disaggregation of seismic vulnerability functions in terms of structure-
hazard. Bull Seismol Soc Am 89(2):501–520 independent spectral acceleration. Earthq Spectra
Cornell CA (1968) Engineering seismic risk analysis. Bull 25(3):607–618
Seismol Soc Am 58(5):1583–1606 Porter KA (2010) Cracking an open safe: uncertainty in
European Seismic Commission Working Group— HAZUS-based seismic vulnerability functions. Earthq
Macroseismic Scales (1998) European macroseismic Spectra 26(3):893–900
scale 1998 EMS-98. Luxembourg. http://www.gfz- Porter KA (2014) A beginner’s guide to fragility vulnera-
potsdam.de/pb5/pb53/projekt/ems/eng/index_eng.html. bility and risk, with solved exercises. Resource
Accessed 17 July 2006 document. http://spot.colorado.edu/
porterka/Porter-
Faenza L, Michelini A (2010) Regression analysis of MCS beginners-guide.pdf. Accessed 28 Apr 2014
intensity and ground motion parameters in Italy Porter KA, Kennedy RP, Bachman RE (2007) Creating
and its application in ShakeMap. Geophys J Int fragility functions for performance-based earthquake
180:1138–1152 engineering. Earthq Spectra 23(2):471–489
Field EH, Dawson TE, Felzer KR, Frankel AD, Gupta V, Porter K, Farokhnia K, Vamvatksikos D, Cho I (2014)
Jordan TH, Parsons T, Petersen MD, Stein RS, Weldon Analytical derivation of seismic vulnerability func-
RJ II, Wills CJ (2007) The Uniform California tions for highrise buildings, global vulnerability
260 Behavior Factor and Ductility
Behavior Factor and Ductility, Fig. 1 (a) Baron wikipedia.org/wiki/ Baron_Baker, also: http://www-g.
J. Baker demonstrating the effectiveness of a model Mor- eng.cam.ac.uk/125/1925-1950/baker6.html). (b) Picture
rison shelter to survive a collapsed building floor follow- of a Morrison shelter deployed inside a house for
ing a bomb hit, with controlled inelastic deformation in civilian protection (http://en.wikipedia.org/wiki/Morrison_
order to ensure survival of its occupants (http://en. shelter#cite_note-21)
materials, was well developed in the early nine- recognizing (and thereby enforcing) the fact that
teenth century, it was at this time that through the such an intentionally (or accidentally)
formulation of the plasticity theorems that limit underdesigned structure with adequate ductility
analysis design methods for reduced loading capacity will survive the extreme loading event
capacity were adopted, for the design of simple scenario with tolerable inelastic deformations, to
steel structures for civilian protection under a level that is acceptable for the safety of its
extreme blast loads (Fig. 1). Specifically for seis- occupants and equipment. As an extension of
mic loading and following the observation/eval- this initial collapse limitation design approach
uation of the response of simple structures adopted in the 1970s, additional intermediate
following severe earthquakes in California in performance requirements for occupants and
the 1950s, it was demonstrated by Housner equipment have been adopted, leading to a com-
(1956) that a sufficient inelastic deformation sup- plete performance-based design (PBD) philoso-
ply by the structure through controlled phy for the safety and/or operability of the
(hierarchical) underdesign of the elements and structure, occupants, and equipment.
local detailing resulted in these structures behav- In order to reduce the design load for elastic
ing in an acceptable manner during the extreme response, a factor greater than 1.0 is adopted,
earthquake event. which, depending on the code, is denoted as: the
Taking therefore into account the transient behavior factor q (Europe), the response reduction
time-varying nature of the loading and of the coefficient or response modification factor
response, the type of the loading and inertia R (United States and the Americas), the structural
effects, and the extreme magnitude of the load performance factor Sp (New Zealand), or the struc-
intensity, a reduction factor was proposed to be tural characteristic factor Ds (Japan), all these fac-
used in order to reduce the design force (therefore tors denoted as the behavior factor q herein, for
resistance) of the structure to levels (often con- simplicity. The magnitude of q, obtained for
siderably) lower than those required for full elas- (design) or by (assessment) the quantification of
tic response, at the expense of increased inelastic the limiting structural resistance versus the mag-
global (structural) and local (element, section) nitude of the corresponding inelastic global or
deformations of a controlled magnitude and dis- local element, section, or material inelastic defor-
tribution. This reduction is enforced through mations (axial or shear strains, flexural rotations,
262 Behavior Factor and Ductility
and so on) – expressed as the corresponding duc- Behavior Factors for Seismic Design or
tilities m – and the interrelation of these two, is the Evaluation and Redesign of Structural
topic of discussion of this entry. Systems, qd and qs
The earliest use of a ductility reduction coeffi-
cient can be traced to the seismic design regula-
The Behavior Factor: Its Types, tions in California, in which a force reduction
Definitions, and Uses in Seismic Design coefficient was adopted depending on the build-
ing properties, namely, seismic resisting system
Behavior Factor for Ductility qm of SDOF and construction material. The 1959 edition of
Equivalent Structural Systems the Structural Engineers Association of Califor-
The behavior factor for ductility (denoted as qm) nia (SEAOC) Blue Book and its successive edi-
of an inelastic SDOF system is defined, for a tions (e.g., SEAOC 1974) used this horizontal
particular base excitation g(t), as the ratio of its force factor K, in the context of an allowable
yield resistances Ry in order to develop two duc- stress design procedure, for the estimation of the
tility levels, a prescribed ductility level m = mi and design base shear. ATC (1978) was the first seis-
a ductility m = 1.0, namely, entirely elastic mic design code that adopted an R factor greater
response (Eq. 1): than 1.00 as a divisor of the elastic design
response spectrum (EDRS) forces, which
qm ðT 0 , x, f ðR, mÞ, gðtÞ, mi Þ depended on the type of structural system but
was, as yet, independent of the building period.
Ry ðmaxjmj ¼ 1:0Þ
¼ (1) The early uses of R were based not only on SDOF
Ry ðmaxjmj ¼ mi Þ analysis but, primarily, on consensus values
through the earthquake engineering design com-
where qm, besides g(t), depends on the initial munity, with typical values of R and comparisons
elastic response parameters, namely, elastic among different codes given in ATC (1995).
period T0 and critical damping ratio x, and on Since then, the reduction coefficient R – initially
the parameters characterizing the inelastic as Rw, for codes based on working stress design
response, namely, Ry and the cyclic hysteretic methods such as UBC (1988) – has been adopted
response shape f(R, m). in all ultimate level design codes (e.g., NEHRP
Both terms of the fraction above are evaluated 1985; FEMA-356 2000; ASCE 2013). In Europe,
from time history integration of the equation of Eurocode 8, from the initial stages of develop-
motion under the base excitation g(t). Solution of ment (EC8 ENV 1988) up to the final completion
the equation over a range of periods T0 yields the (EC8 2004), also adopted the behavior factor q.
ductility spectrum for the given earthquake exci- To date, most modern force-based seismic design
tation, also denoted as the qm–m–T spectrum of codes adopt the behavior factor approach for
this excitation. For design applications, such quantifying the seismic design forces, with dif-
spectra are evaluated for an ensemble of base ferent levels of rationalization for the magnitude
excitation inputs characterizing the local seismic- of q (R) (see, for instance, an overview of
ity of the region and local soil effects; the R factors adopted in South American seismic
resulting average plus percent fractile qm–m–T design codes in Chavez et al. 2012).
spectrum is obtained for the design and evalua- Based on the way the behavior factor is
tion of structures in the region exhibiting com- applied in seismic design, one can identify two
patible hysteretic response characteristics and uses of q, namely: (i) qd – specified for the design
initial damping. The definition of qm and its of new buildings in order to control the ductility
dependence on structural and excitation parame- demand (and thus damage level) for a given site
ters is discussed in section “Evaluation of the seismicity and performance level. This is speci-
Behaviour Factor Due to Ductility qm of SDOF fied in the seismic codes as the maximum allowed
Systems” herein. behavior factor for the design of modern
Behavior Factor and Ductility 263
Ved
×1/qμ
B
Base Shear V Vud
×1/qd,s
×1/qΩ (×Ω)
Vyd
Vd ×1/qm (×γm)
Vd,w ×1/qw (×ε)
Performance
δI δII δIII δIV levels (I-IV)
δy δu
1.0 μII μIII μIV
Roof Deformation δ
Behavior Factor and Ductility, Fig. 2 Definition of the In the context of PBD, qs,d are no longer unique and
design and supplied (available) behavior factors qd and qs depend on the local (and therefore global) ductility
and the contributing q, for non-base-isolated structures: demands mI to mIV for different performance levels I to
(i) q for ductility (qm), (ii) q for structural overstrength IV, respectively
(qO), and (iii) q for material strength reduction (qm or qw).
factor relates the elastic spectral base shear of the structure. Its magnitude and mode of
demand Ved to the equivalent SDOF bilinear evaluation are discussed in section “Build-
yield strength Vud, obtained by assuming an ing Behaviour Factor qs” herein.
elastoplastic equal area approximation of (iii) qx is the behavior factor contribution due to
the inelastic base shear–roof displacement increased critical damping x; this contribut-
response under a given lateral load profile, ing factor is meaningful for base-isolated or
up to the ultimate deformation du. The artificially damped systems, since the con-
value of this factor varies with the base tribution of the initial mass proportional
excitation(s), the dynamic and inelastic hys- damping (rather than the hysteretic
teretic characteristics of the system damping, which is included in qu) is not as
(including structural damping), and its duc- significant in non-base-isolated structures.
tility. It corresponds to the basic design This factor is therefore not considered fur-
behavior qs in EC8 (2004): for concrete ther herein.
structures qs takes values between (iv) qm (or qw) is the behavior factor due to
1.5 < qs < 3.0 and 2.0 < qs < 4.5, for material strength qm = Vyd/Vd or qw = Vyd/
ductility classes medium and high, respec- Vdw, relating the base shear at incipient
tively. qm is established from SDOF analy- member yield Vyd to the design base shear
sis, as described in section “Evaluation of level specified by the code Vd (for ULS
the Behaviour Factor Due to Ductility qm of design) or Vd,w (for working stress design).
SDOF Systems” of this work. This factor is practically the ratio of the
(ii) qO is the behavior factor for structural material strength partial factor in load and
overstrength, qO = Vud/Vyd, relating Vud to resistance factor design, in order to relate
the base shear Vyd at incipient onset of yield the design material stress level (for EC2
within the structural elements, at a roof 2004, the characteristic strength fyk) to the
deformation dy. The value of this factor, yield strength (fyd) of the material control-
which in EC8 (2004) is defined as the ratio ling the response, namely, the steel rein-
au/a1, accounts for the overstrength of the forcement in modern structural RC
structure and depends on the type of struc- designs; for EC2 (2004), qm = fyk/fyd =
tural system and the degree of redundancy 1.15. For buildings in Greece constructed
built in the system: for instance, for a typical during the 1970s using reinforcement with
soft ground story pilotis structure with lim- a yield strength of 400 MPa and designed
ited redistribution of inelastic actions in the for an allowable stress in flexural design
members of the upper stories, this factor is calculations of 230 MPa with a 20 %
close to 1.0. For medium-rise RC frame increase for the seismic load combination,
buildings, however, where seismic loads qm ¼ 400=1:2 230 ¼ 1:45 (Repapis
govern the design and members exhibit suf- et al. 2006a). It has been proposed that the
ficient local ductility for redistribution, qO is factor qm may also include strength gain due
of the order of 1.5–1.8, while for fully to strain rate effects, although this increase
infilled RC frames with good-quality is not a factor to take into account since it is
masonry construction, qO can easily reach counterbalanced by strength reduction by
values of 2.50 and higher (Repapis cyclic degradation.
et al. 2006b). For low-rise RC frame build-
ings, overstrengths of 4.0 have been This behavior factor is implicitly taken into
reported. This factor can be quantified account in the design of new buildings and is
through static or dynamic inelastic analysis of primary importance since it defines the
Behavior Factor and Ductility 265
transition from allowable design level to peak Evaluation of the Behavior Factor Due to
member resistance in the assessment of Ductility qm of SDOF Systems
existing buildings, in which their members
may behave in a brittle manner with very little The response of nonlinear SDOF systems under
deformability reserves. seismic excitation has traditionally been formu- B
lated in the form of qm–m–T (or Rm–m–T) spectra,
Behavior Factor qs Supplied by an Existing representing an extension of earlier uses of such
Structure Under a Given EDRS spectra, relating the ratio of yield to elastic resis-
In the context of performance-based design tance of the SDOF system (qm) to the
(PBD), the available behavior factor qs supplied corresponding displacement ductility m, under
by an existing structure can also be established blast-type loads (Biggs 1964). Such inelastic
for a given EDRS demand and at specific perfor- response spectra (also known as shock spectra)
mance limits of damage. The structure may have have been formulated for different time-varying
been designed using a prescribed design behavior forcing shapes, such as impulse, triangular, or
factor qd (different than qs) or none at all, as the half sine, and are widely used in military and
case is for older existing buildings which have protective designs for blast. As discussed in the
been designed using past force-based design phi- introduction, the use of such design load reduc-
losophies (e.g., working stress). tions has been well established in the design of
The available behavior factor qs of the struc- protective structures against blast, where the
ture as is (often reevaluated after strengthening introduction of a factored elastic design for
and/or rehabilitation of the structure at hand) is controlled inelastic deformation was initially
evaluated (i) through dynamic analyses of the introduced.
structure until the establishment of a collapse The mathematics for the evaluation of the
base excitation, which is then compared to the dynamic response of SDOF systems under seis-
design base excitation of the currently enforced mic excitation is well established in textbooks of
EDRS for this type of structural system, or structural dynamics (Clough and Penzien 1975)
(ii) through static inelastic analyses of the struc- and is not covered herein. The peak ductility
ture, using an often idealized bilinear approxima- demand m of a given inelastic system under a
tion of the resulting base shear resistance–roof certain type of base (or force) excitation is
deformation capacity curve under an imposed obtained by step-by-step integration using
lateral deformation. Following an initial explicit or implicit time integration procedures,
idealization of the structure as an equivalent depending on the duration of the forcing function.
SDOF system, with a yield resistance equal to From the way the problem is posed in Eq. 1, given
the base shear at the onset of yielding in the the SDOF characteristics (Ry, T0, x, f, g), only the
building and an effective period equal to the peak ductility can readily be estimated. Conse-
initial secant stiffness from the lateral pushover quently, the numerical estimation of the qm–m–T
curve, qs is the behavior factor which results in a for a prescribed ductility m = mi is necessarily an
ductility demand equal to the ductility actually iterative process, involving successive iterative
supplied by the system equivalent SDOF estimates of Ry and solution of the equation of
system. Contributions to qs in this case are motion to establish Ry ðmaxjmj ¼ mi Þ, for which
accounted from the structure overstrength or the the SDOF system attains a peak ductility m = mi,
member overstrength due to material qm, relating e.g., using a variable bound bisection iterative
the yield strength (also obtained from in situ solution scheme, to within a user-specified toler-
evaluation of material capacity) to the design ance. Subsequent application of the procedure for
strength. m = 1 as well (elastic response) will establish
266 Behavior Factor and Ductility
VELOCITY (IN./SEC.)
8 8
the graphic methodology 6 6PGD
by Newmark and Hall
(1973) 4 4
PGA
2 2
1 1
.8 .8
.6 .6
.4 .4
.2 .2
.1 .1
.01 .02 .04 .06 .08 .1 .2 .4 .6 .8 1 2 4 6 8 10
PERIOD (SECS.)
Ry ðmaxjmj ¼ 1Þ and qm. It should be noted that in the IDRS. Such indirect methods essentially pro-
the case of strongly softening systems which are vide qm(m, T) relations, as a function of target mi,
representative of structures with severe second- for an assumed damping and form of inelastic
order effects or strength degradation, such an hysteretic response (initially an elastic perfectly
iterative solution may not converge, implying plastic (EPP) oscillator was considered).
that the SDOF system may never attain this target Following estimations of SDOF inelastic
ductility. response assuming EPP behavior and using lim-
ited accelerogram data available at the time,
Indirect Evaluation Methods Newmark and Hall (1973) observed that (Fig. 3)
For seismic design, the evaluation of the behavior (i) elastic and inelastic SDOF systems with suffi-
factor through the formulation of an inelastic ciently medium-to-long periods (medium-to-low
design response spectrum (IDRS) shape for a frequencies) tend to displace equally and, conse-
given seismicity condition – expressed in the quently, the behavior factor is equal to the
form of a peak ground acceleration (PGA), required ductility, (ii) elastic and inelastic
ground velocity (PGV), and ground displacement SDOF systems with very short period (very high
(PGD) at the site – has initially been based on frequency) have spectral accelerations (and thus
indirect methods. Such methods specified ampli- force demands) equal to the PGA and should be
fication factors for the establishment of an EDRS designed to remain elastic, and (iii) in between,
and subsequent spectral reduction factors (i.e., the energy absorbed by an inelastic system hav-
period-dependent behavior factors) for ing a yield strength 1/qm of the elastic demand
constructing the IDRS. This approach was ini- and, as a consequence, deforming to a target mi is
tially recognized by Veletsos and Newmark nearly equal to the kinetic energy absorbed by the
(1960), yet, later on, Newmark and Hall (1973) elastic SDOF system, thereby leading to an equal
formulated the procedure for the construction of energy criterion between qm and mi (Fig. 3).
Behavior Factor and Ductility 267
Behavior Factor and Ductility, Table 1 Period dependent response reduction coefficients qm following the
Newmarka and Hall (1973) procedure
Soil class I II III
EDRS region D V A D V A D V A
mi = 1.5 1.62 1.70 1.44 1.58 1.66 1.50 1.78 1.94 1.47 B
mi = 2.0 2.28 2.21 1.76 2.16 2.19 1.83 2.46 2.70 1.79
mi = 5.0 5.63 4.59 2.84 5.04 4.63 2.91 5.82 5.96 2.85
They therefore proposed the indirect construction ductility mi, using the following period-dependent
of an IDRS from an EDRS, for a given target response reduction coefficients qm (Eq. 4):
0 T < T 1 =10 qm ¼ 1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi T 1 2:513log pffiffiffiffiffiffiffi
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T 1 =4 T < T 01 qm ¼ 2mi 1 (4)
T
T 01 T < T 1 qm ¼ mi
T1
T 1 T < T 2, T 2 T < 10 s q m ¼ mi
where T1, T0 1, and T2 are the characteristic corner Successive improvements of the indirect qm–m–T
periods of the EDRS, defined from the peak construction model were proposed by
ground response parameters (Newmark and Hall Newmark and colleagues (for a review of the
1973): literature, see Miranda and Bertero 1994)
(i) using additional ground motions and including
fev PGV f PGD the vertical excitation, (ii) for wider sets of criti-
T 1 ¼ 2p , T 2 ¼ 2p ed , cal damping ratios x (2–10 %) and target ductil-
fea PGA fev PGV
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (5) ities mi (1–10), and (iii) for different hysteretic
2mi 1
T 01 ¼ T 1 shapes f(u,t).
mi
Riddell (1995) extended the indirect method
where fea, fev, fed are the amplification factors of evaluating qm for different soil classification
defining the EDRS acceleration, velocity, and conditions, by considering 72 accelerograms
displacement from the corresponding PGA, which were obtained during the 1985 Chile earth-
PGV, and PGD for the site, defined by local quake, initially classified according to the
seismicity considerations. Thus, in a log–log for- prevailing soil conditions as records on rock (I),
mat, the Newmark and Hall qm–m–T relation firm (II), and medium stiffness (III) soil sites.
consisted of two constant qm regions (Eq. 4) Following statistical analysis, soil-dependent
followed by intermediate linear transitions, as corner frequencies and amplification values Cm
shown in Fig. 3. Based on a statistical parameter were proposed for obtaining the 5 % damping
evaluation, they recommended averaged rela- EDRS and the ductility-dependent IDRS, given
tions of PGV/PGA (48 in/s/g) and PGD/PGV2 the site PGA, PGV, and PGD, following the
(6 s/in.) and fea, fev, fed equal to 2.6, 1.9, and Newmark and Hall (1973) procedure. Based on
1.4, respectively, for the construction of the the amplification factors reported, typical qm,
EDRS at 5 % damping; different amplification evaluated as qm = Cm=1/Cm for mi = 1.5, 2.0,
values were given for stiff or soft soil or, in and 5.0, are given below for the constant dis-
addition to the average values above, for placement (D), velocity (V), and acceleration
average plus or minus one standard deviation. (A) regions of the EDRS (Table 1):
268 Behavior Factor and Ductility
Behavior Factor and Ductility, Fig. 4 Comparison of different qm–m–T models for evaluating the behavior factor, for
target ductilities of (a) mi = 4 and (b) mi = 6
model the hysteresis curve of the SDOF oscilla- degrading model. The qm–m–T expression pro-
tor, namely, an EPP and a bilinear stiffness posed is given in Eq. 6:
The authors concluded that, with the exception of influence of the hysteretic response, they
the short-period range, the stiffness degrading performed a statistical analysis of qm–m–T
models with no hardening exhibited a 20 % using 40 ground motion records taking into
higher behavior factor compared to the EPP sys- account five different SDOF oscillator hyster-
tem, with the effect diminishing for the 10 % etic shapes, namely, (i) EPP oscillator,
hardening slope EPP and degrading oscillators. (ii) bilinear hysteretic oscillator, (iii) degrading
Furthermore, in addition to the quantification of strength hysteretic oscillator, (iv) degrading
the oscillator bias and based on the attenuation stiffness oscillator (centered upon reversal),
studies and the evaluation of the corresponding and (v) pinching hysteretic oscillator (in all
qm for the subject datasets, they further cases initial damping of 5 % was assumed).
established that qm of SDOF systems for given They concluded that in the absence of the soil
mi is not sensitive to the epicentral distance of the bias, additional factors that affect the qm
earthquake and the ground motion; this observa- demand (given target ductility and period) are
tion, however, is considered in more detail the hardening slope and the form and extent of
further on. pinching of the oscillator cyclic characteristic,
Lee and Han (1999) developed direct expres- in a statistically independent manner to each
sions of qm for stiff soil and rock sites only, other. They subsequently proposed a qm–m–T
namely, soils exhibiting a shear-wave velocity functional in terms of the hysteretic parameters
larger than 750 m/s. In order to evaluate the a1, a2, a3, and a4 (Eq. 7):
270 Behavior Factor and Ductility
0:83
qm ¼ ð0:99 m þ 0:15Þ 1 e23:69Tm Ca1 Ca2 Ca3 Ca4 , with :
where a1 measures the hardening slope, a2 mea- conservative designs based on the qm–m–T rela-
sures the strength degradation upon reversal, a3 tions obtained on firm ground or alluvia proposed
measures the stiffness degradation, and a4 mea- by other studies, while for lower T/Ts values, the
sures the pinching response. opposite was observed, with the reduction factors
Influence of the local soil conditions. in this case being lower than mi (e.g., qmi=3 = 2 or
Miranda (1993) and Miranda and Bertero (1994) qmi=5 = 2.8 at T/Ts=0.5). Finally, for normalized
considered a wide dataset of ground motion exci- periods above Ts (T/Ts > 2), the behavior factor
tations which had been previously classified was close to the target ductility, as for stiff soils
according to the relevant earthquake parameters and rock, yielding similar or conservative designs
(magnitude, distance from the fault) and the local compared to the use of rock spectra.
soil conditions prevalent at the site. Statistical In order to investigate the influence of stiff-
equations of qm in terms of the local soil condi- ness degradation on soft soils, Miranda and Ruiz-
tions were subsequently obtained, proposing the Garcia (2002) extended this study by expressing
following qm–m–T evaluation function for differ- qm in terms of the structural period normalized
ent soil classifications (Eq. 8): by the predominant soil period Ts (namely,
qm–m–T/Ts), using a set of 116 ground motions
mi 1
qm ¼ 1 þ on soft-to-medium soils and two hysteretic sys-
FðT Þ
1 1 3ðlnT3=5Þ2 tems, a bilinear EPP and a stiffness degrading. It
Rock sites Fð T Þ ¼ 1 þ e2 was shown that the influence of stiffness degra-
10T mi T 2T
1 2 2ðlnT1=5Þ2 dation was significant for soft soil excitations,
Alluvium sites FðT Þ ¼ 1 þ e
12T mi T 5T since for period ratios lower than soil resonance,
T s 3T s 3ðlnðTTs Þ1=4Þ2 in the range of T/Ts~2/3 (also depending on mi),
Soft soil sites FðT, T s Þ ¼ 1 þ e
3T 4T the average qm of stiffness degrading systems was
(8) about 25 % lower than qm of EPP systems, leading
where, apart from the parameters defined already, to unconservative designs based on the qm–m–T
the dominant soil period Ts was introduced for the relations obtained on firm ground for EPP sys-
soft soil sites. The analysis showed that there is a tems. On the contrary, qm of stiffness degrading
distinct trend in the shape of the qm–m curves for systems at soil resonant periods and above was
(primarily soft) soils that exhibit a predominant higher than that for EPP systems, indicating that
period, whereby in the vicinity of the oscillator the earlier onset of nonlinearity in these struc-
resonance with the soil dominant period tures induced a drop in the required Ry for a
Ts (T/Ts~1), the behavior factor qm is much larger given target ductility compared to EPP structures.
than the target ductility (e.g., qmi=3 = 8.5 or Vidiç et al. (1994) expressed qm versus ductil-
qmi=5 = 18.0 was obtained); this effect led to ity relations in site-specific form, expressed in
Behavior Factor and Ductility 271
terms of the accelerogram’s frequency parameter Italy, Montenegro, and the 1985 Mexico City
T1, obtained at the period of change in slope of earthquake, analyzing SDOF oscillators with
the idealized bilinear approximation of the either EPP or stiffness degrading characteristics,
pseudovelocity response spectrum PSv(T, x) of yielding the more general qm–m–T exponential
the accelerogram. For their analyses they expression of Eq. 9 (a tangent stiffness propor- B
considered 24 records from California, Chile, tional damping equal to 5 % critical was used):
T
T < T0 q m ¼ c 1 ð m i 1 Þ cR
þ 1 1, 0
T0
cR
T0 T qm ¼ c1 ðmi 1Þ þ 1
where : T 0 ¼ c 2 m cT T 1 (9)
and : c1 c2 cR cT
stiffness degrading SDOF : 0, 75 1, 0 0, 65 0, 30
elastoplastic SDOF : 1, 10 0, 95 0, 75 0, 20
In the context of PBD, Fajfar (1999) proposed at a given qm (Eq. 10), by adapting the statistical
more simplified conservative design envelopes qm–m–T estimations by Vidiç et al. (1994) above
for the estimation of the required target ductility in the N2 method:
T
T < T0 q m ¼ ð mi 1 Þ þ 1 1, 0
T0
T0 T q m ¼ mi (10)
where : T 0 ¼ 0, 65 m0, 3 T c T c
or, in a simplified form : T0 ¼ Tc
This expression has been adopted with refinements expressions for spectral amplification factors C
(soil effect, softening or hardening oscillator char- (T, m) and therefore qm in terms of ductility and
acteristics) as the displacement coefficient method period were also proposed, as being more reliable
(DCM) for PBD (NEHRP 1985; FEMA-356 and less prone to statistical scatter than the former
2000), where qm is equal to the coefficient C1 in indirect method. The function proposed was of the
the target roof estimation equation. exponential form given in Eq. 11 using
Riddell (1995) examined the influence of local five statistical parameters (a1–a5) – although a
soil conditions on qm using 72 records obtained dominant soil period was not explicitly considered
during the Magnitude 7.8 1985 Chile earthquake (type IV soft soil records were not available).
and aftershocks on rock (I), hard (II), or medium
soil (III), as also previously described in the indi- Sa ðT, m ¼ 1Þ
qm ¼ , where
rect evaluation method for qm: following the indi- Sa ðT, m ¼ mi Þ
rect method analysis and recognizing that the a1 þ a2 T a 3
CðT, m ¼ mi Þ ¼
variation in qm was high, direct evaluation 1 þ a4 T a5
272 Behavior Factor and Ductility
Soil Type mi ¼ 2 mi ¼ 5
a1 a2 a3 a4 a5 a1 a2 a3 a4 a5
I 0, 45 5, 81 0, 66 18, 0 2, 04 0, 60 4, 75 0, 86 37, 3 2, 03 (11)
II 0, 88 7, 77 1, 39 24, 2 2, 79 0, 84 1, 41 1, 44 13, 6 2, 52
III 33, 96 35, 65 0, 005 1, 4 2, 62 0, 86 0, 06 2, 967 2, 303 2, 72
The parameters were obtained from includes local soil effects. They therefore pro-
regression analysis after relaxing, for the sake of posed a functional for qm in terms of the
improved statistical correlation, the requirement spectral displacement amplification Sd/PGD
that C(T,m) should be equal to 1.0 for rigid instead of T (Eq. 12) since, implicitly, it is
structures. a function of the period T. For their statistical
Ordaz and Pérez-Rocha (1998) considered analysis, they considered 445 records over a
the direct evaluation of qm as a force reduction wide range of magnitudes, epicentral distance,
parameter from the elastic relative displace- and local soft soil to rock conditions.
ment spectrum Sd, since both the inelastic
and elastic displacements of the SDOF system
bðmÞ
Sd ðT Þ
qm ¼ 1 þ ðmi 1Þ PGD , where
are related to the ductility and qm, the varia-
0, 173
tion of Sd with period being the controlling bðmÞ ¼ 0:388 ðm 1Þ
factor for the variation of qm, and also Sd
Soil Type mi ¼ 2 mi ¼ 4 mi ¼ 6
a b c a b c a b c
I 1, 29 2, 77 0, 0218 1, 12 2, 18 0, 0777 2, 35 1, 69 0, 0080 (12)
II 1, 12 2, 18 0, 0416 , 989 1, 62 0, 2037 1, 52 1, 05 0, 1334
III 2, 35 1, 69 0, 0418 1, 03 1, 24 0, 2707 1, 85 , 821 0, 1184
Their expression qm–m–Sd(T)/PGD yielded code classification. They showed that their pre-
similar results as the proposed direct qm–m–T dictions of the average qm approach reasonably
relations by other investigators (e.g., Nassar and well the equal energy and equal displacement
Krawinkler (1991) or Miranda (1993)), for either
the average
approximation (Fig. 3). However,
firm or soft soil, yet, unlike previous studies, is of minus one standard deviation sqm values differ
a more general applicability over the entire range
significantly from this simplification and on the
of soil types considered. unconservative side, since sqm increases dis-
Watanabe and Kawashima (2002) investi- proportionately, and therefore, the use of qm for
gated in detail the scattering of qm from the design may be unconservative, especially for
mean value as well as the influence of initial larger target ductilities. They then proposed an
damping under elastic and EPP inelastic exponential qm–m–T relation and the
response, using 70 free-field records from a rela- corresponding expression for sqm in Eq. 13:
tively large dataset from Japanese earthquakes:
ground motions were classified according to dif- qm ¼ 1 þ ðmi 1Þ CðT Þ, where
ferent soil conditions with predominant soil Ta 1
CðT Þ ¼ 1 þ c bðTaÞ , c ¼
period Tg, following the Japanese seismic design e ða eab Þ
Behavior Factor and Ductility 273
Soil Type mi ¼ 2 mi ¼ 4 mi ¼ 6
a b c a b c a b c
I T g < 0:2s 1, 29 2, 77 0, 0218 1, 12 2, 18 0, 0777 2, 35 1, 69 0, 0080 (13)
II T g < 0:6s 1, 12 2, 18 0, 0416 , 989 1, 62 0, 2037 1, 52 1, 05 0, 1334
III T g > 0:6s 2, 35 1, 69 0, 0418 1, 03 1, 24 0, 2707 1, 85 0, 821 0, 1184 B
and sqm
0:4 mi 0:3 ðall soil classesÞ
Additional soil-dependent qm–m–T functions hysteresis, (iii) a shear slip model, and (iv) a
were proposed by Genshu and Yongfeng (2007) bilinear elastic oscillator, in order to establish
with a range of application of mi between 1.0 and the influence of energy absorption. Three critical
6.0 and a period range between 0.1 and 6.0 s. damping ratios were considered, namely,
A dataset of 370 records was used, divided undamped, 3.5 % critical, and 5 % critical. Fol-
according to the prevailing soil conditions at the lowing a sensitivity study, the resulting qm–m–T
site and classified in accordance with the four spectra were correlated to the ratio of T/Tg, where
soil-type classifications adopted in the Chinese Tg is the characteristic ground motion period of
seismic code. Four different hysteretic oscillator each record, defined as the period at which qm was
characteristics were considered, namely, (i) an maximized. They therefore proposed expressions
EPP model, (ii) a bilinear with hardening (5 %, with T for the estimation of qm with 90 % confi-
10 %, and 20 % hardening stiffness) and pinching dence, qm,90 (Eq. 14):
8 h i
ð0, 6T =Tg Þ
>
> 1 þ 0:36 ð2:5 mi 2Þ0:75 1 TTg , 0 < T 0:6 T g
>
>
>
> h i
2
>
< 1 þ ð2:5 mi 2Þ0:75 1 TTg , 0:6 T g < T T g
qm, 90 ¼¼
>
> h i
T 12=mi
>
> 0:75 0:75
>
> ð2:5mi 2Þ ð2:5mi 2Þ 0:5 0:6mi Tw1
g
, T g < T wT g
>
:
0:5 þ 0:6 m , w Tg < T
Soil Type A B C D
(14)
w 1:5 2:1 2:5 3:0
where w is a coefficient depending on the soil studies reported has been (and the statistics
class. Their statistical analysis confirmed that proved it so) that the earthquake magnitude was
expressing qm in terms of the ratio T/Tg reduced not a governing parameter for the quantification
the statistical scatter, making the governing of qm, thereby the effort for parameter identifica-
parameter the target ductility; little influence of tion concentrated on the target ductility, the oscil-
the critical damping and the hysteretic shape, lator hysteretic shape, the damping, and,
apart from the short-period range, was observed primarily, the soil characteristics and its predom-
on the magnitude of qm. inant period.
Influence of seismological and ground Mavroeidis et al. (2004) investigated further
motion characteristics for near-field excita- the seismicity bias by considering the pulse
tions. So far, the overall consensus of the SDOF duration Tp of near-field ground motions and
274 Behavior Factor and Ductility
its effect on the strength demands for a given mi correlated to the earthquake moment magni-
of EPP systems. The pulse duration observed in tude Mw. Subsequent inelastic SDOF analyses
near-field records was shown on geophysical using near-field pulses demonstrated that Tp
(source mechanism) terms to be about twice was an important parameter for near-field
the rise time of the rupture process generating events, particularly through the establishment
the earthquake, namely, the time it takes for a of amplification factors along the lines of a
point on the fault to reach the maximum fault modified Eq. 4 (Newmark and Hall 1973), in
displacement, which, therefore, was well terms of T/Tp (Eq. 15):
8
>
> T T
>
> 1 <
>
> Tp Tp a
>
< pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
T 2mi 1 T
qm ¼ 2mi 1 Tp b < T <
T
>
> p
mi Tp c
>
>
>
> T T
>
: mi >
Tp Tp c
1200
V max failure capacity curve
1000 bilinear
approximation
Ω
m equivalent
B
800
system
Vu = b · e · W
V (KN)
m'
400
Ve = e · W
200
δy' δy δu δu'
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
δroof (m)
Behavior Factor and Ductility, Fig. 5 Evaluation of the overstrength, the ductility, and the behavior factor of an
existing RC structure (From Repapis et al. 2006a)
code or structural system under scrutiny. used for the corresponding design environment
Although most studies (especially related to over which they have been calibrated for.
code design parameter calibration and reliability
analysis) have been aimed toward the evaluation Evaluation of the Behavior Factor
of the overall design behavior factor qs, studies of Using SPO Analysis
specific structural systems (e.g., RC shear wall or The evaluation of qs using SPO methods involves
steel braced frame structures) have also been the estimation of the capacity curve of the struc-
concerned with the evaluation of the behavior ture under a prescribed lateral load and/or dis-
factor due to overstrength qO. placement profile (modal, triangular, constant,
Generally, the evaluation of qs of a new or an adaptive) and crucial modeling assumptions
existing structure is based on either static push- (among others) of the strength and stiffness
over (SPO) or time history inelastic analyses dependence of the member characteristics (also
under an earthquake excitation set. The resulting the finite element formulation itself), of the foun-
q is particular to the building type and structural dation and the joint models, including or not the
material under scrutiny, the EDRS, and the per- second-order effects at the global and local ele-
formance level criteria adopted, as well as the ment level. Following the definition of a set of
entire set of underlying code-specific design local and global failure criteria (the damage indi-
requirements inherent in the seismic design of ces), the available global ductility of the struc-
the structure. Such requirements include, among tural model is evaluated at the minimum roof
others, the design load level during earthquake, deformation at incipient satisfaction of any of
the relative magnitude of gravity and seismic these indices, over the onset of yield roof defor-
loads under the seismic load combination, limit- mation. Repapis et al. (2006a) have adopted SPO
ing section geometry or reinforcement require- procedures in order to evaluate the available qs
ments, the degree of conservatism in the design, and also qO of existing RC plane irregular frames,
the code prescribed detailing rules, and so typical of structural designs between the 1960s
on. Hence, these behavior factors can only be and 1990s in Greece (Fig. 5). Their results were
276 Behavior Factor and Ductility
compared to dynamic analysis estimates of qs, as (b) Direct evaluation of qs. A set of collapse
described in the next section. criteria is adopted together with an EDRS
(shape, PGA, and damping) and a set of base
Evaluation of the Behavior Factor Using excitation records matching the specific EDR-
Dynamic Analysis S. The structure is designed for the assumed
Several analytical estimations of qs based on EDRS in a repetitive manner, initially assum-
dynamic analysis have been published, in order ing a design qd as proposed by the code and,
to calibrate or verify the reliability of the behav- subsequently, for increasing (or possibly
ior factor as a static analysis parameter in force- decreasing) values of qid for each ith design
based seismic design, against local or global iteration; decreasing values may need to be
damage predictions. Two methods, the direct considered in irregular or non-ductile struc-
and the indirect method, have been proposed for tures in regions of high seismicity. For each
evaluating qs of a building using dynamic analy- ith design, an inelastic model of the structure is
sis (EC8 ENV 1988). Recently, in FEMA P-695 formulated and analyzed over the design
(2009), the procedure is formalized for a uniform earthquake set. This design – analysis
reliability of evaluating qs and qO in PBD. procedure – is repeated until any of the preset
collapse criteria is incipiently violated: this
(a) Indirect evaluation of qs. A building ini- limiting qid is the behavior factor qs.
tially designed for a given EDRS (shape,
PGA equal to Ad, critical damping, and a qs ¼ maxðor minÞqid (17)
design behavior factor qd) is analyzed in the
time domain under a base excitation whose In this case, a direct redesign method is
response spectrum matches the EDRS adopted for qs at the expense of several structural
entirely or locally in the vicinity of the initial (re)designs. As before, the resulting behavior
period of the structure for the same critical factor is a function of the design assumptions as
damping. The PGA is linearly increased until well as the performance level evaluation criteria
incipient violation of any of a set of failure and the design record set.
criteria, at the nominal collapse value of Ac. Kappos (1991) applied the direct method for
The behavior factor qs is equal to the evaluation of qs for two typical RC plane
frames, namely, a moment frame and a dual
Ac frame-wall system, designed according to EC8
qs ¼ qd (16)
Ad ENV (1988) for medium ductility class using a
qd of 3.50 and 2.10, respectively. Using perfor-
The process is repeated for different earth- mance limits of total and interstory drift, member
quake input sets, and a statistical estimate of shear capacity, and local curvature ductility m’,
qs is established for this type of building, the resulting available qs were evaluated to be 4.9
design procedure, and performance criteria. and 3.4 of the bare frame and the dual system,
The method suffers from the drawback that respectively. Different criteria governed for each
the performance indices do not depend line- structure and base excitation: for the bare frame
arly on the PGA (Kappos 1991); however, it system, the story drift limit governed, while for
has been commonly used since qs is obtained the dual system, member failure governed the
at the expense of only one initial structural response (the ground story columns). Zeris
design and model formulation. This method et al. (1992) applied the direct method to
falls closely with the application of the incre- medium-height RC frames with a relatively taller
mental dynamic analysis method (IDA) pro- ground story designed according to EC8 ENV
posed by Vamvatsikos and Cornell (2003) for (1988) for ductility class II and a qd of 3.50,
evaluating existing structural systems. observing design code limitations (e.g., member
Behavior Factor and Ductility 277
Behavior Factor and Ductility, Fig. 6 Variation of the structural behavior factor qs of a tall first-story building with
base excitation amplitude (Zeris et al. 1992)
Behavior Factor and Ductility, Table 2 Comparison of building qs using static and dynamic analysis for three
excitations (minimum value, mean, and standard deviation) (Zeris et al. 2005)
Frame building type q (SPO) q (IDA)
q qmin sq
Nonconforming ordinary frame, Zone I 2.03 2.18 1.92 0.33
Nonconforming ordinary frame, Zone II 1.77 3.15 2.43 1.25
Nonconforming frame with tall first story 2.34 2.82 2.41 0.70
Nonconforming frame with a recess 1.55 2.78 2.36 0.39
Nonconforming frame with discontinuous column 1.98 4.31 2.87 2.48
Nonconforming frame with discontinuous beam 2.14 2.12 1.81 0.43
Conforming ordinary frame, Zone I 5.44 6.78 5.22 1.84
factors are well below currently adopted and earthquake event happens. Consequently, until
(ii) dynamic analysis predictions of qs for irregu- sufficient seismic data are available, structures
lar buildings tend to be more conservative will continue to be constructed in regions in
than those using dynamic analysis, being closer which, historically or otherwise, a fault was not
to the minimum predictions obtained in the known to exist or the existing faults’ ability to
latter case. induce a given seismic magnitude has been
Chryssanthopoulos et al. (2000) evaluated the underestimated. Such famous examples, among
reliability of the design behavior factor qd others, were (i) the unexpected 1983 Coalinga
adopted by the Eurocode for the case of a regular earthquake in a quiet region in California;
ten-story RC frame designed for ductility class (ii) the 1971 Sylmar earthquake with its surpris-
M. Using the indirect method of evaluation of the ing (at the time) Pacoima Dam record exceeding
available qs, they assumed a statistical variation 1.0 g; (iii) the numerous strong motion records
of the structural material properties (concrete and since then that have well exceeded 1.0 g in Chile,
steel yield strength and ultimate failure strain of New Zealand, Taiwan, and elsewhere; and
the reinforcement), the uncertainty in confine- (iv) for Greece, the 1999 Athens earthquake that
ment (and therefore supplied m’ in the critical ruptured an unknown extension of a known fault
regions), and a set of spectrum compatible base in a low- to medium-seismicity zone with a
excitation records, in order to evaluate the reli- design effective PGA of 0.16 g, yielding esti-
ability of ULS design following EC8 recommen- mated PGAs of 0.5 g, and the 2014 earthquakes
dations for the subject building. At the same time, in a well-known seismically active region in
they established the variation of the available qs Kefallinia, exceeding the design effective PGA
of the subject frame under the base input set. by as high as 200 %. Such consistent overshoots
They concluded that for a seismic hazard model of our impression of the expected level of shaking
with a return period of 475 years for Southern in a seismically affected region demonstrate the
Europe, the violation of the initial design value need for introducing to our designs reasonable
(qd) of 3.75 was to be expected in 0.6 % of the reduction factors coupled with excess ductility
frames exhibiting this type of building and design capacity and structural redundancy, rather than
characteristics and the material property variabil- adopting qd = 1.0-type designs with no built-in
ity assumed. ductility.
Extensive statistical studies have demon-
strated that the q(or R) factors of SDOF systems
Summary and Conclusions are not sensitive to the earthquake magnitude
unless near the causative fault, while they show
It should be noted at the outset that such an a reasonable similarity (in terms of the estimated
intentional underdesign approach (i.e., using a qm–m–T relations) between earthquake faults of
behavior factor) is by no means a socially irre- similar tectonic characteristics. Furthermore,
sponsible approach from the point of view of the they do not seem to be sensitive to the hysteretic
engineers’ role to the society. It is well accepted shape of the system, for EPP and hardening sys-
that economic limitations and design feasibility tems of conventional new buildings. They are
will define the intensity level that is socially sensitive, however, to softening cyclic response,
acceptable for the design of structures for the such as the case of structures with excessively
extreme earthquake; furthermore, it is well softening or brittle failure mechanisms, such as
known that the gradual increase of records existing RC frame or frame-wall structures,
through more dense earthquake monitoring buildings with strong P-d effects, or infilled RC
grids of digital instruments has proven more and frames with weak infills. They also depend
more often that the actual seismic load intensity strongly on the EDRS (de)amplification in the
at a site will surprise us, when an extreme case of a resonant (or not) structural period near
Behavior Factor and Ductility 279
Kappos A (1999) Evaluation of behaviour factors on the earthquake motions. In: Proceedings second world
basis of ductility and overstrength studies. Eng Struct conference on earthquake engineering, Japan,
21(9):823–835 pp 895–912
Lee LH, Han SW (1999) Determination of ductility factor Vidiç T, Fajfar P, Fischinger M (1994) Consistent inelas-
considering different hysteretic models. Earthq Eng tic design spectra: strength and displacement. Earthq
Struct Dyn 28(9):957–977 Eng Struct Dyn 23(5):507–521
Mahin SA, Bertero VVB (1981) An evaluation of inelastic Watanabe G, Kawashima K (2002) An evaluation of the
seismic design spectra. J Struct Div ASCE forced reduction factor in the force-based seismic
107(9):1777–1795 design. In: Proceedings of 39th joint meeting, panel
Mavroeidis GP, Dong G, Papageorgiou AS (2004) Near- on wind and seismic effect, UJNR, NIST Special Pub-
fault ground motions, and the response of elastic and lication, Gaithersburg
inelastic single-degree-of-freedom (SDOF) systems. Whittaker A, Hart G, Rojahn C (1999) Seismic response
Earthq Eng Struct Dyn 33(9):1023–1049 modification factors. J Struct Eng 125(4):438–444
Miranda E (1993) Site-dependent strength reduction fac- Zeris C, Tassios TP, Lu Y, Zhang GF (1992) Influence of
tors. J Struct Eng ASCE 119(12):3503–3519 irregularity on the q factor of RC frames. In: Proceed-
Miranda E, Bertero VVB (1994) Evaluation of strength ings, 10th world conference of earthquake engineer-
reduction factors for earthquake-resistant design. ing, Madrid
Spectra 10(2):357–379 Zeris C, Repapis K, Vintzeleou E (2005) Seismic perfor-
Miranda E, Ruiz-Garcia J (2002) Influence of stiffness mance of existing irregular RC buildings. Paper
degradation on strength demands of structures built No. 32, Proceedings of the 4th European workshop
on soft soil sites. Eng Struct 24(10):1271–1281 on the seismic behaviour of irregular and complex
Nassar AA, Krawinkler H (1991) Seismic demands for structures, Thessaloniki
SDOF and MDOF systems, TR 95, The J. Blume
Earthquake Engineering Center, Stanford University,
Palo Alto
NEHRP (1985) NEHRP Recommended provisions
for the development of seismic regulations for
new buildings. Building Seismic Safety Council, Blind Identification of Output-Only
Washington, DC Systems and Structural Damage
Newmark NM, Hall WJ (1973) Seismic design criteria for via Sparse Representations
nuclear reactor facilities. Report No. 46, Building
Practices for Disaster Mitigation, National Bureau of
Standards, US Department of Commerce, pp 209–236 Satish Nagarajaiah1,2 and Yongchao Yang1
1
Ordaz M and Pérez-Rocha LE (1998) Estimation of Department of Civil and Environmental
strength-reduction factors for elastoplastic systems: a Engineering, Rice University, Houston, TX, USA
new approach. Earthq Eng Struct Dyn 27(9):889–901 2
Repapis K, Vintzeleou E, Zeris C (2006a) Evaluation of
Department of Mechanical Engineering, Rice
the seismic performance of existing RC buildings: University, Houston, TX, USA
I suggested methodology. Eur J Earthq Eng
10(2):265–288
Repapis K, Zeris C, Vintzeleou E (2006b) Evaluation of
the seismic performance of existing RC buildings: II a
Synonyms
case study for regular and vertically irregular build-
ings. Eur J Earthq Eng 10(3):429–452 Blind source separation; Modal identification;
Riddell R (1995) Inelastic design spectra accounting for Sparse representation; Structural health monitor-
soil conditions. Earthq Eng Struct Dyn
24(11):1491–1510
ing; Time-frequency analysis
SEAOC (1974) Recommended lateral force requirements
and commentary, 4th edn. Struct. Engineers Assoc. of
California, San Francisco Introduction
Uang CM (1991) Establishing R (or Rw) and Cd factors for
building seismic provisions. J Struct Eng ASCE
117(1):19–28 Traditional model-based structural modal analy-
UBC (1988) Uniform building code. International Confer- sis and damage identification methods are typi-
ence of Building Officials, Washington, DC cally parametric and user involved; as such, they
Vamvatsikos D, Cornell CA (2003) Applied incremental
dynamic analysis. Earthquake Spectra 20(2):525–533
are usually associated with demanding computa-
Veletsos A, Newmark NM (1960) Effect of inelastic tional resources and require quite a lot of
behaviour on the response of simple systems to prior knowledge of structures. For practical
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations 281
applications, it would be useful to seek efficient BSS technique called independent component
structural identification methods that may be able analysis (ICA), as detailed in the following.
to extract the salient information directly from ICA is popularly used to estimate the blind
the measured structural signals. The recently source separation (BSS) model (Hyvärinen and
widely deployed advanced structural health mon- Oja 2000), B
itoring (SHM) systems in structures with dense
sensors also support such an effort: the massive X
n
xðtÞ ¼ AsðtÞ ¼ ai s i ð t Þ (1)
recorded data especially call for efficient data- i¼1
driven algorithms (Yang and Nagarajaiah
2014c, d) for further structural assessment. using only the observed mixture vector
Recently, blind source separation (BSS) has xðtÞ ¼ ½x1 ðtÞ, x2 ðtÞ, . . . , xm ðtÞT ; sðtÞ ¼ ½s1 ðtÞ,
emerged as a new unsupervised machine learning
s2 ðtÞ, . . . , sn ðtÞT and A denote the latent source
tool (Hyvärinen and Oja 2000) and has been
vector and the unknown constant m n linear
extensively studied in structural dynamics and
mixing matrix, respectively, to be simultaneously
output-only modal identification (Antoni 2005;
estimated. ai is the ith column of A and is associ-
Kerschen et al. 2007; Yang and Nagarajaiah
ated with the corresponding source si(t). The
2013a, b, c, 2014a, b; Poncelet et al. 2007; Zhou
assumption of m ¼ n is imposed herein, i.e., the
and Chelidze 2007; McNeill and Zimmerman
number of mixtures equals that of the sources and
2008; Hazra et al. 2010; Hazra and Narasimhan
A is square. With only x(t) known, Eq. 1 may not
2010; Sadhu et al. 2011, 2012; Abazarsa
be mathematically solved by classical methods;
et al. 2013; Antoni and Chuahan 2013; Ghahari
additional assumption is thus needed to estimate
et al. 2013). Essentially, BSS techniques are able
the BSS model.
to recover the hidden source signals and their
The principle of ICA estimation is based on
underlying factors using only the observed mix-
the classical central limit theorem (CLT), which
tures; it is thus suitable to perform output-only
states that a sum of independent random variables
structural identification when structural input or
tends to distribute toward Gaussian, i.e., a mix-
excitation is usually extremely difficult or expen-
ture of independent random variables is always
sive to obtain.
more Gaussian than any one of the original vari-
This entry presents the authors’ recent work
ables (except that the mixture only contains one
on data-driven output-only modal identification
random variable). As seen in Eq. 1, mixtures
and damage detection of structures. It is found
are expressed as a weighted sum of
that exploiting the sparse essences of modal
the sources themselves; they are thus always
expansion and damage information can effi-
more or equally Gaussian than the sources. ICA
ciently and effectively address some challenging
therefore searches for proper demixing matrix
problems in output-only modal identification
W such that the recovered independent compo-
and damage detection via BSS. A series of
nents (ICs) yðtÞ ¼ ½y1 ðtÞ, y2 ðtÞ, . . . , yn ðtÞT
novel algorithms are developed with experimen-
obtained by
tal and real-world structure examples for
demonstrations.
yðtÞ ¼ WxðtÞ (2)
The sparse damage features hidden in the struc- with wi denoting the ith row of W. By seeking
tural information can be blindly extracted via a those ICs which maximize non-Gaussianity, the
282 Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations
sources (and simultaneously the mixing matrix) efficient algorithms implementing ICA estima-
can therefore be recovered by ICA. tion and is adopted in this study.
Non-Gaussianity of a random variable can be Damage may behave as pulse-like information
measured by some contrast function, e.g., hidden in the structural vibration response signals
negentropy. The
entropy of a discrete random once they are processed further such as in the
variable u ¼ u1 , u2 , . . . , ui , . . . is defined by wavelet domain. The pulse-like feature containing
damage information may be buried in the noisy
X
HðuÞ ¼ p u ¼ ui logp u ¼ ui (4) wavelet-domain signals on a certain scale. Because
i ICA biases to extract sparse components from the
observations, feed the wavelet-domain responses
where pðÞ is the probability mass operator. xl(t) at the lth scale as mixtures into the BSS model,
Entropy measures the uncertainty or randomness
of a random variable. For example, for a random xl ðtÞ ¼ AsðtÞ (7)
variable with impulse probability mass function,
its entropy is zero, i.e., it is completely determined. If there is any pulse-like feature hidden in xl(t),
The Gaussian random variable has the largest then ICA will extract it, which is to be revealed in
entropy among all other random variables with the recovered sparse component yj(t) with sharp
equal variance (Hyvärinen and Oja 2000), i.e., it spikes indicating damage
is the most random or uncertain one. On the other
hand, a random variable with sparse representa-
tion has small entropy as it is less random or yðtÞ ¼ Wxl ðtÞ (8)
easier to be predicted. This conclusion yields
the definition of negentropy as a measure of Such yj(t) is proposed as the “interesting” source
non-Gaussianity given by within the damage identification framework.
Note that xl(t) inherits the temporal information
JðuÞ ¼ H ugau HðuÞ (5) of the responses; this implies that the recovered
“interesting” source yj(t) retains temporal signa-
in which ugau is a standardized Gaussian random tures of the inflicted damage, which is indicated
variable (zero mean and unit variance); it quanti- by the time instant location of the sharp spike.
tatively evaluates the entropy distance of a Expanding the WT-BSS model Eq. 7 as
(standardized) random variable from a Gaussian
X
n
variable. Finding the ICs that maximize the xl ðtÞ ¼ ai si ðtÞ
negentropy by ICA thus yields random variables i¼1
with sparse representation (Yang and X
n
xl1 ¼ a1i si ðtÞ ¼ a11 s1 ðtÞ þ a12 s2 ðtÞ þ . . . þ a1n sn ðtÞ
Nagarajaiah 2014a). This finding turns out very i¼1
useful for damage identification, as subsequently Xn
described. xlj ¼ aji si ðtÞ ¼ aj1 s1 ðtÞ þ aj2 s2 ðtÞ þ . . . þ ajn sn ðtÞ
i¼1
A simplified approximation to the negentropy X n
is the classical kurtosis, which is defined by xln ¼ ani si ðtÞ ¼ an1 s1 ðtÞ þ an2 s2 ðtÞ þ . . . þ ann sn ðtÞ
i¼1
2 (9)
kurtðuÞ ¼ E u4 3 E u2 (6)
Observe that the mixing coefficient aji locates the
where E½ denotes the expectation operator. The mixture and the source by its indices j and i,
kurtosis of a Gaussian random variable is zero, respectively, and the columnwise vector
T
and that of a non-Gaussian random variable is ai ¼ a1i , a2i , . . . , aji , . . . , ani contains the spa-
nonzero. It is easy to estimate and computation- tial signature of the corresponding source si(t).
ally efficient. The FastICA is one of the most Herein, ai and its element aji are proposed as the
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations 283
Blind Identification of Output-Only Systems and experimental structure with four sensors embedded sub-
Structural Damage via Sparse Representations, ject to white noise excitation at the base and (b) the
Fig. 1 Sparse ICA simultaneous identification of both WT-ICA identification results
structural damage instant and damage location: (a) the
source distribution vector (SDV) and source dis- The proposed concepts of SDV and SDF are
tribution factor (SDF), respectively; they readily extended to damage localization issues.
describe how source si(t) is distributed among As the structural response in the vicinity of dam-
n mixtures. Specifically, if aji has the largest age naturally contains most spike-like features,
(absolute) value among ai ¼ ½a1i , a2i , . . . ; damage can be localized by tracking the spike in
T
aji , . . . , ani , then xj(t) contains most si(t) compo- the SDV of the recovered impulse-like IC, which
nents among all the mixtures (Yang and is the “interesting” source. Figure 1 shows an
Nagarajaiah 2014a). experimental structure application of WT-ICA
284 Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations
~ f t ¼ WX
Q ~ ft (13) B
q ~ ðtÞ
~ ðtÞ ¼ Wx (14)
whereby readily estimating the modal frequen- Blind Identification of Output-Only Systems and
cies and damping ratios in free vibration are Structural Damage via Sparse Representations,
readily estimated by Fourier transform (FT) and Fig. 2 The experimental highly-damped 3-story steel
Hilbert transform (HT) or logarithm-decrement frame and the zoomed fluid damper
technique (LT) (Nagarajaiah and Basu 2009;
Basu et al. 2008; Nagarajaiah and Li 2004;
Dharap et al. 2006; Koh et al. 2005a, b; Li
(system response) always lies between that of
et al. 2007).
the simplest source (simplest modal response)
Figures 2, 3, 4, and 5 as well as Tables 1 and 2
and most complicated source (most complex
present the successful application of STFT-ICA
modal response).
in identification of the modal parameters of an
In statistics, the complexity of a signal, say,
experimental structure (Fig. 2) and a real-world
yi ¼ wi x (the temporal index is made implicit), is
structure subject to the Northridge Earthquake
rigorously measured by Kolmogorov complexity.
1994 (Nagarajaiah and Dharap 2003;
Given that Kolmogorov complexity is not intui-
Nagarajaiah and Sun 2000, 2001), both of which
tive and difficult to approximate in practice,
are highly damped. The details are referred to
Stone (2001, 2004) provided a simple yet robust
Yang and Nagarajaiah (2013a).
complexity measure of a signal, temporal predict-
ability, which is defined by
Complexity Pursuit on “Independent”
Modal Coordinates X
N
ðyi ðtÞ yi ðtÞÞ2
Vðyi Þ
Exploiting the properties of the system responses Fðyi Þ ¼ log ¼ log t¼1
Uðyi Þ X N
and modal responses can lead to an efficient ðy^i ðtÞ yi ðtÞÞ2
output-only modal identification method (Yang t¼1
Power 1
Accel. 1
via Sparse
Representations, 0 2
Fig. 3 (a) System
responses (free vibration)
of the experimental −0.05 0
0 0.5 1 1.5 2 0 5 10 15 20
structure (b) the recovered
modal responses by 0.02 0.015
STFT-ICA
Power 2
Accel. 2
0 0.01
−0.02 0.005
−0.04 0
0 0.5 1 1.5 2 0 5 10 15 20
0.02 0.02
Accel. 3
Power 3
0
0.01
−0.02
−0.04 0
0 0.5 1 1.5 2 0 5 10 15 20
Time (sec) Frequency (Hz)
Power 1
0 2
−1 1
−2 0
0 0.5 1 1.5 2 0 5 10 15 20
0.5 10
Modal Resp. 2
Power 2
0
5
−0.5
−1 0
0 0.5 1 1.5 2 0 5 10 15 20
0.5 6
Modal Resp. 3
Power 3
4
0
2
−0.5 0
0 0.5 1 1.5 2 0 5 10 15 20
Time (sec) Frequency (Hz)
Roof
B
6 @ 14.5’
6 th
15 23
4 th
2 nd 13 21
15’ 15’
Ground
Lower Level
Foundation Level
North Elevation
214’ 253’ 14 22
27 26
76.5’ 50’ 126.5’ 4th Floor Plan Roof Plan
5’
25 8 Lead-Rubber 12
Free
Isolators
20
Field
126.5’
Nref
303’
25’
1 11 19
3 7 2
4
5 9 17
Elastomeric Structure Reference
151.5’
Isolators Orientation: N = 5°
6 10 18
Foundation Level Lower Level 6th Floor Plan
(Above Isolators)
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations,
Fig. 4 Sensor outline of the highly-damped USC hospital building
Power 1
Accel. 1
via Sparse
Representations, 0 5000
Fig. 5 (a) System
responses (during the −200 0
Northridge Earthquake 0 10 20 0 10 20
1994) of the USC hospital
200 16000
building (b) the recovered
Power 2
modal responses by STFT-
Accel. 2
ICA 0 8000
−200 0
0 10 20 0 10 20
500 16000
Power 3
Accel. 3
0 8000
−500 0
0 10 20 0 10 20
Time (sec) Frequency (Hz)
b Recovered Modal
Responses PSD
1 0.1
Power 1
Resp. 1
0 0.05
−1 0
0 10 20 0 10 20
5 2
Power 2
Resp. 2
0 1
−5 0
0 10 20 0 10 20
1 1
Power 3
Resp. 3
0 0.5
−1 0
0 10 20 0 10 20
Time (sec) Frequency (Hz)
predictability using the gradient ascent tech- The gradient of F reaches zero in the solution,
nique. Stone (2001) proposed a more elegant where
algorithm that can efficiently extract all the hid-
den sources simultaneously, described as 2wi 2wi ^
∇wi F ¼ R R¼0 (21)
follows. Vi Ui
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations 289
Blind Identification of
Output-Only Systems
and Structural Damage
via Sparse
Representations,
Fig. 6 (a) The system
responses of a 3-DOF
numerical model and (b)
the modal responses
recovered by CP in free
vibration (closely-spaced
coupled with high-damped
modes)
basis cos 2pft in Eq. 24. The DCT is adopted in qj ðj ¼ 1, . . . , nÞ is active and qi ¼ 0 for i 6¼ j ,
the proposed SCA method, then x(f) and q(f) are Eq. 24 becomes
understood as real-valued cosine transform
coefficients. xð f k Þ ¼ ’ i qj ð f k Þ (26)
Using the disjoint sparsity property of
modal responses with distinct frequencies, at Therefore, the points of x(f) will cluster to the
some fk where only one modal response direction of the jth modeshape ’j ðj ¼ 1, . . . , nÞ
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations 291
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations, Fig. 7 (a)
The experimental 3-story structure, (b) the modeshapes estimated by CP, and (c) the modeshapes estimated by SCA
e
such that the estimated vibration mode matrix F X
n
can be extracted by the automatic fuzzy C-means where kqðf Þk‘1 ¼ jqi ðf Þ. The validity of this
i¼1
clustering algorithm.
e is square, and the strategy resides in the ability of the ‘1-
For determined cases, F
minimization to recover the sparsest solution to
time-domain modal responses can be recovered
Eq. 28, which is exactly the desired monotone
directly by
frequency-domain modal responses since they
are the sparsest solution among all feasible solu-
e 1 xðtÞ
~ ðtÞ ¼ F
q (27) tions to Eq. 28.
Using the inverse cosine transform, the
time-domain modal responses can be readily
from which the frequency and damping ratio can recovered by
be estimated by straightforward Fourier trans-
form and logarithm decrement, respectively. In
underdetermined cases, F e is rectangular. The ~ ðtÞ ¼ F c1 ðq~ ðf ÞÞ
q (29)
frequency-domain modal sources q~ ðf Þ can first
be recovered using ‘1-minimization (P1): at each
f O, thereby estimating the frequency and damping
ratio.
q e ðf Þ
~ ðf Þ ¼ arg minkqðf Þk‘1 subject to Fq For demonstrations, Fig. 10 shows that
SCA is able to recover the close modes
¼ xð f Þ coupled with high damping using only two
(28) sensors of the 3-DOF numerical model.
292 Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations
Blind Identification of
Output-Only Systems
and Structural Damage
via Sparse
Representations,
Fig. 8 (a) The measured
system responses and (b)
the modal responses
recovered by CP of the
experimental model subject
to white noise excitation
Blind Identification of
Output-Only Systems
and Structural Damage
via Sparse
Representations,
Fig. 9 (a) The measured B
system responses and (b)
the modal responses
recovered by CP of the
USC hospital building in
the Northridge
earthquake 1994
Blind Identification of Output-Only Systems and Blind Identification of Output-Only Systems and
Structural Damage via Sparse Representations, Structural Damage via Sparse Representations,
Table 3 CP identified results of the experimental model Table 4 CP identified results of the USC building
Frequency (Hz) Frequency (Hz)
Mode Peak picking CP MAC Mode Analytical CP MAC
1 2.550 2.600 1.0000 1 0.746 0.768 0.9751
2 7.330 7.395 0.9993 2 1.786 1.907 0.9054
3 10.460 10.720 0.9997 3 3.704 3.941 0.7874
294 Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations
Blind Identification of
Output-Only Systems
and Structural Damage
via Sparse
Representations,
Fig. 10 The modal
responses recovered by
SCA using only two sensors
(close modes coupled with
high damping case)
effectively address some challenging prob- both damage time instants and damage loca-
lems in output-only modal identification and tions and then further employed to exploit the
damage detection via the unsupervised blind sparse nature of modal expansion to handle
source separation (BSS) method. Sparse ICA the problem of identification of highly
is first introduced to simultaneously identify damped structures. What is more, two new
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations 295
Blind Identification of
Output-Only Systems
and Structural Damage
via Sparse
Representations,
Fig. 11 The free-vibration B
system responses of the 3-
story experimental model
(Fig. 7a)
Blind Identification of Output-Only Systems and system responses of (a) all three sensors; (b) Sensor
Structural Damage via Sparse Representations, 1 and 2 in the 3-story experimental model (shown in
Fig. 12 The scatter plot in frequency domain of the Fig. 7a)
output-only modal identification methods are sensors. The interpretations of CP and SCA
presented: the time-domain CP method and in output-only modal identification are
the SCA method which can handle the presented, and the successful implementations
underdetermined problems with limited are also demonstrated using both
296 Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations
Blind Identification of
Output-Only Systems
and Structural Damage
via Sparse
Representations,
Fig. 13 The modal
responses recovered by
SCA (a) using all the three
sensors and (b) using
Sensor 1 and 2 of the
experimental model
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations,
Table 5 Identified modal parameters by SCA of the experimental system
Frequency (Hz) Damping ratio (%) MAC
Mode PP Determined Under- Determined Under- Determined Under-
1 2.55 2.66 2.66 1.12 1.04 0.9997 0.9995
2 7.33 7.51 7.51 1.08 1.03 0.9997 0.9996
3 10.46 10.80 10.80 0.68 0.67 0.9988 1.0000
experimental and real-world structure exam- efficient computation and little user involve-
ples. The established data-driven output-only ment, which is expected to have potential for
modal identification and damage identification real-time structural identification and health
framework is nonparametric and enjoys monitoring.
Blind Identification of Output-Only Systems and Structural Damage via Sparse Representations 297
Yang Y, Nagarajaiah S (2013c) Output-only modal iden- earthquakes with magnitudes greater than seven.
tification with limited sensors using sparse component Examples include the 1964 M9.2 Alaskan Earth-
analysis. J Sound Vib 332:4741–4765
Yang Y, Nagarajaiah S (2014a) Blind identification of quake, the 1991 M7.6 Costa Rica Earthquake, the
damage in time-varying system using independent 1999 M7.8 Izmit-Turkey Earthquake, the 2010
component analysis with wavelet transform. Mech M8.8 Chile Earthquake, the 2011 M9.0 Tohoku-
Syst Signal Process 47(1):3–20 Oki Earthquake in Japan, and the 2012 M7.4
Yang Y, Nagarajaiah S (2014b) Structural damage identi-
fication via a combination of blind feature extraction Guatemala Earthquake. In all these earthquakes,
and sparse representation classification. Mech Syst spread footings failures were due to loss of bear-
Signal Process 45(1):1–23 ing capacity and lateral spreading due to soil
Yang Y, Nagarajaiah S (2014c) Blind denoising of struc- liquefaction. Figure 1 depicts photo of a col-
tural responses with outliers via principal component
pursuit. Struct Control Health Monit 21(6):962–978 lapsed bridge during the Limon Province, Costa
Yang Y, Nagarajaiah S (2014d) Data compression of Rica Earthquake. The piers have disappeared in
structural seismic responses using principled indepen- the river a result of liquefaction and both bridge
dent component analysis. ASCE J Struct Eng spans have collapsed during the earthquake.
140(7):04014032
Zhou W, Chelidze D (2007) Blind source separation based The seismic response of bridge foundations
vibration mode identification. Mech Syst Signal Pro- depends on: the type, geometry, and embedment
cess 21(8):3072–3087 of the foundation; the characteristics of the sur-
rounding soil; and the interaction between the
foundation and the surrounding soil.
Soil–structure interaction is the influence of the
Bridge Foundations soil and structure both on each other simulta-
neously. Seismic ground motions that are not
Ayman A. Shama1 and Hesham El Naggar2 influenced by the presence of the structure are
1
Ammann & Whitney, New York, NY, USA known as free-field motions. If the structure
2
Geotechnical Research Centre, Department of exists, the deformations of the foundations devi-
Civil and Environmental Engineering, ate from the deformations of the free-field motion
The University of Western Ontario, London, and the dynamic response of the structure induces
ON, Canada deformation of the supporting soil. The process in
which the response of the soil influences the
motion of the structure and the motion of the
Introduction structure influences the response of the soil is
known as soil–structure interaction (SSI). There
The poor performance of bridges during the 1971 are two approaches used to incorporate the effect
San Fernando earthquake has urged the engineer- of SSI in seismic analyses of structures, direct
ing community to further investigate methods of methods and simplified methods. In the direct
analysis and evaluation of seismic loads and their approach, the entire soil-foundation-structure
effects on these structures. Later, earthquakes system is modeled and analyzed simultaneously
such as Northridge, 1994, and Kobe, 1995, have in one step using finite element or finite differ-
further increased engineers’ awareness of the ence methods with the seismic waves applied at
effects of earthquakes on bridges. The past two the model boundaries. This approach requires
decades have witnessed a great evolution in seis- specific programs that can idealize the nonlinear
mic design codes, advance modeling, and analy- behavior of the soil and large CPU time for anal-
sis. Moreover, the importance of the interaction ysis. Although it is conceptually very attractive, it
between the structure and the surrounding soil on is fairly complex and very seldom been applied
the seismic response of bridge structures during for practical bridge problems. Simplified
earthquakes has been fully realized. methods include the substructure approach and
Past performance of foundations was satisfac- the Winkler approach. The substructure approach
tory in moderate earthquakes and poor in large includes subdividing the structure into two
Bridge Foundations 299
Bridge Foundations,
Fig. 1 A bridge collapse
due to foundation failure
Costa Rica
Earthquake 1993
B
or j
a0 ¼ (2)
Vs
of 4 is taken between the upper and lower bound
(ATC-1995).
The dynamic impedance model is based on in which, o is the circular frequency of excitation
earlier studies on machine foundation vibrations, in rad/s, rj is equivalent radius for a specific
in which, it is assumed that the response of rigid degree of freedom, and VS is the shear wave
foundations excited by harmonic external forces velocity of the soil medium. Both the dynamic
can be characterized by the impedance or stiffness kj and the damping cj are obtained in
dynamic stiffness matrix for the foundation. The terms of the static stiffness Kj as:
impedance matrix depends on the frequency of
excitation, the geometry of foundation and the K j rj
kj ¼ aj K j and cj ¼ bj (3)
properties of the underlying soil deposit. VS
The evaluation of the impedance functions for a
foundation with an arbitrary shape has been solved where aj and bj are dynamic modifiers are evalu-
mathematically using a mixed boundary-value ated as function of the dimensionless frequency
problem approach or discrete variation problem a0 and Poisson’s ratio n.
approach. Currently, there are two commonly Evaluation of the static stiffness coefficients
used approaches for evaluating the dynamic can be obtained using the equivalent circular
impedance functions for a shallow foundation. footing, wherein it is assumed that the footing is
The first is based on the approximate solution for rigid and founded on top of a semi-infinite elastic
a circular footing rigidly connected to the surface half space. The rectangular footing is converted
of isotropic homogeneous elastic half space. The to an equivalent circular footing for each degree
second approach is more general and is applicable of freedom. The stiffness coefficients are evalu-
to a foundation with an arbitrary shape. ated for the circular footing and then multiplied
Bridge Foundations 301
Bridge Foundations, 1
Fig. 3 Procedure for 3
evaluating equivalent
radius of a rectangular
footing B
2B
2
RECTANGULAR
FOOTING
D
2L
EQUIVALENT R
2B
CIRCULAR 2
FOOTING
EQUIVALENT RADIUS:
TRANSLATIONAL: Ro = 4BL
π
1/4
ROTATIONAL: (2B) (2L)3 (x−AXIS ROCKING)
R1 =
3π
1/4
(2B)3 (2L) (y−AXIS ROCKING)
R2 =
3π
1/4
4BL (4B2+ 4L2)
R3 = (z−AXIS TORSION)
6π
by shape and embedment factors to obtain the Bridge Foundations, Table 1 Static stiffness coeffi-
solution for the rectangular base. The solution is cients for a circular footing resting on the surface of soil
carried out in five steps as follows: Displacement degree-of-freedom k0
Vertical translation 4Gr
1u
Step 1: Determine the equivalent radius for each Horizontal translation 8Gr
2u
degree of freedom, which is the radius of a Torsional rotation 16 3
3 Gr
circular footing with the same area as the Rocking rotation 8Gr 3
3ð1uÞ
rectangular footing as shown in Fig. 3.
Step 2: Calculate the stiffness coefficients K0 for
the transformed circular footing (Table 1), Step 3: Multiply each of the stiffness coefficients
where, G and n in the table are the shear values obtained in step 2 by the appropriate
modulus and Poisson’s ratio for the soil- shape correction factor a from Fig. 4 (Lam and
foundation system. Martin 1986). This figure provides the shape
302 Bridge Foundations
SHAPE FACTOR,∝
AT −A AX
SL (X Y−
AN N) NG IN G(
R
T IO KI CK
AL T C RO
NT REC RO
R IZO X-DI
O (
H
1.10
ION
TAT N
RO IO
LAT
NS
T RA ION)
AL T
TIC IREC
VER (Z−D
1.05
ION
SLAT
TRAN
IZ O NTAL CTION)
HOR (Y−DIRE
1.00
1 2 3 4
L/B
factors for different aspect ratios L/B for the the foundation are determined and compared to
foundation. acceptable values. The foundation must be eval-
Step 4: Multiply the values obtained from step 3 uated for bearing capacity and sliding resistance
by the embedment factor b using Fig. 5. D in due to the seismic loads.
these figures is the footing thickness.
Step 5: Calculate the dynamic stiffness and the Seismic Design of Spread Footings
damping coefficients using Eq. 3 and the Earthquakes will induce moments and horizontal
charts displayed in Figs. 6, 7, and 8 (Veletsos loads in addition to the traditional vertical loads
and Verbic 1973), applied to a spread footing. In low seismic hazard
areas, spread footings may be proportioned to
The static stiffness can be calculated using the resist overturning, sliding, flexure, and shear
second approach for calculating the impedance due to forces obtained from the seismic analysis
functions for the soil-foundation system (Gazetas of the bridge. In high seismic regions, the flexural
1991). Using Fig. 9, a two-step calculation pro- and shear demands are those associated with the
cess is required. First, the stiffness terms are over-strength plastic moment capacity of the col-
calculated for a foundation at the surface. Then, umn or pier attached to the footing. To represent
an embedment correction factor is calculated for the combined effect of the seismic forces and
each stiffness term. The stiffness of the embed- moments a resultant load that may have to be
ded foundation is the product of these two terms. inclined and applied eccentrically can be applied
According to Gazetas, the height of effective in lieu of the seismic vertical forces, seismic
sidewall contact, d, in Fig. 9 should be taken as horizontal forces, and seismic moments. There-
the average height of the sidewall that is in good fore, a procedure is established to account for
contact with the surrounding soil. load inclination and load eccentricity of footing
Considering the range of frequencies and through which the general bearing capacity equa-
amplitudes in earthquake ground motions com- tion of shallow footing is adjusted to account for
pared to machine foundations, it is reasonably to these effects. This procedure is carried out in
ignore the frequency dependence of the stiffness three steps as follows:
as well as the damping parameters.
Based upon the results of the dynamic analy- Step 1: Compute the seismic vertical loads, seis-
sis, the peak dynamic loads and deformations of mic horizontal loads, and seismic moments
Bridge Foundations 303
2.5 7.0
)
XIS
IGH AL
TA
(US TATION
TRANSLATIONAL (VERTICAL AND HORIZONTAL)
ER
RO
6.0
L
IS)
LE NTA
AX
SE IZO
FT
R
HO
(U
L ) 4.0
CA XIS
R TI A
T
VE LEF
SE
(U
1.5 3.0
2.0
1.0 1.0
0 0.5 1.0 1.5 2.0 2.5
D/R
imposed to the footing. These seismic loads Mb and Ml = seismic moments about the short
and moments can be taken directly from the and long axes of the footing; and eb and el =
results of a global dynamic response analysis eccentricities of the load Q about the centroid
of the structure with the soil–foundation inter- of the footing in the direction of the short and
action SFI effects represented in the structural long axes respectively.
model. For each direction, these forces are It is known from basic principles of
then combined into a single resultant force strength of materials that if the eccentricity
with an inclination angle b with respect to in one direction is less than 1/6 of the founda-
the vertical. tion’s length in that direction, the footing is in
Step 2: Calculate the equivalent dimensions for compression throughout. As eccentricity
the footing to account for the load eccentricity, exceeds this value, a loss of contact occurs.
which is caused by the seismic moments The concept of effective width was introduced
applied to the foundation in both directions. by Meyerhof (1953) who proposed that at the
The vertical load can be transferred to an ultimate bearing capacity of the foundation, it
eccentric position defined by eb = Mb/Q and could be assumed that the contact pressure is
el = Ml/Q, where Q is the central vertical load identical to that for a centrally loaded founda-
due to seismic load plus other service loads; tion but of reduced width. The reduced
304 Bridge Foundations
Values of k z
1/
3
0.5
V=1/2
Exact
Approximate
a0
0 2 4 6 8
V=0
1/3
1.0
1/2
Values of c z
0.5 Exact
Approximate
a0
0 2 4 6 8
Values of k x
0.5 Exact
Approximate
a0
0 2 4 6 8
1.0
V=0
1/3
Values of c x
1/2
0.5
Exact
Approximate
a0
0 2 4 6 8
reactions are usually mobilized along the top demand, it is imperative to estimate precisely the
5–10 pile diameters. The axial soil resistances, foundation stiffness to be included in the overall
however, develop at greater depths. Hence, the structural model for determination of the
axial and lateral capacities of piles are considered demands.
to be uncoupled.
In general, seismic design of pile foundations Pile Foundation Stiffness
is evaluated by the substructure method. First, the Soils are inherently nonlinear, starting from
total structure is divided into two substructure incredibly small load levels. Lateral loads on
models with a convenient interface between the piles are resisted by the surrounding soil. There-
two at the pile cap level. Next, the stiffness matrix fore, piles exhibit nonlinear load deflection char-
of the pile group, which implies soil effects, is acteristics. This behavior is represented assuming
established using the foundation substructure the pile as a beam supported on Winkler springs
model or simple methods. The stiffness matrix that are characterized by nonlinear p-y curves for
is then implemented in the superstructure model lateral loading, or t-z and q-z for vertical loading.
and seismic analysis is conducted using the These curves characterize the lateral soil resis-
superstructure model to determine the demands tance per unit length of pile as a function of the
at the interface point. Finally, capacities of indi- displacement. These relationships are generally
vidual piles are evaluated and compared to the developed on the basis of semiempirical curves,
demands by back substitution in the substructure which reflect the nonlinear resistance of the
model. For a proper determination of the seismic local soil surrounding the pile at a certain depth.
306 Bridge Foundations
Bridge Foundations, 0
Fig. 8 Dynamic modifiers 1.0 Exact
for footing in rocking Approximate
motion
Values of k θ
V=0
0.5 1 /3
1/2
a0
0 2 4 6 8
0.5
V=0
1 /3
Values of cθ
1 /2
Exact
Approximate
a0
0 2 4 6 8
The two most commonly used p-y models are at 0.5–0.65 of peak deflection (Lam et al. 1998).
those proposed by Matlock (1970) for soft clay A stiffness matrix can be obtained by prescribing
and by Reese et al. (1974) for sand. The most a unit deformation vector for each degree of the
commonly used t-z and q-z models are those six degrees of freedom, while keeping the other
developed by McVay et al. (1998). five degrees zero. The resultant force vector
There are two methods for modeling the corresponding to each unit deformation vector
behavior of pile groups for seismic response stud- can be used to form the corresponding column
ies, the soil-pile stiffness as will be explained in vector in the stiffness matrix. The stiffness matrix
the following subsections. must be positive definite otherwise, numerical
problems may be expected when the stiffness
Coupled Pile Foundation Stiffness Matrix matrix is implemented in the overall structural
In this method, a quasi-dynamic analysis for the model. One way to ensure that the stiffness
pile group is conducted by applying loading matrix is positive definite is to invert it and
(either as forces or displacements) at the interface check that the diagonal elements in the inverted
node between the superstructure and foundation (compliance) matrix are positive values. Pro-
model using linearized properties for the soils. grams such as LPILE (2010), FLPIER (Hoit and
Linearized properties for a single pile can be McVay 2010), and SWM (Ashour and Norris
achieved by assuming secant foundation stiffness 2000) may be used to establish the foundation
Bridge Foundations 307
Bridge Foundations,
Fig. 10 Effective footing
dimensions due to eccentric
loads (After AASHTO
LRFD 2012)
308 Bridge Foundations
stiffness matrix of the pile group through an iter- stiffness design charts for lateral loading as
ative process. presented in Figs. 11, 12, and 13. These charts
are applicable for piles up to 24 in. Two parame-
Simplified Procedure for Pile Group Stiffness ters are required to define the soil-pile system: the
Matrix pile bending stiffness, EI, and the coefficient of
The simplified method assumes coupling only variation f of soil reaction modulus Es with depth.
between shear and overturning moment for the The coefficient f has units of force/unit volume.
single pile. The general form of a single pile
stiffness matrix can take the form: Step 2: Single Pile Under Axial Load The
recommended LRFD guidelines for the seismic
2 3
K xx 0 0 0 K xyy 0 design of highway bridges proposed the follow-
6 0 K yy 0 K yyx 0 0 7 ing simple equation for the determination of the
6 7
6 0 0 K zz 0 0 0 7 axial stiffness of the pile:
6 7
6 0 K yx y 0 K yx 0 0 7
6 7
4 K y x 0 0 0 K yy 0 5 Ep A
0
y
0 0 0 0 K yz K v ¼ 1:25 (5)
L
(4)
Wherein, Ep is the modulus of elasticity of pile
in which, Kxx, Kyy are the lateral stiffnesses; Kzz is material; A is cross sectional area of pile; and L is
the axial stiffness; and K xyy and K yyx are length of pile.
corresponding coupled stiffnesses between shear
and overturning moment. Step 3: Pile Group Stiffness The stiffness of
The simplified method involves five basic the single pile can be used to establish the pile
steps (Lam et al. 1991) as follows: group stiffness matrix. If the pile group consists
of vertical piles, the stiffness summation proce-
1. Determine the stiffness coefficient of a single dure is relatively straight forward. The stiffness
pile under lateral loading. for the translational displacement terms (the two
2. Determine the stiffness coefficient of a single horizontal and the vertical displacements) and the
pile under axial loading. cross-coupling terms can be obtained by multi-
3. Superimpose the stiffness of individual piles plying the corresponding stiffness components of
to obtain the pile group stiffness. an individual pile by the number of piles.
4. Solve for the stiffness contribution of the In general, the axial stiffness of the piles will
pile cap. dominate the rotational stiffness of the group.
5. Superimpose the stiffness of the pile cap to the Therefore, the rotational stiffness terms require
pile group. consideration of this additional stiffness compo-
nent. Similarly, there is an interaction between
These steps are described herein. the translational stiffness of piles and the tor-
sional stiffness of the group. The following equa-
Step 1: Single Pile Under Lateral Load Lam tion (Lam et al. 1991) can be used to develop the
and Martin (1986) came to a realization that lat- rotational terms of a pile group:
eral load-deflection characteristics representing
the overall stiffness of the soil pile system are X
N
dominated by the elastic pile stiffness over the K RG ¼ NK RP þ K dn S2n (6)
nonlinear soil behavior. Moreover, the localized n¼1
Bridge Foundations,
107
Fig. 11 Pile translational Embedment
stiffness 0'
5'
105
104 f = 10
f=1
f = 0.1
Coeff. of Variation of Soil Reaction
Modulus with Depth, f (Ib/in3)
103
Kdn is the translational stiffness coefficient of an Step 5: Superimpose the Stiffness of the Pile
individual pile (axial for group rocking stiffness Cap to the Pile Group The resultant pile cap
and lateral for group torsional stiffness); and Sn is stiffness obtained from step 4 can be added to the
the distance between the nth pile and the axis of diagonal lateral translational stiffness coeffi-
rotation. cients in the pile group stiffness matrix for the
total pile group–pile cap stiffness matrix.
Step 4: Stiffness Contribution of the Pile The above procedure does not account for
Cap In addition to the component of soil resis- group effects which relate to the influence of the
tance acting on piles, the passive pressure soil adjacent piles in affecting the soil support char-
resistance on the vertical pile cap face can be acteristics. Full-scale tests by a number of inves-
added to the stiffness and resistance obtained tigators demonstrate that the lateral capacity of a
from the pile members. This is contingent on pile in a pile group may be less than that of a
stable level ground conditions. The pile cap stiff- single pile due to the interaction between closely
ness can be estimated as the ultimate soil capacity spaced piles in the group (group efficiency). As
divided by an estimated displacement to mobilize the pile spacing reduced, the reduction in lateral
this capacity. Centrifuge tests (Gadre 1997) capacity becomes more pronounced. In general,
showed that the deflection level to reach the ulti- pile spacing of less than 3–5 pile diameters are
mate pile cap capacity occurs at about 0.02 times necessary before the effects of pile interaction
the embedment depth. This equivalent linear becomes significant in practical terms. Type and
secant stiffness can be added to the stiffness of strength of soil, number of piles, and loading
the piles. level are other factors that may affect the
310 Bridge Foundations
109
f = 100
f = 10
f=1
108 f = 0.1
efficiency and lateral stiffness of the pile. More- pile group will govern the design. For a fixed-
over, in addition to group effect, gapping, and head pile group, the piles must be designed for the
potential cyclic degradation were also subject of plastic hinging moment capacity of the column/
many investigations (e.g., Brown et al. 1987; pier connected to the pile cap. The maximum
McVay et al. 1995). It has been shown that a negative moment occurs at the base of the pile
concept based on p-multiplier applied on the cap while the maximum positive moments occur
standard stiffness matrix can work reasonably to in the pile at a short depth below the base. In this
account for pile group and cyclic degradation regard, the pile cap has to be designed to behave
effects. rigidly. The rigid response of the footing may be
assumed if the following equation is satisfied
Seismic Design of Pile Footings (AASHTO 2011):
Design of pile footings for seismic effects is often
relatively a complex process. The strategy of the
L
seismic design is to determine capacity and 2:5 (7)
deflection of the piles under the action of the D
seismic lateral loads and ensure the integrity of
the pile group against liquefaction. where, L is the cantilever overhang length mea-
sured from the face of pier or column to the
Moment Capacity of Pile Groups outside edge of the pile cap and D is the depth
Typically, the foundation system is designed to of the pile cap. Due to the interaction between the
be capacity protected, which ensures that damage rocking response and the vertical response of a
occurs above ground, where it can be accessed pile group, the axial demands on an individual
and repaired. The seismic moment capacity of the pile can be estimated as:
Bridge Foundations 311
107
f = 100
106 f = 10
f=1
f = 0.1
Coeff. of Variation of Soil Reaction
Modulus with Depth, f (Ib/in3)
105
Bridge Foundations, a b
Fig. 14 Drilled shafts
configuration
Drilled Shafts
embedded 24 in. into a pile cap can produce a using drilling (excavating) equipment capable of
sufficient moment capacity for the connection auguring or coring 3000 to 12000 diameter excava-
without the need for a reinforcement detail. tions into soil and rock. After the excavation is
These observations agree with the results of completed, a reinforcing cage is placed in the
cyclic tests on steel HP piles (Shama et al. 2002). excavation and the excavation is filled with high
slump concrete. Drilled shafts are usually
Lateral Capacity of Pile Footings constructed with steel shells.
The lateral capacity of piles requires compiling of Drilled shafts are attached to the superstruc-
both geotechnical and structural engineering prin- ture in two configurations as illustrated in Fig. 14.
ciples. For a successful design, it is important to In the first configuration a group of drilled shafts,
get the soil resistance mobilized before structural usually 4–6 are connected to a cap that supports
failure of the pile. The relative importance of the the bridge pier or column. In the second configu-
pile cap lateral capacity to the individual piles ration, a large diameter shaft is extended as a
needs to be considered. Therefore, the lateral structural unit to support the bridge superstruc-
capacity evaluation must include the resistance ture. The second configuration comes in two
developed by the pile cap in addition to the lateral alternatives as illustrated in Fig. 15 (Priestley
shear resistance of the piles. Including the lateral et al. 1996). In the first alternative, the section
capacity of the pile cap is contingent upon efficient of the column above the ground is similar to that
compaction of the soils around the pile cap to of the shaft. The maximum moment forms at a
improve the cap resistance. The lateral capacity depth of typically 1.5–2.5 shaft diameter.
of the pile cap must include the passive pressure A disadvantage of this alternative is that the
mobilized at the front face of the cap perpendicular extent of damage to the plastic hinge region will
to the seismic load and the interface shear resis- be below ground after an earthquake and needs
tance developed along the sides of the cap parallel excavation of the soil material for inspection.
to the application of the seismic load. With the alternative detail of Fig. 15b, the
moment capacity of the shaft below ground is
increased above that of the column above ground
Drilled Shafts to ensure that hinging occurs at the base of the
column and damage can be expected after an
Drilled shafts are large diameter piles that are earthquake. The disadvantage of this detail is
designed to sustain high axial loads and that it is more expensive with respect to the alter-
overturning moments. They are constructed native of Fig. 15a.
Bridge Foundations 313
a P b P
M M
V Capacity
V Capacity
B
Plastic
Hinge Critical
Section
Plastic Critical
Hinge Section
Moment Moment
Bridge Foundations, Fig. 15 Extended shaft seismic design alternatives (Priestley et al. 1996)
Bridge Foundations,
Fig. 16 Methods of
representing shaft
foundation stiffness
pushing the shaft in the lateral direction. For a soil-pile system. The effective length of the shaft
group, the strain wedges for different shafts will shall be equal to the laterally unsupported length
interfere together, depending on the spacing, LC plus an embedded depth to fixity LF. The depth
diameter, and location of the shaft within the to fixity below the ground in ft. may be taken as
group leading to strength reduction and (AASHTO-LRFD 2012):
softening. For clays:
The nonlinear spring modeling technique can
0:25
include the Masing hysteretic behavior of the EP lw
soil, which when implemented includes automat- 1:4 (10)
Es
ically in the model the energy dissipation associ-
ated with material damping. The radiation and for sands:
damping component can be included by attaching
additional viscous dashpots. Recent studies 0:20
(Wang et al. 1998) showed that placing the Ep l w
1:8 (11)
viscous dashpots (representing radiation damping nh
in the far field) in series with the hysteretic
component of the soil–structure element Where Ep is the modulus of elasticity of the
(representing the nonlinear soil-pile response in shaft material; lw is the weak axis moment of
the near field) is technically preferable to a par- inertia for the shaft; Es is the soil modulus for
allel arrangement of the viscous and hysteretic clays; Su is the undrained shear strength of clay;
damping components. and nh is the rate of increase of soil modulus with
depth for sand.
Equivalent Cantilever Model
The equivalent cantilever model is the simplest Coupled Foundation Stiffness Matrix
approach to represent the effects of the surround- This approach is usually employed in regions of
ing soil, wherein the sectional properties of the low to moderate seismicity. As illustrated in
cantilever are the same as that of the shaft, but its Fig. 16, a 6 6 can be determined for a single
length (depth to fixity) is adjusted to provide the shaft at the ground line. As indicated in section
same maximum bending moment as in the actual “Coupled Pile Foundation Stiffness Matrix,”
Bridge Foundations 315
linearization of p-y curves is required for the and they often become advantageous where water
development of a 6 6 coupled stiffness matrix. depths exceed 30–36 ft. Caissons are divided into
three major types: (1) open caissons, (2) box cais-
Seismic Design of Drilled Shafts sons (or closed caissons), and (3) pneumatic cais-
The following design requirements shall apply sons. A common feature of large caissons B
(ATC-32 1996) to bridges on drilled shaft produced by the three methods is that they are
foundations: massive structures that respond to seismic loads
in a primarily rocking mode about the base plus
(I) Seismic lateral foundation design forces some translations.
shall be based on either plastic hinging of
the bridge column or pier; linear dynamic Seismic Evaluation of Large Caissons
analysis; or more advanced nonlinear The behavior of caissons under lateral seismic
dynamic response analysis. loads is essentially nonlinear. Geometric
(II) Foundation stiffness must be accounted for nonlinearity dominates this behavior due to
in the dynamic response analysis of the rocking of the caisson and gapping at the
overall bridge. soil–caisson interface. The weight of caissons
(III) The capacity of shaft foundations and their constitutes the major portion of the entire struc-
individual components shall be based on ture due to its large volume. Hence, the funda-
the safety evaluation earthquake. mental natural period of vibration of the large
(IV) Earth pressures generated by lateral ground caisson is short and very different from the long
displacements due to liquefaction shall be period response of the bridge. This difference in
accounted for at poor soil sites. the two modes of vibrations enables uncoupled
(V) Strong connection details shall be evaluated seismic analyses of the two structures. However,
at poor soil sites. when evaluating the bridge structural response,
the caisson as an element must be included in the
total bridge model. Two approaches are usually
Large Gravity Caissons used for representing large caissons for seismic
design.
Large caissons are very large concrete boxes that The uncoupled spring approach represents the
are excavated or sunk to a predetermined depth. caisson as a lumped mass and establishes
They are usually used for the construction of the interaction between the caisson and soil by
bridge piers or other heavy waterfront structures, the impedance approach. Figure 17a displays a
Bridge Foundations, Fig. 17 Modeling techniques of caisson foundations: (a) Uncoupled spring approach; and (b)
Winkler approach
316 Bridge Foundations
Bridge Foundations,
Fig. 18 Continuum
models of caissons
schematic of the caisson model using this between the soil and the caisson at the base and
approach, wherein a rigid link is established side walls. Nonlinear static pushover analyses
between the node located at the center of mass shall be carried out to establish the soil–structure
of the caisson where its mass is lumped and interaction behaviors for implementation in the
another node at the bottom where a set of springs global model. Pushover analyses of the local
is attached. This method has been presented in model shall be performed by applying a point
section “Spread Footing Stiffness.” This method load at the center of gravity of the rigid caisson
overlooks the propensity of gapping at the base for each mode of soil resistance. The soil
and at the embedded sides of the caisson during response results from the different pushover ana-
the seismic event and may lead to irrelevant lyses of the local model in each direction shall be
results in the large displacement range of loading. extracted and distributed to the soil spring ele-
Hence, it is only recommended for regions of low ments at various nodal points in accordance with
seismicity. the discretization scheme of the global model.
The nonlinear spring Winkler approach The behavior of the caissons under seismic
(Fig. 17b), as mentioned earlier, treats the soil loads shall be assessed through nonlinear time
supporting the foundation as a series of springs. It history analysis of the global model. The perfor-
represents linear and nonlinear soil properties and mance of the caissons is evaluated by comparing
hence favorable for life safety performance based the maximum drifts from the results of the time
design. It also accounts for the separation and history analysis to the permissible levels
gapping between the caisson walls and the sur- according to the performance based design
rounding soils. The properties of these nonlinear criteria of the project.
springs are established through static pushover
analyses of three dimensional finite element con-
tinuum models as. Figure 18 depicts an example Cross-References
of such continuum model, where both the caisson
and surrounding soil are modeled by three dimen- ▶ Dynamic Soil Properties: In Situ Characteriza-
sional brick elements. The detailed 3-D finite tion Using Penetration Tests
element shall include constitutive relationships ▶ Seismic Analysis of Masonry Buildings:
for the nonlinear behavior of the soil and special Numerical Modeling
interface elements that can capture gapping ▶ Soil-Structure Interaction
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings 317
Buckling- Link
a Concentric Brace
b Restrained Brace c Beam
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings, Fig. 1 Types
of steel braced frames
Fig. 1c shows a link in the middle of the beam brace and the BRB is the symmetry of the
span although it is also commonly adjacent to the response. A BRB is specially designed to prevent
column. The diagonal braces used in EBFs are buckling during the compression cycle. This
designed to remain elastic for the maximum results in highly ductile behavior in both tension
forces and moments that can develop in the link and compression and prevents the force unbal-
beam. An EBF system is more flexible than a ance present in other brace configurations due to
CBF and has a symmetric response as the yield- the significantly lower buckling strength as com-
ing under shear and flexure is symmetric. The pared to the tension yield (see Fig. 1b). Brace
primary drawback for EBFs is that replacing the fracture is not a problem due to the fact that the
link beam is more difficult and costly than large stresses in the core are primarily axial.
replacing a diagonal brace. There has been recent There is no strength or stiffness degradation for
work in developing and testing replaceable link BRBs that are detailed correctly. The stiffness of
beams to overcome this problem (Mansour a BRBF is less than a CBF as the BRB strength is
et al. 2011). directly related to the tension strength, whereas in
The third type of braced frame system is the a CBF the area generally must be increased to get
buckling-restrained braced frame, which is a type the required capacity in compression. Buckling-
of concentrically braced system. The BRB is the restrained braces are unique in that they are a
primary structural element in the BRBF provid- manufactured device rather than a fabricated
ing the necessary ductile energy dissipation. The steel section. Since their development and intro-
difference between a typical fabricated steel duction into the industry, BRBs have taken a
320 Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings
Connection to Gusset
Plate (Bolted Option Shown)
large market share in braced steel frame systems runs through the center of the tube which is then
due to their cost competitiveness and their reli- filled with a mortar paste to fill the space allowing
able, resilient performance in seismic events. the steel tube to provide the required flexural
resistance to brace the core. It is not desirable
for any axial load to go through the mortar or
Buckling-Restrained Brace Description the steel outer tube. This is accomplished by
providing a bond breaker between the mortar
Because braced frames have very beneficial char- and the steel core. Several different options
acteristics, significant work was undertaken to have been used for this including a lubricant or
determine ways to overcome the negative aspects a material wrapped around the steel core prior to
of brace buckling. The early beginnings of a filling the outer tube. In addition to the BRB
brace restrained against buckling were in the shown in Fig. 2, several other systems have
1970s in Japan where steel plates were encased been developed by researchers to create a BRB
in a precast concrete wall (Wakabayashi system for lower cost, to simplify manufacturing,
et al. 1973). Since then, significant improvements or to replace damaged cores more simply after an
and modifications have been made to the original event. Several of the BRB systems developed
configuration that has enabled the reliable and through these research efforts are shown in
ductile BRB commonly used today (Black Fig. 3, although this is a sample of the different
et al. 2004; Fahnestock et al. 2006; Xie 2005). variations that have been investigated. Figure 4
The key to the mechanism is the axial load going shows a photograph of a building retrofitted with
through the steel core while sufficient buckling BRBs.
resistance (flexural rigidity) is provided in The geometry and yield strength of the core
another component. This has been accomplished section play an important role in the performance
in several ways although the most common is of BRBs. From a material specification, it is
shown in Fig. 2. The basic premise of the BRB desirable to have a known value or at least a
is that the core steel element has a cross section small range where yielding will occur. An A36
designed to yield in tension and compression. In material with controlled yield is commonly used
order for this to occur, sufficient flexural rigidity for this purpose. The yielding core section can be
must be provided so the overall buckling load of fabricated to any area specified by the structural
the composite device is greater than the ultimate engineer. The area of the yielding core is the
strength of the steel core. This is accomplished prime cross-sectional characteristic used to deter-
through the use of a steel tube which is most mine strength. There are two primary shapes
commonly square or circular. The core steel for the core region, flat plate or cruciform.
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings 321
High-Strenth Lateral B
Threaded Restraint
Steel Core Steel Core
Steel Core Rod
Steel Tube
Bolted Mortar
Connection Confinement Welded
Filler Plate Connection
Steel Bar
Guide Plate Steel Core Steel Channel Steel Core
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings, Fig. 3 Various
BRB cross sections
Buckling-Restrained
Braces and Their
Implementation in
Structural Design of
Steel Buildings,
Fig. 4 Photograph of
building retrofitted with
BRBs (Courtesy of Core
Brace, LLC)
In addition, some BRBs use multiple cores within core axial strength. The equation for the equiva-
the restraining tube to achieve the desired core lent elastic stiffness of the brace is shown in Fig. 4
area. The length of the yielding core is an impor- and is based on a combination of springs in series
tant part in determination of the stiffness of the representing the axial stiffness of each segment.
brace. Figure 5 shows the breakdown of a typical The overall stiffness of the brace is important for
core with the flat plate yielding core, transition accurate analysis of the frame with BRBs and can
region, and connection region. The flexural stiff- be modified if the brace stiffness is too low. This
ness of the transition region and connection modification is typically done by shortening the
region is important as the flexural rigidity of yielding core length (i.e., increasing the transition
these regions must be sufficient to sustain the length) which is the most flexible segment.
322 Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings
keff
Buckling-Restrained
Braces and Their Brace Axial Force
Implementation in
Structural Design of
Steel Buildings, PY
Fig. 6 Hysteresis plot for
Concentric
buckling-restrained brace
and concentric brace Brace
Brace
Deformation
Buckling-
PCR Restrained
Brace
PY
However, it is important to recognize that a shorter deformation divided by the yield deformation.
yielding length results in higher inelastic strains in There is a small increase in strength on the com-
the core which will result in higher forces in the pression side, typically about 10 % larger than the
braces, an important design criterion. brace tension strength. This is due to friction
The overall representative hysteretic behavior between the core and the mortar. Even with a
of a BRB and a concentric brace is shown in bond breaker, the core will dilate under compres-
Fig. 6. The BRB has full hysteresis loops in sion resulting in greater frictional force and some
both tension and compression including strain axial load transfer. By contrast, a concentric
hardening. The capacity for significant deforma- brace performs well in tension but experiences
tion ductility in both directions is apparent where buckling in compression which typically results
deformation ductility is defined as the total in formation of a flexural plastic hinge near the
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings 323
center of the brace. Once the brace has buckled Buckling-Restrained Brace Design
one time, it quickly loses strength in compression
upon subsequent cycles. The formation of the Buckling-restrained braces were first used in the
hinge can result in brace fracture in tension due USA in 1999, over a decade after the first usage in
to the damage occurring during compression Japan in 1987 (UBB 2014). Part of the motivation B
buckling. The lower compression capacity and for this new system was due to the failure of steel
potential brace fracture significantly limit the moment frames that occurred in the 1994
energy dissipation and deformation ductility Northridge, California, and the 1995 Kobe,
capacity of CBFs when compared to BRBs. Japan, earthquakes. As with other new develop-
As was stated previously, BRBs are a ments in structural engineering practice, it was
manufactured product. In the United States there implemented in building practice in the USA
are three primary brace producers, Core Brace prior to inclusion in building codes or reference
(http://www.corebrace.com/), Star Seismic standards. The first inclusion of BRB design pro-
(http://www.starseismic.net/), and Unbonded visions was in the NEHRP Recommended Pro-
Brace (http://www.unbondedbrace.com/). Each visions for Seismic Regulations for New
of these companies has tested their braces to Buildings and Other Structures (FEMA 2003),
meet the requirements put forth in the applicable FEMA 450. The information included in FEMA
building codes and reference documents. In addi- 450 was developed by a joint committee includ-
tion to the braces, they provide design and engi- ing AISC and the Structural Engineers Associa-
neering guidance associated with their products. tion of California (SEAOC). FEMA 450 is not a
While the basic premise of the BRB is the same building code or reference standard, but inclusion
for each manufacturer, many specific details of is typically the first step in the process. The
the various brace manufacturers differ as they buckling-restrained braced frame was included
compete based on price and performance in the in the 2005 version of ASCE/SEI 7, Minimum
marketplace. Standard for Design Loads for Buildings and
BRBs, as manufactured products, have a high Other Structures (2005), as an acceptable seismic
degree of quality control requirement placed on force-resisting system. In ASCE/SEI 7-05, a
them in the American Institute of Steel Construc- BRBF had the following seismic performance
tion (AISC) Seismic Provisions for Structural factors (SPF): R = 7, Oo = 2, Cd = 5.5. A dual
Steel Buildings Section K3 (AISC 2010), which system with BRBs and moment-resisting connec-
is a reference standard in the United States (USA) tions at the beam-column joints without a bracing
governing the design of steel buildings for seis- connection had slightly different SPFs: R = 8,
mic loads. The requirements include testing Oo = 2.5, Cd = 5. One change of note that
which includes both axial demands on the brace appeared in ASCE 7-10 was that the SPFs for
and the consideration of rotations at the connec- BRBs with and without moment frames were
tions due to frame deformations. The primary made the same as the dual system based on
requirements of the testing are to show (1) the research that had taken place in the interim. In
required ductility, both for a single cycle and addition to ASCE 7, BRB provisions were
cumulatively; (2) repeatable stable hysteresis included in AISC Standard 341-05, Seismic Pro-
with positive post-yield slope; and (3) avoidance visions for Structural Steel Buildings (2005). This
of core rupture, brace instability, and connection document deals with more detailed requirements
failure. The required testing for a wide variety of for design of not only the BRBs but also the other
brace sizes has been completed by the three pri- elements of the lateral force-resisting system.
mary producers. The specification however This document also lists the required testing pro-
allows for these tests to be run on a project- tocol for BRBs. The subsequent versions of the
specific basis if a new or unique brace configura- ASCE/SEI 7 and AISC 341 standards continued
tion is specified. to include provisions for BRBs with
324 Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings
modifications based on continuing research. An plot of how these values would be determined.
early and still applicable reference for the design This is one of several items that will require input
of BRBs is by Lopez and Sabelli (2004) which from a brace manufacturer. The adjusted brace
has detailed examples of design provisions. strength is used to ensure that the beams, col-
A more recent textbook (Bruneau et al. 2011) umns, and connections in the frame have the
also has a chapter on BRBFs. required strength to remain elastic through the
From the design perspective, sizing of BRBs is expected strength of the BRB. The design of
a very straightforward process. The size of the brace connections must be designed to 1.1 times
core is based on the maximum axial demand the adjusted brace strength in compression.
according to the limit state of yielding based on The other component of design that is impor-
the load combinations from the applicable build- tant is accurate modeling of brace stiffness. As
ing code. Equation 1 shows the basic premise discussed previously, the stiffness of the brace
where Pysc is the nominal axial yield strength, represents a combination of the stiffness of the
Asc is the area of the steel core, and Fysc is either different segments. This versatility provides
the minimum specified yield stress or the actual additional mechanisms to control response by
yield stress from a coupon test. If the manufac- modifying the stiffness of the brace. For example,
turer provides coupon tests from the steel used to if the brace is too flexible, the yield length could
fabricate the core, this value can be used in be shortened. However, this will have an effect on
design. Otherwise, a range of values should be other components as the inelastic strains in the
used to determine the upper and lower limits that brace for the same deformation will increase
could occur. which will also increase the adjusted brace
strengths. Brace manufacturers will provide
Pysc ¼ Fysc Asc (1) assistance to create the appropriate balance of
stiffness and required strength.
Once determined, the nominal yield strength One of the final design considerations is that of
value is then modified by a resistance factor of the connection of the BRB to the framing. There
0.9 for LRFD design to determine the design are three typical options for connecting BRBs to
axial strength. One of the benefits of BRBs is gusset plates although the specific details of the
that any cross-sectional area can be provided so connections vary based on the different manufac-
each floor can have a BRB sized to the demand. turer’s brace geometry. The three options are
Typically the area will be rounded to some degree bolted, welded, and pinned connections. Exam-
based on the manufacturer’s practice, but this is ple details are shown in Fig. 8 although many
not generally a significant change. In addition to variations can be found and have been used. As
nominal yield, the adjusted brace strength must opposed to CBFs where the gusset plate is
also be established as part of the design. The designed to allow for the out-of-plane rotation
adjusted brace strength has two components, the due to brace buckling, BRBs can sustain signifi-
tension (oRyPysc) and compression adjusted cant moments. These large moments are devel-
brace strength (boRyPysc), where o is the strain oped based on the large deformations in the frame
hardening adjustment factor, b is the compres- inducing rotations in the beam and columns. The
sion strength adjustment factor, and Ry is the ratio pin allows for these rotations to occur in the
of expected to specified minimum yield stress. Ry connection rather than the structural element.
can be set to 1 if coupon tests are used to deter- This allowance for rotation can also be developed
mine the yield strength. The values for b and o using welded and bolted gusset connections by
are taken from coupon tests that are carried out to providing a beam stub off the column with a
the strain equivalent to that occurring in the brace hinge connection to the remainder of the beam
at the expected deformation. The expected defor- as shown in Fig. 8c. It is also of critical impor-
mation is defined as the larger of 2 % story drift or tance to ensure that the connection (gusset
two times the design story drift. Figure 7 shows a place and end section of BRB) has sufficient
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings 325
Buckling-Restrained
Braces and Their
Implementation in
Structural Design of PMaxT PMaxT
Steel Buildings, w= PYSC
PYSC
Fig. 7 Determination of B
adjusted brace strength PMaxC
from test data b=
PMaxT
Axial
ΔMax Deformation
ΔY ΔMax
PYSC = FYSC A SC
Δmax = Max(2% Drift, 2*Design Drift)
P MaxT = Tensile Force at ΔMax
PYSC PMaxC = Compressive Force at ΔMax
PMaxC
Axial
Force
Buckling-Restrained a b
Braces and Their
Implementation in
Structural Design of
Steel Buildings,
Fig. 8 Typical gusset plate
connection details
out-of-plane flexural resistance to resist buckling on the existing concept. One of the biggest weak-
up to the full strength of the core section. Typical nesses of BRBs that has been identified in ana-
gussets for BRBFs are markedly smaller than lytical and experimental research is residual
those used in CBFs. displacements following significant earthquakes.
In addition to typical frame structures, BRBs The modifications and/or additions to BRBs have
can be used anywhere a ductile fuse element is primarily aimed to reduce residual deformations
required. Examples include bridges as ductile in BRBFs. This movement is also centered on a
connections between the superstructure and the larger goal of providing a higher performance
substructure, stadiums as part of the lateral standard than life safety in buildings for major
resisting systems, and dam intake towers to pre- earthquakes. Since BRBs are an effective lateral
vent lateral motion while limiting the forces system, additions and modifications to a high-
imparted to the dam. They have also been used performance seismic system represent an ideal
more recently in the San Francisco Airport Air way to move toward damage reduction in addi-
Traffic Control Tower which is a self-centering tion to life safety.
system where the BRBs provide hysteretic The first of these innovations is called the
energy dissipation (Muthukumar and Sabelli hybrid passive control device (HPCD). The
2013). HPCD consists of a BRB in series with a visco-
It is interesting to note that the design for elastic damping device. For small deformations,
BRBs in Japan is different from the USA. In only the viscoelastic damper is active. After a
Japan, braced frames are designed to higher specified displacement, the viscoelastic device
force levels than moment frames (i.e., the equiv- locks out and the BRB becomes active. The
alent of the R value penalizes braced frames) due device was tested in the laboratory and
to the potential for buckling. Because of this fact, implemented in analytical models of a nine-
most low-rise construction in Japan is designed story structure. Experimental testing showed the
with moment frames. However, for taller build- concept worked as designed. The building
ings, BRBs are used significantly due to the response to a suite of earthquakes was improved,
increased stiffness. In order to use BRBs for and residual displacements were reduced when
their stiffness and energy dissipation without the compared to a BRB frame (Marshall and Charney
force penalty, they are used as hysteretic energy- 2010a, b).
dissipating devices as part of a dual system with One of the ways that has been explored to
moment-resisting frames (Xie 2005). The first improve building response and reduce damage
reported use of a BRB in Europe was in 2006 at is through the use of self-centering systems. The
the University of Ancona in Italy. Use of BRBs in idea of these systems is to provide an element that
Europe is much less than in Japan and the USA remains elastic while a fuse element yields and
(Della Corte et al. 2011). dissipates energy. Analytical and experimental
research was completed on a self-centering
buckling-restrained brace (SC-BRB) using
Buckling-Restrained Brace Innovations nickel-titanium-based shape memory alloys.
The shape memory alloy’s unique properties pro-
Buckling-restrained braces are one of the more vide the capacity to self-center in addition to
impactful developments in earthquake engineer- providing additional energy dissipation. The
ing. They are considered today a standard option SC-BRB consists of typical BRB with two addi-
for both new designs and retrofits of existing tional outer concentric tubes and free-floating end
buildings in significant seismic hazards. How- plates so the shape memory alloy rods were
ever, as with any innovation, additional improve- always loaded in tension. The experimental test-
ments are constantly explored to further improve ing showed significant capacity for energy
Buckling-Restrained Braces and Their Implementation in Structural Design of Steel Buildings 327
Marshall JD, Charney FA (2010b) A hybrid passive con- urban growth in risk-prone areas due to scarcity
trol device for steel structures II: physical testing. of land, and global warming. Along with increas-
J Constr Steel Res 66(10):1287–1294
Miller DJ, Fahnestock LA, Eatherton MR (2012) Develop- ing frequency, recent disasters show an increase
ment and experimental validation of a nickel-titanium in magnitude and resulting destruction according
shape memory alloy self-centering buckling-restrained to studies by the Red Cross. Both natural and
brace. Eng Struct 40:288–298 technological/man-made disasters have seen
Muthukumar S, Sabelli R (2013) Nonlinear seismic anal-
ysis of a round concrete tower with a post-tensioned nearly exponential rises in the number of disas-
self-centering system. ASCE/SEI Structures Congress ters over time.
2013, American Society of Civil Engineers, Pitts- Despite the increasing number of disaster
burgh, 2–4 May 2013 experiences, post-disaster activities remain inef-
UBB (2014) http://www.unbondedbrace.com/facts.htm.
Accessed Jan 2014 ficient and poorly managed and need to be
Wakabayashi M, Nakamura T, Kashibara A, Morizono T, improved according to Halvorson and Hamilton
Yokoyama H (1973) Experimental study of elasto- (2010). Traditionally, post-disaster reconstruc-
plastic properties of precast concrete wall panels with tion consisted of simply repairing the physical
built-in insulating braces. Summaries of technical
papers of annual meeting, Architectural Institute of damage that has been induced by a disaster. How-
Japan, pp 1041–1044 (in Japanese) ever, Kennedy et al. (2008) pointed out that
Xie Q (2005) State of the art of buckling-restrained braces rebuilding the built environment and infrastruc-
in Asia. J Constr Steel Res 61(6):727–748 ture exactly as they were prior to a disaster often
re-creates the same vulnerabilities that existed
earlier. If restored to pre-disaster standards,
disaster-affected communities would face the
“Build Back Better” Principles for same difficulties if exposed to another disaster
Reconstruction event in the future. The reconstruction and recov-
ery period following a disaster poses an opportu-
Sandeeka Mannakkara, Suzanne Wilkinson and nity to address and rectify vulnerability issues
Tinu Rose Francis found in communities.
Department of Civil and Environmental As a result of witnessing the ongoing impacts
Engineering, The University of Auckland, of disasters on communities, a concept started to
Auckland, New Zealand emerge where post-disaster reconstruction was to
be taken as an opportunity to not only reconstruct
what was damaged and return the community to
Synonyms its pre-disaster state but to also seize the oppor-
tunity to improve its physical, social, environ-
BBB; Reconstruction; Recovery mental, and economic conditions to create
a new state of normalcy that is more “resilient”
(Boano 2009). This concept was termed “Build
Introduction Back Better,” suggesting that successful recovery
of communities following disasters needs to
“Build Back Better” signifies an ideal reconstruc- amalgamate the rehabilitation and enhancement
tion and recovery process that delivers resilient, of the built environment along with the psycho-
sustainable, and efficient recovery solutions to logical, social, and economic climates in
disaster-affected communities. The motivation a holistic manner to improve overall community
behind the Build Back Better concept is to make resilience. The phrase “Building Back Better”
communities stronger and more resilient follow- became popular during the large-scale recon-
ing a disaster event. Statistics from the United struction effort following the Indian Ocean Tsu-
Nations Environment Programme in 2008 show nami disaster in 2004 after which it became more
an increase in the number of natural disasters officially embraced with the creation of sets of
over time attributing to growing populations, BBB Guidelines to steer recovery and
“Build Back Better” Principles for Reconstruction 329
reconstruction activities toward achieving this infrastructure, insufficient focus given to certain
goal (Clinton 2006). aspects of the recovery process such as livelihood
This chapter reviews what BBB entails and development programs and small business sup-
presents the key elements required to improve port programs, overruling of local government
post-disaster reconstruction and recovery prac- agencies, and neglecting vulnerable groups of B
tices to build back better. First, existing guide- people in the community.
lines and reports providing recommendations for Complete recovery requires attention to many
BBB are introduced. Key information from the different elements. BBB was defined by
guidelines and reports is used to then identify the Khasalamwa (2009) as a way to utilize the recon-
key concepts which represent Building Back Bet- struction process to improve a community’s
ter. Finally, each concept and its importance for physical, social, environmental, and economic
building back better are reviewed. conditions to create a more resilient community,
where resilience is defined as “the capacity to
recover or ‘bounce back’ after an event” (Twigg
The Need for Building Back Better 2007). Therefore, what the concept of BBB pro-
poses is a broad holistic approach to post-disaster
The South Asia Disaster Report (DNS and PA reconstruction in order to address the wide range
2005) states that disasters are produced due to the of prevalent issues including those mentioned
weaknesses and vulnerabilities of communities, above and ensure that affected communities are
countries, and structures to withstand encoun- regenerated in a resilient manner for the future.
tered hazards. Wisner et al. (2004) defines vul-
nerability as the lack of capacity to anticipate,
cope with, resist, and recover from the impact of Existing Guidelines for Building
a hazard. The destruction and loss of human lives Back Better
from the 2005 Kashmir Earthquake in Pakistan
was primarily due to the collapse of inappropri- Clinton’s (2006) “Key Propositions for Building
ately built structures constructed on earthquake- Back Better” was the earliest known official
prone land using substandard building materials document to be published which attempted
and designed with little earthquake resistance. to provide a comprehensive guideline for
Poorly planned and sometimes illegal develop- implementing BBB practices in post-disaster
ments and their resulting impacts on the environ- environments. The report was based on and
ment worsened the damage from the Mumbai aimed at the Indian Ocean Tsunami disaster. He
Floods in 2005. A similar situation was seen in introduced ten propositions for building back
Sri Lanka after the Indian Ocean Tsunami. better.
Restoration of the damaged physical, social, Clinton’s propositions were:
economic, and environmental impacts of disas-
ters is a complicated and drawn-out process. • Proposition 1: Governments, donors and aid
Reconstruction and recovery projects often agencies must recognize that families and
focus on quick restoration of affected communi- communities drive their own recovery.
ties which can replicate and worsen existing vul- • Proposition 2: Recovery must promote fair-
nerabilities faced by the community. The ness and equity.
Tsunami Evaluation Commission Synthesis • Proposition 3: Governments must enhance
Report in 2007 provided examples where esca- preparedness for future disasters.
lated pressures and the need for fast rebuilding • Proposition 4: Local Governments must be
and recovery processes following a disaster can empowered to manage recovery efforts, and
further increase the vulnerability of a community. donors must devote greater resources to
Examples include: nonadherence to design and strengthening Government recovery institu-
construction policies for buildings and tions, especially at the local level.
330 “Build Back Better” Principles for Reconstruction
offer easy access to recovery-related information care in New Orleans, which were the foundations
for the community is also recommended. of the city’s economy, eventually leading to
Upholding a sense of community spirit and a decline in population numbers as people
improving community cohesion through organiz- moved away in search of better opportunities.
ing group activities are recommended for social Post-disaster recovery efforts to date have B
recovery. The Canterbury Earthquake Recovery shown support for economic recovery with strat-
in Christchurch proposed sports, recreation, arts, egies such as: “cash-for-work” programs, provi-
and cultural programs to engage the community sion of business grants, “asset replacement”
and provide a sense of normality. programs to provide industries with necessary
One of the first steps to be taken in post- resources, and training programs to up-skill
disaster recovery efforts in order to build back locals and help them find work. In Aceh, Indone-
better is to understand the local context of the sia, tsunami-affected people were trained and
affected community through needs assessments employed in reconstruction to provide them
and surveys in order to provide appropriate assis- with a source of income alongside the opportu-
tance to satisfy the community (Khasalamwa nity to become involved in their own recovery
2009). The reconstruction and recovery policies (Kennedy et al. 2008). In Japan following the
must then be developed based on local require- 2011 earthquake and tsunami, the government
ments to support and preserve the local culture decided to consolidate smaller fishing markets
and heritage. Batteate (2006) stated that into large fishing centers to enable fishermen to
maintaining community involvement throughout support each other (Okuda et al. 2011). The
recovery is integral for BBB success. The impor- Christchurch City Council’s Central City Plan
tance of decentralization to empower disaster- proposes fast-tracking of building consents for
affected communities by enabling them to take businesses to allow faster repair and construction
responsibility of the recovery effort and become work. Despite the implementation of such initia-
involved in decision-making has been stressed by tives, post-disaster economic recovery is report-
literature. The establishment of community edly slow and below pre-disaster levels. The lack
consultation groups is an effective way to com- of success in economic recovery initiatives can
municate with the community. Community con- be attributed to insufficient backing from policies
sultation groups consisting of community leaders and legislation for employment creation and lack
from preexisting community groups and reputed of consideration given to the needs of affected
members of the community to liaise between the communities.
wider community and governmental authorities Clinton (2006) said in his BBB propositions
have been successful in Sri Lanka and India. that “a sustainable recovery process depends on
Existing community groups can also be called reviving and expanding private economic activity
upon to assist with recovery activities. and employment and securing diverse livelihood
opportunities for affected populations.” Thus, the
Economic Recovery uniqueness of BBB comes from the integrated
Supporting economic recovery of the community approach it proposes by giving economic recov-
and supporting livelihood regeneration and entre- ery as much importance as reconstruction and
preneurship are also an important part of recov- aiming to provide solutions to suit local dynamics
ery. Disasters cause damage to the economy of and preferences.
communities with the disruption of businesses Monday (2002) stated that one of the first steps
and income-generating industries leading to needed for effective economic recovery is to
issues such as high inflation rates and poverty. obtain accurate information about the local pop-
The adverse effects of disasters on the economy ulation through data collection and consultation
can also impede the overall recovery of a city. with local governmental authorities, and
Hurricane Katrina displayed a disaster’s long- a comprehensive economic recovery strategy
term impacts on higher education and health must be created that is tailor-made to suit each
334 “Build Back Better” Principles for Reconstruction
knowledge between organizations to help recov- due to unawareness (Boano 2009). The National
ery activities (FEMA 2000). Knowledge from Post-Tsunami Lessons Learned and Best Prac-
past disasters should be retained and transferred tices Workshop held in Sri Lanka in 2005
to the government and other relevant organiza- highlighted the importance of training stake-
tions who will be involved in future post-disaster holders (especially external NGOs) about B
efforts. Twigg (2007) recommends that local existing and newly introduced legislation and
government should be included as a key stake- regulations. The community’s support can also
holder in the recovery effort and also given the be obtained by educating them about legislation
responsibility to manage local-level activities. and regulations that must be adhered to in recon-
Another obstacle preventing successful struction and recovery.
BBB-centered recovery is the absence of proper Post-disaster legislation can also be used to
controls to enforce BBB principles. Having BBB simplify and assist recovery activities to speed
knowledge and producing recovery plans in-line up the recovery process. Legislation that is cus-
with these principles are futile without proper tomarily used to impose security and safety con-
legislation and regulations in place to ensure trols (such as building consents) can become an
they are implemented. A common challenge in obstacle in high-pressure post-disaster environ-
post-disaster environments is the sudden ments. Time-consuming procedures, insufficient
increased work load, especially in the building resources to process permits, and the lack of fast-
industry, along with a drop in the workforce tracked methods delay reconstruction. Delays in
across local organizations which slow down and permits were a major reason for the holdup in
impede recovery activities. Post-disaster recon- housing repair and rebuilding following the 2005
struction requires time-consuming activities such Bay of Plenty storm in New Zealand (Middleton
as hazard analysis, land selection, infrastructure 2008). Fast-tracked consenting procedures, col-
development, and rebuilding to be done in laboration with other local councils, and open
a relatively short period of time. It is important access to information between stakeholders can
to facilitate recovery-related activities by simpli- help speed up recovery.
fying, fast-tracking, and exempting certain rules Legislation can be used to remove unneces-
and regulations using special legislation. sary red tape to facilitate recovery activities.
Post-disaster legislation can be used to ensure Meese III et al. (2005) reported a good example
compliance with BBB-based activities as well as in the recovery following the 1994 Northridge
to facilitate normal operations to improve the Earthquake, USA, where legislative suspensions
efficiency of recovery efforts. The lack of and emergency powers greatly reduced highway
enforcement of hazard-related laws and adequate reconstruction time. The construction work pro-
risk-based building controls contributed to the vided employment and opening up the highways
large-scale devastation caused by the 2004 Indian soon after the disaster helped boost the economy.
Ocean Tsunami (DNS and PA 2005). The same
was seen in countries like Pakistan, Turkey, Monitoring and Evaluation
Samoa, and Haiti. Enforcing updated risk-based The effectiveness and efficiency of post-disaster
building design standards through the use of com- reconstruction and recovery activities is crucial
pulsory building codes and maintaining construc- to the success of a community’s restoration fol-
tion standards through careful inspections is an lowing the impact of a disaster event. Having the
important regulatory requirement in reconstruc- knowledge of Build Back Better concepts in
tion (James Lee Witt Associates 2005). Lack of designing recovery programs is insufficient with-
awareness and understanding of new legislation out systems in place to overlook and monitor
can also lead to noncompliance. In the post- implementation. The creation of a recovery strat-
tsunami recovery effort in Sri Lanka, external egy to assist in conducting post-disaster recon-
nongovernmental organizations (NGOs) who struction and recovery activities is a common
took part did not comply with local standards response following disaster events. Despite
336 “Build Back Better” Principles for Reconstruction
having recovery strategies and revisions in legis- disaster interventions and establish lessons learnt.
lation and regulation to improve recovery activi- Lessons learnt should be incorporated into revis-
ties, the findings by Tas (2010) indicated that ing policy and procedures for future disaster man-
compliance was not monitored in the respective agement practices. Bakir (2004) recommends
recovery efforts in Sri Lanka and Turkey, leading that public education campaigns should be run
to poorly executed recovery projects. The lack of on lessons learnt, including the community in
properly trained professionals who were compe- participatory disaster management. Public semi-
tent in post-disaster environments and disaster nars have been held and advice notes have been
management activities poorly affects the out- distributed in Australia during the Victorian
come of recovery efforts. The shortage of effec- Bushfires recovery to keep the community
tive information and knowledge sharing and informed about revised guidelines and standards.
dissemination are also reasons for unsatisfactory Workshops have been held in the Philippines,
disaster management practices. Findings from Japan, and California involving the community
the Business Civic Leadership Center in 2012 in vulnerability identification which have been
on “What a Successful Recovery Looks Like” successful (Batteate 2006).
raised concerns that long-term recovery beyond
reconstruction often does not take place due to
the lack of mechanisms and expertise which pre- Summary
vents affected communities from satisfactorily
“building back better” in the long run. “Build Back Better” is an important concept for
Recommendations to improve post-disaster post-disaster reconstruction and recovery, signi-
recovery efforts through monitoring and evalua- fying the need to use reconstruction as an oppor-
tion have been provided in many sources of liter- tunity to not only recover from the encountered
ature. The role of monitoring and evaluation is disaster but to improve the resilience of commu-
twofold: (1) to monitor and ensure compliance of nities to face and withstand future disaster events.
recovery activities in accordance with the recov- BBB represents adopting a holistic approach
ery strategy in place and relevant guidelines and toward recovery by addressing risk reduction of
regulations (Clinton 2006) and (2) to obtain les- the built environment, psychosocial recovery of
sons for the future and improve future disaster affected people, and rejuvenation of the economy
management and post-disaster reconstruction and in an effective and efficient manner. Risk reduc-
recovery efforts (Monday 2002). tion can be achieved primarily through the
The 2003 Bam earthquake reconstruction pro- improvement of structural designs in buildings
vided a good example where rebuilding was mon- and infrastructure and through better risk-based
itored by providing construction supervision land-use planning. BBB requires improved build-
which assisted in assuring the quality of the ing codes and land-use plans to be enforced using
rebuild. Clinton (2006) stated that the Tsunami a strong legal framework along with financial
Recovery Impact Assessment and Monitoring backing to encourage adoption. Quality assur-
System (TRIAMS) was put in place during the ance of the rebuild is also integral for building
Indian Ocean Tsunami recovery for the most back better.
affected countries. The recovery strategy in Psychosocial recovery needs to be addressed
Christchurch, New Zealand, has also been to assist communities with moving forward with
equipped with monitoring mechanisms. Clinton their lives as an important part of overall commu-
(2006) suggested that long-term recovery should nity recovery. Psychosocial recovery of affected
be monitored through continued data collection people needs to be assisted through the provision
to ensure that recovery efforts do not leave com- of support services such as personal case man-
munities with residual issues. agement, counseling, and social activities. Inclu-
Monday (2002) pointed out that monitoring sion of community members in recovery
can be used to identify problems with post- activities is another way to support psychosocial
“Build Back Better” Principles for Reconstruction 337
recovery and provide recovery solutions that are practical guidelines to design future reconstruc-
in-line with community needs as part of building tion and recovery efforts including all these key
back better. facets to effectively build back better. Further
Economic recovery is essential for the recov- comprehension of how these strategies for build-
ery of communities. An informed economic strat- ing back better can be more successfully B
egy to address and support community-specific implemented in different environments can be
issues is the first step toward BBB-based eco- gained by studying different disaster events in
nomic recovery. Financial assistance, training, the future. It is also suggested that criteria for
and business rebuilding support need to be pro- measuring levels of resilience should be
vided to assist with economic recovery. established which can serve as indicators to mea-
Reconstruction and recovery requires effec- sure progress and effectiveness of build back
tive management of stakeholders and the use of better practices.
post-disaster legislation and regulation in order to The long-term sustainability of resilience in
build back better. The creation of a recovery communities instilled by using BBB concepts
authority to allocate roles and coordinate and depends on how they are linked with on-going
manage stakeholders is recommended. Success- developmental strategies. It is therefore impor-
ful recovery requires local-level partnerships and tant for the key concepts identified in this chapter
contribution to provide locally viable recovery to be incorporated into local and national govern-
solutions. Compliance of BBB-based concepts ment policies for community planning and devel-
in recovery needs to be ensured through the use opment even during non-disaster periods.
of appropriate post-disaster legislation and regu-
lation to enforce risk reduction and community
recovery initiatives. Legislation and regulation Cross-References
can also be used to facilitate post-disaster recov-
ery activities by fast-tracking and exempting nor- ▶ Building Codes and Standards
mal procedures. ▶ Building Disaster Resiliency Through Disaster
The effective implementation of risk reduc- Risk Management Master Planning
tion and community recovery initiatives concur- ▶ Community Recovery Following Earthquake
rently will result in building back better. Disasters
Recovery efforts also need to be monitored con- ▶ Economic Recovery Following Earthquakes
tinuously through short-term and long-term Disasters
recovery to ensure compliance with BBB con- ▶ Land Use Planning Following an Earthquake
cepts and to obtain lessons to improve future Disaster
disaster management efforts. ▶ Legislation Changes Following Earthquake
Disasters
Boano C (2009) Housing anxiety and multiple geogra- Monday JL (2002) Building back better: creating
phies in post-tsunami Sri Lanka. Disasters a sustainable community after disaster. Natural
33(4):762–785 Hazards Informer [Online], 3. Available http://www.
Clinton WJ (2006) Lessons learned from tsunami recov- colorado.edu/hazards/publications/informer/infrmr3/
ery: key propositions for building back better. Office of informer3b.htm. Accessed 27 Feb 2013
the UN Secretary-General’s Special Envoy for Tsu- Okuda K, Ohashi M, Hori M (2011) On the studies of the
nami Recovery, New York disaster recovery and the business continuity planning
Davidson CH, Johnson C, Lizarralde G, Dikmen N, for private sector caused by Great East Japan earth-
Sliwinski A (2007) Truths and myths about commu- quake. Commun Comput Inf Sci 219:14–21
nity participation in post-disaster housing projects. Pathiraja M, Tombesi P (2009) Towards a more “robust”
Habitat Int 31(1):100–115 technology? Capacity building in post-tsunami Sri
DN & PA (2008) Disaster and development in South Asia: Lanka. Disaster Prev Manag 18(1):55–65
connects and disconnects, South Asia disaster report. Ruwanpura KN (2009) Putting houses in place: rebuilding
Duryog Nivaran & Practical Action, Colombo communities in post-tsunami Sri Lanka. Disasters
DNS & PA (2005) Tackling the tides and tremors: South 33(3):436–456
Asia disaster report 2005. Duryog Nivaran Secretariat Tas M (2010) Study on permanent housing production
& Practical Action – South Asia Programme, Colombo after 1999 earthquake in Kocaeli (Turkey). Disaster
FEMA (2000) Rebuilding for a more sustainable future: Prev Manag 19(1):6
an operational framework. FEMA report. Federal Twigg J (2007) Characteristics of a disaster-resilient
Emergency Management Agency, Washington, DC community – a guidance note. DFID Disaster Risk
Florian S (2007) Housing reconstruction and rehabilita- Reduction Interagency Coordination Group, London
tion in Aceh and Nias, Indonesia – Rebuilding lives. Wisner B, Blaikie P, Cannon T, Davis I (2004) At risk:
Habitat Int 31(1):150–166 natural hazards, people’s vulnerability and disasters.
Halvorson SJ, Hamilton JP (2010) In the aftermath of the Routledge, New York
Qa’yamat: the Kashmir earthquake disaster in northern
Pakistan. Disasters 34(1):184–204
James Lee Witt Associates (2005) Building back better
and safer: private sector summit on post-tsunami
reconstruction. James Lee Witt Associates, Washing-
ton, DC Building Codes and Standards
Kennedy J (2009) Disaster mitigation lessons from “build
back better” following the 26 December 2004 Tsu- P. G€ulkan1 and Robert K. Reitherman2
namis. In: Ashmore J, Babister E, Kelman I, Zarins J 1
Department of Civil Engineering, Çankaya
(eds) Water and Urban Development Paradigms, Tay-
lor and Francis Group, London University, Ankara, Turkey
2
Kennedy J, Ashmore J, Babister E, Kelman I (2008) The Consortium of Universities for Research in
meaning of ‘build back better’: evidence from post- Earthquake Engineering, Richmond, CA, USA
tsunami Aceh and Sri Lanka. J Conting Crisis Manag
16(1):24–36
Khasalamwa S (2009) Is ‘build back better’ a response to
vulnerability? Analysis of the post-tsunami humanitar- Synonyms
ian interventions in Sri Lanka. Nor J Geogr
63(1):73–88 Codes of practice; Guidelines; Regulations
Mannakkara S, Wilkinson S (2012a) Build back better
principles for economic recovery: the Victorian bush-
fires case study. J Bus Contin Emerg Plan
6(2):164–173
Mannakkara S, Wilkinson S (2012b) Build back better Introduction
principles for land-use planning. Urban Des Plan
166(5):288–295 Selected key elements of modern seismic codes
Meese E III, Butler SM, Holmes KR (2005) From tragedy
are reviewed here, along with a brief review of
to triumph: principled solutions for rebuilding lives
and communities, Heritage special report. The Heri- their historical development. What can be con-
tage Foundation, Washington, DC sidered a single seismic code pertaining to
Middleton D (2008) Habitability of homes after a disaster. a nation or a subdivision thereof consists of
In: 4th international i-REC conference on building
dozens of pages of provisions and by reference
resilience: achieving effective post-disaster recon-
struction. International Council for Research and Inno- includes thousands of pages of standards that
vation in Building and Construction, Christchurch relate to loads, testing procedures, manufactured
Building Codes and Standards 339
boundaries provides a better technical basis for its created between innovative designs and tradi-
seismic and other provisions, the more important tional safety concerns. It also requires a high
motive for such a code of European scope is degree of professionalism, because it gives the
economic. The Euro economic block can com- architect and engineer more leeway, as compared
pete more effectively against the nations outside to more prescriptive requirements, and also
it if its design, construction, and building mate- requires a higher level of building code enforce-
rials industries are guided by consistent ment review. Seismic isolation, inclusion of
provisions. damping devices, response history analyses, and
displacement-based design are some of the inno-
vative approaches currently in use in some places
Prescriptive and Performance-Based or where this higher level of design and review
Objective-Based Codes capacity is present. Each of these represents
challenges in analysis and design that require an
Building code requirements are usually intimate knowledge of the underlying mechanics-
a combination of prescriptive requirements that based mathematical theory and its limitations. As
spell out exactly how something is to be done, on such, they are best performed by professionals
the one hand, and performance requirements that with considerable experience because there are
just outline what the required level of perfor- alternative approaches that require deep insight
mance is and leave it up to the designer how on the part of the engineer.
this is achieved, on the other. An example of the Seismic codes begin with the goal of provid-
former would be a rule of thumb for spacing of ing safety, and many stop there in most respects,
anchor bolts in house construction; an example of but some include requirements for protecting the
the latter would be to have the engineer calculate functionality of essential buildings, such as fire
interstory drift and then design concrete cladding stations, hospitals, and emergency communica-
to accommodate that building-specific distortion. tions centers or data processing centers. This is
In recent years there has been a move among discussed in the separate chapter on Essential
many building codes to move to more perfor- Facilities and is only mentioned in passing here.
mance requirements and less prescriptive Some of the most stringent regulations of this
requirements. Performance-based code require- type were passed in California after the 1971
ments still require tight definitions so that ade- San Fernando Earthquake, when the Hospital
quate performance can be evaluated by the Seismic Safety Act of 1972 was passed. The
building regulatory agency. The fire protection Veterans Administration adopted its own regula-
field has developed performance-based design tions after that earthquake, regulations that seek
approaches for many years, in which test or to not provide not only safe hospitals but also
other data can be used to provide alternate more functional ones. Some voluntary above-
means of fire protection instead of following the code (“performance-based design”) approaches
prescriptive requirements of a code. lead an owner to invest in the cost of greater
In recent years, several countries, beginning seismic protection to achieve less property dam-
with Australia, have moved to much shorter age in earthquakes, but the most common seismic
objective-based building codes. Rather than pre- design criteria that go beyond the goal of provid-
scribing specific details, objective-based codes ing safety are related to protecting essential
list a series of objectives all buildings must meet functions.
while leaving open how these objectives will be
met. When applying for a building permit, the
designers must demonstrate how they meet each Geologic and Geotechnical Topics
objective. This makes it necessary for approving
authorities to employ correspondingly qualified Mentioned briefly here are geologic and geotech-
personnel so that a productive synergy can be nical engineering aspects to the seismic
Building Codes and Standards 341
provisions of a building code. Construction reg- earthquake nonstructural damage was done for
ulations that deal effectively with earthquake the 1964 Alaska Earthquake (Ayres et al. 1973),
vulnerabilities must include within their purview with a similar report on the 1971 San Fernando
geologic hazards such as surface faulting and Earthquake (Ayres and Sun 1973). Nonstructural
liquefaction. Geologic and geotechnical engi- provisions are only briefly mentioned here but are B
neering provisions to deal with those hazards an essential part of the modern seismic building
came later than structural provisions. In 1972, code. Approximately three-quarters of the initial
the state of California passed the Alquist-Priolo value of a modern building’s construction, dwell-
Fault Zoning Act (the current name, which has ings excepted, is composed of nonstructural com-
changed over the years), and slightly before that ponents, such as enclosure systems, elevators,
the city of Portola Valley, which is bisected by partitions, fire sprinklers, ceilings, and heating-
the San Andreas Fault, passed its own legislation. ventilating-air-conditioning equipment. In gen-
Surface fault zoning is the most clearly delin- eral it can be said that the leading seismic codes
eated of the earthquake ground failure hazards, of the world have increasingly included more and
and thus it was logically the first to be subject to more detailed regulations for the calculations of
regulations. Liquefaction was not a word that loads that are used to design nonstructural brac-
soils engineers (then the term for geotechnical ing and anchorage, but a lack of coordination of
engineers) used prior to the 1964 Alaska and the the implementation of these rules still persists.
closely following 1964 Niigata Earthquake, but Different design professionals work on their own
in the following decades, areas of a jurisdiction systems – architects on partitions and ceilings,
that were suspect in terms of poorly compacted fire protection engineer on fire sprinkler system,
granular soils and high water table were fre- mechanical engineer on the HVAC system, and
quently zoned as requiring special geotechnical so on – and yet it is the structural engineer with
investigations. Geotechnical engineers now have the most expertise for such work. And then dif-
at their disposal field and laboratory investigation ferent contractors and trades work on the imple-
techniques and a knowledge base about how soils mentation phase, often with little guidance or
behave in earthquakes that did not exist in the oversight. Recent earthquakes have demonstrated
1960s. Codes and standards of practice can now that simply having nonstructural damage protec-
require geotechnical evaluations that would have tion rules on the building code’s books is not
been impossible to perform a few decades ago. sufficient to actually protect that myriad of com-
One tool that developed out of non-seismic ponents. A common example is that some codes
slope stability concerns that later was seen to for many years have contained a clause stating
have a seismic benefit is the grading ordinance. that nonstructural components must be designed
The city of Los Angeles passed such a law in for the motions and deformations that the struc-
1952 controlling how excavation and cut and fill ture imposes upon them, but in actual practice,
could be performed. Such ordinances have partitions and other components were
become common, and the geotechnical engineer- constructed without this engineering.
ing evaluations required in the design stage now
typically include earthquake considerations in
regions where a seismic code is enforced as well Historic Context
as non-seismic landslide hazards.
Some accounts of historic building code evo-
lution claim that the Code of Hammurabi
Nonstructural Code Provisions (a Babylonian king of some 3,800 years ago
who was considered by his subjects to be a god
Coming later than structural regulations were as well) engraved on stone is the first building
provisions to protect the nonstructural features code ever to have been put into effect. That
of a building. The first detailed report on notion needs to be dispelled because the Code
342 Building Codes and Standards
analysis, applying engineering theory to a given load was sometimes thought to be a surrogate for
case and calculating loads and resistances. It was seismic load.
an attempt at earthquake-resistant construction, A few years later, after the 1910 Cartago
not earthquake engineering per se, because the Earthquake in Costa Rica, a national code was
engineering techniques of the nineteenth and passed that phased out adobe and trapia (rammed B
twentieth centuries were far off in the future. It earth) structures, substituting bahareque con-
was decreed in royal fashion for use in a city that struction (sawn lumber or bamboo framing with
had experienced one of the most destructive plaster). This was a similar but more regulatory
earthquakes in Europe and was based on the approach to the substitution of wood for masonry
construction tradition in the Mediterranean that occurred in construction styles after some
basin and may have borrowed construction lore other earthquakes, such as the 1855 Wairarapa
developed by shipbuilders. It was used on the (Wellington, New Zealand) Earthquake.
interior of buildings that consisted of timber
frames with vertical and horizontal timbers of
approximately 10–12 cm square, with internal The First Seismic Codes with
braces, forming an “X,” referred to in Italy and Engineering Content
Portugal as the “Cross of St. Andrew.” The tim-
bers for the cross were 9 cm by 11 cm in section. Soon after the 1906 California Earthquake, the
The frame was then “nogged” (i.e., filled with 1908 Messina-Reggio Earthquake occurred in
brick) in the triangular spaces formed by the southern Italy, but it had a different and more
crosses with a mixture of stone rubble, broken beneficial impact on building codes. Sorrentino
brick, and square pieces of Roman brick in dif- (2007, 2011) relates how a committee of engi-
ferent patterns in each panel. The interior walls neers developed an equivalent lateral force anal-
were then covered with plaster, hiding the infill ysis method that was adopted into the building
and the timber frame. The exterior facades of the code of the disaster region. It even had the refine-
Baixa buildings were reconstructed with load- ment of applying a different lateral force coeffi-
bearing masonry walls of about 60 cm in thick- cient (rapport sismica) to the ground story (1/12)
ness, some of which had a timber frame on the than the story above (1/8), accounting for greater
inside face (G€ulkan and Langenbach 2004). acceleration in the second or third stories. After
The 1906 San Francisco Earthquake was the 1915 Avezzano (or L’Aquila) Earthquake, the
something of a nonevent in the history of seismic code was changed to have factors of 1/8 for the
codes. Although the cause of the earthquake was ground story and 1/6 for the upper stories. These
accurately noted – the San Andreas Fault had Italian seismic provisions applied only to the area
been mapped in several areas before the earth- affected by the earthquakes, a pattern that was
quake and its surface rupture was well common in early codes prior to the availability of
documented afterward – and although civil engi- national maps depicting the hazard of seismic
neers trained in universities were in existence, no ground shaking.
engineering consensus existed around provisions At the turn of the nineteenth to twentieth cen-
that would define how to calculate loads and then tury, Riki Sano was getting his PhD from the
calculate the lateral resistance of a structure. University of Tokyo on “Seismic Design Concept
Unreinforced masonry was not prohibited, and for Building Structures,” along with Tachu Naito,
by the time the city took stock of its unreinforced his student; Sano was to be instrumental in the
masonry building seismic hazards in the 1980s, development of the Japanese equivalent lateral
over 90 % of the 2,000 buildings that required force method. Their work came to fruition after
retrofitting were dangers that were built after the the 1923 Great Kanto or Tokyo Earthquake of
earthquake, not before it. The San Francisco 1923, when the 1924 Urban Building Law
building code even reduced the wind load a few Enforcement Regulations were passed. It used
years after the earthquake. In that era, the wind a 10 % lateral force factor (shindo) applied
344 Building Codes and Standards
uniformly up the height of the building. As an building confirmation and construction inspec-
example of how engineering practice in some tion works, (c) deregulating urban land use, and
cases is in advance of the codes, when the 1923 (d) allowing public survey of design and inspec-
earthquake occurred, three large buildings tion documents. New technical specifications in
designed by Tachu Naito had already been built the form of the Law Enforcement Order and
(and performed well in the earthquake, giving the a series of Notifications of the Minister of Con-
new equivalent lateral force method a boost into struction were issued in June 2000, including the
the code). definition of performance objectives at design
An important threshold was crossed in Japan limit states and the specifications for verification
toward the implementation of scientifically deter- methods.
mined national seismic provisions (Otani 2008). California’s contribution to the development
The objectives of the Japanese Building Standard of the modern seismic code based on engineering
Law, proclaimed in May 1950, were “to safe- calculations had a small start after the 1925 Santa
guard the life, health, and properties of people Barbara Earthquake, and then after the 1933
by providing minimum standards concerning the Long Beach Earthquake, statewide regulations
site, structure, equipment, and use of buildings.” were passed that were derived from the Japanese
The law outlined the basic requirements, and the code. The 1930s was the decade when the lineage
technical details were specified in the Building of seismic codes in several countries began. Pro-
Standard Law Enforcement Order (Cabinet visional earthquake regulations were enacted in
Order) and in a series of Notifications by the Chile in 1930 after the Talca Earthquake and
Minister of Construction. became more institutionalized after the 1939
The seismic design provisions of the Building Chillán Earthquake. In New Zealand, following
Standard Law Enforcement Order were signifi- the 1931 Hawke’s Bay Earthquake, the 1935 New
cantly revised in 1981, including the following Zealand Standard Model Building By-Law
three major changes: included earthquake regulations. India adopted
its first seismic code in 1935 following destruc-
1. Structural calculations are required to exam- tive earthquakes in 1931 and 1935 in Quetta (now
ine (a) maximum story drift under design part of Pakistan), and the 1939 Erzincan Earth-
earthquake forces, (b) lateral stiffness distri- quake in Turkey led to its first engineering regu-
bution along the height, (c) eccentricity of lations for the earthquake hazard.
mass and stiffness in plan, and (d) story
shear-resisting capacity at the formation of
a collapse mechanism. Equivalent Static Elastic Lateral Force
2. Earthquake resistance is specified (a) in terms Method
of story shear rather than horizontal floor
forces, (b) as a function of fundamental period A seismic load on a building is actually dozens of
of a building and soil type, and (c) separately different significant seismic loads that occur dur-
for the allowable stress design and the exam- ing the 20–60 s during which the ground is
ination of story shear-resisting capacity. strongly shaking. The earliest practical way to
3. Required story shear-resisting capacity is var- “put a number” on that bewildering set of loads,
ied for construction materials and with the which are not well known in advance of the
deformation capacity of hinging members earthquake, was to represent the worst loading
under earthquake forces. effect as a percentage of the weight of the struc-
ture, recognizing that seismic shaking causes
The framework of the law was also signifi- inertial loads, and inertia is a product of mass
cantly revised in 1998, (a) introducing and acceleration. The full descriptive name of
performance-based regulations wherever feasi- the method is the equivalent static elastic seismic
ble, (b) allowing private agencies to execute the lateral force analysis method. It is intended to be
Building Codes and Standards 345
equivalent to or adequately represent the actual enough to track rapid and severe motion of the
earthquake forces; the method computes a single underlying strata, and soil can behave nonlinearly
static force to represent the changing dynamic just as structural materials do. The limiting case
forces; that force is applied to an analytical struc- of microzonation is the site-specific evaluation,
tural model that remains elastic, even though where the construction to be placed on a site is B
inelastic behavior obviously occurs; and it is an designed for a customized set of earthquake
analysis method for determining only the design criteria, a more expensive approach usually
loads, not for distributing them through the struc- reserved for major facilities.
tural components and connections that provide On the structural side, knowledge of different
the resistance to the loads. And finally, the only structural systems and materials had to evolve for
design loads were lateral or horizontal, whereas codes to seismically regulate them, and a growing
strong motion records routinely measure some body of structural laboratory research provided
vertical motions as well. Each of these limitations that information. The steel and concrete indus-
was to be worked on in the following decades. tries in particular were the settings for a large
From these first incarnations, the equivalent lat- amount of research conducted in the 1950s and
eral force method was refined to include several later. Research on reinforced concrete in great
necessary considerations. part conducted at the University of Illinois was
Instead of applying seismic regulations to the technical basis for an important document
a region after it had an earthquake, earth scientists that advanced the use of that material in seismic
compiled maps showing where earthquakes codes (Blume et al. 1961). The other kind of
should be expected. This was first done with research that provided this knowledge was actual
a small number of large-scale zones for earthquake performance. Earthquake reconnais-
a country, each zone defining the force factor to sance or field surveys of the effects of earth-
be applied. Then beginning in the 1950s in Japan, quakes became increasingly common in the
a probabilistic basis to the maps was developed, 1960s.
Kawasumi (1951) developed three maps, The early codes provided loads computed with
depicting the shaking that should be expected to a slide rule that were low enough to keep the
occur in exposure periods of 75, 100, and structural model that was on the engineer’s draw-
200 years. The Applied Technology Council in ing board in the elastic range. (Computers did not
its ATC-3 document (Applied Technology Coun- become widespread in seismic design or other
cil, 1978) included national maps produced by civil engineering practice until the 1970s.)
S.T. Algermissen that had a probabilistic basis. Slowly, inelastic behavior was included in codes
In China in 1977 a national seismic design map by including requirements for ductility. Classify-
was based on shaking with 3 % and 10 % proba- ing structural systems and materials was an
bilities of being surpassed. Formats such as these approximate first approach. A structure thought
for tying the probability of occurrence to the to have greater ductility was designed for lower
severity of shaking have become standard in seis- elastic-level forces, because there was confidence
mic codes today. Another refinement has been in how it would perform when pushed into the
microzonation: zoning small areas as having dif- inelastic range. The ability of ductile connections
ferent ground shaking severities, or ground fail- and members to absorb punishment without fail-
ure hazard (e.g., liquefaction), based on local ure, and also the softening effect (period length-
soils. It became increasingly known that in gen- ening) and its beneficial effect on response,
eral soft soils amplified ground shaking, though became increasingly recognized in the 1960s,
the large differential response between soft and 1970s, and 1980s. Today, the choice of
hard grounds as measured in low-level shaking a structural system is heavily influenced by how
was not found to be linear up through high levels the code defines its ductility. Seismic structural
of ground motion: a column of soft soil 30 m high codes are all aware that ground motions are ran-
simply cannot move back and forth rapidly dom and can be determined only approximately
346 Building Codes and Standards
in advance. For this reason they contain provi- Concerning the growing number of earth-
sions to prevent unexpected surprises leading to quake records, the city of Los Angeles played
catastrophic failures. Lessons that have been an important role when it passed a revision to its
taught by past earthquakes for how building com- building code in 1965 requiring that three strong
ponents behave under ground shaking are embed- motion instruments be installed in buildings six
ded in their verbiage. As theory and experiment stories or taller (one at the base, one at
combine to lead to more refined ways of seismic mid-height, one at the roof). This resulted in the
protection, this knowledge must be incorporated large harvest of ground motion and structural
into the newer versions of codes for continuous motion records from the 1971 San Fernando
improvement. Earthquake. The city of Los Angeles also was
precocious in its 1943 adoption of a formula in
its code that related period of vibration as
The Response Spectrum Method represented by number of stories to the base
shear coefficient calculation in its equivalent lat-
Yet another improvement in seismic codes was eral force method: the taller the building, the
consideration of the dynamic properties of the lesser the base shear. In 1957, Los Angeles also
structure, its period or periods of vibration, and had a strong effect on the development of earth-
the amount of damping. The tendency of the quake codes, even though it was not an earth-
structure to respond at its natural frequency quake code revision: the zoning code was
was only useful information if the frequency revised to allow buildings taller than 150 ft. Engi-
content of the ground shaking was also known: neers in California thought that taller, more flex-
in structural dynamics, it takes two to tango. ible (longer period) buildings could be designed
Because the first strong motion seismograph for lower force coefficients, and they also knew
(accelerograph) was only deployed in 1932, it that if the same coefficients as were used for
took years for records to accumulate to provide low-rise buildings (typically 10 % to 13 %)
a statistical picture of earthquake severity in were applied to every story in a tall building,
terms of frequency content. Trifunac (2006) the large amount of strength, and thus area
traces the history of the response spectrum taken up by shear walls and other structural mate-
method of analyzing earthquake ground rial, would have very negative architectural and
motions and producing design spectra that real estate implications.
engineers could use to proportion required When it seemed that engineers in the two
resistance of the structure. Two of the most large urban regions of California, centered on
influential developers of the method were Mau- Los Angeles in the south and San Francisco in
rice Biot and George Housner, whose doctoral the north, could be diverging toward signifi-
theses at the California Institute of Technology cantly different provisions for dealing with the
(1932 and 1941, respectively) were devoted to tall building question and the modernization
this topic. The theory became widely applied in of seismic codes, the statewide structural engi-
engineering practice in the 1970s and 1980s neering organization Structural Engineers
when several conditions were favorable. First, Association of California (SEAOC) began its
there were more earthquake records. Second, influential set of editions of the “Blue Book,”
engineers could afford to have the new, more the Recommended Lateral Force Requirements
powerful electronic computers on their desks and Commentary (SEAOC 1959). Its require-
that could do the extensive mathematical work. ments became the seismic provisions of the
Housner (1997, p. 33) notes a third factor: Uniform Building Code used in various edi-
“Because of the practicing engineers’ reluc- tions and adaptations throughout California
tance to employ the design spectrum, I think and the Western United States until the year
it was essentially the nuclear power business 2000 when the UBC was folded into the IBC,
that got the spectrum into widespread use.” as previously described.
Building Codes and Standards 347
The Response History Method capable of substituting for the simulation of phys-
ical testing.
The equivalent static lateral force method leads the A 1973 benchmark for the development of the
engineer to calculate a single quantity, the total response history method is provided by the seis-
shear at the base of the building, the base shear, mic design manual of the US military B
and then proportion that load up the height of the (Departments of the Army, the Navy, and the
structure according to procedures that attempt to Air Force 1973, pp 3–3): “Since the mechanics
represent the dynamic response of the building in of dynamic analysis requires that a separate solu-
a simple way. While there is dynamic thinking tion must be obtained for each instant of time
underlying the method, it still represents the series during the entire history of interest, computation
of motions that occur during the earthquake with by computers is necessary. This kind of analysis
a static view or “snapshot” of the overall effect of is generally beyond the kind of effort that can be
those motions. The response spectrum method afforded in the design of almost all but the most
similarly is a way to give the engineer a base critical structures.” Today, the method is more
shear to use in design. A given overall representa- frequently used, though it still is applied to
tion of one record, for example, the peak ground a small minority of all the structures being seis-
acceleration, might exceed that of another record mically designed. Chopra (2005, p. 107) provides
and yet that other record might be more damaging an assessment of the future of the method:
to a particular structure. A record with greater At the present time, nonlinear RHA [response his-
duration, more pulses above a threshold such as tory analysis] is an onerous task, for several rea-
the elastic limit of a portion of the structure, can sons. First, an ensemble of site-specific ground
create more demand than a record of smaller dura- motions compatible with the seismic hazard spec-
trum for the site must be simulated. Second, despite
tion, even if the latter has a greater peak value increasing computing power, inelastic modeling
(Johnson 2013). The logical improvement was and nonlinear RHA remains computationally
seen to be to calculate the various forces and demanding, especially for un-symmetric-plan
deformations that occurred to a structural model buildings—which require three-dimensional anal-
ysis to account for coupling between lateral and
split second by split second in response to the torsional motions—subjected to two horizontal
changing ground motion. components of motion. Third, such analyses must
Three sources of research information that be repeated for several excitations because of the
developed from the 1960s to the present day wide variability in demand due to plausible ground
motions, and the statistics of response must be
provided the basis for incorporating a fully considered. Fourth, the structural model must be
dynamic method into seismic codes, subjecting sophisticated enough to represent a building real-
the analytical model of the structure to the istically, especially deterioration in strength at
motions of several recorded earthquakes. One large displacements. Fifth, commercial software
is so far not robust enough to predict response
came from completely outside the earthquake with high reliability. Sixth, an independent peer
engineering field, the development of the modern review of the results of nonlinear RHA is required
computer. A second prerequisite was a large by the FEMA-356 guidelines, adding to the project
library of earthquake records, conveniently sup- duration and cost. With additional research and
software development, most of the preceding
plied by growing coverage of highly seismic issues should be resolved, and nonlinear RHA
areas with strong motion instruments and, unfor- may eventually become the dominant method in
tunately, by Earth’s frequent earthquakes. The structural engineering practice.
third was improved knowledge of structural
behavior, in particular of how structures behaved
in the inelastic range. While computers played Growth in Adoption of Seismic Codes
a role in that research on structural behavior, it
has also required the traditional approach of Figure 1 provides a quick way to see how seismic
subjecting specimens to simulated seismic load- codes have spread from an initially small number
ing in the laboratory. Computer simulation is not of countries. The numbers are drawn from the
348 Building Codes and Standards
Building Codes and Standards, Fig. 1 Growth in the adoption of seismic codes (Reitherman 2012, p. 582)
various editions of Regulations for Seismic listed only a few specific nonstructural compo-
Design: A World List, published by the Interna- nents that needed anchorage and then added
tional Association for Earthquake Engineering a general phrase such as “all other equipment
(1960 and later editions). and machinery.” That sounds comprehensive,
One could argue that because seismic codes but in the absence of a well-defined definition of
now cover most of the significantly seismic areas that all-encompassing phrase, it was not particu-
of the world, seismic safety has now been larly meaningful. Another example is the accu-
achieved on a global scale. This would be overly rate calculation of interstory drift, which in some
optimistic for two reasons. First, there are always codes was seemingly covered many decades ago
older, pre-code buildings that did not benefit from but in practice was not well-implemented in stan-
a modern seismic code. Secondly, even today in dard practice. The engineer who diligently tried
countries with regulations “on the books,” legally to meet that code requirement had inadequate
binding laws that include the latest earthquake analytical tools available because of the lack of
engineering thinking, those regulations are not underlying knowledge of the displacements that
always thoroughly carried out. would actually occur. Conversely, some engi-
The seismic provisions in a code must be neers incorporated design measures that were
complemented with reliable implementation of only required by the code much later. In the
the code, its effective enforcement. This remains words of George Housner (1986, p. 25), “in
one of the hardest issues to solve, especially in some instances, earthquake requirements were
poorer countries but also in the most developed. adopted in building codes but were not used by
Merely making seismic codes more sophisticated architects and engineers. And in other instances
as each new edition is promulgated every few earthquake design was done by some engineers
years is not a solution, unless comprehensive before seismic requirements were put in the
implementation also occurs. This includes the code.”
education and professional training of the engi-
neers and other design professionals involved as
well as the thoroughness of the quality control Summary
measures of building regulation agencies, such as
in the plan review and construction inspection Model building codes are common around the
phases. In some instances, the comprehensive world, although the consistency with which they
protection that the provisions in a building code are applied throughout a jurisdiction can vary
seem to provide exists mostly on paper. Up until greatly. In New Zealand and Japan, there is
the 1990s, even the most advanced seismic codes a high degree of consistency in how a nationally
Building Codes and Standards 349
management of major natural disasters, as well as important factor that makes satellite data able to
humanitarian crisis situations. Compared to face and provide useful information during the
15 years ago the availability of satellite imagery, crises phase soon after an earthquake.
amount, timeliness and the capability to cover a Both optical and SAR data have characteris-
certain crisis situation or disaster event with high tics to be used as a source of information to B
degree of detail have all improved substantially. provide a damage map after an earthquake
There are several factors which have led to this. (Yonezawa and Takeuchi 2001; Matsuoka and
First of all ground pixel spacing of civil Earth Yamazaki 2004; Chini et al. 2008a; Chini
Observation (EO) systems has developed to the et al. 2013), with particular attention to very
meter domain (or less) for optical and radar sys- high-resolution data, which has the potential to
tems, such as synthetic aperture radar (SAR) monitor wide areas, but at the same time giving
(Voigt et al. 2007). Indeed on September 24 information on single structures present in the
1999 the commercial EO satellite IKONOS was scene (Stramondo et al. 2008; Turker and Sumer
launched, the first able to collect openly available 2008; Brunner et al. 2010; Chini et al. 2011;
high-resolution imageries that were both pan- Dell’Acqua et al. 2011; Ferro et al. 2013). The
chromatic and multispectral with a spatial reso- huge amount of information carried by these
lution of 1 and 4 m, respectively. Two years later, kinds of data can hamper the visual inspection,
on October 18, 2001, another important step for- which is time consuming during a crisis phase,
ward was made with the launch of one more high- fostering the implementation of more automatic
resolution commercial EO satellite, QuickBird, approaches.
with an increased spatial resolution up to 0.6 m
for the panchromatic sensor and 2.4 m for the
multispectral one. The first earthquake where Methodology
this kind of data proved its usefulness in conjunc-
tion with automatic image-analysis techniques The automatic or semiautomatic classification of
was the one that struck the city of Bam (Iran) in remote sensing data is a tool to help understand
2003, provoking a lot of damage to man-made and monitor a large variety of scenarios,
structures and causing the death of many people converting them into tangible data, which can
(Yamazaki et al. 2005; Gusella et al. 2005; be utilized in conjunction with other data sources
Stramondo et al. 2006; Gamba et al. 2007; Chini to provide crucial and accurate information. Even
et al. 2008b; Chini et al. 2009; Bignami if metric or submetric resolution images provide
et al. 2011). Concerning SAR satellite sensors, new opportunities in the damage mapping field,
the spatial resolution breakthrough arrived in new problems arise at the same time in automatic
2007 with the launch of the first two satellites of classification procedures. Indeed, the aim of
the two X-band satellite constellations, COSMO- image classification analysis is to extract infor-
SkyMed and TerraSAR-X, on June 8 and on June mation and display the results in a simple and
15, 2007, respectively, both having the capability effective way for possible further studies. In met-
to acquire images in the so-called SpotLight ric resolution images a single pixel represents a
image mode with a spatial resolution of 1 m. small area (0.5
10 m2) of the acquired scene
TerraSAR-X mission is composed of two satel- and is usually smaller than the size of the objects
lites, the second one was launched on June of interest. Hence the information of a single
21, 2010, while COSMO-SkyMed has four satel- pixel is strongly linked to the information carried
lites, the fourth one launched on November by its neighbors, which are part of the same
5, 2010. The fact that they are constellations has object, pointing out the necessity to perform an
reduced the site access time down to two and a object-oriented analysis, although the
half days with TerraSAR-X and to 12 h with measureable statistics typically exploited for
COSMO-SkyMed. This latter issue is the second classification purposes are necessarily pixel
352 Building Damage from Multi-resolution, Object-Based, Classification Techniques
based (Blaschke 2010). This aspect is very inter- It is worth recalling the optical and SAR image
esting, because it makes possible to focus on a properties that are usually used for damage detec-
single object, for example, a building or also tion purposes, highlighting characteristics that
other man-made structure, having the advantage are relevant to damage mapping analysis such
of a valuable amount of pixels to characterize the as the physical properties of the measured signal
object itself. and the spatial resolution that can have an impact
Object-based techniques recognize that on the final products. Basically, the process for
important semantic information is not always forming a SAR image is different from that used
represented in single pixels but in meaningful for an optical image for two main aspects: the
image objects and in their contextual relations. physical properties of the imaged targets that
For example, it is more likely that a pixel adjacent contribute to the measured signal and the signal
to a private garden should be classified as a house, processing steps used to create the image. Optical
road, or private garden rather than as a forest. sensors measure the radiometric properties of
Usually, object-based classification comprises reflected sunlight in spectrally distinct regions
two steps: the image segmentation and the object of the visual and near-infrared spectrum or inte-
classification. Image segmentation subdivides the grated in a single panchromatic band. The mate-
image into groups of contiguous pixels called rial properties of objects, the illumination
objects or segments that correspond to meaning- conditions of the scene and the sensor perspective
ful features or targets in the field (Blaschke determine the radiometric and geometric appear-
2010). The images are segmented into homoge- ance of distinct targets in optical imagery. The
neous objects based on the spectral information SAR system is an active sensor and measures the
and local patterns or textural information that are backscatter of a transmitted signal, typically in a
included in groups of neighboring pixels. As a narrow microwave frequency band and sampled
result, the object-based classifications can con- in the range direction, i.e., the line of sight (LOS)
sider a wide range of variables such as mean from SAR sensor to target. Backscattering is pri-
reflectance, texture, shape and size of objects marily determined by the geometry and dielectric
and can potentially produce more accurate and properties of an object and the transmit/receive
detailed maps than conventional classification configuration of the SAR sensor (Brunner
strategies (Mathieu et al. 2007; Blaschke 2010; et al. 2010).
Chini et al. 2014). Concerning optical images it is easier for the
All algorithms exploiting metric resolution reader to grasp how different spatial resolutions
data for damage mapping purposes have to deal make buildings appear and more in general, cit-
with object-oriented approaches even if the final ies, where if the resolution is higher, the amount
target is not the object itself, but maybe just a of detail composing land use classes of interest
classification useful for restricting the analysis will be greater. Indeed, in high-resolution optical
and the statistics to classes we are interested images the radiance of each pixel is not directly
in. Such an example could be a change detection related nor can be considered representative of
approach to identify a collapsed building using the class it belongs to. However, many studies are
one pre-earthquake image and one post- able to demonstrate that the advantage of being
earthquake image by a threshold approach. At able to aggregate pixels to an object and then
this resolution many regions will be affected by address object characteristics, is that it can dis-
changes even those not relevant for damage cern between different kinds of buildings and
purposes. Thus to fix the threshold used to different kinds of asphalted areas such as roads
identify damaged buildings (pixels that have or parking lots, turning all of this augmented
changed their radiometric values because of the complexity of understanding into an information
earthquake) we should restrict the analysis to luxury (Pacifici et al. 2009). Such an example
objects of interest such as buildings (Chini could be the shadow from a building, which has
et al. 2009). at this resolution a clear signature in the image,
Building Damage from Multi-resolution, Object-Based, Classification Techniques 353
giving us information about the building height, the distance between SAR sensor and the
which could be also an important source of infor- facade base (all different scattering points are
mation for damage detection purposes. As a mat- positioned at a unique distance, corner
ter of fact, Turker and Sumer (2008) detected reflector).
damaged buildings from the big earthquake in 3. The contribution from the roof that could be of B
Turkey in 1999, applying a watershed segmenta- two types: (a) flat roof, a unique gray area
tion on a post-event optical image utilizing the corresponding to backscattering from the
relationship between the buildings and their roof, and (b) gamble roof, the roof has a dif-
shadows, labeling 80 % of buildings correctly as ferent scattering signature with respect to a flat
either damaged or undamaged. Another example one. Basically it is different due to the pres-
of how object-oriented approaches are useful is ence of a second bright scattering feature,
presented in Gusella et al. (2005) that quantified which is closer to the sensor than to the
the number of collapsed buildings of the Bam double-bounce contribution (Brunner
earthquake in 2003, starting from the inventory et al. 2010). This behavior concerns only the
of buildings as objects in QuickBird satellite direct backscattering from the part of the roof
imagery taken before the event. that is oriented toward the sensor.
Pertaining to SAR data an in depth description 4. A dark area corresponding to the SAR build-
of how a building appears is necessary. For this ing shadow which should be not confused with
purpose, it is useful to describe how radar returns the shadow concept in an optical image.
from different parts of the building are mapped Shadow in radar imagery is called the region
on the SAR image due to the fact that they reach where an object point is not reached by any
the receiver at different times (Kropatsch and radar beam. Such object points produce a
Strobl 1990; Franceschetti et al. 2002; “zero” signal in the image. Therefore, shadow
Franceschetti et al. 2003; Guida et al. 2008; regions appear as dark areas corrupted by
Guida et al. 2010). Given an isolated building noise (Kropatsch and Strobl 1990)
on ground and proceeding in the SAR image
with a constant azimuth from near to far range, Of course this is a simplistic schematization of
we can identify four main singular contributions: the phenomenon; other parameters also play an
important role in the backscattering values of the
1. A bright stripe corresponding to a mixture four building parts such as geometric parameters
between the backscattering from the ground, (height, length, and width of the building and the
facade, and roof, which is a typical example of roughness of the ground) and the electromagnetic
layover in a SAR image. Layover is called parameters (dielectric constants of facade, roof,
“radar mapping,” where different object and ground). The four types of building responses
points having the same time and same range with a distinct backscattering values can be
are mapped into one image point, i.e., more detected as separate objects if the resolution is
than one point on the Earth’s surface are higher than the dimension of a building itself,
mapped into one image point (many to one while this distinction is not possible if the SAR
mapping). SAR mapping is mainly an integra- resolution is lower than the buildings being iden-
tion of reflected signals having the same tified. As a matter of fact an urban area imaged by
Doppler frequency (azimuth or along-track a SAR sensor with a resolution in the order of tens
measurement) and same distance (range or of meters appears uniformly brighter than other
across-track measurement) (Kropatsch and land cover classes since the double-bounce
Strobl 1990). effects are the dominant scattering component
2. A very bright line corresponding to in the SAR resolution cell (i.e., the image pixel),
facade–ground and ground–facade double- that one characterizing the backscattering for
bounce scattering, because double-bounce such scenario. Whilst in metric resolution SAR
ray paths all have the same length, equal to sensors such as COSMO-SkyMed and
354 Building Damage from Multi-resolution, Object-Based, Classification Techniques
TerraSAR-X, we still have very bright values in as a smallest image object the pixel. Of course an
the city, principally double-bounce pixels, but we object can be defined in different ways,
have gray and dark regions as well, roofs and depending on our scope and at which scale of
shadows respectively, that are well identified detail we want to provide the information. An
and with a remarkable spatial extension. On the object can be a building or a neighborhood in a
contrary big uniform surfaces, such as bare soil or city with a particular type of structure and both of
asphalted industrial areas, appear the same in them can give us important information in terms
both high- and low-resolution SAR since their of damage mapping. The following are some
extension is bigger than both sensor resolutions. examples of applied object-oriented methodolo-
Change detection approaches with lower- gies in real scenario, where very high-resolution
resolution SAR stem from the concept that the optical and SAR data have been used for the
double-bounce effect fails when a building col- seismic events that occurred in Bam (Iran) in
lapses (which is the dominating scattering mech- 2003 and in the city of L’Aquila (Italy) in 2009.
anism in the resolution cell); thus in the
comparison between two images, one acquired
before and one after the seismic event, a decrease Case Studies
in the backscattering value is expected
(Yonezawa and Takeuchi 2001; Matsuoka and On December 26, 2003, the southeastern region
Yamazaki 2004; Stramondo et al. 2006). It is a of Iran was hit by a 6.5 moment magnitude
different case with the high-resolution SAR data (Mw) earthquake whose epicenter was located
because we can see a decrease of the backscatter- very close to the city of Bam. The event had a
ing for the pixel where the double-bounce is tremendous impact all over the world, due to a
located, but also an increase in the roof part, heavy loss of lives, injuries, and destruction; the
since rubble can cause an increase of the back- earthquake caused the death of more than 26,000
scattering as compared to a smooth roof. Further- residents and injured another 30,000. Addition-
more, where the shadow is present in the ally, the city of Bam was a UNESCO World
pre-event image, some changes are expected in Heritage Site and as such it plays an important
the post-event image, depending on the structures historical role. The Bam earthquake was the
shadowed before the collapse. worst seismic event in recent Iranian history.
As it appears from the aforementioned The damage was concentrated in a relatively
description of the optical and SAR data, the small area around Bam. Ninety percent of the
high resolution has drastically changed the structures in the city of Bam were left more or
semantic of objects as they were known so far. less heavily damaged or totally collapsed. In par-
Indeed the lower resolution caused them to ticular, the main historical and traditional adobe
appear differently than the metric or submetric buildings collapsed. Contemporary houses dem-
spatial resolution, thus the necessity to introduce onstrated better resistance to earthquake shake,
new characteristic features and a new vocabulary. even though many collapsed due to poor con-
For this reason the object-oriented approach struction practices and the presence of weak
seems more promising than per pixel, since with stories in the buildings (EERI 2004; Chini
submetric resolution image without having a et al. 2009). As already highlighted this event
well-defined semantic for identifying classes of was the first one to be imaged by an optical sensor
interest we can just identify primitive objects, with a submetric spatial resolution (60 cm) such
which are different from objects of interest. as QuickBird, so for this event we have many
Indeed only objects of interest match real-world papers where the damage mapping was treated
objects, e.g., the building footprints. Primitive making use of these new satellite sensors. Within
objects are usually the necessary intermediate these, three publications have been selected to
step before objects of interest can be found by a show the different scales of products that can be
segmentation and classification process, having provided and the synergy between the optical and
Building Damage from Multi-resolution, Object-Based, Classification Techniques 355
SAR sensors, but not only and also the impact been labeled as belonging to one of three possible
that the very high resolution can have in the damage levels: light or no damage, medium and
lower-resolution products. heavy. Two thresholds have been set in order to
In Chini et al. (2008b) pre- and post- identify the three classes. The satellite-based
earthquake QuickBird panchromatic images damage map has been compared with a map B
have been used to show the capability of this derived from ground survey, provided by Geo-
data to map damage at building scale by means logical Survey of Iran, which reports the map of
of segmentation approach based on the applica- collapse ratio. This latter concerns the percentage
tion of mathematical morphology operators of completely collapsed buildings with respect to
(Pesaresi and Benediktsson 2001; Soille 2003). the total number of buildings within a city block.
In this article, a classification procedure was It is worth noting that the information about dam-
applied to the two panchromatic available data, age from field survey was given at district scale,
one taken before (September 30, 2003) and one while the results from object-oriented approach
after (January 3, 2004) the seismic event. The provided information at building scale. Although
buildings’ extraction from one single panchro- the different scales of the maps did not allow a
matic image is not an easy task. Many objects in direct comparison, a qualitative accuracy evalu-
the imaged scene can be confused because they ation was done counting the percentage of the
have very close radiance values: cars with some three classes from remote sensing data in each
buildings, shadow with some asphalt roads, soil regional class obtained from ground survey. It is
with some kinds of roofs and so on. Due to the interesting to highlight that the percentage of
lack of multispectral data with a resolution of heavy damaged buildings (from the object-
60 cm, a contextual analysis is needed to extract oriented analysis) increased with the increase of
geometric information of objects/classes within ground survey damage degree.
the images. Morphological profiles from the orig- In Chini et al. (2008b) the semiautomatic clas-
inal panchromatic images taken before and after sification approach aimed to identify completely
the earthquake have been carried out in order to or partially collapsed buildings based on the opti-
classify all the buildings in the scene. The classi- cal satellite images vocabulary; it was basically a
fication was done so as to avoid false alarms in change detection approach at building scale,
the change detection process caused by shadow which evaluated how much the radiance value
or other temporary objects like cars or recovery for all pixels of each building changed as an
tents. For this purpose, the open and close mor- effect of the earthquake. A step forward was
phological operators have been applied and the attempted in the next article we will describe
derived profiles have been used as inputs to an (Bignami et al. 2011) where an analysis was
unsupervised classifier in order to extract the performed on the sensitivity of textural features
entire map of buildings before and after the earth- at object scale with respect to the damage levels
quake. Hence, after the classification process, according to the European Macroseismic Scale
each single building has been recognized and 1998 (EMS98). Textural features were clearly
then labeled. Finally, the damage estimation can demonstrated to be a method to overcome the
be computed, by comparing the pixel belonging lack of spectral resolution for classification pur-
to the same building before and after the seismic poses. Indeed texture analysis plays an increas-
event. The percentage of the damaged pixels with ingly important role in digital image processing
respect to the total for each building is an estima- and interpretation, especially for very high-
tion of the damage level (ratio between the num- resolution data (Haralick et al. 1973; Haralick
ber of pixels after and before the earthquake 1979; Pacifici et al. 2009). This is mainly due to
forming the single object). The higher the number supplementary information about image proper-
of damage pixels, the higher damage level of the ties it can provide. Such information can be fruit-
buildings. The map deals with a classification fully exploited for change detection applications
criteria where each building of Bam city has as well. In Bignami et al. (2011), the analysis was
356 Building Damage from Multi-resolution, Object-Based, Classification Techniques
focused on the sensitivity of some textural fea- pre-seismic at object scale has been computed
tures for change detection purposes; in particular in order to analyze their sensitivity to the damage
it was focused on earthquake damage mapping. levels for those building labeled by ground sur-
This approach was preferred in order to give vey. A total of 367 buildings were accounted for
information strictly related to each building and in the sensitivity analysis. For each object tex-
to mitigate false alarms weighting per-pixel clas- tural feature the mean value within a damage
sification error. Instead of the usual textural fea- class from the ground survey was evaluated and
tures extraction by considering the co-occurrence the difference between post-seismic and
matrix on a moving window, the proposed pre-seismic features was calculated. These values
method was based on the calculation of textural have been plotted in respect to the five ground
parameters at object scale. Indeed, the survey damage grade. By observing the plots,
co-occurrence matrix is evaluated by taking into some evidence of sensitivity with respect to dam-
account all the pixels belonging to a single object, age levels are observed. A positive trend in the
i.e., the building, and considering a shift of one difference of textural features has been obtained
pixel in the horizontal and vertical directions. In for the contrast, the dissimilarity and the entropy,
this way for each building a co-occurrence matrix while a negative trend is visible for the homoge-
was obtained. The algorithm makes use of the neity. The trends reflect the expected behavior of
same dataset as Chini et al. (2008b) and also of these textural parameters. In fact, as the damage
some outcomes of the algorithm such as the level of an object increases, the numbers of pixels
buildings map, where each building/object was within the object, which appear scattered,
singled out. Stemming from this building map, increase. From a textural point of view, the
the textural parameters for each specific object is homogeneity of the object should decrease, i.e.,
derived, overcoming the dependency from the the more scattered pixels are, the less homoge-
window size, which makes the analysis test case neous is the object. On the contrary, the contrast,
dependent. Following the formulation of the tex- the dissimilarity, and the entropy increase with
tural parameters described in Haralick the increasing number of scattered pixels.
et al. (1973) and Haralick (1979), for each one A comparison with textural parameter extracted
of the selected building in the two images, pre- by using a moving window with a fixed size has
and post-earthquake, four different object tex- also been carried out, highlighting that the trends
tural parameters have been calculated, such as were not as expected. This work showed that to
contrast, dissimilarity, homogeneity and entropy. better understand some discrepancies present
In order to investigate the sensitivity of these between satellite information and ground survey
parameters to damage degree from EMS98 dam- damage grade, an object-by-object analysis is
age scale, some buildings were selected by the needed. It was stated that the damage level
ground truth map described in Hisada and assigned to an object should be validated also
Shibaya (2004), where some buildings, located from the remote sensing point of view because
around eight strong motion stations, were sur- sometimes the damage level cannot be identified
veyed and classified. The EMS98 classifies five from satellite data. A typical false alarm case is
damage levels: G1 (negligible to slight), G2 the pancake effect which occurs when severe
(moderate), G3 (substantial to heavy), G4 (very damage affects a building (G3 and G4 levels),
heavy) and G5 (destruction). The G1 level can or sometimes it collapses (G5 level), but the roof
also include the non-damage class, because the remains almost undamaged, preventing its iden-
difference between G1 and no damaged building tification from space images.
cannot be appreciated using this kind of satellite The Bam earthquake was also studied in Chini
data. Once the four object textural parameters et al. (2009), where the integration between
have been derived on both pre-seismic and post- medium resolution SAR data was analyzed,
seismic data, the difference between post- and from 20 m resolution C-band sensor onboard of
Building Damage from Multi-resolution, Object-Based, Classification Techniques 357
Envisat satellite mission and very high-resolution with the increase of damage level (Yonezawa
optical data, with the aim of producing damage and Takeuchi 2001; Stramondo et al. 2006). In
maps at district scale with three different levels of addition to Envisat SAR data, the information
damage. In this article additional parameters, extracted from a high-resolution optical
namely, the InSAR complex coherence and the QuickBird image has been used, by extracting B
intensity (or backscattering) correlation, have the building maps before the earthquake carried
been computed, since couples of pre- and post- out with the adoption of object classification
seismic interferometric images were available approaches. The building map was used as a
(two images before and one or two after the mask to restrict the SAR intensity correlation
event). Both parameters can be derived by com- change detection only to built-up areas, in order
bining the pre-seismic pair, the post-seismic, and to reduce possible false alarms. Indeed, in Bam
the coseismic (i.e., one pre-seismic and one post- City, there were a lot of vegetated areas, which
seismic image) ones. Note that these two features could affect the value of the correlation. The
contain slightly different information concerning intensity correlation difference showed an
changes in the scene. The complex coherence is increase of the dynamic range of about 25 %
primarily influenced by the phase difference with respect to the non-masked data, using the
between radar returns, which is a distinctive information on buildings from optical data.
parameter measured by a coherent sensor like On April 6, 2009, a Mw = 6.3 earthquake hit a
SAR. It is particularly related to the spatial wide portion of the Abruzzi region in Italy, with
arrangement of the scatterers within the pixel the epicenter located 4 km southwest of the city
and, thus, to their possible displacements. Con- of L’Aquila, whereas the highest damages
versely, the intensity correlation is more related occurred about 8 km to the southeast. In L’Aquila
to changes in the magnitude of the radar return. and in the neighboring villages, the earthquake
Unfortunately, for this event, SAR interferomet- caused the collapse or irreparable damage to over
ric image pairs have very high values of the 15,000 buildings, killing 308 people and causing
perpendicular baseline, around 500 m for both the relocation of over 65,000 (Stramondo
pairs and the resulting high spatial decorrelation et al. 2011). For this event very high-resolution
prevents their employment from detecting dam- SAR images, acquired in SpotLight mode (1 m
age levels through the InSAR phase coherence, per pixel on ground) from the X-band COSMO-
which has almost the same values both in dam- SkyMed satellite mission, were available. The
aged and undamaged areas when the baseline is very high resolution images from optical sensors
too large (Zebker and Villasenor 1992). such as QuickBird and IKONOS were available
Concerning the use of the backscattering correla- as well. It is interesting to see how information
tion as an indicator of surface changes, the has been exploited in the work of Dell’Acqua
approach is funded on the idea that the correlation et al. (2011), for estimating the damage at dis-
coefficient is high when two images are similar, tricts levels from a post-event SAR image and the
while it becomes low when changes have fusion with optical images for increasing the
occurred in the surface target. Thus, in the latter accuracy. It is commonly acknowledged that
case, a decrease of correlation of coseismic due to speckle effects, single-pixel classification
images is expected to be relative to pre-seismic of SAR images leads to unsatisfactory results,
ones and this phenomenon is expected to increase and this seems to hold true also when damage
with the increase of damaged buildings on the assessment is concerned. Satisfying results may
ground. The difference between the pre- and be achieved if the damage is assessed at a block
coseismic intensity correlations has been aver- level, somehow averaging the unreliable results
aged within areas of homogeneous damage levels of pixel-wise comparing pre- and post-event
provided from a ground survey map, and it is images due to speckle noise. In the Dell’Acqua
clearly shown how this difference increases et al. (2011) article, damage was considered at
358 Building Damage from Multi-resolution, Object-Based, Classification Techniques
pixel scale, in optical data and at block scale, in Relevant case studies, where object-based dam-
SAR data, separately. Indeed, the pixel-based age mapping approaches have been considered,
approach by using optical data provided useful are here presented.
information on the discrimination between dam-
aged or non-damaged areas. Even though no fur-
Acknowledgment The author would like to thank R. De
ther classification with respect to damage levels Meo for reviewing this manuscript.
can be retrieved, the information derived from
optical imagery can be effective where appar-
ently very high-resolution SAR data is least Cross-References
with the proposed method, and it could be prob-
ably the reason why fusion of optical and SAR ▶ Building Disaster Resiliency Through Disaster
data in this method is promising. Indeed in this Risk Management Master Planning
work the SAR textures, which were the features ▶ Building Monitoring Using a Ground-Based
used as damage indicator, were somehow corre- Radar
lated to low damage levels (or affected by other ▶ Damage to Ancient Buildings from
factors than damage), whereas optical data were Earthquakes
more suitable to distinguish between damage and ▶ Damage to Buildings: Modeling
non-damage classes. Definitely, the SAR texture ▶ Damage to Infrastructure: Modeling
was more effective to distinguish between two ▶ Earthquake Damage Assessment from VHR
different levels of damage, in the regions that Data: Case Studies
were identified as damaged by optical data. ▶ Earthquake Disaster Recovery: Leadership and
Some important issues remained still open; such Governance
an example could be finding suitable threshold ▶ Estimation of Potential Seismic Damage in
levels for classification in future cases, although Urban Areas
significant links between satellite-derived indica- ▶ Features Extraction from Satellite Data
tors and damage levels have been discovered. ▶ Hyperspectral Data in Urban Areas
It seems that some texture maps computed on ▶ Remote Sensing in Seismology: An Overview
SAR data can provide an estimate of damage ▶ Resilience to Earthquake Disasters
levels, only if independent information concerning ▶ SAR Images, Interpretation of
presence or absence of damage in a given block ▶ Seismic Collapse Assessment
would be available. It is evident from the results ▶ Seismic Loss Assessment
that further studies are required and they should be ▶ System and Damage Identification of Civil
aimed at finding sufficiently independent statistics, Structures
extracted from SAR data, which may possibly be ▶ Urban Change Monitoring: Multi-temporal
linked to presence or absence of damage. SAR Images
Summary References
This entry highlights the characteristics of very Bignami C, Chini M, Stramondo S, Pierdicca N, Emery
high-resolution optical and SAR images that are WJ (2011) Objects textural features sensitivity for
earthquake damage mapping. In: IEEE GRSS and
pertinent for earthquake damage mapping pur- ISPRS Joint Urban Remote Sensing Event – JURSE
poses. Automatic or semiautomatic classification 2011 proceedings, pp 333–336
algorithms, which make use of this kind of data, Blaschke T (2010) Object based image analysis for remote
have been described as well. Particular attention sensing. ISPRS J Photogramm 65:2–16
Brunner D, Lemoine G, Bruzzone L (2010) Earthquake
has been paid to object-based classification
damage assessment of buildings using VHR optical
approaches, which are more suitable to deal and SAR imagery. IEEE Trans Geosci Remote
with metric or submetric resolution data. 48:2403–2420
Building Damage from Multi-resolution, Object-Based, Classification Techniques 359
Chini M (2009) Earthquake damage mapping techniques buildings. IEEE J Selected Top Appl Earth Observa-
using SAR and optical remote sensing satellite data. tions Remote Sens 1:107–119
In: Jedlovec G (ed) Advances in geoscience and Guida R, Iodice A, Riccio D (2010) Height retrieval of
remote sensing. InTech, Vukovar, pp 269–278 isolated buildings from single high-resolution SAR
Chini M, Bignami C, Stramondo S, Pierdicca N (2008a) images. IEEE Trans Geosci Remote Sens
Uplift and subsidence due to the December 26th, 2004, 48:2967–2979 B
Indonesian earthquake and tsunami detected by SAR Gusella L, Adams BJ, Bitelli G, Huyck CK, Mognol
data. Int J Remote Sens 29:3891–3910 A (2005) Object-oriented image understanding and
Chini M, Bignami C, Emery WJ, Pierdicca N, Stramondo post-earthquake damage assessment for the 2003
S (2008b) QuickBird panchromatic images for map- Bam, Iran, earthquake. Earthq Spectra
ping damages at building scale caused by the 2003 21(S1):225–238
Bam earthquake. In: 2008 I.E. International Geosci- Haralick RM (1979) Statistical and structural approaches
ence and Remote Sensing Symposium to texture. Proc IEEE 67:786–804
(IGARSS) – proceedings, vol 2, Issue Haralick RM, Shanmuga K, Dinstein IH (1973) Textural
1, II1029–II1031 features for image classification. IEEE Trans Syst Man
Chini M, Pierdicca N, Emery WJ (2009) Exploiting SAR Cybern 3:610–621
and VHR optical images to quantify damage caused by Hisada Y, Shibaya A (2004) Building damage and seismic
the 2003 Bam earthquake. IEEE Trans Geosci intensity in Bam city from the 2003 Iran, Bam, earth-
47:145–152 quake. Bull Earthquake Res Inst Univ Tokyo 79:81–93
Chini M, Cinti FR, Stramondo S (2011) Co-seismic sur- Kropatsch WG, Strobl D (1990) The generation of SAR
face effects from very high resolution panchromatic layover and shadow maps from digital elevation
images: the case of the 2005 Kashmir (Pakistan) earth- models. IEEE Trans Geosci Remote Sens 28:98–107
quake. Nat Hazard Earth Syst Sci 11:931–943 Mathieu R, Aryal J, Chong AK (2007) Object-based clas-
Chini M, Piscini A, Cinti FR, Amici S, Nappi R, De sification of Ikonos imagery for mapping large-scale
Martini PM (2013) The 2011 Tohoku-Oki (Japan) vegetation communities in urban areas. Sensors
tsunami inundation and liquefaction investigated by 7:2860–2880
optical, thermal and SAR data. IEEE Geosci Remote Matsuoka M, Yamazaki F (2004) Use of satellite SAR
Sens Lett 10:347–351 intensity imagery for detecting building areas dam-
Chini M, Chiancone A, Stramondo S (2014) Scale object aged due to earthquakes. Earthq Spectra 20(975):994
selection (SOS) through a hierarchical segmentation Pacifici C, Chini M, Emery WJ (2009) A neural network
by a multi-spectral per-pixel classification. Pattern approach using multi-scale textural metrics from very
Recognit Lett 49:214–223 high resolution panchromatic imagery for urban
Dell’Acqua F, Bignami C, Chini M, Lisini G, Polli D, land-use classification. Remote Sens Environ
Stramondo S (2011) Earthquake rapid mapping by 113:1276–1292
satellite remote sensing data: L’Aquila April 6th, Pesaresi M, Benediktsson JA (2001) A new approach for
2009 event. IEEE J Selected Top Appl Earth Observa- the morphological segmentation of high-resolution
tions Remote Sens 4:935–943 satellite imagery. Trans Geosci Remote Sens
Earthquake Engineering Research Institute – EERI 39:309–320
(2004) Preliminary observations on the Bam, Iran, Soille P (2003) Morphological image analysis – principles
earthquake of December 26, 2003. EERI Special and applications, 2nd edn. Springer, Berlin
Earthquake Report, [Online]. Available at http:// Stramondo S, Bignami C, Chini M, Pierdicca N,
www.eeri.org/lfe/pdf/iran_bam_eeri_preliminary_ Tertulliani A (2006) Satellite radar and optical remote
report.pdf sensing for earthquake damage detection: results from
Ferro A, Brunner D, Bruzzone L (2013) Automatic detec- different case studies. Int J Remote Sens
tion and reconstruction of building radar footprints 27:4433–4447
from single VHR SAR images. IEEE Tran Geosci Stramondo S, Chini M, Salvi S, Bignami C, Zoffoli S,
Remote Sens 51:935–952 Boschi E (2008) Ground deformation imagery of the
Franceschetti G, Iodice A, Riccio D (2002) A canonical may Sichuan earthquake. EOS Trans Am Geophys
problem in electromagnetic backscattering from build- Union 89:341–342
ings. IEEE Trans Geosci Remote Sens 40:1787–1801 Stramondo S, Chini M, Bignami C, Salvi S, Atzori
Franceschetti G, Iodice A, Riccio D, Ruello G (2003) SAR S (2011) X- C- L-Band DInSAR investigation of the
raw signal simulation for urban structures. IEEE Trans April 6th, 2009, Abruzzi earthquake. IEEE Geosci
Geosci Remote Sens 41:1986–1995 Remote Sens Lett 8:49–53
Gamba P, Dell’Acqua F, Trianni G (2007) Rapid damage Turker M, Sumer E (2008) Building-based damage detec-
detection in the Bam area using multitemporal SAR and tion due to earthquake using the watershed segmenta-
exploiting ancillary data. IEEE Trans Geosci Remote tion of the post-event aerial images. Int J Remote Sens
Sens 45:1582–1589 29(11):3073–3089
Guida R, Iodice A, Riccio D, Stilla U (2008) Model-based Voigt S, Kemper T, Riedlinger T, Kiefl R, Scholte K,
interpretation of high-resolution SAR images of Mehl H (2007) Satellite image analysis for disaster
360 Building Disaster Resiliency Through Disaster Risk Management Master Planning
and crisis-management support. IEEE Trans Geosci and its contributory factors such as poverty and
Remote Sens 45:1520–1528 rapid urbanization (Khazai and Bendimerad
Yamazaki F, Yano Y, Matsuoka M (2005) Visual damage
interpretation of buildings in Bam city using 2008). Contrary to popular misconception of
QuickBird images following the 2003 Bam, Iran, local authorities in urban areas having the requi-
earthquake. Earthq Spectra 21(S1):329–336 site capacity to address risk on their own, most
Yonezawa C, Takeuchi S (2001) Decorrelation of SAR cities, particularly in the developing world, have
data by urban damage caused by the 1995 Hoyogoken-
Nanbu earthquake. Int J Remote Sens 22:1585–1600 actually been ineffective in managing their risk,
Zebker HA, Villasenor J (1992) Decorrelation in interfer- which remains high and is continuously rising.
ometric radar echoes. IEEE Trans Geosci Remote Sens The relative negligence of disaster risk manage-
30:950–959 ment as a local responsibility is also exacerbated
by the fact that many local authorities are already
overburdened by their presently mandated
responsibilities.
Building Disaster Resiliency Through Disaster risk management (DRM) is defined
Disaster Risk Management Master by the United Nations International Strategy for
Planning Disaster Reduction (UN-ISDR) as “the system-
atic process of using administrative directives,
Fouad Bendimerad1, Jerome Zayas1, organizations, and operational skills and capac-
Bijan Khazai2 and Moses Kent Borinaga1 ities to implement strategies, policies and
1
Earthquakes and Megacities Initiative, Quezon improved coping capacities in order to lessen
City, Philippines the adverse impacts of hazards and the possibility
2
Karlsruhe Institute of Technology, Center for of disaster” (UN-ISDR 2009). Current
Disaster Management and Risk Reduction approaches to managing disaster risk are often
Technology (CEDIM), Karlsruhe, Germany characterized by (a) the prevalence of a response
or humanitarian perspective rather than a devel-
opment perspective, (b) the lack of cross-
Introduction disciplinary and cross-sectoral processes
demanded by a more integrated mainstreaming
Since impacts of disasters are most felt at the approach that captures the parameters of risk that
local level, they require proactive action and are relevant to each sector, and (c) the absence of
intervention directly from local authorities and policymaking processes based on consensus
affected communities. However, the extent of building and effective engagement of the
impacts varies from one area to another. Highly concerned stakeholders. These factors are corre-
urbanized areas are typically more vulnerable to lated, calling for a holistic approach to urban
natural hazards as a result of environmental and resilience, one that integrates disaster risk reduc-
socioeconomic stresses brought about by rapid tion (DRR) with developmental policy and plan-
and often unplanned urbanization. Because of ning, delivery of core services, natural and
the concentration of population, infrastructure, environmental resource management, and pov-
and resources in urban areas, they stand to lose erty reduction. A methodology that could define
more in the event of actual disaster. Thus, it is needs to be met and why and how to do it is
becoming increasingly imperative for city gov- necessary to generate support for allocating
ernment authorities and local actors in urban resources to disaster risk reduction, especially in
areas to develop and enhance their capacity to view of overwhelming demands from citizens for
manage disaster risks in order to protect their vital services and to meet day-to-day needs.
development. The key to manage disaster risks in urban
Cities have largely and chronically been areas without draining the resources of local
neglected by national governments and interna- authorities is to mainstream DRM in the daily
tional organizations in terms of dealing with risk functions and services of the city. By integrating
Building Disaster Resiliency Through Disaster Risk Management Master Planning 361
risk reduction parameters in planning processes, Mainstreaming requires a deliberate and well-
such as land use and urban planning, public founded strategy and consistent applications of
works, delivery of core services, and emergency policies that strive to pinpoint and enforce roles
response planning, among others, the task of and responsibilities. Inherent in this process is a
managing disaster risks becomes more attainable. policy of decentralization of authority and B
However, mainstreaming DRM in the operations accountability. The current practice shows an
of urban local authorities is easier said than done. overemphasis on centralization at the national
Disaster risk management in general is a new level of all DRM functions. The reasons are
field of practice for most city officials and per- twofold:
sonnel, not to mention that disaster risk reduction
implementation itself is complex. It takes time, (a) Lack of decentralization and coordination
effort, training, and, most importantly, scientifi- structures. The DRM practice is an emerging
cally designed methods and tools in order to function of government; sublevel govern-
effectively and fully assimilate disaster risk ment institutional structure and coordination
reduction in city functions and ongoing opera- mechanisms have yet to be devised and put in
tions. DRM mainstreaming is also closely in place.
line with the commitment of countries to several (b) Lack of capacity of local authorities. The
international conventions and agreements such as central authorities of the government are
the Hyogo Framework for Action for 2005–2015 reluctant to provide local government with
(HFA) and the UN Millennium Development the authority for DRM out of prudence.
Goals (MDG), to name a few.
This entry further elaborates the concept of Despite these limitations, a distribution of
DRM mainstreaming, shows examples of tools roles and responsibilities between the various
that are used for mainstreaming, and provides a levels of government is critical to achieve
case study of an actual application of DRM mainstreaming. One approach to establish a
mainstreaming using the Disaster Risk Manage- mainstreaming strategy is to follow the scheme
ment Master Planning (DRMMP) process devel- shown in Fig. 1 (EMI 2009). The three
oped by the Earthquake and Megacities Initiative encompassing boxes provide a descriptive
(EMI). assignment of roles and responsibilities. Funda-
mentally, the model represented in the figure
indicates that, following other more conventional
The Concept of DRM Mainstreaming functions of government, DRM implementation
has three principles of mainstreaming: central
Mainstreaming is a core concept of DRM but is coordination, local implementation, and partici-
seldom understood. Mainstreaming is essential to pation (Bendimerad and Zayas 2015).
achieving effectiveness in a DRM structure. It is
about defining and distributing roles and respon- • Local implementation: Implementation should
sibilities between national and local authorities as much as possible be delegated to local
and civil society in a way that would result in authorities and local actors. Local authorities
greater outcome for lower amount of resources. It cannot mainstream DRM without a clear
is about determining the parameters of risk that authority and resources.
are most relevant to each sector and function of • Central coordination: The role of the central
government and making these parameters an inte- authorities should focus on establishing pol-
gral part of the governance and management of icy, enacting regulation, putting in place con-
urban life. An organization that has accomplished trol processes, allocating resources, and
mainstreaming has effectively integrated the exercising oversight.
practice of risk management within its gover- • Participation: Government (central and local)
nance, its functions, and operations. must open the door to the participation of civil
362 Building Disaster Resiliency Through Disaster Risk Management Master Planning
Building Disaster
Resiliency Through
Disaster Risk
Management Master
Planning,
Fig. 1 Framework for
mainstreaming disaster risk
reduction. Functions are
indicative (Source:
EMI. State of the practice
report on urban disaster:
open file report. June 2009)
society, which collectively groups all the that they undertake such as construction and
active agents of society. building licensing, land use and urban develop-
ment planning, environmental management, and
The central government delegates aspects of social welfare as well as in the services that they
policy, regulatory powers, and other functions provide or regulate. To provide necessary author-
and allocates additional resources to local gov- ity, additional reforms are often needed, particu-
ernment. It should in turn monitor the perfor- larly in aligning DRM laws and policies with
mance of local authorities based on sanctioned other acts that regulate local government func-
guidelines and criteria. The boxes do indicate a tions. The point is that wherever responsibility is
strict division between the various entities. In located, authority and resources should also be
fact, flexibility and coordination are keys to suc- provided.
cess. Within this mainstreaming framework, an
optimization of resources and outcome is accom-
plished when local authorities are strategically Review of Tools for DRM Mainstreaming
establishing internal policies and process where
planning, programming, and action for risk In line with global efforts to reduce disaster risks
reduction is integral to each of their jurisdictional by focusing on local-level action through
responsibilities either as a service provider or as a mainstreaming, several disaster risk management
regulator. This process is sometimes referred to models have been developed that integrate disas-
as “integrated” or “holistic” approach to DRM. ter risk in local development. A review has been
This framework also suggests that risk reduc- conducted of 76 tools, 10 of which were
tion can be mainstreamed in local governance by shortlisted for detailed assessment because of
harnessing existing legal mechanisms (e.g., local their applicability for local application in urban
government acts, environmental management settings (Berse et al. 2009). “Tools” are used
acts, DRM acts and related laws and regulatory interchangeably with models and instruments
mechanisms), processes (e.g., licensing, land use, and in this context refer to a framework that
etc.), and service delivery systems (e.g., health, lays out the methodology for undertaking DRM
education, transport, utilities, revenue genera- including the mechanism for implementing the
tion) and making use of such resources. For elements in the methodology.
local authorities, this would translate into having Each tool’s relevance to disaster risk reduction
processes and practices that inherently incorpo- and their applicability or feasibility to the local
rate urban risk reduction in these core functions government’s capacity to institutionalize disaster
Building Disaster Resiliency Through Disaster Risk Management Master Planning 363
risk reduction efforts within its functions and these constituencies in the DRM processes.
operations were carefully considered and Although designed to be implemented at the
weighed in the assessment. Each tool has its national level, Emergency Management
own strengths and weaknesses that local author- Australia’s (EMA) Emergency Risk Manage-
ities in urban areas can learn from in ment Process (Commonwealth of Australia B
mainstreaming DRM. Evaluation of DRM tools 2004) can also be developed for local government
were based on a set of criteria indicators DRM mainstreaming implementation.
representing the three principles of
mainstreaming, i.e., central coordination, local Central Coordination
implementation, and participation. The criteria The reviewed DRM mainstreaming models are
indicators are meant to assess effectiveness and consistently aligned with the disaster risk reduc-
relevance to DRR processes, feasibility require- tion strategies formulated by the DRM authorities
ments, and applicability to the urban setting espe- of their respective countries. The TDRM as in a
cially of developing countries. Through this few other cases was developed by
study (which is not exhaustive), a set DRM nongovernmental organizations for application
instruments have been developed—or at least in different cities and countries. The Total Disas-
have the potential—for replication in areas out- ter Risk Management implementation strongly
side of their respective origins and were rated for encourages endorsement from the central DRM
local adoption in the urban context. A summary authority and adaptation to the DRM institutional
of the results of this study is provided here. framework of the country of application. The
Total Disaster Risk Management developed by
Effectiveness the Asian Disaster Reduction Center- Office for
the Coordination of Humanitarian Affairs
Local Implementation (ADRC-OCHA) was endorsed by the govern-
The paradigm shift in the disaster management ment of Thailand and its central DRM authority
policies has elicited the participation and com- and adapted for implementation in the country
mitment of local governments to a more proac- (ADRC 2005).
tive disaster risk management. It is argued that
since local stakeholders are the ones greatly Participation
affected in the events of disasters, it is critical The need for stakeholder participation is a well-
for local governments to play a major role in the established approach used in development work.
DRM. This is the very reason for giving emphasis Its main tenet is that stakeholders are collabora-
on the local implementation of DRM tors in a development project at every stage of the
mainstreaming models. process. Through this approach, stakeholders
Of the frameworks reviewed, the following influence and share control over the decisions
have evolved around the concept of DRM that affect them and their resources. In essence,
mainstreaming: Comprehensive Hazard and the stakeholders drive the development process.
Risk Management (CHARM) Programming They are the ones who decide in the planning,
Approach (SOPAC 2002), Total Disaster Risk design, implementation, monitoring, and evalua-
Management (TDRM) (ADRC 2005), Central tion of the project. In this process, they are guided
United States Earthquake Consortium’s by experts and specialists who also processed the
(CUSEC’S) Disaster Resistant Community outcome in a scientifically rigorous output. With
(DRC) Model (CUSEC 1997), and the Bangla- this, participation does not only ensure stake-
desh DRR Mainstreaming model (Siddiqui holders’ ownership but also encourage commit-
2007). CUSEC’s DRC Model was developed by ment. In general, the reviewed models emphasize
a consortium that includes both the local commu- multistakeholder participation at every stage of
nity and business sectors and not surprisingly the DRM planning process. They also recognize
gives the most emphasis on the involvement of the relevance of the participatory process as
364 Building Disaster Resiliency Through Disaster Risk Management Master Planning
enabling communication mechanisms. Models disaster hit (Berse et al. 2009). The Asian Disas-
provide more relevance to particular stake- ter Reduction Center- Office for the Coordination
holders: national authorities (e.g., Comprehen- of Humanitarian Affairs’ Total Disaster Risk
sive Hazard and Risk Management and Total Management model, like the Comprehensive
Disaster Risk Management), local authorities Hazard and Risk Management model, is
(e.g., Central United States Earthquake Consor- anchored on risk management. However, the
tium), business sector (e.g., Central United States TDRM also focuses on “total” approach of
Earthquake Consortium’s Disaster Resistant balancing the ability to respond to the conse-
Community), or community (e.g., BDRRM, Cen- quences of disaster with predisaster actions that
tral United States Earthquake Consortium). At lessen the need for response (ADRC 2005).
the same time, other tools such as the BDRRM
provide specific indications of the targeted Feasibility/Resource Requirements
stakeholders. Financial and resource requirement are crucial
factors for assessing the implementation potential
Applicability of a particular DRM mainstreaming model.
Oftentimes, the availability of resource dictates
Relevance the scope of DRM mainstreaming implementa-
The reviewed DRM mainstreaming models gen- tion and the strategies or tools that will be uti-
erally adhere to the main goal of integrating risk lized. Acknowledging this, most of the shortlisted
management in the development and governance DRM models signified the use of available
processes in order to reduce risks and mitigate the resources and the development or enhancement
impact of disasters, albeit employing distinct of existing resources whether it be financial/
strategies and aiming at varied focus. material or human resources. While some DRM
The CUSEC model, for example, prioritizes mainstreaming models identified low-cost tools
achieving disaster-resistant communities. With and strategies for their implementation, others
this, the model aims to reduce community vul- opted for more sophisticated and high-cost tools
nerability to natural hazards in order to minimize or mechanisms.
losses and accelerate recovery (CUSEC 1997). Apart from relying on the engagement of dif-
The Emergency Management Australia’s Emer- ferent stakeholders and the local government for
gency Risk Management Process, on the other human resources, interestingly Central United
hand, focuses on reducing risk by making modi- States Earthquake Consortium uses the strategy
fications in the causes of risks (Commonwealth of of providing incentives to promote steps and
Australia 2004). Similarly, South Pacific Applied measures to reduce the vulnerability of commu-
Geoscience Commission Disaster Management nities to natural hazards (CUSEC 1997). The
Unit’s Comprehensive Hazard and Risk Manage- Food and Agricultural Organization’s DRM
ment Programming Approach aspires to develop Mainstreaming model, South Pacific Applied
preparedness and reduce vulnerabilities while Geoscience Commission-DMU’s Comprehen-
increasing resilience of the community (SOPAC sive Hazard and Risk Management, and the Ban-
2002). The added value of this particular model is gladesh DRM Mainstreaming model all give
that it engages the implementing institution in importance on strengthening the capacities of
hazard and risk management in a more holistic local stakeholders in disaster management and
manner as it intrinsically links DRR to national risk reduction and improving advocacy and
development (South Pacific Applied Geoscience awareness (SOPAC 2002). They do not specify
Commission SOPAC 2002). Food and Agricul- what would be the required resources for
tural Organization’s DRM Framework (Baas implementation.
et al. 2008) follows the same line and aims to Emergency Management Australia’s Emer-
reduce underlying factors of risk and to prepare gency Risk Management did more than this by
for and initiate an immediate response should identifying a package of special tools for risk
Building Disaster Resiliency Through Disaster Risk Management Master Planning 365
management (i.e., Unique Identifier System, the implementation also strongly encourages formal
Risk Register Database, and the Geographic endorsement from the central DRM authority and
Information Systems) and at the same time by adaptation to the DRM institutional framework of
carrying out emergency risk management train- the country of application. Finally, the DRMMP
ing for local stakeholders (Commonwealth of process has been designed as an approach for B
Australia 2004). organizing the stakeholders in thematic Focus
Groups to ensure efficiency and sustainability
using a participatory process with targeted
Disaster Risk Management Master Plan stakeholders.
(DRMMP) The DRMMP methodology follows similar
principles and processes as other more conven-
The Disaster Risk Management Master Plan tional planning processes, e.g., land use planning,
was developed by Earthquake and Megacities namely,
Initiative (EMI) to support local-level long-
term planning and programming of disaster • It is structured and systematic and follows a
risk reduction activities. It follows the approach consistent step-by-step methodology.
set out in Australian/New Zealand Standard • It is science based and data driven.
4360-2004, a standard in risk management • It is multisectoral and multidisciplinary.
developed jointly by Australia and New • It applies to all hazards.
Zealand and recently adopted by the Interna- • It is participatory and consensus driven.
tional Standards Organization as ISO31000 for
risk management. The DRMMP is based on By anchoring on mainstreaming, the DRMMP
EMI’s mainstreaming framework that optimizes puts in place a structured approach for setting up
resources by defining and distributing roles and a DRM system that is based on science, under-
responsibilities across various stakeholders at stood by all, and provides the rationale for invest-
all levels of government. ments in DRR programs and projects. As of 2014,
The DRMMP provides opportunity for local the DRMMP has already been applied to eight
authorities to systematically and rigorously iden- (8) local authorities, namely, Istanbul (Turkey),
tify programs, projects, and activities (PPA) Kathmandu (Nepal), Metro Manila (Philippines),
aimed at reducing risk caused by natural and Amman (Jordan), Mumbai (India), Pasig City
man-made hazards and to define processes for (Philippines), Quezon City (Philippines), and
implementing these projects and activities. Dhaka (Bangladesh).
Local authorities are familiar with the master
planning process as it is an integral part of their
planning activities in particular for land use plan- Process to Implement the DRMMP
ning and transport. These plans constitute the
basis for budgeting and investments. The Disaster Risk Management Master Plan pro-
Embedding the three principles of vides a road map for a proactive DRM approach,
mainstreaming (i.e., local implementation, cen- which recognizes that disasters are not just “set-
tral coordination, and participation), the backs” or “roadblocks” to development but result
DRMMP process is designed to align along core from the paths that development is pursuing.
planning processes of local governments and is Thus, by changing planning processes, and incor-
strategically developed for local government porating disaster risk parameters explicitly in the
DRM implementation. It is anchored on the reg- planning of processes and projects, it can be
ulatory framework that defines mandate and ensured that in future natural hazards will
authority of government. It is a methodology for encounter resilient communities that are capable
integrating risk parameters into various local of withstanding their impact, reduce their losses,
government functions. EMI’s DRMMP cope, and recover normal life faster. Ultimately,
366 Building Disaster Resiliency Through Disaster Risk Management Master Planning
PHASE 1-
Project Organization
By Month 1
Project Streering Committee; Technical Working Group;
Project Secretariat; Consultative Committee
Strengths, Weaknesses,
Situational Opportunities &
Resilient City Challenges (SWOC)
Analysis
analysis
Geographic Information
System – Information &
Communication Programs Projects Activities
Technology (GIS-ICT) 1. .... 1. .... 1. ....
2. .... 2. .... 2. ....
Information, Education 3. .... 3. .... 3. ....
and Communication (IEC)
Others Sector:
Social
Economic DRRMP Implementation Process
Infrastructure-
Utilities PHASE 4-
Physical Monitoring and Evaluation; Indicators By Month 12
Environment
Building Disaster Resiliency Through Disaster Risk Management Master Planning, Fig. 2 The DRMMP
workflow (Source: Urban Resilience Master Planning: A Guidebook for Practitioners and Policy Makers, March 2015)
impacts of hazards become mere emergencies • Legal and Institutional Arrangements (LIA)
rather than disasters. • Land Use Planning (LUP) and development
The undertaking of a DRMMP follows a series controls
of analyses along each of four phases, namely, • DRM Database and Geographic Information
System (GIS)
• Phase 1: Organization and Preparation • Hazard, Vulnerability, and Risk Assessments
• Phase 2: Diagnosis and Analysis (HVRA)
• Phase 3: Plan Development • Emergency Management (EM)
• Phase 4: Plan Implementation, Monitoring,
and Evaluation
Further elements/sectors (e.g., housing, trans-
Figure 2 provides the workflow process for port, sanitation, health, education, social ser-
developing the DRMMP, which can be com- vices, and others) can be added depending on
pleted in 12–24 months, depending on the ele- the jurisdictional responsibility of the city. This
ments that need to be covered. First, the elements also depends on what is most relevant to local
that need to be included in the master planning authorities in terms of priority, access to data, and
process have to be identified. Some components availability of resources and expertise. For exam-
are core to the DRMMP methodology because ple, in the application of the DRMMP in Greater
they provide the essential scientific information Mumbai Project, there were additional elements
on the DRM practice. These include that were included such as Water, Wastewater
Building Disaster Resiliency Through Disaster Risk Management Master Planning 367
and Storm Drainage Systems, and Transporta- the stakeholders. Throughout the DRMMP pro-
tion. This is due to the fact that the Municipal cess, the stakeholders are made to understand and
Corporation of Greater Mumbai regulates the take ownership of the types of risks that their city
delivery of these services. It is important to note faces as they acquire the competency and tools to
that not all sectoral elements fall under the be able to manage those risks on their own. B
authority of the city or of other agencies of the
government. Many could be provided by the pri-
vate sector or utility companies (e.g., water, Participatory Approach to Develop the
power). Others would be a mix of public and DRMMP
private entities (e.g., transport). Other elements
may also have to be addressed because of the The DRMMP allows for the meaningful partici-
local context and conditions particular to a city. pation of stakeholders. By design, the DRMMP
For example, in the Mumbai application, ele- provides stakeholders with an opportunity for
ments such as Slums, Shelter, and Housing, learning the concepts of hazard, vulnerability,
were likewise addressed because of the unique and risk and to understand their relevance to
characteristics of the city that has more than 60 % planning. It develops consensus among them on
of its population living in the slum areas. the trade-offs and the rationales for investment in
A listing of additional elements included in the urban resilience. With their engagement in the
DRMMP includes development of the plan, the stakeholders not
only provide essential knowledge to the plan but
• Shelter and Housing (including slums and acknowledge their roles and responsibilities in
informal settlements) (S&H) the implementation of the plan, thus taking own-
• Health ership. The underlying principle to EMI’s partic-
• Education ipatory approach is that urban DRR can only be
• Construction Codes and Standards (CCS) made meaningful if stakeholders understand their
• Water Systems risks. They should be well informed in order to
• Waste Water and Storm Drainage Systems meaningfully contribute. In most cases, there is a
(WWSD) gap in how risk is understood and interpreted by
• Power Systems (PS) various stakeholders with diverse backgrounds
• Transportation Systems (TS) and oftentimes competing interests. Once there
• Solid Waste Management (SWM) is shared understanding of risk, stakeholders can
• Telecommunications contribute to the work at hand in terms of data
• Possibly other relevant sectors collection and validating findings and interpreta-
tions of risk assessments that are fundamental to
There are other elements that support the any risk reduction program.
whole planning process and are crosscutting in EMI’s version of participatory approach was
nature. These include first applied in 2008 in the Makati Urban Rede-
velopment Planning Project: Mainstreaming
• Training and Capacity Building (TCB) Disaster Risk Reduction in Megacities, A Pilot
• Information Education and Communication Application in Metro Manila and Kathmandu.
(IEC) It was then further tested and applied in various
cities and has become EMI’s cornerstone strategy
Training and Capacity Building and Informa- to stakeholder engagement in urban DRR. Using
tion Education and Communication components this approach, the participation by stakeholders is
facilitate the meaningful participation of stake- organized along key sectors that allow for gener-
holders since their aim is to bring the gap in ating inputs across a broad range of stakeholders
understanding of DRM concepts and principles with different agenda and interests but at the
between the EMI experts and practitioners and same time facilitate decision and action.
368 Building Disaster Resiliency Through Disaster Risk Management Master Planning
Building Disaster
Resiliency Through
Disaster Risk
Management Master
Planning, Fig. 3 Project
organization structure of
the disaster risk reduction
in Greater Mumbai project
which shows an example of
a flat non-hierarchical
structure that is composed
of stakeholders from
different sectors that are
working together in a
participatory approach
Typically, implementing the DRMMP requires addressed by the project. The members of the
organizing the stakeholders along the following: Scientific Consortium are selected on the basis
of their credentials. They advise on the valid-
• Creation of a Project Management Team ity of the scientific data and approach and help
(PMT) composed of the project managers reach scientific consensus.
and technical experts and specialists who • Creation of the Focus Groups which are com-
make up the main workforce that will develop posed of local practitioners, researchers, and
the DRMMP. It is multidisciplinary and struc- community leaders. These include midlevel
tured as a single unit incorporating EMI’s managers and specialists from the various
experts and counterpart from the host local departments and offices of the city. They are
authority. organized along each of the sectors that are
• Creation of the Advisory Committee which is being analyzed in the DRMMP. Through the
composed of policy-/decision-making offi- Focus Groups, relevant stakeholders get
cials who are endorsed as representatives to engaged in the development of the project by
the project by their respective institutions. It providing inputs and tackling issues, as well as
provides the platform to receive input and validating the assumptions, findings, and rec-
guidance from the stakeholders and share the ommendations relevant to the project.
project issues, findings, and accomplishments
that the members can take back to their respec- Figure 3 below shows an example of the pro-
tive institutions. ject organization structure of the DRMMP that
• Creation of the Scientific Consortium which was put in place in Mumbai. It reveals a flat
is composed of local experts in relevant fields nonhierarchical structure that puts emphasis on
Building Disaster Resiliency Through Disaster Risk Management Master Planning 369
teamwork and consensus building but at the same Building Disaster Resiliency Through Disaster Risk
time assigns responsibility and accountability to Management Master Planning, Table 1 Data typi-
cally collected in the development of the DRMMP
the project team. It requires experts, practitioners,
policymakers, community leaders, and private Administrative and Build environment,
institutional including:
sector working together to provide inputs to
Demographic and socio Buildings
B
develop the elements of the DRMMP. In the economic
Disaster Risk Reduction in Greater Mumbai Pro- Physical profiles Critical and essential
ject, there were more than 100 members of the including: facilities
Advisory Committee representing various orga- Topology; geology; Infrastructure: transport,
nizations of government, civil society, technical hydrology; soil, climate water, drainage, sanitation,
power, communication,
agencies, and the private sector. In the Building a health
Disaster Resilient Quezon City Project, the vari- Land use High loss facilities
ous Focus Groups were organized under a Tech-
nical Working Group (TWG). They are
composed of technical specialists from the differ-
ent departments of the Quezon City Government collected is indicated in Table 1 below. The data
officially assigned by the Office of the Mayor to is compiled in a City Risk Profile shared and
work with EMI to develop the various elements validated with all the stakeholders. Knowledge-
of the DRMMP. sharing mechanism is defined and endorsed to
ensure the data is kept current and is benefiting
other planning functions of the city. In addition to
Strong Emphasis on Data and Science facilitating DRM planning, tax mapping provides
another motivation for cities to have a centralized
The development of the DRMMP requires a database in uniform format.
strong emphasis on data and science. It includes
collecting evidence to understand the DRM con- Analysis of the Legal and Institutional
text and situation, assessing the inherent risks of Arrangements for DRM
the city, developing an information database on The DRMMP process is anchored on the legal
disaster risk management, and identifying the and institutional arrangements that define the
gaps and needs. The following sections describe geopolitical boundaries of the concerned plan-
several key activities that are conducted to pro- ning authority (i.e., the local government), its
duce the required assessments to develop the jurisdictional responsibility, and the potential
DRMMP. elements of the plan. An in-depth analysis of the
legal and institutional context is essential in pro-
Development of the DRM Database viding legitimacy and credibility to the plan and
and City Risk Profile in avoiding potential jurisdictional conflicts. Sev-
A significant part of the effort goes into data eral outputs are generated such as a network
collection, review, and structuring all data into a analysis that shows the organizational chart of
single DRM database. Data must be collected the DRM undertaking in the city with key agen-
from several city departments as well as from cies and their relationships. A functional map-
relevant government agencies and other service ping is also completed to indicate functions,
providers. This is the most challenging task in the roles, and responsibilities of each organization
process as not all departments willingly share in the DRM process. The results are shared and
data. The data is typically organized at the reso- validated with the stakeholders through focus
lution of the lowest geopolitical boundary of the group exercises to ensure that the DRM organi-
city. This will enable an effective use and man- zation is understood by all concerned agencies
agement of the data by the local and sublocal and officials. An example of network analysis is
authorities. A list of the data that needs to be shown in Fig. 4.
370
Building Disaster Resiliency Through Disaster Risk Management Master Planning, Fig. 4 Network representation indicating relationship of DRM stakeholders in
Building Disaster Resiliency Through Disaster Risk Management Master Planning
Building Disaster Resiliency Through Disaster Risk was generated through the conduct of a hazard, vulnera-
Management Master Planning, Fig. 5 Earthquake bility and risk assessment, which is a vital component of
risk profile for Quezon City, Philippines, showing the the DRMMP
hotspot counties. The map and corresponding information
Building Disaster Resiliency Through Disaster Risk Management Master Planning 373
Building Disaster Resiliency Through Disaster Risk order to assess the geospatial systemic impacts ③ from an
Management Master Planning, Fig. 6 The diagram, earthquake, one must first understand the geographic dis-
as an example of a situational analysis exercise, shows the tribution of the system ① being analyzed and the geo-
lifeline system and resiliencies and how they influence graphic distribution of the hazard ②, in this case
community resilience. The elements of resiliency are earthquake ground motions and permanent ground defor-
numbered from ① to ⑦, each with visual representations. mation. Taken from the Disaster Risk Reduction in
The arrows identify prerequisite elements; for example in Greater Mumbai Project. 2011
374 Building Disaster Resiliency Through Disaster Risk Management Master Planning
Building Disaster Resiliency Through Disaster Risk Management Master Planning, Table 2 Results of
consequential analysis of a magnitude 7.2 earthquake to physical resources of Quezon City
Primary damage: Primary loss: life/
Primary hazards building/ injury, repair costs, Secondary hazard/ Secondary loss:
faulting, shaking, structural function, damage: liquefaction. business/operations
liquefaction, non-structural/ communication/ landsliding fire, interruptions, market
landsliding equipment control hazmat, flooding. . . share, reputation
Collapse of Damage to water Trauma (Stampede) Possible trash slide in Loss of jobs
water sam reservoir dumpsites
Extreme flood Building collapse Isolation of victims Loss of property, lives Loss of income
Loss of lives, (Eastwood, Libis due to no road/
physical injuries area) highways access
Damages to Loss of water Operation Physical injuries Loss of opportunities
main roads, supply interruptions in
infrastructure hospitals
Loss of lives Fire Eastwood, malls
Trapped inside the Houses made from Loss of livelihood
building light materials
Damage to Biogas plant explosion Production halts
equipment
Stranded people Informal settlers Businesses, offices
Chaos/stampede Food shortage No classes
(inside buildings)
Contamination of
waterways
the current practice. Through the consequential taking into account key internal and external ele-
analysis, the stakeholders achieve an understand- ments that may affect the attainment of the resil-
ing of the impacts of the current practices and ience vision and mission.
define strategies and rationales for reducing these After reaching consensus on the resilience
impacts. They also identify the “entry points” for strategies, programs, projects, and activities are
mainstreaming DRR in each particular develop- then formulated. Apart from listing down and
ment sector. Table 2 shows an example of the discussing potential investments, stakeholders
result of a consequential analysis for a magnitude also prioritize projects and activities. This prior-
7.2 to the Physical Resources Sector of Quezon itization process is based on a set of criteria that
City. helps stakeholders focus their attention to inter-
Next is the establishment of a vision for resil- ventions that will have the greatest impact in
ience and a mission to reduce risk through a terms of reducing their risk while taking into
visioning exercise. In the Disaster Risk Reduc- account their inherent limitations such as budget
tion and Management Plan (DRRMP) process in capacity or resource constraints. The final form of
Quezon City, the stakeholders took some time to the DRMMP is then validated. An example of a
establish consensus around developing the vision DRMMP program and its related projects are
for the city. There was a long gridlock in identi- shown in Table 3.
fying a common descriptor for the natural envi-
ronment around descriptors such as habitable,
sustainable, resilient, “clean and green.” This is Ensuring Implementation of the DRMMP
followed by the development of resilience strate-
gies through the conduct of Strengths- The time frame for a DRMMP can span anywhere
Weaknesses-Opportunities-Challenges (SWOC) from 6 to 15 years, depending on the planning
exercise to formulate and prioritize strategies by cycles of cities. For the DRMMP to be fully
Building Disaster Resiliency Through Disaster Risk Management Master Planning 375
Building Disaster Resiliency Through Disaster Risk Management Master Planning, Table 3 Example of
DRMMP program, related projects, and priority ranking of each project
Program 2: Mainstreaming DRR in land use planning and land use management
Mitigation and prevention projects Priority
Formulating a communications plan to institutionalize HVRA within Quezon City (with focus on Immediate B
hotspot areas)
Mainstreaming DRR in the 2010–2030 comprehensive land use plan Medium-
term
Consultation workshops for pilot project identification and feasibility of selected land use management Short-term
methods for flood and earthquake risk reduction
Program for improving Quezon City Government’s capacity and expertise in building and Medium-
infrastructure construction project development permitting, monitoring and evaluation term
Restructuring existing M and E program for stakeholder response(s) to and development impacts of Short-term
Quezon City Government’s projects on disaster risk reduction and management
Pilot projects on preferred land use management methods (e.g., development regulations, zoning Short-term
earthquake hazard areas, floodway zoning and set-backs, fire zones)
implementable, there has to be a clear under- benchmarks and can be used to measure progress.
standing of the legal and administrative processes The first is the quantitatively derived Urban
of a city that concerns with approval of mandated Disaster Risk Indicators (UDRI) which provide
plans. This requires knowing who approves, a holistic view of disaster risk in a local authority
when, and through which step in the approval by capturing the direct physical damages and the
process should the DRMMP be incorporated so aggravating social conditions contributing to
that it becomes integrated in the budgeting cycle. total risk. The model and methodology referred
This way, it is ensured that budget for priority to here as the Urban Disaster Risk Indicators was
projects identified in the DRMMP gets allocated originally developed for the Inter-American
in the overall budget of the city. Figure 7 below Development Bank through the IDB-IDEA Indi-
provides an example of the budget approval pro- cators Program by the Institute of Environmental
cess of Quezon City indicating how the Local Studies (IDEA) of the National University of
Disaster Risk Reduction and Management Plan Colombia, Manizales (NUCM) (Cardona 2003;
(LDRRMP) gets integrated in the Annual Invest- Carreno et al. 2006). In the context of EMI’s
ment Plan (AIP). The Annual Investment Plan DRMMP methodology and approach, the indica-
refers to the indicative yearly expenditure tors were used as an innovative risk communica-
requirements of cities that will be integrated tion tool to engage stakeholders in understanding
into their annual budget. their involvement and taking ownership of the
Understanding who approves, when, and risk factors in the city. Besides the initial EMI
through which steps in the approval process of implementation in Metro Manila (Fernandez
cities helps ensure that DRMMP projects and et al. 2006), the risk indicators have also been
activities are incorporated in the budget and, as successfully implemented with stakeholders in
such, get funded. Istanbul (Khazai et al. 2009), Amman
(Bendimerad et al. 2009), Mumbai (EMI and
Municipal Corporation of Greater Mumbai
Monitoring, Evaluation, and 2011a), and Quezon City (Bendimeard
Communication et al. 2013). The application of the Urban Disas-
ter Risk Indicators in a city will prove its value if
To track the progress (or lack of progress) of it is “mainstreamed” within the DRM practices
implementation, monitoring and evaluation and other planning processes that impact DRM in
tools are developed. One of these tools is to the city. To reach this objective, the indicators
develop a set of Indicators that set initial have been linked to the quantitative outputs and
376 Building Disaster Resiliency Through Disaster Risk Management Master Planning
Acronym
CPDO+QC Department
CPDO City Planning and Development LDRRMC+
Office+LDC+City
office LDRRMO+LDC+City Council+
Council+Mayor+HLURB+DILG
LCD Local Development Council OCD
MMDA
HLURB Housing and Land Use Regulatory
Board
DILG Department of Interior and Local
Government
CLUP and Zoning Plan
MMDA Metro Manila Development LDRRMP
Authority (2012-2042)
CLUP Comprehensive Land Use Plan
CDP Comprehensive Development Plan
LDIP Local Development Investment CDP (9 years)
Plan
LDRRMC Local Disaster Risk Reduction and
Management Council
LDRRMO Local Disaster Risk Reduction and LIDP (9 years)
Management Office
Building Disaster Resiliency Through Disaster Risk highlighting how the local disaster risk reduction and
Management Master Planning, Fig. 7 Example of a management plan gets integrated into the annual
budget approval process in Quezon City, Philippines investment plan
analyses of the DRMMP. Guided by recommen- of stakeholders in the city, such that they are
dations following implementation of the risk used as an effective risk communication tool
indicators by EMI in different cities, the current that inherently relates to the city-level DRM
methodology encompasses the following three practices; as a planning tool that aids in
broad areas: correcting, reviewing, and deciding on where
to invest resources; and as Disaster Risk Man-
• To develop the system of indicators in close agement benchmarks which assist in policy
collaboration with the technical staff and offi- decision-making and monitoring of different
cials in a city, to engage them in the develop- risk reduction practices implemented at the
ment of the data and the understanding of the local level.
conceptual framework and methodology, and
to ensure ownership over the indicator system The second group of indicators is the qualita-
and its periodic updating and upgrading tive Disaster Risk and Resiliency Indicators
• To work closely with the Focus Group, in (DRRI) that is a set of ten (10) indicators that
evaluating the indicators, such that they are establishes initial benchmarks to measure to what
relevant to the megacity scale of analysis, extent risk reduction approaches have been
represent the conditions and reality of socio- mainstreamed in the organizational, functional,
economic vulnerability in the city, and reflect operational, and development systems and
the outcomes of the various studies under- processes of local governments (EMI and
taken as part of the DRMMP (e.g., earthquake MCGM 2011b). The Disaster Risk and Resil-
risk assessment, transportation study, housing iency Indicators are linked to the priorities of
and shelter, land use planning, construction the Hyogo Framework for Action 2015–2015.
codes and standards, water supply and water The aim of these indicators is to understand how
treatment analysis, etc.) well the local authority is performing in
• To implement the system of indicators mainstreaming DRR into different sectors based
through a process of engaging a wider group on predefined benchmarks and performance
Building Disaster Resiliency Through Disaster Risk Management Master Planning 377
B
strategic Goals in Mainstreaming DRR
Systemic integration of risk reduction Mainstreaming
Development and strengthening of
approaches into critical services and disaster risk reduction
institutions, policies and capacities for
infrastructure, and emergency preparedness, into development
mainstreaming disaster risk reduction
response and recovery policies and planning
Planning
LIA IT SDRR ESF HVRA & LUP-CSS
Legal and Training Shelter and Housing Emergency Support Hazard,
Institutional Risk Communication Transport-Water- Functions Vulnerability and
Sanitation Risk Assessment
Land use Planning
Building Disaster Resiliency Through Disaster Risk Management Master Planning, Fig. 8 Disaster risk and
resiliency indicators (DRRI) (Source: EMI and MCGM (2011b))
targets. The rationale for selecting the 10 DRRRI perspective of their institutions and not only the
indicators can be traced in Fig. 8 by following the indicators that relate most closely to their
information from top to bottom of the chart. The activities.
main aim of the DRRRI indicators is to track The scores for the DRRI are derived from a
progress on the mainstreaming of risk reduction self-assessment along key functional activities/
approaches in the city’s organizational, func- policies undertaken by the DRMMP Focus
tional, operational, and development systems Groups. A handbook is developed to describe
and processes. The mainstreaming goal is further the organization of the indicators, a rationale for
divided into three strategic goals shown in the their selection, and provide guidelines for their
chart. Each of the strategic goals corresponds to implementation and scoring that is contextual-
one or more key areas analyzed in the DRMMP ized to the city for which it is implemented by
where these goals are to be implemented. Finally, providing “guiding questions” and “evidence for
two indicators corresponding to each of the discussion” derived from in-depth interviews
five key areas of mainstreaming are shown. with stakeholders along each indicator (EMI
Mainstreaming risk reduction and achieving risk and MCGM 2011b). With five Performance Tar-
resiliency cuts across all the key areas and all the get Levels of attainment: little or no awareness,
10 indicators shown below. Thus, the stake- awareness of needs, engagement and commit-
holders using and scoring the indicators should ment, policy engagement and solution develop-
look at all 10 indicators together from the ment, and full integration, the DRRI allocates a
378 Building Disaster Resiliency Through Disaster Risk Management Master Planning
ranking for the 10 indicators that fall under the city has to make continued investments in the
main 5 areas of mainstreaming, resulting in a medium to long term to make sure that it con-
graphic visualization of the mainstreaming disas- tinues on the path toward risk reduction.
ter risk reduction at the Municipality at a point
in time.
The Indicator System is also used as a risk Summary
communication and planning tool to assist in
policy development, decision-making, and in Local authorities have difficulty in dealing
monitoring the effectiveness of programs, pro- with the risks from disasters due to the fact
jects, and activities. In actual practice, however, that they are overburdened by their development
there is no clear motivation for cities to monitor aspirations and mandated responsibilities. Main-
performance and outcome of DRR investments streaming DRM in the daily functions and
beyond accounting for the number of projects and services of the city is the key to manage disaster
activities conducted. This is also because there risks in urban areas without exacerbating cost for
are no existing tools as yet for monitoring and local authorities. By incorporating disaster risk
evaluating performance of DRR programs, pro- parameters explicitly in the planning of processes
jects, and activities that have been developed for and projects, it can be ensured that in future,
cities. natural hazards will encounter resilient cities
that are capable of withstanding their impact,
reduce their losses, cope, and recover faster. Ulti-
Lessons Learned mately, impacts of hazards become mere emer-
gencies rather than disasters.
Building resilience, especially within the context Mainstreaming necessitates the distribution of
of urban areas, requires not just raising awareness roles and responsibilities between the various
on risk reduction but actual investments on pro- levels of government and with the society.
grams, projects, and activities that reduce risk. It gives emphasis to the local authority’s imple-
For the plan to be fully understood, accepted, and mentation role, central government’s coordina-
supported, it has to engage a broad range of tion function (manifested in the control
stakeholders at all stages of the planning process. processes, allocation of resources, and exercising
However, they have to be equipped with the oversight on the performance of the local author-
knowledge and tools to be able to participate ity), and the participation of the civil society
meaningfully and contribute to resilience build- which is crucial in all facets of the process.
ing. Stakeholders have to understand what types Thus, it optimizes resources from these stake-
of risk they face, what are the implications of holder institutions to reduce the risks.
these risks to their own mandated responsibili- Following the mainstreaming tenet, the
ties, what options are available to them to reduce DRMMP provides a road map for a proactive
their risk, and what possible trade-offs they need DRM approach to support local-level long-term
to make in order to prioritize limited resources planning and programming of disaster risk reduc-
and address inherent constraints. tion activities. It follows a series of analyses in
Full implementation of the DRMMP, how- four phases, namely, (1) Organization and Prep-
ever, requires continued technical assistance. aration, (2) Diagnosis and Analysis, (3) Plan
Even if funding is available from cities to imple- Development and Plan Implementation, (4) Mon-
ment DRR projects, oftentimes, they lack the itoring and Evaluation. The DRMMP process is
resources to manage and undertake DRR pro- designed to align along core planning processes
jects. It is a resource-intensive approach. The of local authorities. Furthermore, it provides
DRMMP only provides the roadmap for change stakeholders with an opportunity for learning
in practice from a reactive to a more proactive the concepts of hazard, vulnerability, and risk
approach. For the actual change to take place, the and to understand their relevance to planning.
Building Disaster Resiliency Through Disaster Risk Management Master Planning 379
The DRMMP was developed taking into consid- ▶ Earthquake Protection of Essential Facilities
eration other related DRM models that focus on ▶ Earthquakes and Their Socio-economic
local-level action through mainstreaming. Consequences
With DRMMP’s strong emphasis on data and ▶ Economic Recovery Following Earthquakes
science, evidences are collected to understand the Disasters B
DRM context and situation, assessing the inher- ▶ Emergency Response Coordination Within
ent risks of the city, developing information data- Earthquake Disasters
base on DRM, and identifying the gaps and ▶ Land Use Planning Following an Earthquake
needs. These tasks are achievable through a par- Disaster
ticipatory approach involving a broad range of ▶ Resilience to Earthquake Disasters
stakeholders. Implementing the DRMMP
requires organizing these stakeholders into
References
(1) Advisory Committee, (2) Scientific Consor-
tium, and (3) Focus Groups. Guiding these Asian Disaster Reduction Center (ADRC) (2005) Total
groups is the Project Management Team. disaster risk management: good practices. ADRC,
The DRMMP provides opportunity for these Kobe
stakeholders to understand the parameters of vul- Baas S, Ramasamy S, De Pryck J, Battista F (2008) Disas-
ter risk management systems analysis. FAO, Rome
nerability and risk in the context of their own Bendimerad F (2009) State-of-the-Practice Report on
responsibilities and functions. With this knowl- Urban Disaster Risk Management. EMI internal
edge, they can systematically and rigorously report. http://emi-megacities.org/?emi-publication=
identify and prioritize programs, projects, and state-of-the-practice-report-on-urban-disaster-risk-
management. Quezon City
activities to implement in managing and eventu- Bendimerad F, Zayas J (2015) Urban Resilience Master
ally reducing their risk. These projects and activ- Planning: A Guidebook for Practitioners and Policy
ities are institutionalized by integrating these in Makers. EMI, Quezon City
the budgeting cycle. This way, it is ensured that Bendimerad F, Mattingly S, Seva V, Kustu O, Khazai B,
Reyes M, Berse K, Al-Soub W, Quran S, Al-Sous G,
budget for priority projects identified in the Khalili H, Dabaseh A (2009) Amman – disaster risk
DRMMP gets allocated in the overall budget of management master plan, support to building national
the city. Monitoring and evaluation tools are then capacities for earthquake risk reduction in Amman.
developed to track the progress (or lack of pro- volume 1, Earthquake and Megacities Initiative
(EMI), main report. UNDP Jordan – 00051485,
gress) of implementation. 31 Mar 2009, p 130
With their engagement in the development of Bendimerad F, Khazai B, Zayas J, Daniell JE, Salunat J,
the plan, the stakeholders not only provide essen- Perez B, Lingad L, Dakis K, Bergonia L, Magtaas B,
tial knowledge to the plan but acknowledge their Barba B, Padao I, Pino L, Lanuya E, Valera A, Villa P,
Hisanan T (2013) Building a disaster resilient Quezon
roles and responsibilities in the implementation City project: hazard, vulnerability and risk assessment
of the plan, thus taking ownership. By anchoring report. Earthquake and Megacities Initiative, Quezon
on mainstreaming, the DRMMP puts in place a City
structured approach for setting up a DRM system Bendimerad F, Zayas J, Padao I (2015) Legal and institu-
tional arrangements guidebook, Urban resilience
that is based on science, understood by all, and guidebook series. EMI, Quezon City
provides the rationale for investments in DRR Berse K, Bendimerad F, Cruz J, Dagli W, Deocariza M,
programs and projects. Khazai B, Lopez IG, Pulmano N, Seva V (2009)
Desk review of international disaster risk
management models and tools intended for local insti-
tutional application. GFDRR, p 118, Makati City,
Cross-References Philippines
Cardona OD (2003) The need for rethinking the concepts
▶ Community Recovery Following Earthquake of vulnerability and risk from a holistic perspective: a
necessary review and criticism for effective risk man-
Disasters agement. In: Bankoff G, Frerks G, Hilhorst D (eds)
▶ Earthquake Disaster Recovery: Leadership and Mapping vulnerability: disasters, development and
Governance people. Earthscan Publishers, London
380 Building Monitoring Using a Ground-Based Radar
Carreno L, Cardona O, Barbat A (2006) Urban seismic South Pacific Applied Geoscience Commission (SOPAC)
risk evaluation: a holistic approach. Nat Hazards (2002) Comprehensive Hazard and Risk Management
40(1):137–172 (CHARM): guidelines for pacific Island countries.
Central United States Earthquake Consortium (CUSEC) SOPAC report DM0044, Suva, 32 pp
(1997) Disaster resistant communities: a community- UNISDR (2009) 2009 UNISDR terminology on disaster
based approach to hazard mitigation. CUSEC J 4(1):2–4 risk reduction. Available at: http://www.unisdr.org/
Commonwealth of Australia (2004) Emergency risk man- we/inform/terminology. Cited Mar 2014
agement application guide. Emergency Management
Australia, Canberra
EMI (2009) State-of-the-practice report on urban
disaster risk management. Quezon City. Retrieved
Mar 2014, from http://emi-megacities.org/home/ Building Monitoring Using a
components/com_booklibrary/ebooks/StateofPractice_
UrbanDRM.pdf
Ground-Based Radar
EMI (2014) Legal and Institutional Arrangements Frame-
work Guidebook. EMI, Quezon City Guido Luzi and Michele Crosetto
EMI, Municipal Corporation of Greater Mumbai CTTC Division of Geomatics, Av. C.F. Gauss,
(MCGM) (2011a) Mumbai DRMMP framework
report and implementation work outputs. EMI,
Castelldefels, Spain
Quezon City
EMI, MCGM (2011b) Final technical report supplement.
EMI, Quezon City Synonyms
EMI, Quezon City Government (QCG) (2014) Disaster
risk reduction and management plan 2014–2020. EMI,
Quezon City Displacements; Measurement; Natural frequen-
Fernandez J, Mattingly S, Bendimerad F, Cardona O cies; Real aperture radar; Structural health
(2006) Application of indicators in urban and mega- monitoring
cities disaster risk management: a case study of Metro
Manila. EMI topical report TR-07-01. p 30. A report
for this application is available from the earthquakes
and megacities Initiative website. http://emi- Introduction
megacities.org/upload/3cd_2007_MISReport_TR0701.
pdf. Cited Mar 2014
Khazai B, Bendimerad F (2011a) Megacity Indicator Sys-
This entry is focused on a terrestrial remote sens-
tems (MIS) for DRM in Greater Mumbai. In: ing technique that can provide a valuable input
Bendimerad F, Daclan JM, Dagli W, Zayas J (eds) to earthquake engineering: the real aperture
Mumbai Disaster Risk Management Master Plan radar interferometry. In the last decades
(DRMMP). Earthquake and megacities initiative final
technical report No CR-1 1-0, 31, MCGM
non-contacting measurement techniques based
Khazai B, Bendimerad F (2011b) Risk and resiliency on optical sensors that can be employed for vibra-
indicators. Earthquake and megacities report tion monitoring have been largely described in
No. TR-1 1-03. MCGM, p 103 the literature (Passia et al. 2008; Nassif
Khazai B, Wenzel F, Kilic O, Basmaci A, Konukcu B,
Mentese EY, Sungay B (2009) Megacity indicator
et al. 2005; Gueguen et al. 2010). As an example,
systems (MIS) for disaster risk management in Istan- techniques like the Scanning Laser Doppler
bul. International conference on megacities: risk, vul- Vibrometry (SLDV) appear to be an appropriate
nerability and sustainable development, 7–9 Sept non-contacting alternative to accelerometers,
2009. Helmholtz Centre for Environmental
Research – UFZ, Leipzig
offering wide frequency bandwidths
Nardo M, Saisana M, Saltelli A, Tarantola S, Hoffman A, (0–200 kHz for the most common models) and
Giovannini E (2005) Handbook on constructing com- extreme sensitivity to velocity and displacement
posite indicators. OECD Statistics working paper. (down to 2 pm). On the other hand, they suffer
OECD, Paris
Siddiqui KM (2007) National institute of disaster manage-
important limitations, due to constraints in the
ment: 2nd Asian ministerial conference on DRR maximum observation angle and the maximum
high level round table 1. Available at the National working range. In addition, they might need
Institute of Disaster Management in India. http:// external acoustic stimulation to carry out mea-
www.nidm.gov.in/amcd_presentations/HighLevel_
Table1/BanglaDesh.pdf. Cited Mar 2014
surements with adequate sensitivity, as described
Building Monitoring Using a Ground-Based Radar 381
in Esposito et al. (2004). Finally, they measure active. This can be achieved through the mea-
velocity, not displacements. Although less fre- surement of dynamic parameters of buildings
quently, the use of microwave sensors to evaluate and structures, which can be done using a RAR
the vibration state of structures as buildings, brid- system, and the dynamic parameters of the under-
ges, and wind turbine towers has been investi- lying soil in order to assess possible resonance B
gated (Tarchi et al. 1997; Farrar et al. 1999; phenomena.
Pieraccini et al. 2003; Pieraccini et al. 2006; This entry is organized in four main sections.
Bartoli et al. 2008; Pieraccini et al. 2008; Luzi The first one explains the RAR working principle
et al. 2012a). Microwave techniques measure and the main characteristics of the RAR measure-
directly the displacement and have demonstrated ments. The second to fourth sections describe
to be an effective tool when the monitored dis- three examples of vibration measurement: a tall
placements are down to tens of microns. Its capa- tower, a building, and a bridge.
bility to provide input data that can be used in
a modal testing has been assessed, and the
obtained results finally encouraged the produc- Working Principle
tion of radar instrumentation aimed at this appli-
cation (Coppi et al. 2010). A coherent Real- Radar is a well-known instrument able to detect
Aperture Radar (RAR) is able to measure and and range objects. It sends a modulated electro-
monitor the natural vibrations of buildings, struc- magnetic signal, the transmitted pulse, along the
tures, and infrastructures. The main observations line of sight (LOS), and receives the portion of
provided by the sensor are the displacements the signal reflected by the illuminated area. Dif-
between the radar sensor and the observed target. ferent objects or parts of the illuminated area will
The displacements can be observed at relatively reflect the microwave radiation with different
high frequencies, up to 100 or 200 Hz with cur- intensities: the basic signal provided by a radar
rently available sensors, thus providing a dense is a profile of the intensity backscattered from the
sampling of the displacements over time and scene as a function of the distance, usually
space. For this reason, the technique is able to referred as “range profile.” The radar acquires
measure and monitor the vibrations of the simultaneously the contributions from the differ-
observed objects. It is possible to derive key ent parts of the structure included in its antenna
information of the observed objects as the main field of view (FOV), and different amplitude
vibration frequencies, the main vibration modes, peaks correspond to contributions from parts of
etc. from the displacement time series. There are the observed structure located at different dis-
two main potential applications of the described tances. The FOV of the radar is dictated by the
technique. The first and more consolidated is pulse width and the antenna characteristics. The
structural health monitoring, where RAR can elementary sampling volume of a radar measure-
provide input for the characterization of engi- ment is usually called radar bin and different bins
neering structures, especially of their structural are used to isolate different parts of the monitored
systems. This input can be provided periodically structure. Such vision of the radar prevents to see
to determine the current state of system health details as optical systems do, but at the same time,
and, in particular, to monitor its aging and degra- this averaging allows to enhance the general
dation. In addition, the RAR can provide behavior of the structure. In addition, the radar
a valuable input after extreme events, such as acquires all the bins simultaneously (with respect
earthquakes or blast loading, for rapid condition to the observed phenomenon) thus allowing to
screening and assessment of the integrity of the achieve a spatial sampling useful to estimate
structures at hand. The second potential applica- modal shapes. Reflecting properties of a given
tion is the seismic risk evaluation of populated structure are strongly related to its geometry and
and industrialized areas that are seismically the dielectric characteristics of the medium: for
382 Building Monitoring Using a Ground-Based Radar
Displacement
0.05
Bin n 0 B
−0.05
−0.1
−0.15
−0.2
450 455 460 465 470 475 480 485 490 495
Time
e
ng Bin selection
Ra Echo intensity
80
70
FOV 60
50
Range profile
40
α 30
20
Radar 10
0
0 2 4 6 8 10 12 14 16 18 n 22 m 28 30 32 l 36 38 40
Building Monitoring Using a Ground-Based Radar, range, sampled at regular spatial steps, the radar bins.
Fig. 1 Scheme of a standard procedure aimed at retriev- A displacement time signal is obtained from the interfer-
ing displacement time series from a radar acquisition. The ometric phase of the echo of the selected bin using Eq. 1
radar collects an amplitude profile as a function of the
amplitude profile as a function of the range, sam- manufactured and marketed by IDS (Ingegneria
pled at regular spatial steps (radar bins); dei Sistemi SpA), are summarized. It is important
(2) obtaining the interferometric phase of the to note that the results here discussed do not
echo of the selected bin; and, finally, basically depend on the radar system used.
(3) transforming the phase temporal variations IBIS-S is designed to monitor large structures as
into displacement time signal for analysis and bridges and wind turbine towers. The system
processing using Eq. 1, as, for example, to calcu- consists of a sensor module, a control PC, and
late the spectral content of the signal. a power supply unit. The maximum acquisition
The acquired data can be processed through rate is currently 200 Hz. This parameter depends
conventional signal processing tools. Due to the on the selected maximum range and decreases as
statistical characteristics of the sampled phenom- the distance increases. The sensor is installed on
ena, random and stationary, a typical approach a tripod equipped with a rotating head used to
consists in calculating the Power Spectral Den- adjust the bearing of the sensor towards the inves-
sity (PSD) using the Welch method (Welch tigated structure. The radar equipment is a
1967). continuous-wave (CW) step-frequency
Nowadays only a few companies market (SF) transceiver, which transmits continuous
radars for vibration monitoring; as a reference, waves at discrete frequency values, sampling
the main characteristics of one of the interfero- a bandwidth B at a constant interval. The trans-
metric radar most frequently cited in the literature ceiver transmits an electromagnetic signal at
and available at the CTTC, the IBIS-S, a central frequency of 17.2 GHz with a maximum
384 Building Monitoring Using a Ground-Based Radar
268m
ΔS
dLOS
Rrange
D a
0m
Radar
Building Monitoring Using a Ground-Based Radar, indicating the fraction of the displacement (bold arrow),
Fig. 2 Photography of the tower and the radar sensor dLOS, detected by the radar (right)
(left) and a simple scheme of the measurement geometry
60
55
Bin 571
50
Bin 648
45
B
40 Bin 618
Radar Intensity [dB]
35
Bin 743
30 Bin 694
25 Bin 703
20
15
10
Building Monitoring Using a Ground-Based Radar, a tower monitoring. A selection of bins, to be processed
Fig. 3 Range profile, intensity of the radar echoes as and analyzed, is marked
a function of the range expressed in bins, obtained from
interpretation. Finally, its geometry and compos- analysis of the vibration behavior of the moni-
ing materials, metal and concrete, offer a strong tored structure. The curves retrieved from this
radar reflectivity and this allows acquiring data plot indicate that the amplitudes of the displace-
from ranges up to hundreds of meters. It is worth ment increase with the height; this trend fits the
noting that an analogous monitoring carried out main mode of vibration of the tower, similar to
with conventional point-wise sensors, e.g., accel- the first mode expected from a clamped beam.
erometers, would be unadvisable due to its high The period of oscillation is also easily calculable
costs and the strong electromagnetic noise of this in approximately 3.7 s (frequency 0.27 Hz).
environment. A spectral analysis has been carried out through
Different contributions from parts located at the estimation of the PSD calculated using the
different heights can be clearly identified and Welch method (Welch 1967) to look for further
selected for processing and observing a range vibration modes and deepen the analysis of the
profile acquired at a distance of 266 m from the vibrating characteristics of the tower. The PSD
tower and with an observation angle of 37 calculated with the following configuration is
(Fig. 3). A typical approach consists in selecting plotted in Fig. 5: window duration = 50 s,
a set of bins sampling the entire height of the type = Hamming, overlapping between win-
structure and drawing the variation of the mea- dows: 66 % for different bins. The results
sured displacement as a function of the time. obtained indicate the presence not only of the
Figure 4 represents the starting point of the 0.27 Hz peak but also further peaks at 0.49 and
386 Building Monitoring Using a Ground-Based Radar
5
Rbin571 [mm] Rbin618 [mm]
Rbin648 [mm] Rbin694 [mm]
4 Rbin703 [mm] Rbin743 [mm]
LOS Displacement (mm)
0
70 80 90 100 110 120 130 140
Time(s)
Building Monitoring Using a Ground-Based Radar, Fig. 4 Example of a 70 s sample of the displacement of
different bins retrieved from radar measurements as a function of the time
Rbin 571
0.26 Hz Rbin 618
Rbin 648
Rbin 694
0.48 Hz Rbin 703
Rbin 743
2
10
Line of sight Displacement [mm*mm/Hz]
0.72 Hz
1
10
0
10
−1
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.2 1.4 1.8 2 2.4 2.8 3 3.4 4 4.6 5
Frequency [Hz]
Building Monitoring Using a Ground-Based Radar, Fig. 5 PSD calculated for the bins shown in Fig. 4. Different
colors refer to different bins. The three peaks related to tower vibration are marked
Building Monitoring Using a Ground-Based Radar 387
80
107
Radar, Fig. 6 Location of Radar bin
the seismometers (#104 and
70
#107) with respect to the 104 69
building, and radar bins B
60
(#53 and #69) selected from 58
the range profile, to
50
perform a radar data
Range
validation
40
30
20
RAR
10
0
80
70
60
50
40
30
20
10
0
Radar Intensity
0.72 Hz. Some minor peaks, probably caused by placed, in two different configurations, on the
the cables and its interaction with the tower, are roof and at different heights of the structure.
also present. The experiment was carried out with the aim of
Case studies aimed at estimating vibration comparing seismometer data with the displace-
characteristics of towers of different shapes and ment retrieved from several radar bins. In partic-
heights, selected among architectural heritages, ular, a direct comparison was carried out between
have been carried out by different authors the data obtained using two radar bins and the
(Pieraccini et al. 2005; Atzeni et al. 2010; Gentile closest sensors installed in the building,
and Saisi 2011). whose position is shown in Fig. 6, where the
geometry of the measurement is also depicted.
The seismometer acquisition set includes eight
Comparison of Radar Observations and sensors and a wireless networking connection
Conventional Measurements: The Case was used during the measurements. The normal-
of a Building ized PSD calculated for the displacement signal
retrieved from the radar measurements and that
Radar observations can be integrated with some obtained from microtremor signal are compared
standard point-wise measurements as a support to in Fig. 7.
model calculation of the modal shapes of the The location of the main peaks is almost coin-
investigated structures. This is the case described cident and the radar signal shows some minor
in Negulescu et al. (2013), where a building, part peaks too. The amplitude is a little different;
of a sportive complex located in Font Romeu this can be imputed to the fact that the radar
(France), was monitored by means of radar and only measures the LOS component of the dis-
several accelerometers, and the results obtained placement and the two sensors are not exactly
were compared with a model calculation. The positioned in the same point. The agreement
building has a reinforced concrete structure, between radar observations and seismometer
with double basement, ground floor, 10 storeys, measurements was also confirmed by numerical
and a total height of 33 m (not including the model calculation and modal shape estimation
basement). In this study, several sensors were (Negulescu et al. 2013).
388 Building Monitoring Using a Ground-Based Radar
mm2/Hz 9.00E-04
Radar Radar bin #69
8.00E-04
7.00E-04
6.00E-04
5.00E-04
4.00E-04
3.00E-04
2.00E-04
1 2 3 4 5 6 7 8 9 10
2.14 GHz
2.72 GHz
2.47 GHz
mm2/Hz
0.012
0.008
4.05 GHz
6.8 GHz
8.8 GHz
0.006
0.004
0.002
0
1 2 3 4 5 6 7 8 9 10
Frequency (Hz)
Building Monitoring Using a Ground-Based Radar, Fig. 7 Comparison of the normalized Power Spectral Density
(PSD) functions from the records obtained using the radar technique, bin #69 (top) and the seismometer #107 (down)
40
30
a
20
10
0 50 100 150 200 250 300
Range bin
6
12
10
81
66
47
25
n
n
Bi
Bi
Bi
Bi
Bi
Bi
Pier
Radar
Building Monitoring Using a Ground-Based Radar, geometry of acquisition indicating the selected bins: bin
Fig. 8 Photography of the radar pointing towards the 25, 47, 66, 81, 106, and 121 belong to the bridge deck,
deck (a); range profile acquired (b); scheme of the while bin 260 corresponds to the pier (c)
Building Monitoring
Using a Ground-Based
Radar, Fig. 9 Photos and
geometry of the
acquisitions with the radar
pointing towards the inner
stay cables positioned
within the two pillars (left)
and watching the set of
guys tied to the soil (right)
while the part close to the pier (bin#25) experi- tension force of the cable (applying a formula
ences a 500 mm lowering. Using the second geo- based on the tau string model; see Caetano and
metrical configuration (Fig. 9), the radar Cunha 2011; Gentile and Ubertini 2012; Luzi et
observation is focused towards the cables. In al. 2014). The calculation of the PSD of the
this case, the measurement of the natural fre- displacement signal allows detecting the main
quency of the cable can be used to estimate the vibrating frequencies of the cables. Figure 11
390 Building Monitoring Using a Ground-Based Radar
−1.5
−2 Rbin 25
Rbin 47
−2.5 Rbin 66
Rbin 81
−3
−3.5
−4
725 730 735 740 745 750 755 760 765
time [sec]
103
Rbin 22
Line of Sight Displacement [mm2/Hz]
Rbin 23
X: 7.04
102 X: 3.52 Y: 44.49
Y: 133.5 X: 10.58
Y: 16.76 X: 14.16
Y: 11.6 X: 17.78
Y: 7.174
101
X: 21.52
Y: 1.722
100
10−1
0 2 4 6 8 10 12 14 16 18 20 22
frequency [Hz]
Building Monitoring Using a Ground-Based Radar, Fig. 11 PSD calculated for two bins corresponding to two stay
cables. Different colors refer to different bins
shows the results obtained from the simultaneous expected from the tau string model, thus allowing
measurement of the displacements of two bins the application of the model and the estimation of
corresponding to two cables. the tension force of the cables. It is worth noting
The values of the main peaks displayed in that obtaining the same information using standard
Fig. 11, 3.52, 7.04, 10.58, 14.16, 17.78, and local sensors as accelerometer should be
21.52, show that they are related to each other by unfeasible. Besides, the knowledge of the spectral
a linear relationship. In fact, they are natural mul- behavior of all the stay cables can also be used for
tiples of the fundamental frequency, 3.52 Hz, evaluating the health state of the structure or com-
although the higher values gradually deviate. The paring these experimental results with model anal-
first frequencies are in agreement with the values ysis provided by structural engineering analysis.
Building Monitoring Using a Ground-Based Radar 391
Other devices that could be classified as passive damping systems where the expected perfor-
dampers include tuned mass and tuned liquid mance is similar or higher than that of buildings
dampers. with a conventional lateral load resisting system.
Passive dampers are typically used within Recent analytical and experimental research
conventional lateral load resisting systems to dis- showed that steel moment-resisting frames B
sipate a significant portion of the seismic input (MRFs) with elastomeric dampers can be
energy (Christopoulos and Filiatrault 2006). The designed to be lighter and perform better than
degree to which a passive damper is able to conventional steel MRFs under the design basis
achieve this goal depends on the inherent proper- earthquake (DBE; 475-year return period)
ties of the lateral load resisting system, the prop- (Karavasilis et al. 2011a, 2012a). However, it
erties of the passive damper and its connecting was shown that it is generally not feasible to
elements, the characteristics of the ground design steel MRFs with passive dampers at a
motion, and the limit state being investigated practical size to eliminate inelastic deformations
(Symans et al. 2008). Over the past decades, in main structural members under the DBE
passive dampers have been specified for applica- (Karavasilis et al. 2011a). To address this issue,
tion to buildings and bridges with a wide variety a seismic design strategy for steel MRFs that
of structural configurations. The growth in appli- isolates damage in removable steel dampers and
cation of passive dampers has been steady to the uses in parallel viscous fluid dampers to reduce
extent that there are now numerous applications drifts has been proposed by Karavasilis
(Soong and Spencer 2002). et al. (2012b). A study shows that supplemental
The first design/analysis procedure for build- viscous damping does not always ensure ade-
ings with passive dampers was published in quate reduction of residual drifts (Karavasilis
SEAONC (Whittaker et al. 1993). The intent of and Seo 2011). A recent work evaluates the seis-
these guidelines was to direct engineers in locat- mic collapse resistance of steel MRFs with vis-
ing dampers in the lateral load resisting system to cous fluid dampers and shows that supplemental
reduce repair costs and business interruption due viscous damping does not always guarantee a
to a severe earthquake. In the mid-1990s, the better seismic collapse resistance when the
Federal Emergency Management Agency strength of the steel MRF with dampers is lower
(FEMA) funded the development of guidelines or equal to 75 % of the strength of a conventional
for the seismic rehabilitation of buildings propos- steel MRF (Seo et al. 2014).
ing, among others, basic principles for the seis- In the following section, two major real-world
mic design with passive dampers. The outcome applications of passive dampers in a tall building
of this effort was the NEHRP Guidelines for the and a bridge are briefly discussed. The most com-
Seismic Rehabilitation of Buildings, FEMA mon types of passive dampers and the models
Reports 273 and 274 (ATC 1997a, b). Later on, used to simulate their hysteretic behavior are
the 2003 NEHRP Recommended Provisions then presented. Finally, the analysis methods
(BSSC 2004) included a chapter on design and that can be utilized to predict the response of
analysis methods for buildings with passive structures with passive dampers are discussed.
dampers. These methods were reformatted and
included in the 2005 edition of the ASCE/SEI
7-05 standard Minimum Design Loads for Build- Application of Passive Dampers in
ings and Other Structures (ASCE 2005). The Buildings and Bridges
2003 NEHRP Recommended Provisions (and
the current 2010 ASCE/SEI 7 standard (ASCE Two relatively recent applications of passive
2010)) allow a reduced design base shear force dampers in a building and a bridge are discussed
for the seismic design of buildings with passive below.
394 Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis
Buildings and Bridges Equipped with Passive construction showing partial view of megabraces; (b)
Dampers Under Seismic Actions: Modeling and view of installed viscous fluid dampers (Symans
Analysis, Fig. 1 (a) Torre Mayor Tower under et al. 2008)
Torre Mayor Tower, Mexico magnitude and epicenter 500 km from the build-
The Torre Mayor Tower is a 77,000 m2, 57-story ing site (Rahimian and Romero 2003).
steel and reinforced concrete office/hotel tower in
Mexico City. Its construction was completed in Rion-Antirion Bridge, Greece
2003 and its height is 225 m. The lower half of the The Rion-Antirion Bridge (Fig. 2), that crosses
superstructure is a rectangular tower which con- the homonymous strait in Greece, is located in the
sists of steel framing encased in concrete, while very active seismic region of the Gulf of Corinth.
the upper half consists of steel framing (Fig. 1a). The deck of this multi-span cable-stayed bridge is
The seismic design of the structure had the chal- continuous and fully suspended from four pylons
lenging objective of achieving “operational” per- (total length of 2,252 m). Its approach viaducts
formance for an expected 8.2 magnitude seismic comprise 228 m of concrete deck on the Antirion
event (Rahimian and Romero 2003). The seismic side and 986 m of steel composite deck on the
design employs 96 large nonlinear viscous fluid Rion side. Said structures are designed to with-
dampers with 2,670 and 5,340-kN force output stand seismic events generating ground accelera-
capacities (Fig. 1b). The dampers on the faces of tions of up to 0.48 g through the use of viscous
the building are installed in megabraces, which fluid dampers and other seismic devices. The
are diagonal braces that span over more than one Main Bridge seismic protection system consists
story, as shown in Fig. 1a. Note that, a first design of fuse restraints and viscous fluid dampers that
of the building without dampers resulted in act in parallel, connecting the deck to the pylons
weight of the building that was excessive for the (Fig. 2b). The fuse restraints are designed as a
soil capacity. The addition of the dampers rigid link intended to withstand high wind loads
resulted in less steel weight and acceptable soil up to a predetermined force. Under the action of
bearing pressure. The structure experienced no the design earthquake, the fuse restraints will fail
damage during a seismic event with 7.8 and leave the dampers free to dissipate the semi
Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis 395
Buildings and Bridges Equipped with Passive under construction; (b) arrangement of viscous fluid
Dampers Under Seismic Actions: Modeling and dampers and fuse restraints on the main piers (Infanti
Analysis, Fig. 2 Rion-Antirion Bridge: (a) main bridge et al. 2004)
induced energy. The approach viaducts are seis- commonly described using various nonlinear
mically isolated utilizing elastomeric isolators hysteretic models. Typical dampers falling in
and viscous dampers. this category include metallic dampers and fric-
tion dampers (Symans et al. 2008; Christopoulos
and Filiatrault 2006).
Passive Dampers and Constitutive Motion-activated passive dampers consist of
Models secondary systems that disturb the flow of energy
in the structure. These systems have the same
Passive dampers can be classified in three main resonance period with the main structure.
categories: (1) velocity-activated or rate-depen- Tuned-mass dampers (TMDs) are an example of
dent, (2) displacement-activated or rate-indepen- motion-activated dampers. This last category is
dent, and (3) motion-activated (Symans not discussed further in this contribution, but the
et al. 2008; Christopoulos and Filiatrault 2006). reader is referred to documents such as the one by
Velocity-activated or rate-dependent dampers Soong and Spencer (2002) for more details.
have force output that depends on the relative A summary of velocity-activated and
velocity across the damper. Therefore, their effi- displacement-activated passive dampers is pro-
ciency to dissipate energy depends on the fre- vided in Fig. 3 (reproduced from Symans
quency of the input ground motion. The et al. (2008)). Figure 3 lists the basic construc-
behavior of such dampers is commonly described tion, the idealized hysteretic response and the
using various models of linear viscoelasticity. associated physical model, and the major advan-
Typical dampers falling in this category include tages and disadvantages of each passive damper.
viscous fluid dampers and viscoelastic dampers
(Symans et al. 2008; Christopoulos and Viscous Fluid Dampers
Filiatrault 2006). Viscous fluid dampers consist of a hollow cylin-
Displacement-activated or rate-independent der filled with fluid (Fig. 3), the fluid typically
dampers have force output that does not depend being silicone based. As the damper piston rod
on the relative velocity across the damper but on and piston head are stroked, fluid is forced to flow
the magnitude of the displacement and possibly through orifices either around or through the pis-
the direction of motion. Their behavior is ton head. The resulting differential in pressure
396 Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis
Viscoelastic Solid
Viscous Fluid Damper Metallic Damper Friction Damper
Damper
BRB
Basic
Construction
ADAS
Force
Force
Force
Force
Idealized
Hysteretic
Behavior
Displacement Displacement Displacement Displacement
Disadvantages - Possible fluid seal - Limited deformation - Device damaged - Sliding interface
leakage (reliability capacity after earthquake; may conditions may
concern) - Properties are require replacement change with time
frequency and - Nonlinear behavior; (reliability concern)
temperature- may require nonlinear - Strongly nonlinear
dependent analysis behavior; may excite
- Possible debonding higher modes and
and tearing of VE require nonlinear
material (reliability analysis
concern) - Permanent
displacements if no
restoring force
mechanism provided
Buildings and Bridges Equipped with Passive hysteretic behavior, physical models, advantages, and dis-
Dampers Under Seismic Actions: Modeling and advantages of passive dampers for seismic protection
Analysis, Fig. 3 Summary of basic construction, applications (Symans et al. 2008)
across the piston head can produce very large fluid particles and the piston head, which leads to
forces that resist the relative motion of the energy dissipation in the form of heat. Interest-
damper. The fluid flows at high velocities, ingly, although the damper is called a viscous
resulting in the development of friction between fluid damper, the fluid typically has a relatively
Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis 397
Fd (kN)
Analysis, 0
Fig. 4 Hysteretic behavior B
of a nonlinear viscous −250
damper from nonlinear
dynamic analysis
−500
(Karavasilis et al. 2012b)
−0.06 −0.03 0 0.03 0.06
u (m)
low viscosity, e.g., silicone oil with a kinematic bonded to steel plates, which are attached to the
viscosity in the order of 0.001 m2/s at 20 C. The structure within chevron or diagonal bracing
term viscous fluid damper is associated with the (Fig. 3). As one end of the damper displaces
macroscopic behavior of the damper, which is with respect to the other, the viscoelastic material
essentially the same as that of an ideal linear or is sheared resulting in the development of heat,
nonlinear viscous dashpot, i.e., the resisting force which is dissipated to the environment. Visco-
is directly related to the velocity (Symans elastic dampers provide both a velocity-
et al. 2008). dependent force, which provides supplemental
Experimental testing (Seleemah and viscous damping to the system (similar to viscous
Constantinou 1997) has shown that the behavior dampers), and a displacement-dependent elastic
of viscous fluid dampers corresponds to a restoring force (Symans et al. 2008).
nonlinear viscous dashpot and can be modeled Experimental testing (Bergman and Hanson
by the following nonlinear force-velocity 1993) has shown that, under certain conditions,
relation: a suitable mathematical model for simulating the
mechanical behavior of viscoelastic dampers is
PðtÞ ¼ Cju_ðtÞja sgn½u_ðtÞ (1) the Kelvin model of viscoelasticity. This model is
described by the following equation:
where P(t) is the force developed by the damper,
u(t) is the displacement across the damper, C is
PðtÞ ¼ KuðtÞ þ Cu_ðtÞ (2)
damping coefficient, a is an exponent whose
value is determined by the piston head orifice
design, sgn[] is the signum function, and the where K is the storage stiffness of the damper and
overdot indicates differentiation with respect to C is the damping coefficient which is equal to the
time, t. For earthquake protection applications, ratio of the loss stiffness to the frequency of
the exponent a typically has a value ranging loading. The physical model corresponding to
from about 0.3 to 1.0. For a equal to unity, the Eq. 2 is a linear spring in parallel with a linear
damper can be described as an ideal linear vis- viscous dashpot.
cous dashpot (Symans et al. 2008). Figure 4 A constitutive model that can be used to
shows the behavior of a nonlinear viscous damper describe with more accuracy the behavior of vis-
installed in a prototype steel building subjected to coelastic dampers is the generalized Maxwell
a strong earthquake ground motion from (GM) model. The GM model consists of a paral-
nonlinear dynamic analysis conducted in lel combination of a linear spring, a dashpot and
Karavasilis et al. (2012b). multiple in-series combinations of springs and
dashpots as shown in Fig. 5 (Karavasilis
Viscoelastic Solid Dampers et al. 2011b). Figure 5 shows the GM model in
Viscoelastic solid dampers generally consist of terms of shear stress, t, and shear strain, g, of the
solid elastomeric pads (viscoelastic material) viscoelastic material. Under harmonic loading of
398 Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis
cyclic frequency o, the GM model provides stor- software, simply multiply G0 and Gm in Fig. 5
age shear modulus equal to with the ratio Ad/td (Ad: horizontal area of the
viscoelastic material; td: thickness of the visco-
X
n
ðobm Þ2 elastic material) to transform t-g behavior to
G0 ðoÞ ¼ G0 þ Gm (3) force (F = t Ad) – deformation (d = g td)
m¼1 1 þ ðobm Þ2
behavior.
Figure 6 shows the hysteretic behavior of a
and loss factor equal to
typical viscoelastic damper from nonlinear
X dynamic analysis conducted in Fan (1998).
n
ðobm Þ
ðob0 ÞG0 þ Gm
m¼1 1 þ ðobm Þ2 Metallic Dampers
ðoÞ ¼ (4)
X
n
ðobm Þ2 Metallic dampers are designed to yield, based on
G0 þ Gm
m¼1 1 þ ðobm Þ2 the ability of steel to dissipate energy. Two major
types of metallic dampers are the buckling-
These equations can be used to calibrate the GM restrained brace (BRB) and the added damping
model against experimentally obtained values of and stiffness (ADAS) device.
G0 and for different cyclic frequencies. To use A BRB consists of a steel brace with a cruci-
the GM model as a combination of linear springs form cross section that is surrounded by a stiff
and dashpots within a structural analysis steel tube. The region between the tube and brace
is filled with a concrete-like material and a spe-
cial coating is applied to the brace to prevent it
from bonding to the concrete. Thus, the brace can
slide with respect to the concrete-filled tube. The
Go boGo G1 G2 Gm
confinement provided by the concrete-filled tube
b1G1 b 2G 2 bmG allows the brace to be subjected to compressive
loads without buckling. Under compressive
loads, the BRB behavior is essentially identical
to its behavior in tension. Since buckling is
t, g
prevented, significant energy dissipation can
occur over a cycle of motion (Symans
Buildings and Bridges Equipped with Passive
Dampers Under Seismic Actions: Modeling and et al. 2008).
Analysis, Fig. 5 Generalized Maxwell model for visco- The ADAS damper (Whittaker et al. 1991)
elastic dampers consists of a series of steel plates wherein the
0.2
Analysis,
Fig. 6 Hysteretic behavior
of viscoelastic damper from
nonlinear dynamic analysis 0
(Fan 1998)
−0.2
25°C
−0.4
−0.04 −0.02 0 0.02 0.04
VE Damper Deformation (in)
Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis 399
bottom of the plates are attached to the top of a elastic-perfectly plastic component. The force
chevron bracing arrangement and the top of the output F of the model is
plates are attached to the floor level above the
bracing. As the floor level above deforms later- F ¼ pku þ ð1 pÞFy z (5)
ally with respect to the chevron bracing, the steel B
plates are subjected to a shear force. The shear where u is the deformation across the model, Fy
force induces bending moments over the height the yield force, k the elastic stiffness, p the post-
of the plates, with bending occurring about the yield stiffness ratio, and z a dimensionless hys-
weak axis of the plate cross section. The geomet- teretic parameter governed by
rical configuration of the plates is such that bend-
ing moments produce a uniform flexural stress k
z_ ¼ _ þ gÞ
u_½1 jzjn ðbsgn½uz (6)
distribution over the height of the plates so that Fy
inelastic action occurs uniformly over the full
height of the plates. For example, in the case where b and g are parameters controlling the
that the plates are fixed-pinned, the geometry is shape of the hysteresis, n is a parameter that
triangular; however, if the plates are fixed-fixed, controls the sharpness of the smooth transition
the geometry is an hourglass shape. from the elastic to the inelastic region of the
Other types of metallic dampers have been hysteresis, sgn[.] is the signum function, and the
developed for beam-column connections, braces overdot denotes derivative with respect to time.
and base isolation systems. Early developments Equation 5 shows that the Bouc-Wen model
include the U-strip hysteretic dampers and the accounts for kinematic hardening (i.e., post-yield
T-ADAS damper. Other examples include the force increases with increasing deformation) due
honeycomb damper used as seismic isolation sys- to the post-yield stiffness ratio p. However, the
tem in bridges, C-shaped and E-shaped hysteretic model does not account for the isotropic harden-
dampers for bridges, slit-type dampers applied to ing (i.e., yield force Fy increases due to cyclic
beam-column connections or brace members, inelastic deformation) in the hysteresis of steel
yielding shear panels, cast-iron yielding fuses energy dissipation devices.
installed in braces and hourglass shape pins To account for isotropic hardening in the
installed in beam-column connections. Bouc-Wen model, the mathematical formulation
A complete reference list for the aforementioned developed by Karavasilis et al. (2012b) is
metallic dampers is provided in Vasdravellis described below:
et al. (2014). The yield force Fy needs to be updated by
Various mathematical models that describe considering the history of the imposed cyclic
yielding behavior of metals can represent the deformation u. Examination of the constitutive
hysteretic behavior of metallic dampers such as Eqs. 5 and 6 reveals that a change in the
the standard Bouc-Wen model (Wen 1976). The yield force Fy can be achieved by appropriately
standard Bouc-Wen model results from the par- including a third shape control parameter Ф in
allel combination of an elastic component and an Eq. 6:
k h
i
z_ ¼ _ Þ þ g Fsgnðu_ÞðsgnðzÞ þ sgnðu_ÞÞ
u_ 1 jzjn bsgnðuz (7)
Fy
upl, c
The parameter Ф quantifies isotropic hardening Fp ¼ Fmax, p
1 exp pF, p (8a)
and is calculated using functions that cause Ф to uy
increase exponentially with increasing cumula-
tive plastic deformation upl,c, i.e., or
400 Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis
400 2000
Test Test
Model Model
200 1000
F (kN)
F (kN)
0 0
−200 −1000
SP-6 BRB
−400 −2000
−15 −10 −5 0 5 10 15 −100 −50 0 50 100
u (mm) u (mm)
Buildings and Bridges Equipped with Passive model proposed in Karavasilis et al. (2012b) and experi-
Dampers Under Seismic Actions: Modeling and mental results for a low-yield-strength shear panel and
Analysis, Fig. 7 Results from the modified Bouc-Wen a BRB
upl, c reflects a 25 % increase in the initial yield
Fn ¼ Fmax, n
1 exp pF, n (8b)
uy strength Fy due to isotropic hardening. The
term sgn½u_ after the parameter F in Eq. 7 ensures
where uy ¼ Fy =k is the yield deformation, pФ,p that the above calculations apply to the case of a
and pФ,n are parameters that control the isotropic negative deformation increment and a negative
hardening rate due to cumulative plastic defor- ultimate value of z.
mation, and Fmax,p and Fmax,n are the maximum The state determination procedure requires as
possible values of F for the fully saturated iso- an input the previous force and deformation, the
tropic hardening condition, i.e., for upl.c ! 1, Fp previous z value, and the current deformation.
! Fmax,p and Fn ! Fmax,n. On the other hand, The current value of the parameter F is then
when upl.c = 0.0, Fp = 0.0 and Fn = 0.0. calculated based on the following rules: Eq. 8a
Fp and Fn are used to independently capture is used to update Fp when the deformation incre-
isotropic hardening in different loading direc- ment changes from negative to positive within
tions (positive and negative). Typically, yielding the plastic region of the hysteresis; Eq. 8b is
devices exhibit the same isotropic hardening used to update Fn when the deformation incre-
in different loading directions, and therefore, ment changes from positive to negative within
Fmax,p = Fmax,n and pФ,p = pФ,n. However, the the plastic region of the hysteresis; F equals to
model can simulate different isotropic hardening Fp when a positive deformation increment
in different loading directions (e.g., compressive occurs; and F equals to Fn when a negative
and tensile loading in BRB hysteresis) by using deformation increment occurs. With the current
different parameter values in Eqs. 8a and 8b. F value known, Eq. 7 is numerically integrated to
To understand the effect of F, consider the obtain the current value of z which is simply used
proposed Bouc-Wen model with p = 0.0 in Eq. 5 to provide the current force output F of
(i.e., without kinematic hardening), n = 1 and the model.
b + g = 1, under a positive deformation incre- Figure 7 shows how the aforementioned mod-
ment, i.e., sgn½u_ ¼ 1. Assume that in the previ- ified Bouc-Wen model captures the behavior of
ous deformation increment, z has reached its low yield strength shear panels showing signifi-
positive ultimate value zu, and therefore, sgn cant isotropic hardening in their hysteresis as well
[z] = 1 and z_ ¼ 0:0. For this case, Eq. 7 yields as the behavior of BRBs showing different iso-
zu ¼ 1=ð1 2 FÞ and Eq. 5 yields tropic hardening in tension and compression.
F ¼ Fy zu ¼ Fy =ð1 2 FÞ . When F = 0.0
(i.e., without isotropic hardening), F is equal Friction Dampers
to Fy =ð1 0:0Þ ¼ Fy . When F 6¼ 0.0 (e.g., Friction dampers dissipate seismic energy by
F = 0.1), F is equal to Fy =ð1 2 FÞ ¼ friction that develops between two solid bodies
Fy =ð1 2 0:1Þ ¼ 1:25 Fy . In that case, F sliding relative to one another. Typical friction
Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis 401
X
dampers are slotted-bolted (Grigorian Ci ðfi fi1 Þ2
et al. 1993), where a series of steel plates are T1
xeq ¼
i
X (10)
bolted together with a specified clamping force. 4p mi fi 2
The clamping force is such that slip occurs at a i
pre-specified friction force. Another configura- B
tion is the Pall friction damper, which consists where fi and fi1 are the first modal displace-
of cross bracing that connects in the center to a ments of stories i and i 1, respectively, and mi is
rectangular damper (Pall and Marsh 1982). The the mass of story i. Equation 10 assumes that the
damper is bolted to the cross bracing and, under braces supporting the dampers are stiff enough
lateral load, the structural frame distorts such that (essentially rigid) so that story deformation
two of the braces are subjected to tension and the results in damper deformation rather than brace
other two to compression. This force system deformation. Alternative equations that take into
causes the rectangular damper to deform into a account the effect of the brace flexibility on xeq
parallelogram, dissipating energy at the bolted are provided for various brace-damper configura-
joints via sliding friction (Symans et al. 2008). tions in Hwang et al. (2008).
Experimental testing (Pall and Marsh 1982) xeq is used to calculate the damping reduction
has shown that the idealized Coulomb friction factor, which is then utilized to appropriately
model can model the behavior of friction reduce the ordinates of the elastic response spec-
dampers, i.e., the force output of friction dampers trum. This reduced highly damped spectrum is
is given by used to estimate the response of the structure with
dampers. Lin and Chopra (2003) have evaluated
PðtÞ ¼ mNsgn½u_ðtÞ (9) the premise that the peak displacement of an
elastic SDOF system with a natural period Tn
where m is the coefficient of dynamic friction and and equipped with a viscous damper (damping
N is the normal force at the sliding interface. coefficient C) in series with a brace of stiffness
Kb, can be estimated by a linear viscous SDOF
system with the same period Tn along with a
Analysis Methods for Buildings and simplified calculation of the added damping
Bridges with Passive Dampers ratio (see Eq. 10). This premise was found valid
for SDOF systems with t=T n < 0:02, where t is
The analysis methods used to predict the response the relaxation time, i.e., t ¼ C=K b . As discussed
of buildings and bridges equipped with passive by Lin and Chopra (2003), the relation
dampers are classified in two categories: (1) linear t=T n < 0:02 is satisfied for the practical range
analysis methods and (2) nonlinear analysis of values for the bracing axial stiffness Kb and the
methods. damping coefficient C. Therefore, the damper
and brace properties necessary to ensure that
Linear Analysis Methods elastic drift demands do not exceed a prescribed
All linear analysis methods for structures design limit can be determined using the reduced
equipped with rate-dependent passive dampers elastic design spectrum associated with the
involve a simplified calculation of the supple- equivalent damping ratio.
mental (equivalent) damping ratio, xeq. For In the 2003 NEHRP Recommended Provisions
example, given the damping coefficients Ci at (BSSC 2004) for buildings with passive dampers,
each story i of a building and by assuming linear and for the equivalent lateral force (ELF) analysis
fluid viscous dampers (a = 1 in Eq. 1) positioned (linear static analysis) specifically, the response
in a horizontal configuration, xeq at the funda- is defined by two modes: the fundamental mode
mental period of vibration T1 under elastic con- and the residual mode, which is used to approx-
ditions can be estimated as (Whittaker imate the combined effects of higher modes. For
et al. 2003): response spectrum analysis, higher modes are
402 Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis
Base Shear
Ω0Cd
et al. 2008) Ω0 reduced by CD/R
R
ratio when used
with elasto-plastic
V1 curve)
g g 2
D1D = G1 SD1T1D ≤ G1 SDST 1D
2 B1D 2 B1D
4p 4p
DY D1D
Roof Displacement
explicitly evaluated. For both the ELF and the of analysis used for conventional lateral load
response spectrum analysis procedures, the shape resisting systems in seismic codes and adopts
of the fundamental-mode pushover capacity the use of the strength reduction factor to calcu-
curve is not known and an idealized elastoplastic late the required design base shear and the dis-
pushover curve is assumed, as shown in Fig. 8. placement amplification factor to estimate drifts.
Note that, in Fig. 8, the parameters G1 and SDS,
which are used to compute the design earthquake Nonlinear Analysis Methods
displacement, D1D, represent the modal partici- The 2003 NEHRP Recommended Provisions
pation factor for the fundamental mode and the (BSSC 2004) specify procedures for nonlinear
5 % damped design spectral response accelera- static analysis and nonlinear dynamic (response-
tion at short periods, respectively. The idealized history) analysis. The nonlinear static analysis
pushover curve permits defining the effective procedure is similar to the ELF procedure, in
global ductility demand due to the design earth- that the pushover capacity curve is used to define
quake, mD, as the ratio of design roof displace- the nonlinear behavior of the structure. However,
ment, D1D, to the yield displacement, DY. This in the nonlinear static analysis procedure, the
ductility factor is used to calculate various design actual nonlinear force-displacement relation is
factors and to limit the maximum ductility used, rather than an idealized elastoplastic
demand, mmax, in a manner that is consistent curve. In addition, since actual pushover strength
with conventional building response limits. Pas- is known from the nonlinear pushover analysis,
sive dampers should be designed for actual the force reduction for design of the seismic-
fundamental-mode design earthquake forces force-resisting system is based on overstrength
corresponding to a base shear value of VY, while alone with no additional reduction (Symans
the elements of the seismic-force-resisting sys- et al. 2008).
tem be designed for a reduced fundamental-mode In general, the nonlinear dynamic analysis
base shear, V1, where the force reduction is based procedure is the most robust procedure available
on system overstrength, O0 (Symans et al. 2008). for evaluating the behavior of structures with
A simplified linear method of analysis for passive dampers. It allows explicit modeling of
building frames with elastomeric or viscoelastic individual devices, the elements connecting the
dampers has been proposed by Lee et al. (2009) devices to the structure, and the structure itself.
and has been recently extended to the case of fluid If the connecting elements or the structural fram-
viscous dampers by Seo et al. (2014). This ing yields during the response, this behavior must
method is fully compatible with linear methods be incorporated into the analytical model.
Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis 403
It is noted that accurate modeling of the flexibility ▶ Behavior Factor and Ductility
of the floor diaphragm and of the connecting ▶ Buckling-Restrained Braces and Their
elements (braces) is essential since a loss of Implementation in Structural Design of Steel
effective damping may occur if these elements Buildings
are overly flexible. ▶ Earthquake Response Spectra and Design B
Nonlinear dynamic analysis may be Spectra
performed using a variety of commercially avail- ▶ Earthquake Return Period and Its
able software. In addition, there are several aca- Incorporation into Seismic Actions
demic programs available, such as DRAIN-2DX ▶ Equivalent Static Analysis of Structures
(Prakash et al. 1993) and OpenSees (Mazzoni Subjected to Seismic Actions
et al. 2006). Most of these programs can readily ▶ European Structural Design Codes: Seismic
be used to model the behavior of linear fluid Actions
viscous dampers, viscoelastic dampers, friction ▶ Mixed In-Height Concrete-Steel Buildings
dampers, or metallic yielding dampers. However, Under Seismic Actions: Modeling and
modeling of some damping devices (e.g., Analysis
nonlinear viscous dampers and dampers with ▶ Modal Analysis
temperature-dependent or frequency-dependent ▶ Plastic Hinge and Plastic Zone Seismic
mechanical properties) can be more challenging Analysis of Frames
or, in some cases, not possible with a given ▶ Response Spectrum Analysis of Structures
program. Subjected to Seismic Actions
When nonlinear dynamic analysis is used, it is ▶ Seismic Analysis of Masonry Buildings:
often beneficial to investigate the sensitivity of Numerical Modeling
the structure response to one or more systemic ▶ Seismic Analysis of Steel and Composite
parameters. Examples of parameters to vary Bridges: Numerical Modeling
include ground motion scaling parameters and ▶ Seismic Analysis of Steel–Concrete
parameters of the passive dampers (Symans Composite Buildings: Numerical Modeling
et al. 2008). ▶ Soil-Structure Interaction
▶ Steel Structures
▶ Strengthening Techniques: Code-Deficient
Summary Steel Buildings
▶ Structural Design Codes of Australia and New
The scope of this contribution was the presenta- Zealand: Seismic Actions
tion of numerical models and analysis methods ▶ Time History Seismic Analysis
used for structures equipped with passive
dampers. The principal function of passive
dampers and their application in buildings and
bridges were discussed. The most common
References
types of passive dampers and the constitutive
American Society of Civil Engineers [ASCE] (2005) Min-
models used to simulate their hysteretic behavior imum design loads for buildings and other structures.
were presented. Finally, the analysis methods ASCE/SEI, Reston, pp 7–05
used for seismic design and assessment of struc- American Society of Civil Engineers [ASCE] (2010) Min-
imum design loads for buildings and other structures.
tures with passive dampers were covered. ASCE/SEI 7–10. ASCE, Reston
Applied Technology Council [ATC] (1997a) NEHRP
commentary on the guidelines for the seismic rehabil-
Cross-References itation of buildings. 1997 ed, report no FEMA-274,
prepared for the Building Seismic Safety Council
(BSSC) by the Applied Technology Council (ATC).
▶ Assessment of Existing Structures Using Federal Emergency Management Agency (FEMA),
Response History Analysis Washington, DC
404 Buildings and Bridges Equipped with Passive Dampers Under Seismic Actions: Modeling and Analysis
Applied Technology Council [ATC] (1997b) NEHRP Lin W-H, Chopra AK (2003) Earthquake response of
guidelines for the seismic rehabilitation of buildings. elastic single-degree-of-freedom systems with
1997 ed, report no FEMA-273, prepared for the Build- nonlinear viscoelastic dampers. J Eng Mech ASCE
ing Seismic Safety Council (BSSC) by the Applied 129(6):597–606
Technology Council (ATC). Federal Emergency Man- Mazzoni S, McKenna F, Scott MH, Fenves GL
agement Agency (FEMA), Washington, DC (2006) Open system for earthquake engineering simu-
Bergman DM, Hanson RD (1993) Viscoelastic mechani- lation: user command language manual. University of
cal damping devices tested at real earthquake displace- California at Berkeley, Pacific Earthquake Engineer-
ments. Earthq Spectra 9(3):389–418 ing Research Center, Berkeley
Building Seismic Safety Council [BSSC] (2004) NEHRP Pall AS, Marsh C (1982) Seismic response of friction
recommended provisions for seismic regulations for damped braced frames. J Struct Div 108(6):1313–1323
new buildings and other structures. 2003 ed, report no Prakash V, Powell GH, Campbell S (1993) DRAIN-2DX
FEMA-450/1 and FEMA-450/2, prepared by the base program description and users guide. Department
Building Seismic Safety Council (BSSC) for the Fed- of Civil Engineering, University of California at
eral Emergency Management Agency (FEMA), Wash- Berkeley, Berkeley
ington, DC Rahimian A, Romero EM (2003) Super structure. Civ Eng
Christopoulos C, Filiatrault A (2006) Principles of sup- NY 73(6):21–69
plemental damping and seismic isolation. IUSS Press, Seleemah A, Constantinou MC (1997) Investigation
Milan of seismic response of buildings with linear and
Fan C-P (1998) Seismic analysis, behavior, and retrofit of nonlinear fluid viscous dampers. Report no NCEER
non-ductile reinforced concrete frame buildings with 97–0004. State University of New York at Buffalo,
viscoelastic dampers. Phd Lehigh University, National Center for Earthquake Engineering Research,
Bethlehem Buffalo
Grigorian CE, Yang TS, Popov EP (1993) Slotted bolted Seo C-Y, Karavasilis TL, Ricles JM, Sause R (2014) Seis-
connection energy dissipators. Earthq Spectra mic performance and probabilistic collapse resistance
9(3):491–504 assessment of steel moment resisting frames with fluid
Hwang J-S, Huang Y-N, Yi S-L, Ho S-Y (2008) Design viscous dampers. Earthq Eng Struct Dyn 43
formulations for supplemental viscous dampers to (14):2135–2154
building structures. J Struct Eng ASCE 134(1):22–31 Soong TT, Spencer BF Jr (2002) Supplemental energy
Infanti S, Papanikolas P, Benzoni G, Castellano MG dissipation: state-of-the-art and state-of-the-practice.
(2004) 13th world conference on earthquake engineer- Eng Struct 24(3):243–259
ing, paper no 2174, Vancouver Symans MD, Charney FA, Whittaker AS, Constantinou
Karavasilis TL, Seo C-Y (2011) Seismic structural and MC, Kircher CA, Johnson MW, McNamara RJ
non-structural performance evaluation of highly (2008) Energy dissipation systems for seismic appli-
damped self-centering and conventional systems. Eng cations: current practice and recent developments.
Struct 33:2248–2258 J Struct Eng ASCE 134(1):3–21
Karavasilis TL, Ricles JM, Sause R, Chen C (2011a) Vasdravellis G, Karavasilis T, Uy B (2014) Design
Experimental evaluation of the seismic performance rules, experimental evaluation, and fracture models
of steel MRFs with compressed elastomer dampers for high-strength and stainless-steel hourglass shape
using large-scale real-time hybrid simulation. Eng energy dissipation devices. J Struct Eng 140
Struct 33(6):1859–1869 (11):04014087. doi: 10.1061/(ASCE)ST.1943-
Karavasilis TL, Blakeborough T, Williams MS (2011b) 541X.0001014
Development of nonlinear analytical model and seis- Wen YK (1976) Method for random vibration of hyster-
mic analyses of a steel frame with self-centering etic systems. J Eng Mech Div 102(2):249–263
devices and viscoelastic dampers. Comput Struct Whittaker AS, Bertero VV, Thompson CL, Alonso LJ
89(11–12):1232–1240 (1991) Seismic testing of steel-plate energy dissipating
Karavasilis TL, Sause R, Ricles JM (2012a) Seismic devices. Earthq Spectra 7(4):563–604
design and evaluation of steel MRFs with compressed Whittaker AS, Aiken ID, Bergman D, Clark PW, Cohen J,
elastomer dampers. Earthq Eng Struct Dyn Kelly JM, Scholl RE (1993) Code requirements for
41(3):411–429 design and implementation of passive energy dissipa-
Karavasilis TL, Kerawala S, Hale E (2012b) Model for tion systems. In: Proceedings of ATC-17–1 seminar on
hysteretic behaviour of steel energy dissipation seismic isolation, passive energy dissipation, and
devices and evaluation of a minimal-damage seismic active control, vol 2. ATC, Redwood City, pp 497–508
design approach for steel frames. J Constr Steel Res Whittaker AS, Constantinou MC, Oscar MR, Johnson
70:358–367 MW, Chrysostomou CZ (2003) Equivalent lateral
Lee KS, Ricles J, Sause R (2009) Performance based force and modal analysis procedures of the 2000
seismic design of steel MRFs with elastomeric NEHRP provisions for buildings with damping sys-
dampers. J Struct Eng ASCE 135(5):489–498 tems. Earthq Spectra 19(4):959–980
C
Cast Fuse Braces: Design and bolted to a fabricated splice plate assembly
Implementation which is connected to a standard beam-column
gusset plate connection. A YBS prototype is
Michael Gray illustrated in a full-scale frame test configuration
Cast Connex Corporation, Toronto, ON, Canada in Fig. 1.
Much like the TADAS system (Tsai
et al. 1993), the fingers are triangular to promote
Synonyms yielding along their entire length. The bending
moment diagram of each finger is triangular, like
Cast steel yielding brace system a cantilever. The triangular shape of the finger
ensures that moment resistance diagram matches
the applied moment diagram and the extreme
Introduction fiber of the finger yields simultaneously at all
points along the length of the finger.
The yielding brace system (YBS) is a seismic- The yielding fingers are connected to a fabri-
resistant bracing system for concentrically braced cated plate assembly via slotted holes. The slotted
frame building structures designed at the Univer- holes enable the end of the finger to translate in
sity of Toronto (Gray et al. 2013). The YBS the direction that is normal to the brace axis as the
consists of a special yielding connector, or fuse, finger is deformed. A standard hole would
at the end of a concentric steel brace. When large restrain that movement and cause large axial
lateral earthquake forces are imposed on a build- forces, and strains, to develop in the fingers,
ing designed with the YBS, the connector’s spe- which would decrease their displacement
cially designed triangular fingers yield in flexure. capacity.
The yielding of the fingers dissipates seismic
energy and limits the forces that the remaining
structural elements must resist. This is analogous Nonlinear Response
to the fuse in an electric circuit, which is why the
yielding connector is sometimes called a “fuse.” The YBS yielding connector exhibits a full, duc-
Yielding connectors are comprised of two tile hysteretic response. A sample hysteresis of
castings which are connected to the end of a the YBS is presented in Fig. 2. The stiffness and
brace member. Each casting includes an elastic displacement capacity of the braces and connec-
arm portion with a series of triangular yielding tors have been demonstrated to be very similar to
fingers. The ends of the yielding fingers are other high-ductility braces, like the buckling
# Springer-Verlag Berlin Heidelberg 2015
M. Beer et al. (eds.), Encyclopedia of Earthquake Engineering,
DOI 10.1007/978-3-642-35344-4
406 Cast Fuse Braces: Design and Implementation
1000
Post-yield
500 stiffness increase
0
−500
−1000
−1500
−2000
−2500
−100 −80 −60 −40 −20 0 20 40 60 80 100
Brace Elongation [mm]
restrained brace. However, unlike the buckling deformed shape. The small axial force appears
restrained brace (▶ Buckling-Restrained Braces in the connector’s global response as distinct
and Their Implementation in Structural Design of increase in strength and stiffness at large dis-
Steel Buildings) which has a higher strength in placements. The benefit of such a response,
compression than in tension, the YBS has a sym- when considered in the structural design, is the
metric response. ability to limit the amount of inelastic demand
Another difference between the two systems is that can concentrate on a single story of a
the post-yield response of the YBS. Although the multistory building. As the lateral deformation
slotted holes, to which the fingers are bolted, of a single story increases, the brace eventually
reduced the axial forces that develop in the fin- increases in strength, thereby forcing adjacent
gers at large displacements, some axial force stories to become engaged and inelastic demand
develops due to the finger’s second-order to spread up the height of the building.
Cast Fuse Braces: Design and Implementation 407
Introduction
Cross-References
Characterization of damping forces in a vibrating
▶ Buckling-Restrained Braces and Their Imple- structure has long been an active area of research
mentation in Structural Design of Steel in structural dynamics. The most common
Buildings approach to model viscous forces is to use the
410 Classically and Nonclassically Damped MDOF Structural Systems
so-called viscous damping, first introduced by may suffice, based on the spectral characteristics
Lord Rayleigh (1877), which relies on the of the processes under study (Priestley 1987).
assumption that the damping matrix is a linear In particular, spectral characteristics can be
combination of the mass and stiffness matrices. employed to evaluate the time-variant central
Since its introduction, this model has been used frequency and bandwidth parameters, which
extensively and is now usually known as Ray- characterize in a synthetic way a nonstationary
leigh damping or classical damping (Argyris and stochastic process. The spectral characteristic has
Mlejnek 1991; Clough and Penzien 1993). With been proved appropriate for describing
such a damping model, the modal analysis pro- nonstationary stochastic processes and can be
cedure, originally developed for undamped sys- effectively employed in structural reliability
tems, can be used to analyze damped systems in applications, such as the computation of the
a very similar manner. time-variant probability that a random process
In reality, physical structures or systems are outcrosses a given limit-state threshold, and is
generally comprised of many substructures tied useful for reliability of earthquake-sensible struc-
together in various fashions. These substructures tures (Barbato and Conte 2008).
can be made up of a variety of materials, i.e.,
metals, plastics, and woods. Furthermore, these
substructures may be connected to one another by Classically Damped MDOF Structures
rivets, bolts, dampers, springs, friction, etc. Also,
the spatial geometry of the structure may be very Free Vibrations of MDOF Structures
complicated. All of these factors influence the The free vibration equations of motion of an
inherent dynamical properties of the structure. For n-degree-of-freedom elastic structure are
these structures, mass, damping, and stiffness dis-
tribution (matrices) of the system are rather com- €ðtÞ þ KUðtÞ ¼ 0
MU (1)
plicated. In general, for real-life structures, the
damping matrix for such a system will not always where M and K are the n n mass and stiffness
be proportional to the mass and stiffness matrix. symmetric and positive definite matrices, while
For nonproportional damping, the well-known U(t) is an n vector whose component collects the
modal coordinate transformation, which diago- time-dependent response components of the
nalizes the system mass and stiffness matrices, structure. Equation 1 is a coupled system of n dif-
will not diagonalize the system damping matrix. ferential equations to be integrated with initial
Therefore, when a system with nonproportional conditions
damping exists, the equations of motion are
coupled when formulated in n dimension physi- U_ ð0Þ ¼ U_ 0 , Uð0Þ ¼ U0 : (2)
cal space. Fortunately, the equations of motion
can be uncoupled when formulated in 2n dimen- If the mass and stiffness matrices are positive
sion state space, by means of complex analysis. definite, the solution of Eq. 1 can be written as
When dealing with earthquake excitation, the
dynamic behavior of structural and mechanical X
n
systems subjected to uncertain dynamic excita- UðtÞ ¼ Cj exp ioj t fj (3)
tions can be described, in general, through ran- j¼1
h i
K o2j M fj ¼ 0 (4) Furthermore, giving an initial condition in dis-
placement proportional to the j-th modes with
zero velocity, the structure oscillates maintaining
It can be shown that the orthogonality of the the shape of the given modes with circular fre-
normal modes fj respects the stiffness and mass quency equal to oj.
matrices (see, e.g., Clough and Penzien 1993). By
setting with f = [f1 . . . fn] the n n modal Free Vibrations of Classically Damped C
matrix, whose column collects the normal modes MDOF Structures
of the structure, the following identity holds: The free vibration equations of motion of an
n degree of freedom elastic structure with viscous
fT Mf ¼ I, fT Kf ¼ V2 (5) damping are
where the aj coefficients are related to damping q€j ðtÞ þ 2xj oj q_j ðtÞ þ o2j qj ðtÞ ¼ 0 (18)
ratios and frequencies by
with associated solution
1Xn1
xk ¼ aj ok2j1 : (16)
2 j¼0
q_j ð0Þþxj oj qj ð0Þ
qj ðtÞ¼ exj oj t qj ð0Þcos oj t þ sin oj t
oj
As for the free vibrations, the motion of an n-dof
classically damped structure can be thought as (19)
a linear combination of normal modes fj, each
of them vibrating with the associated circular where oj is the reduced or damped frequency of
frequency oj and damping xj. Indeed, using the the j-th damped natural frequency and is related
modal coordinates q(t) as defined in Eq. 6 and to the j-th undamped natural frequency and
Eq. 11 and left multiplying by fT, we obtain damping ratio by
n uncoupled differential equation of motion in
modal coordinates: qffiffiffiffiffiffiffiffiffiffiffiffi
oj ¼ oj 1 xj : (20)
q€ðtÞ þ Sq_ ðtÞ þ V2 qðtÞ ¼ 0 (17)
which are the n uncoupled equations of motion in The response of the entire structure can be explic-
modal coordinates: itly written as follows:
" #
X
n U_ 0 þ xj oj U0
UðtÞ ¼ fTj M U0 cos oj t þ sin oj t exj oj t fj (21)
j¼1
oj
Equation 21 shows that the classically damped following n uncoupled forced differential equa-
vibration of an n-dof structure can be thought as tion of motion in modal coordinates are obtained:
a linear combination of normal modes fj, each of
them vibrating with the associated circular fre- q€ðtÞ þ Sq_ ðtÞ þ V2 qðtÞ ¼ pðtÞ (23)
quency oj and decaying with damping xj.
having set
Forced Vibrations of Classically Damped
MDOF Structures pðtÞ ¼ fT f ðtÞ (24)
Let us consider the forced vibration equations of
motion of an n degree of freedom elastic structure to be solved with initial conditions
with classical viscous damping:
qð0Þ ¼ fT Uð0Þ, q_ ð0Þ ¼ fT U_ ð0Þ: (25)
€ðtÞ þ CU_ ðtÞ þ KUðtÞ ¼ f ðtÞ
MU (22)
Equation 23 in explicit form becomes
where f (t) is the time-dependent n dimensional q€j ðtÞ þ 2xj oj q_j ðtÞ þ o2j qj ðtÞ ¼ pj ðtÞ: (26)
force vector. Equation 16 represents a system of
n coupled linear differential equations to be
Solution of Eq. 26 can be split in the homogenous
solved with the initial conditions (Eq. 2). Using
q0j (t) and particular qpj (t) integrals
the modal coordinates q(t) transform, Eq. 6 and
Eq. 22, and left multiplying by fT, because of the
qj ðtÞ ¼ q0j ðtÞ þ qpj ðtÞ: (27)
orthogonality conditions (Eqs. 5 and 12), the
Classically and Nonclassically Damped MDOF Structural Systems 413
d U 0 I U
In Eq. 31, HM(o) is the transfer modal matrix, ¼ (37)
dt V M1 K M1 C V
diagonal of order n, whose j-th element is given
by
The matrix equation of motion Eq. 37 can be recast
1 into the following first-order matrix equation:
H Mj ðoÞ ¼ : (32)
o2j o2 þ 2ixj oj o
Z_ ¼ GZ (38)
Once HM(o) is known, the nodal response
where
transform
ð þ1 U
Z¼ (39)
UðoÞ ¼ f ðtÞeiot dt (33) U_ 2n1
1
414 Classically and Nonclassically Damped MDOF Structural Systems
1
0nn Inn HM ðoÞ ¼ K o2 M þ ioC : (46)
G¼ (40)
M1 K M1 C 2n2n
If the damping matrix C is decoupled in the sum
The subscripts in Eqs. 39 and 40 indicate the dimen- of a diagonal matrix CD collecting all the ele-
sions of the vectors and matrices to which they are ments of the principal diagonal of C and a
attached. Solution of Eq. 38 can be written as non-diagonal matrix CF collecting the elements
out of the principal diagonal of C
Z ¼ Celt : (41)
C ¼ CD þ CF (47)
Substituting Eq. 41 in Eq. 38 returns the eigen-
value problem: then defining the diagonal matrix HD(o)
1
GC ¼ lC (42) HD ðoÞ ¼ K o2 M þ ioCD , (48)
The eigenvalue problem in Eq. 42 admits 2n the following approximate representation holds:
complex eigenvectors Cj and complex eigen-
" #
values lj(j = 1,.., 2n) occurring in pair complex X
n
j j
conjugate. It follows that each complex eigenvec- HM ðoÞ ¼ HD ðoÞ I þ ðioÞ ½CF HD ðoÞ
tor Ck with complex eigenvalue lk is associated j¼1
AX(o, t)) of the stationary process YS(t) defined as Response Spectral Characterization of
the Hilbert transform of the embedded stationary Nonclassically Damped MDOF Linear Systems
process XS(t), i.e., Using Complex Modal Analysis
The state-space formulation of the equations of
ð1 motion is also advantageous for the computation
Y ðtÞ ¼ j signðoÞAX ðo, tÞejot dZðoÞ: (62) of the NGSCs of response processes of both clas-
1
sically and nonclassically damped linear MDOF
systems. If only Gaussian input processes are
The cross variance between the response pro- considered, only few spectral characteristics are
cess X(t) and the modulation process Y (t) is used needed to fully describe the response processes of
in the definition of the time-variant central fre- linear elastic MDOF systems, since the response
quency, oc(t), and bandwidth parameter, q(t), of processes are also Gaussian. In particular, if Ui(t)
the NSSP X(t) as denotes the i-th DOF displacement response pro-
cess of a linear elastic MDOF system subjected to
Gaussian excitation, the only spectral character-
sXY_ðtÞ
oc ðtÞ ¼ (63) istics required, e.g., for reliability applications,
s2X ðtÞ
are s2Ui ðtÞ, s2U_ ðtÞ, sUi U_i ðtÞ and sUi Y_i ðtÞ , where
i
Y
complex-valued NSSPs, the complex-valued X¼ _ (66)
Y ð2n1Þ
central frequency and bandwidth parameter
defined in Eqs. 63 and 64 lose the simple
Using complex modal decomposition, the cross-
physical interpretation available for real-
covariance matrices of the response processes
valued NSSPs, even though their computation
and the auxiliary processes can be computed as
is instrumental to the solution of problems
(Barbato and Vasta 2010)
requiring a state-space representation. In addi-
tion, the bandwidth parameter q(t) plays an " #
important role in time-variant reliability anal-
UðtÞUT ðtÞ UðtÞU_ ðtÞ
T
E ZðtÞZT ðtÞ ¼
U_ ðtÞUT ðtÞ U_ ðtÞU_ ðtÞ
T
ysis, since it is an essential ingredient of ana-
lytical approximations to the time-variant T
¼T~ E S ðtÞST ðtÞ T
~
failure probability for the first-passage reli-
ability problem. (67)
Classically and Nonclassically Damped MDOF Structural Systems 417
" #
h i FðtÞ ¼ AF ðtÞPðtÞ (71)
UðtÞY T ðtÞ UðtÞY_ ðtÞ
T
T
E ZðtÞJ ðtÞ ¼
U_ ðtÞY T ðtÞ U_ ðtÞY_ ðtÞ
T
where the time-modulating functions AF(t) are
h i T
¼T~ E S ðtÞST ðtÞ T ~ frequency independent, and the process P(t) is
a colored noise with PSD function having the
(68) following general expression (i.e., rational
function): C
where the components of the vector process
S = [S1(t)S2(t) . . . S2n(t)]T are defined as X
N
Ak
ð1 F P ð o Þ ¼ S0 (72)
k¼1
ðo ok Þ
Si ðtÞ ¼ j signðoÞASi ðo, tÞejot dZðoÞ,
1
In the sequel, it is assumed that the time-
i ¼ 1, 2, . . . , n
modulating functions have the following general
(69) expression:
Eqs. 67 and 68 show that all quantities required
X
M
for reliability applications can be computed from AF ðtÞ ¼ HðtÞ aq ebq t (73)
the following spectral characteristics of complex- q¼1
valued nonstationary processes (i, m = 1, 2, . . .,
2n) Substituting Eq. 73 into Eq. 61 yields (i = 1, 2,
( , 2n)
E Si ðtÞSm ðtÞ ¼ sSi Sm ðtÞ
(70) ( " #)
E Si ðtÞSm ðtÞ ¼ sSi Sm ðtÞ X M jðoo1q Þt1
bq t e
ASi ðo, tÞ ¼ j aq e
q¼1
o o1q
Response Statistics of MDOF Linear Systems
Subjected to Modulated Gaussian-Colored (74)
Noise
Time-modulated Gaussian-colored noises consti- where o1q = jli and o1q = o1 + jbq. The
tute an important class of nonstationary dynamic spectral characteristics defined in Eq. 701 can
load processes. The expression describing a gen- be computed using Cauchy’s residue theorem as
eral nonstationary loading process reduces to (i, m = 1, 2, . . ., 2n)
X
M X
N nh i o
eðbq þbs Þt þ 1 I 1iq, ms, k ejo1q t I iq
, ms, k ejoms t I iq, ms, k
sSi Sm ðtÞ ¼ S0 a q as Ak 2 3 (75)
q, s¼1 k¼1
, ms, k ¼ 1
Biq ,
X3 r
o ep o
er o eq
er o (79)
I 2iq, ms, k ¼ 2pj Briq, ms, k ejo~r t I ðo
e rÞ (77)
r¼1 r, p, q ¼ 1, 2, 3 r 6¼ p 6¼ q
418 Classically and Nonclassically Damped MDOF Structural Systems
X
M X
N nh i
aq as Ak eðbq þbs Þt ejðoms o1q Þt þ 1 J iq
, ms, k
sSi Sm ðtÞ ¼ jS0 1
q, s¼1 k¼1
o (81)
, ms, k , ms, k
ejo1q t J iq
2 ejoms t J iq
3
ð1
eu
in which E1 ð x Þ ¼ du, jargðxÞj < p (85)
x u
, ms, k ¼ X
3
J iq Biq, ms, k ½logðo
e r Þ þ logðe
o r Þ A very useful modulating function in earth-
1 r
r¼1 quake engineering is the one of Shinozuka and
(82) Sato, defined as
n o
, ms, k ¼ 2X Biq, ms, k eje
3
J iq
2 r
or t
½E 1 ð o
je r t Þ þ jpI ðoe r Þ AF ðtÞ ¼ C eB1 t eB2 t H ðtÞ (86)
r¼1
(83)
where
X
3 n o
J 3iq, ms, k ¼ 2 Briq, ms, k eje
or t
½E1 ðje
o r tÞ þ jpI ðe
o r Þ
B2 B2
B1 B2 B1 log
r¼1
C¼
B1
e (87)
(84) B 2 B1
where E1() denotes the integral exponential and B2 > B1 > 0. With this position, Eq. 81
function: reduces to (i, m = 1, 2, . . ., 2n)
CS0 X 2 X 4 nh i
ð1Þkþsþq eðBq þBs Þt ejðoms o1q Þt þ 1 J iq
, ms, k
sSi Sm ðtÞ ¼ 1
4p q, s¼1 k¼1 (88)
o
ejo1q t J iq, ms, k ejoms t J iq, ms, k
2 3
where o1q = o1 jBq (q = 1, 2). damping are revised and discussed. Attention is
focused on earthquake excitation acting on
nonclassically damped linear structures. Spectral
Summary characteristics are employed to evaluate the time-
variant central frequency and bandwidth param-
Response analysis of MDOF linear structures eters, useful for assessing the reliability of
showing classical or nonclassical viscous earthquake-sensible structures.
Code-Based Design: Seismic Isolation of Buildings 419
deformation-sensitive and acceleration-sensitive and the amount and type of damping present.
elements. Whereas the amplitude and forcing frequency
Performance-based seismic design of struc- may be a function of the equipment’s operating
tures is currently undergoing significant develop- speed and weight, the natural frequency and
ment in response to consequences experienced in damping of the supporting hardware may be
recent earthquakes. Not only has there been sub- adjusted to limit the total acceleration due to
stantial loss of human life as a result of damage harmonic excitation.
caused by major earthquakes, the economic toll The reduction of transmissibility is also the
resulting from direct losses (repair of infrastruc- goal of seismic isolation. Unlike traditional
ture, replacement of damaged contents) and indi- equipment isolation, however, the excitation is
rect losses (business disruptions, relocation due to ground shaking and therefore cannot be
expenses, supply chain interruption) has also characterized by harmonic input. Ground shaking
been significant. As a result of the significant resulting from a seismic event is stochastic in
socioeconomic turmoil due to the occurrence of nature, and this excitation may contain a rich
earthquakes worldwide, innovative systems such array of frequency content. Data collected from
as seismic isolation have emerged as a reliable numerous historical seismic events has demon-
approach to limiting damage and providing post- strated that the predominant frequency of seismic
earthquake functionality and operability. excitation is generally above 1 Hz, and hence
systems having a lower natural frequency than
this will experience a reduction in transmissibil-
Principles of Seismic Isolation ity of acceleration from the ground into the struc-
ture. The reduction of natural frequency
Seismic isolation is rooted in the need to control (or elongation of natural period) provides the
structural response due to harmonic vibrations. so-called “decoupling” of the motion of the struc-
This need stems from (a) the discomfort caused to ture with that of the ground, as shown in Fig. 1.
occupants as a result of oscillatory motion of This reduction in natural frequency has the unde-
floor-supported equipment and (b) the potential sirable consequence of increasing deformation
for damage to sensitive equipment caused by demand under the input excitation. These unde-
vibrations of the supporting structure. The source sirable deformations can be mitigated to some
of such vibrations has traditionally been rotating extent through the addition of damping, typically
machinery, ambient traffic conditions, and through a combination of velocity-dependent
walker-induced floor vibrations. A common (termed viscous damping) and deformation-
method of reducing the accelerations due to har- dependent (termed hysteretic damping) energy
monic excitation is to provide a compliant base dissipation mechanisms. As a result, a properly
(either from steel springs or elastomeric pads) designed isolation system will have the appropri-
that adjusts the natural frequency of the supported ate combination of stiffness and damping such
equipment such that it is unable to reach reso- that substantial dynamic decoupling is achieved
nance under the operating frequency of the exci- without detrimental deformation demands in the
tation. This idea is behind the concept of isolation hardware.
transmissibility, or the ratio of the response Period elongation is not, however, the only
amplitude to the input amplitude. For example, necessary component of seismic isolation.
in vibration isolation, one can define the trans- Indeed, if adding flexibility were sufficient to
missibility as the ratio of the peak total accelera- reduce seismic response, then an effective design
tion of the equipment to the peak input strategy would be to use the most flexible mem-
acceleration of the support. The transmissibility bers possible given the consideration of stability
is generally a function of the forcing frequency, under gravity. An equally important aspect of
the natural frequency of the supporting hardware, isolation is the change in fundamental mode
Code-Based Design: Seismic Isolation of Buildings 421
0.8
ISOLATED
C
0.6 BUILDINGS
ISOLATION
RESPONSE
0.4
0.2
0
0 0.5 1.0 1.5 2.0 2.5 3.0
BUILDING PERIOD (SEC.)
Code-Based Design:
Seismic Isolation of
Buildings,
Fig. 2 Comparison of
fundamental mode shape
for fixed-base and base-
isolated buildings
shape. The introduction of a layer that is compli- given by kb, the isolated natural frequency is
ant relative to the supported structure introduces defined as
a key modification to the free vibration charac- rffiffiffiffiffiffiffi
teristics of the structure. The more similar kb g
the fundamental mode shape is to rigid-body ob ¼ (1)
W
behavior, the less mass participation is present
in higher modes. Such rigid-body behavior is This well-known expression assumes that the
associated with the relative compliance of the mass above the isolation system is rigid, and all
isolation layer. Hence, as the natural period of lateral deformation takes place in the isolators.
the isolation system increases relative to the This is an approximation and is clearly dependent
natural period of the supported structure, the par- on the natural frequency of the supported struc-
ticipation of higher modes becomes closer to zero ture, os. However, for reasonable large frequency
and the seismic deformation is concentrated at ratios os/ob, this rigid structure assumption is a
the isolation layer and not in the superstructure very good approximation of reality.
(Fig. 2). Another important parameter for isolated
If the isolated structure has total weight W, and structures is the critical damping ratio of the
the total lateral stiffness of the isolation system is isolation system, zb. This damping is defined by
422 Code-Based Design: Seismic Isolation of Buildings
the isolation bearings themselves and any supple- distributed to the superstructure using the first
mental damping devices. The general expression mode shape, and these forces may be used to
for equivalent damping of an isolation system is compute member forces, interstory drifts,
given by etc. While isolated buildings are generally
designed to be essentially elastic for design
Ed ground motion, the base shear in Eq. 3 way be
zb ¼ (2)
4pEs reduced by an overstrength-based force reduction
factor RO. This factor is much less than the force
where Ed is the energy dissipated in one complete reduction factors R used for ductile fixed-base
cycle of displacement u0 and Es is the energy buildings, since both ductility and overstrength
stored at a displacement equal to u0. Both param- are considered, and conventional buildings are
eters of stiffness and damping for specific isola- expected to undergo potentially large inelastic
tion devices will be further described in the deformations in the design earthquake.
following section. The maximum displacement of the isolation
system, Db, is computed from the basic expres-
Preliminary Design of Isolated Structures sion for a spring force, V b ¼ kb Db . Noting that
The preliminary design of isolated structures gen- kb ¼ o2n Wg and the relationship between spectral
erally begins with an assumption of equivalent acceleration and displacement is given by
linear properties for the isolation system and a Aðon , zÞ ¼ o2n Dðon , zÞ , the maximum isolator
rigid superstructure. This design approach is suf- displacement is simply
ficient to obtain, for a given level of seismic
hazard, design forces and interstory drift Db ¼ Dðob , zb Þ
demands for the superstructure and peak isolator
displacements. The preliminary design of the individual isolators
Suppose an isolated structure is being will be based on the required shear stiffness and
designed at a site characterized by a design- damping and the stability of the isolator under the
level elastic acceleration response spectrum maximum axial load at the maximum displace-
A(on, z), where on is the natural frequency and ment Db. The final design process is more
z is the viscous damping ratio. The isolation involved and is based on the requirements of the
system stiffness kb is the sum of the individual governing building code.
stiffnesses of each bearing, and the viscous
damping zb of the system is computed from
Eq. 2. If the total seismic weight supported Historical Development of Seismic
above the isolation system is W, the isolated Isolation
natural frequency ob is given by Eq. 1. As a
result, the design base shear for the isolated struc- The roots of seismic isolation as a means for
ture would simply be computed as protecting structures from earthquakes go back
to the latter part of the nineteenth century. While
Að o b , z b Þ there appeared to be significant interest in creat-
Vb ¼ W (3)
g ing an innovative seismic protective system, few
opportunities were available for practical appli-
This equation treats the superstructure as a rigid cation. The first documented proposal for a seis-
mass single degree-of-freedom system oscillator mic isolation system appears to be in an 1870 US
whose stiffness and damping are entirely defined patent filed by Jules Touaillon that describes a
by the isolation system. This base shear repre- double-concave rolling ball bearing (Fenz and
sents the total lateral force delivered to the super- Constantinou 2006). Other early proposals
structure given the frequency and damping of the include a 1906 US patent by J. Bechtold and a
isolation system. This base shear may be 1909 British patent by J. A. Calantarients (Buckle
Code-Based Design: Seismic Isolation of Buildings 423
and Mayes (1990), Kelly (1986)). Both patents building in San Francisco, USA, that was dam-
advocated base isolation through the separation aged by the 1986 Loma Prieta earthquake.
of a building from its foundation, with Since the initial development of reliable, prac-
low-friction material providing the necessary lat- tical isolation devices in the early 1970s, seismic
eral flexibility. Despite the innovative design isolation has seen worldwide adoption. Detailed
approaches embodied in these patents, none design provisions have been incorporated in sev-
appear to have seen actual implementation. eral model building codes, and isolated buildings C
The first application of seismic isolation was have reportedly been constructed in over 30 coun-
in 1885 by John Milne, a Professor of Mining tries with the majority found in Japan, China,
Engineering at the University of Tokyo (Naeim Russia, Italy, the USA, France, Armenia, Taiwan,
and Kelly 1999). Having previously conducted and New Zealand (Martelli et al. 2012). Japan in
experiments on sliding isolators, Milne designed particular has emerged as the worldwide leader in
and constructed a seismic isolation system for a the application and advancement of seismic iso-
building in Tokyo that incorporated ball bearings lation technology. This was brought on in large
and cast iron plates formed in a spherical shape. part by the 1995 Hyogoken-Nanbu earthquake
The first application of seismic isolation using near Kobe, an event that highlighted the signifi-
elastomeric bearings was in the Pestalozzi School cant seismic risk to buildings and infrastructure in
located in Skopje, Yugoslavia, constructed in even the most developed countries (Clark
1969 (Kelly 1986). The elastomeric bearings et al. 2000).
were constructed as blocks of rubber with no
intermediate layers of steel reinforcement. This
lack of confinement caused substantial vertical Seismic Isolation Devices
deflection under both gravity and lateral loading,
resulting in an undesirable combination of hori- A critical ingredient of seismic isolation is the
zontal and rocking modes of vibration under introduction of a specially designed stratum
earthquake excitation. This early application of between the structure and the foundation that is
rubber in seismic design highlights the develop- both horizontally flexible and vertically stiff.
ment that was underway and identifies the moti- A stratum with these properties is generally
vation for further research. achieved through a series of manufactured isola-
Modern seismic isolation originated in the tion bearings, whose required performance is
mid-1960s with the New Zealand Department of specified on a set of contract documents, and
Scientific and Industrial Research (Skinner acceptance is verified through prototype (quality
et al. 1993). One of the outcomes of this research assurance) and production (quality control) test-
was the development of the lead rubber bearing, a ing. Manufactured bearings primarily fall into
system that relies on a laminated rubber bearing two categories: elastomeric based or sliding
for horizontal flexibility and a press-fit central based. Elastomeric-based bearings take advan-
lead core to provide energy dissipation. Lead tage of the flexible properties of rubber to achieve
bearings have seen extensive use worldwide, the isolation, while sliding-based bearings rely on the
first application being the William Clayton inherently low stiffness of a structure resting on
Building constructed in Wellington, New its foundation with no connection other than fric-
Zealand, in 1981. Another type of elastomeric tion at the interface. Devices that are designed
bearing is the high-damping rubber bearing, and manufactured to effectively achieve the goals
which was first used in the Foothill Communities of seismic isolation possess three general
Law and Justice Center in Rancho Cucamonga, characteristics:
USA, in 1985. The first modern sliding isolation
system, the Friction Pendulum System, was 1. High axial stiffness and strength to resist grav-
developed in 1986 (Zayas et al. 1987) and was ity and other vertical loads (both sustained and
used in the retrofit of a wood-framed apartment transient) without excessive deformation that
424 Code-Based Design: Seismic Isolation of Buildings
a 40 b 60
30
40
20
20
10
Force (kN)
Force (kN)
0 0 C
−10
−20
−20
−40
−30
−40 −60
−80 −60 −40 −20 0 20 40 60 80 −150 −100 −50 0 50 100 150
Displacement (mm) Displacement (mm)
Code-Based Design: Seismic Isolation of Buildings, Fig. 4 Hysteresis for a low-damping rubber bearing (a)
g = 75 %; (b) g = 175 % (Constantinou et al. 2007)
Ec Ab
kv ¼ (7) Elastomeric bearings in tension behave much
Tr differently than in compression. Whereas lami-
nated rubber bearings can resist large compres-
where Ec is the compression modulus of the rub-
sive stress with no damage to the rubber matrix,
ber layer. This compression modulus is a function
tension stresses cause cavitation and significant
of the shape factor S, defined as the ratio of the
loss of axial stiffness. For this reason, the design
bonded area to the area free to bulge. For circular
of isolated buildings is based in part of distributed
bearings with bonded diameter Db, the shape
overturning forces at the isolation layer such that
factor is computed as S ¼ Db =4tr . For a rectan-
uplift forces are minimized (Fig. 4).
gular bearing having dimensions a b, the shape
factor is computed as
High-Damping Rubber (HDR) Bearings
A class of elastomeric bearings that has seen wide
ab implementation is high-damping rubber (HDR).
S¼ (8)
2tr ða þ bÞ These bearings are similar in construction to LDR
bearings; however, the natural or synthetic rubber
If the shape factor is sufficiently low, for circular is specially compounded with carbon black, oils,
bearings, Ec ¼ 6GS2, and for square bearings, resins, and other fillers that increase the inherent
Ec ¼ 6:73GS2 (Naeim and Kelly 1999). Typical damping characteristics of the rubber (Clark and
shape factors for modern elastomeric bearings Kelly 1994). HDR bearings are characterized by
range from 10 to 30 and result in extremely high effective shear modulus Geff and equivalent vis-
compressive stiffness. At these high shape fac- cous damping zeq that is a function primarily of
tors, the bulk modulus K of the rubber becomes shear strain but also exhibits dependence on
important. An expression for the compressive strain rate, strain history, and aging (Morgan
modulus incorporating the effect of the bulk mod- et al. 2001). Several highly sophistical models
ulus is given as have been developed that are capable of
predicting the behavior of HDR bearings over a
wide range of response amplitudes (e.g., Yama-
6GS2 K
Ec ¼ (9) moto et al. (2012), among others). Such models
6GS2 þ K rely on a combination of nonlinear elastic
426 Code-Based Design: Seismic Isolation of Buildings
Code-Based Design: Seismic Isolation of Buildings, Fig. 5 Hysteresis of full-scale HDR bearings indicating (a)
substantial strain crystallization and (b) low strain crystallization (Yamamoto et al. (2012))
4
nomenon known as large strain crystallization,
which causes an evident increase in stiffness 0
above 120–150 % shear strain. Hysteresis loops
for full-scale HDR bearings with and without
−4
significant large strain crystallization are shown
in Fig. 5.
−8
Another unique property of HDR bearing is
the strain history dependence of both effective −200 −100 0 100 200
stiffness and equivalent viscous damping. Exper- Shear Strain (%)
imental programs (e.g., Morgan et al. 2001) have
demonstrated the variability of measured stiff- Code-Based Design: Seismic Isolation of Buildings,
ness and damping depending on the nature of Fig. 6 Hysteresis loop of reduced-scale HDR bearing
showing scragging effect
the input loading. For example, the effective
stiffness measured at 100 % shear strain
(a common design parameter) could be markedly
different for a test in which the bearing first
experienced a shear strain of 250 % compared HDR bearings at various frequencies has indi-
to a test that only includes strain amplitudes of cated that the measured stiffness and damping
100 %. Another related effect is the increased of the HDR compound is a function of both
stiffness measured in the first cycle of displace- shear strain amplitude and velocity. As a result,
ment compared to the stiffness in subsequent any suitable model for an HDR bearing
cycles. This effect is known as scragging and must include both hysteretic and viscous behav-
appears to be a persistent feature of HDR com- ior to properly account for the likely energy dis-
pounds. A hysteresis loop of an HDR bearing sipation characteristics over a broad range of
showing the effects of scragging is presented in seismic excitations. Further information on sim-
Fig. 6. ple combined models capable of capturing this
Strain rate is also an important consideration behavior is developed in Naeim and
in the modeling of HDR bearings. Testing of Kelly (1999).
Code-Based Design: Seismic Isolation of Buildings 427
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Lead Rubber (LR) Bearings u
u W
The lead rubber (LR) bearing consists of a T b ¼ 2pu X (11)
tg keff , j
low-damping natural rubber bearing, as described
j
previously, with a press-fit central lead core.
A cutaway section of a typical LR bearing is
where keff, j is the effective stiffness of the jth
shown in Fig. 7. The bearing has linear behavior
isolator. The equivalent viscous damping of a C
identical to the LRB bearing described above,
single isolator is given as
and the stiffness is described by Eq. 5. The lead
plug behaves as an elasto-perfectly plastic hys-
1 Ed
teretic element with a yield shear force zeq ¼ (12)
qd ¼ ty APb where ty is the yield stress of lead 4p Es
and APb is the area of the lead core. Lead is an
where the energy dissipated per cycle
attractive material for the lead core since it is
Es ¼ 4qd u uy and the elastic energy stored
malleable, has a low yield stress that is stable
for high levels of purity, and is readily available. Es ¼ 12keff u2 . Substituting these expressions into
A bilinear force–deformation loop for an LR Eq. 12 gives an expression for zeq that is a func-
bearing is shown in Fig. 8. tion of bearing properties qd, uy, kb, and the dis-
Most isolation devices are characterized by placement u:
effective stiffness, keff, and equivalent viscous
damping zeq. For LR bearings, the effective stiff- 2qd u uy
zeq ¼ (13)
ness assuming the above bilinear hysteretic puðqd þ kb uÞ
model is computed as
The equivalent damping of a system of LR bear-
q ings is computed as
keff ¼ d þ kb (10)
u
X
zeq:j keff , j
The effective period of a system of LR bearings is j
computed as zeq ¼ X (14)
keff , j
j
kb
qd
uy u
Code-Based Design: Seismic Isolation of Buildings, Code-Based Design: Seismic Isolation of Buildings,
Fig. 7 Typical lead rubber bearing with cutaway showing Fig. 8 Bilinear force–displacement relationship for LR
internal construction (courtesy of Takenaka Corp., Japan) bearings
428 Code-Based Design: Seismic Isolation of Buildings
Articulated slider
Code-Based Design: Seismic Isolation of Buildings, Fig. 10 Photo (left) and section (right) of a typical SP bearing
Code-Based Design: Seismic Isolation of Buildings 429
j
m W/R
R W C
u
V
ft
From this series, it is clear that for R u , Most isolation devices are characterized by effec-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
R2 u2 R . Substituting this approximation tive stiffness, keff, and equivalent viscous
430 Code-Based Design: Seismic Isolation of Buildings
damping zeq. For SP bearings it is appropriate to behavior (R and m), a multistage FP bearing has
normalize effective stiffness by the supported greater design flexibility because the pendulum
weight W, or lengths and friction coefficients are specific for
each independent pendulum mechanism. The
V~ m 1 triple-pendulum (TP) bearing introduced by
k~eff ¼ ¼ þ (21) EPS, Inc. consists of four concave surfaces and
u u R
three independent pendulum mechanisms. The
Note that k~eff has units of [1/L], and the actual outer slider consists of concave surfaces on either
effective stiffness is simply keff ¼ k~eff W . Simi- side of a cylindrical inner slider with a
larly, the equivalent viscous damping is given as low-friction interface on both ends. This forms
one pendulum mechanism and defines the prop-
1 ED erties of the isolation system under low levels of
zeq ¼ (22)
4p ES excitation. The outer slider also consists of slid-
ing interfaces between top and bottom outer
where the energy dissipated per cycle sliders and the major spherical surfaces of the
ED ¼ 4mWu and the elastic energy stored bearing. The bottom sliding surface is in contact
ES ¼ 12keff u2 . Substituting these expressions into with a spherical surface of a particular radius of
Eq. 22 gives an expression for zeq that is a func- curvature, forming the second pendulum mecha-
tion of bearing properties R, m and the displace- nism. This mechanism defines the primary prop-
ment u: erties of the isolation system under moderate
levels of excitation. The upper sliding surface is
2Rm in contact with another spherical surface of a
zeq ¼ (23)
pðRm þ uÞ particular radius of curvature, forming the third
pendulum mechanism. The friction coefficient of
Note that the equivalent damping is independent this third sliding interface is sufficiently large to
of the weight supported by the SP bearing. prevent sliding until an extreme level of excita-
Another way to express the equivalent damping tion occurs. For this type of bearing, shown in
is by recognizing that the friction shear force at Figs. 13 and 14, the parameters characterizing the
slip is V f ¼ mW, and the shear force in the bearing cyclic behavior are (R1, m1) for the inner concave
is V b ¼ keff u. It terms of these physically mean- sliding interfaces, (R2, m2) for the bottom outer
ingful forces, the equivalent damping is given as surface, and (R3, m3) for the top outer surface. The
slider heights define the kinematic relations for
2 Vf the positions of each sliding surfaces. The behav-
zeq ¼ (24)
p Vb ior of the triple-pendulum bearing has been
described by Morgan (2007) and Fenz and
For an SP system with friction coefficient of 0.06 Constantinou (2008a, b).
and a peak base shear coefficient of 0.25, the Given the geometry and properties of the TP
equivalent damping can easily computed as bearing, the force–displacement relationship for
the element can be derived by taking equilibrium
2 0:06 in the deformed configuration. The TP bearing
zeq ¼ ¼ 0:15 (25)
p 0:25 cannot, however, be treated simply as a series of
three single-pendulum elements, as this would
Multi-pendulum Bearings ignore the relative rotation of the inner slider
Recent developments in the design and relative to the outer concave surfaces. A complete
manufacturing of FP bearings have centered derivation of the monotonic force–deformation
around the use of multiple pendulum mecha- relation for the TP bearings has been presented
nisms. Whereas the single-concave FP bearing by Morgan (2007) and Fenz and Constantinou
has two key parameters that characterize cyclic (2008a, b). This relation is depicted in Fig. 15,
Code-Based Design: Seismic Isolation of Buildings 431
Code-Based Design:
Seismic Isolation of
Buildings,
Fig. 13 Triple-pendulum
(TP) bearing showing
internal components
(courtesy of EPS, Inc.)
C
Radius = R2
Radius = R3
Friction = m3
Friction = m1
h3
h12
Radius = R1
h11
h2
Friction = m2
Code-Based Design: Seismic Isolation of Buildings, Fig. 14 Exploded cross section of triple-pendulum bearing
showing geometric and friction parameters
showing each of the five stages of unidirectional each sliding surface. This results in a wide variety
sliding. Here the effective pendulum lengths are of realizable hysteretic relations, and proper opti-
defined as Lj ¼ Rj hj where Rj is the radius of mization of the parameters results in a TP bearing
the jth sliding surface and hj is the height of the jth that is “tuned” for a particular seismic hazard and
slider, as shown in Fig. 14. performance objective (Fig. 16).
The main feature of the TP bearing is that
stiffness and damping properties evolve with Energy Dissipation Devices
increasing displacement. Moreover, this evolu- An important consideration in the design of iso-
tion of stiffness and damping can be controlled lated structures is the energy dissipation of the
by varying the radii and friction coefficients of isolation system. While LR, HDR, and FP
432 Code-Based Design: Seismic Isolation of Buildings
~
F
1 1
=
LVeff 2L1
1 1
=
m3 IV
Leff L1 + L3
1 1
III
=
Leff L2 + L3
1 1
=
m2 LIeff
I L1 + L2
1 1
=
m1 LIeff 2L1
Code-Based Design: Seismic Isolation of Buildings, Fig. 15 Monotonic force–deformation relation for triple-
pendulum (TP) bearing (Morgan 2007)
0.0
−0.1
−0.2
−0.3
−0.4
−150 −100 −50 0 50 100 150
Displacement [mm]
Code-Based Design: Seismic Isolation of Buildings, Fig. 17 Fluid viscous damper at isolation plane
–100
–200
–300
α = 0.3
–400
–40 –30 –20 –10 0 10 20 30 40
Displacement
The force output Fd of a viscous damper is damper is the phase lag in force output of
given by the isolation devices and the damping
devices. Since isolation bearings are deformation
Fd ¼ cd ju_ja sgnðu_Þ (26) dependent, the peak force output occurs at peak
displacement, and zero force output occurs at
where cd is the damping coefficient and a is the zero displacement. This is in contrast to viscous
velocity exponent. In the case than a = 1.0, the dampers, in which the peak force output occurs at
damper is classified as a linear viscous damper. zero displacement (or peak velocity), and zero
Otherwise, the damper is classified as nonlinear. force output occurs at peak displacement
Linear dampers are characterized by an elliptical (or zero velocity). For moderate levels of
force–displacement loop for harmonic motion, damping, isolator displacements are reduced
whereas nonlinear dampers have more rectangu- without an increase in total building base shear
lar loops with rounded corners (see Fig. 17). (Fig. 18).
In both cases, the force output of the damper The design of isolation systems is based on the
at maximum displacement (and hence zero veloc- natural frequency of the isolation system, ob, and
ity) is equal to zero. A major advantage of viscous the equivalent viscous damping, zb. For some
434 Code-Based Design: Seismic Isolation of Buildings
Code-Based Design: Seismic Isolation of Buildings, Fig. 19 Metallic yielding dampers for use at isolation plane
harmonic displacement with amplitude u0 and For pure friction dampers, the equivalent
forcing frequency o, the energy dissipation for damping is given by
one cycle, Ed, is given by (Soong and
Constantinou 1994)
2 cd g
zeq ¼ (31)
pu0 o2b
Ed ¼ lðaÞ Cd u1þa
0 o
a
(27)
where l(a) is defined in terms of the gamma For general nonlinear viscous dampers, a conve-
function and is given by nient and highly accurate approximation is given
by 2p=lðaÞ 0:4a þ 1:6 (Morgan and Mahin
G2 ð1 þ a=2Þ 2011). This results in a simplified expression for
lð aÞ ¼ 4ð 2a Þ , the equivalent damping
Gð2 þ aÞ
ð1 (28)
G ðxÞ ¼ ux1 eu du cd ua1 a2
0 ob g
0 zeq (32)
0:4a þ 1:6
It is useful to define a weight-normalized
damping coefficient, cd ¼ cd =mg , where m is Hysteretic Dampers
the total supported mass of the isolation system. Devices that dissipate energy through metallic
From the equation for equivalent viscous yielding or friction are classified as hysteretic
damping, the equivalent damping zeq. for an dampers. These dampers are characterized by an
isolation system with viscous dampers is given by elastic stiffness ke and a yield force fy and are
generally considered to behave elasto-perfectly
lð aÞ plastic (much like the lead plug in the LR bear-
zeq ¼ cd ua1 a2
0 ob g (29)
2p ing). Two hysteretic dampers that have been
implemented in Japan are shown in Fig. 19.
Two special cases of viscous dampers are linear Unlike viscous dampers, hysteretic dampers
viscous (a = 1.0) and pure friction (a = 0). are displacement dependent, and the force
For linear viscous dampers, the equivalent output is in phase with the forces in the isolation
damping is given by the simple and well-known system. As a result, the addition of hysteretic
expression dampers, while reducing the peak displacement
of the isolators, may introduce substantially
cd g higher forces in the superstructure and
zeq ¼ (30)
2ob foundations.
Code-Based Design: Seismic Isolation of Buildings 435
The presence of hysteretic dampers increases structures designed to those provisions should
both the effective stiffness and equivalent be able to:
damping of the isolation system. Suppose an iso-
lation system without dampers has effective stiff- 1. Resist minor and moderate levels of earth-
ness keff and equivalent damping zeq. quake ground motion without damage to
Furthermore, a group of hysteretic dampers are structural elements, nonstructural compo-
to be added whose aggregate elastic stiffness and nents, or building contents C
yield force are given by ke and fy, respectively. 2. Resist major levels of earthquake ground
For these dampers subjected to a lateral displace- motion without failure of the isolation system,
ment u, the energy dissipated per cycle without significant damage to structural ele-
E0d ¼ 4fy u uy , where uy ¼ fy =ke . The effec- ments and nonstructural components, and
tive stiffness of the dampers is k0eff ¼ fy =u. From without major disruption of the function of
Eq. 2, the added equivalent damping is the facility
Design Provisions for Isolated Buildings 1. Maintain stability when subjected to design
displacements
Throughout the world, there are a number of 2. Provide increasing lateral resistance with
published design provisions devoted to the design increasing displacement
of seismic isolated structures. The principal goal 3. Suffer no strength or stiffness degradation
of these provisions is to ensure that isolated struc- under repeated cyclic loading
tures maintain an acceptable level of safety to the 4. Have quantifiable force–deflection and
public in the event of an earthquake. The discus- damping characteristics
sion herein focuses on ASCE 7–10, which con-
tains provisions governing design of seismic Analysis Methods
isolated buildings in the USA. The analysis of isolated buildings is generally
The commentary to the 2003 National Earth- conducted by one of three possible methods:
quake Hazards Reduction Program (NEHRP) equivalent lateral force (static) analysis, response
Provisions (BSSC 2003) states that isolated spectrum analysis, and response history analysis.
436 Code-Based Design: Seismic Isolation of Buildings
Each analysis method is capable of representing performance objectives and the resulting need to
the peak internal forces and deformations in both maintain essentially elastic behavior in the struc-
the superstructure and isolation system given a ture. This base shear is also subjected to several
design level of seismic intensity and may be used lower-bound limits to prevent yielding of the
to select and proportion both structural elements building prior to activation of the isolation sys-
and isolation devices. tem and excessive movement under the design
wind loads.
Equivalent Lateral Force Procedure Once the base shear Vs has been computed, the
The simplest analysis procedure is the equivalent lateral forces on the building are distributed ver-
lateral force (ELF) procedure, which is analogous tically using the equation
to the static design procedures for conventional
structures. In this procedure, the 5 % damped w x hx
Fx ¼ V s X n (39)
design spectral acceleration for a site at a period wh
i¼1 i i
of 1 s, SD1, represents the design input.
where Fx is the lateral force at story x, wi is the
gSD1 T D
DD ¼ (36) story weight at the ith floor, and hi is the height
4p2 BD above the grade to the ith floor.
The isolation system must be designed such
where TD is the DE effective period, g is the
that the elements of this system remain stable
gravitational constant, and BD is a factor to
given an MCE event. Given a 5 % damped
account for the equivalent damping of the isola-
MCE spectral acceleration for a site at a period
tion system. While BD is based on tabulated
of 1 s of SM1, the displacement of the isolators is
value, this can be approximated by the equation
computed as
BD ¼ 2:4z0:3
eq where zeq is the equivalent damping
computed for the isolation system. The DE effec- gSM1 T M
tive period is computed as DM ¼ (40)
4p2 BM
sffiffiffiffiffiffiffiffiffiffiffiffiffi
W where TM is the MCE effective period and BM is
T D ¼ 2p (37) similar to BD in Eq. 36, computed at the displace-
kDmin g
ment DM. The MCE effective period is computed
as
where W is the total supported weight of the
building and kD min is the lower-bound effective sffiffiffiffiffiffiffiffiffiffiffiffiffi
W
stiffness of the isolation system at a displacement T M ¼ 2p (41)
DD. The base shear of the superstructure is com- kMmin g
puted as
where kD min is the lower-bound effective stiff-
kDmax DD ness of the isolation system at a displacement DM.
Vs ¼ (38) The isolation system must be designed for this
RI
displacement DM plus the additional effects of
where kD min is the upper-bound effective stiff- both actual and accidental torsion.
ness of the isolation system at a displacement DD The ELF procedure is permitted for analysis of
and RI is the response modification factor for the isolated buildings provided the following restric-
superstructure and is based on 3/8 times the value tions are met:
of R for a conventional structure with the same
lateral system (not to exceed RI = 2.0). This RI 1. Building is vertically regular.
factor is substantially smaller than the R factor for 2. MCE spectral acceleration SM1 < 0.6 g.
conventional structures because of the enhanced 3. T M 3:0 sec .
Code-Based Design: Seismic Isolation of Buildings 437
Code-Based Design: Seismic Isolation of Buildings, Fig. 20 Reproduction of Table C13.2–2 in commentary to
ASCE 7–10 (ASCE 2010)
Code-Based Design:
Seismic Isolation of
Buildings,
Fig. 21 Prototype test of a
triple-pendulum
(TP) bearing, conducted by
EPS, Inc.
Acceptance criteria for these tests are devel- projects. These tests are meant to ensure that
oped by the engineer of records and are generally fabricated isolators meet the established range
focused on meeting a specified range of stiffness of stiffness and damping properties prior to
and damping values for the DE tests and installation. A common production test pro-
maintaining stability in the MCE tests. Figure 21 gram requires three fully reversed cycles at
shows a triple-pendulum (TP) bearing during displacement amplitude DD, considering the
prototype testing. average gravity loads acting on the isolator
Production testing, while not explicitly pre- type. These tests are often conducted on each
scribed by ASCE 7–10, is specified on most isolator prior to installation but could be
Code-Based Design: Seismic Isolation of Buildings 439
limited to some sample subset at the discre- Chopra AK (2007) Dynamics of structures: theory and
tion of the engineer and owner. application to earthquake engineering, 3rd edn. Pear-
son Prentice Hall, Englewood Cliffs
Clark PW, Kelly JM (1994) Mechanical properties of
Design Review Requirements BTR/Andre high-damping rubber seismic isolation
ASCE 7–05 mandates a design review of the bearings. EERC lab report 94–701, Earthquake Engi-
isolation system and related testing program by neering Research Center, University of California,
Berkeley
an independent panel composed of qualified Clark PW, Aiken ID, Nakashima M, Miyazaki M,
C
individuals with experience in earthquake engi- Midorikawa M (2000) The 1995 Kobe (Hyogo-ken
neering principles and analysis and design of Nanbu) earthquake as a trigger for implementing new
seismic isolation systems. This process typi- seismic design technologies in Japan. In: Lessons
learned over time, vol 3, Learning from earthquake
cally includes a review of the following: the series. Earthquake Engineering Research Institute,
seismic hazard at the site and any ground Oakland, pp 79–109
motions that may be developed, the prelimi- Constantinou MC, Whittaker AS, Kalpakidis Y, Fenz DM,
nary design of the isolation system, the obser- Warn GP (2007) Performance of seismic isolation
hardware under service and seismic loading.
vation of the prototype and production testing MCEER-07-0012, Multidisciplinary Center for Earth-
programs, and the final analysis and design of quake Engineering Research, State University of New
the entire structural system. The design review York, Buffalo
is often an interactive process where comments Fenz DM, Constantinou MC (2006) Behavior of double
concave friction pendulum bearing. Earthq Eng Struct
and suggestions are provided to the design Dyn 35(11):1403–1424
team, who can then incorporate these in early Fenz DM, Constantinou MC (2008a) Mechanical behavior
schematic phases. This continuous interaction of multi-spherical sliding bearings. MCEER-08-0007,
largely prevents significant divergence in the Multidisciplinary Center for Earthquake Engineering
Research, State University of New York, Buffalo
design approach adopted by the design team Fenz DM, Constantinou MC (2008b) Spherical sliding
and the preferred approach advocated by the isolation bearings with adaptive behavior: experimen-
review panel. A healthy design review process tal verification. Earthq Eng Struct Dyn 37(2):185–205
is generally one that promotes a combination Kelly JM (1986) Aseismic base isolation: review and
bibliography. Soil Dyn Earthq Eng 5(4):202–216
of sound engineering judgment and technically Kelly JM (1993) Earthquake-resistant design with rubber.
accurate analysis methods. As seismic isolated Springer, London
buildings are expected to achieve enhanced Martelli A, Forni M, Clemente P (2012) Recent world-
performance in a design earthquake, the review wide application of seismic isolation and energy dissi-
pation and conditions for their correct use. In:
panel’s principal role is to give the owner their Proceedings of 15th world conference on earthquake
independent assessment of the expected perfor- engineering, Lisbon
mance of the building and whether it meets Morgan TA (2007) The use of innovative base isolation
the building code’s intended performance systems to achieve complex seismic performance
objectives. PhD dissertation, Department of Civil and
objectives. Environmental Engineering, University of California,
Berkeley
Morgan TA, Mahin SA (2011) The use of innovative base
isolation systems to achieve complex seismic perfor-
References mance objectives. PEER 2011/06, Pacific Earthquake
Engineering Research Center, University of Califor-
American Society of Civil Engineers (ASCE) (2010) Min- nia, Berkeley
imum design loads for buildings and other structures. Morgan TA, Whittaker AS, Thompson A (2001) Cyclic
ASCE/SEI 7–10, ASCE, Reston behavior of high-damping rubber bearings. In: Pro-
BSSC (2003) FEMA 440: NEHRP recommended provi- ceedings of 5th world congress on joints, bearings
sions for seismic regulations for new buildings and and seismic systems for concrete structures. American
other structures. Building Seismic Safety Council, Concrete Institute, Rome
Washington, DC Mosqueda G, Whittaker AS, Fenves GL (2004) Character-
Buckle IG, Mayes RL (1990) Seismic isolation: history, ization and modeling of friction pendulum bearings
application and performance – a world view. Earth- subjected to multiple components of excitation.
quake Spectra 6(2):161–201 J Struct Eng ASCE 130(3):433–442
440 Community Recovery Following Earthquake Disasters
Naeim F, Kelly JM (1999) Design of seismic isolated struc- importance of not only strong local government
tures: from theory to practice. John Wiley, New York capacity but also a cohesive system of public,
Skinner RI, Robinson WH, McVerry GH (1993) An intro-
duction to seismic isolation. Wiley, Chichester private, and volunteer groups integrated into the
Soong TT, Constantinou MC (1994) Passive and active community (Olshansky et al. 2012). Research
structural vibration control in civil engineering. consistently shows that community recovery
Springer-Verlag, Wien from disasters can be greatly enhanced by ensur-
Yamamoto M, Minewaki S, Yoneda H, Higashino
M (2012) Nonlinear behavior of high-damping rubber ing that the preexisting social environment sup-
bearings under horizontal bidirectional loading: full- ports the recovery process.
scale tests and analytical modeling. Earthq Eng Struct
Dyn 41(13):1845–1860
Zayas VA, Low SS, Mahin SA (1987) The FPS earthquake
resisting system: experimental report. UCB/EERC-87/ Resilience in a Recovery Context
01, Earthquake Engineering Research Center, Univer-
sity of California, Berkeley In a recovery context, a resilient community is
Zayas VA, Low SS, Mahin SA (1990) A simple pendulum one that can cope with, adapt to, and recover from
technique for achieving seismic isolation. Earthquake
Spectra 6:317–333 the loss and disruption encountered through the
experience of a disaster. How effectively this is
done is a function of how well people, communi-
ties, and societies can work together and use their
Community Recovery Following resources to deal with the problems encountered.
Earthquake Disasters This entry discusses recovery in relation to com-
munity and societal recovery. In the entry we
Douglas Paton1, Sarb Johal2 and David Johnston2,3 draw largely on recent experiences following
1
School of Psychology, University of Tasmania, the 2011 Christchurch, New Zealand, earthquake
Launceston, TAS, Australia from the perspective of people’s accounts of what
2
Joint Centre for Disaster Research, GNS they had to contend with and how they responded
Science/Massey University, Wellington, to the disruption and challenges encountered dur-
New Zealand ing the recovery. We also draw on experiences
3
Risk and Society, GNS Science, Lower Hutt, gained from research in other cultures to illustrate
New Zealand the similarities and differences that culture intro-
duces into how earthquake recovery can be
understood and managed. The first issue
Synonyms addressed concerns what people had to contend
with and how these demands changed over the
Community leadership; Community processes; course of recovery.
Community recovery; Resilience
The effectiveness and efficiency of community The Christchurch recovery experience illustrates
recovery following a disaster is shaped not just by several issues that need to be considered when
nature and scale of the physical impacts of the conceptualizing community recovery following
event but also on how the social environment earthquakes. The first is that recovery is
supports the complex and protracted processes a process rather than any specific activity and
of recovery (Olshansky et al. 2006; Paton one that extends over time as people deal with
et al. 2014). Recent research following demands and challenges that differ depending on
a number of disasters has highlighted the com- the stage of response and recovery they are
plex nature of community recovery and the at. This is depicted in Fig. 1 as the time arrow
Community Recovery Following Earthquake Disasters 441
Community Recovery Following Earthquake Disasters, Fig. 1 A summary of how adaptive capacities at person,
family, household, and societal levels interact to influence earthquake recovery (Adapted from Paton et al. 2014)
illustrates how the relative salience of recovery to employment and livelihood. As the event moves
issues and demands changes with time (Paton into the recovery phase, people can increasingly
et al. 2014). For example, during the immediate find themselves dealing with demands from deal-
impact phase (e.g., 3–14 days depending on ing with, for example, government agencies and
nature and intensity of impact), the destruction insurance companies. The relative salience of
of infrastructure and loss of essential services these “adaptive demands” (i.e., what people have
means that the demands people must cope with to adapt to and cope with) changes over time
and adapt to relate people’s ability (“self- (Fig. 1). A second issue highlighted by the Christ-
reliance”) to meeting survival needs on their church experience was the fact that people may
own or with their neighbors. Over time, as essen- have transition through response and recovery
tial services are progressively restored but while cycles several times as they deal with the conse-
normal societal and economic activities remain quences of multiple aftershocks. This is illustrated
affected, the challenges and issues people have to in the aftershock feedback loop in Fig. 1. A third
contend with change. As people move into the issue arises from the fact that differences were
response phase, they find themselves working evident in how well people and communities
with neighbors and community members to deal dealt with these adaptive demands (Olshansky
with local issues (e.g., effecting basic repairs, et al. 2006; Paton et al. 2014). This introduces
looking after more vulnerable members of their a need to identify the sources of such variability
community, etc.) and dealing with issues relating in “adaptive capacity.”
442 Community Recovery Following Earthquake Disasters
Mitigation and Structural Preparedness include being able to develop self-help groups
with neighbors and others to deal with local prob-
While recovery takes place in the aftermath of the lems (e.g., removing rubble, setting up commu-
hazard event, how well people, communities, and nity meeting places, etc.). Subsequently,
societies recover is a function of what they have reestablishing community groups or forming
to mitigate their risk and how well they have new ones (e.g., organizing local efforts to repair
prepared to deal with those aspects of risk that homes, identifying and meeting local needs)
cannot be readily anticipated (e.g., the degree of affects the quality of recovery (Paton
localized losses). Mitigation activities such as et al. 2014). Differences in how well these emer-
land use planning and building codes and the gent demands are dealt with are influenced by
steps residents have taken to secure the structural family preparedness beliefs and factors such as
integrity of their homes (Fig. 1) have an impor- levels of community competence and community
tant bearing on the speed of earthquake recovery. leadership (Fig. 1). For example, confidence in
With regard to community recovery, the impor- self-help capability and the ability to engage with
tance of these pre-event activities reflects the fact others under novel circumstances to confront
that having a home that survives the initial impact issues is important (Paton et al. 2014).
reduces the risk of injury and death to its inhab- Developing a group identity, by increasing peo-
itants. It also increases the likelihood of people ple’s commitment to the recovery of the commu-
having a habitable building in which to base their nity and its members, which makes an additional
recovery activities and increases people’s avail- and important contribution to recovery
ability to offer assistance to neighbors and others (Winstanely et al. 2011), and willingness to vol-
and to contribute to economic recovery. How- unteer all contribute to explaining differences in
ever, people differ in the degree to which they levels of community earthquake recovery.
take steps to safeguard their homes. This provides Key characteristics of community members
insights into the predictors of preparedness that (Mamula-Seadon et al. 2012; Norris et al. 2008;
can be defined as adaptive capacities (Paton and Paton and Johnston 2006) include willingness to
Johnston 2006). For example, the level of peo- be proactive despite the novelty in their circum-
ple’s earthquake knowledge and family beliefs stances, taking steps to include as many residents
about the importance of structural preparedness as possible, and empowering, encouraging, and
offers insights in differences in community supporting community initiatives. A need for
recovery capacity (Fig. 1). Thus, prevailing new, community-specific ways of communica-
levels of structural and survival readiness provide tion that address specific and diverse community
the foundation for effective recovery when peo- needs and that are available when people need it
ple face the emergent issues that arise. was also identified as playing a significant role in
expediting community recovery (Winstanely
et al. 2011). A common characteristic of commu-
Personal and Community Adaptive nity groups is that the key roles within them are
Capacity filled by volunteers.
NGOs, and businesses rather than waiting for an example of a strategy that contributes to cre-
them to come to the community. Facilitation ating empowered communities and providing
was increased also by agencies working through a platform for social development.
community leaders. However, when it comes to Development will be further facilitated by
the ability and effectiveness of a proactive developing a comprehensive Canterbury Recov-
community-led approach to dealing with govern- ery Strategy. Long-term recovery is facilitated
ment agencies, the extent to which this relation- by planning that covers, for example, built heri-
ship marginalizes or facilitates community tage, education, Maori, and community develop-
recovery is affected by things such as the gover- ment in ways that build social and cultural
nance mechanisms setup as part of the formal capital.
recovery response (Bach 2012; Collins The Christchurch experience also highlights
et al. 2011). Bureaucracies often do not adapt caution in making recovery decisions when
well to this circumstance, and new governmental a community is under stress. There are challenges
organizations typically emerge to provide more in getting communities to participate in complex
resources. It is important to appreciate that the decision-making in times of stress immediately
quality of earthquake recovery is a reciprocal after a disaster event, as well as delaying such
process. That is, it will only arise when commu- decisions during the recovery period. Ensuring
nities and government agencies, NGOs, and busi- that communities are participating in decision-
nesses play complementary roles in facilitating making processes prior to an event can help alle-
comprehensive recovery. Thus, while societal- viate some of these problems. The importance of
level agencies play key roles in passing appropri- the latter is reinforced by the inclusion of a cross-
ate legislation and providing resources, these cultural perspective on earthquake recovery.
activities must be developed in ways that Because earthquakes occur in many different
empower communities. This is depicted in countries, developing a comprehensive under-
Fig. 1 by the two-way arrow linked to empower- standing of earthquake recovery makes it perti-
ment and trust. nent to consider the similarities and differences
that the associated social and cultural diversity
introduces into the recovery process. While this is
Earthquake Recovery and Community an area requiring more research, it is possible to
Development advance this understanding by comparing coun-
tries that differ in their respective cultural char-
Recovery is also about facilitating development. acteristics. For example, in countries with
The concept of “Building Back Better” applies to a relatively individualist cultural orientation
the social and the physical environment. The (e.g., New Zealand), people act consistently
physical issues requiring attention include land across situations in accordance with a self-
use planning, building codes, and architecture. concept that is relatively independent of social
Social rebuilding requires the inclusion of other situation and in which achieving personal goals is
strategies. For example, in Christchurch, a prominent objective. In contrast, in collectivist
a Psychosocial Recovery Advisory Group countries like Taiwan, actions are underpinned
(PRAG) was developed to provide expert, timely, by culturally embedded beliefs that are reflected
quality evidence-based advice on psychosocial in shared purpose and activity involving align-
issues (Mooney et al. 2011). The PRAG adopted ment with social norms, achieving collective
a strengths-based approach for conceptualizing goals, and engaging in activities related to future
recovery and facilitating longer-term social goals that emphasize social relations (Paton and
development. The strengths-based approach Jang 2014). Comparing these countries can pro-
mobilizes cultural, spiritual, and social resources vide a foundation for development of
with an emphasis on development through com- a comprehensive understanding of earthquake
munity participation and engagement. It is thus recovery.
Community Recovery Following Earthquake Disasters 445
Based on some preliminary research compar- accepting that the earthquake and its conse-
ing New Zealand (populations affected by the quences had occurred, and the importance of
2011 Christchurch earthquake) and Taiwan moving on in ways that emphasized adopting
(those affected by the 1999 Chi Chi earthquake) a problem-focused coping approach to their adap-
(Paton and Jang 2014), people in Christchurch tive recovery. Spiritual beliefs were also impli-
and Taiwan (Tung Shia) endorsed the fact that cated. Spiritual beliefs helped people cope with
earthquake recovery was influenced by survival, challenges, be more optimistic, to focus on the C
household planning, neighborhood self-help, positive aspects of their experience (e.g., how
psychological, and community-agency factors. everyone pulled together to promote the recovery
Irrespective of culture or location, people face of the whole community), accept their losses, and
the same challenges. However, both similarities adopt a problem-focused approach to their prob-
and differences were evident in how people lem solving. Spirituality influenced residents’
responded and recovered. acceptance of earthquake consequences,
The work in Taiwan involved an ethnic minor- enhanced their sense of community, and strength-
ity group, the Hakka that account for some 20 % ened their social support networks.
of the Taiwanese population. They were the The comparative New Zealand-Taiwan ana-
group most significantly affected by the Chi Chi lyses illustrate how everyday or mainstream
earthquake. Similarities between the social pro- community and cultural competencies and char-
cesses that influenced earthquake recovery were acteristics (e.g., the community participation and
evident in both Taiwan and New Zealand. Both collective efficacy variables reflect the outcomes
populations endorsed the need for self-reliance, of people’s accumulated experience of commu-
active community involvement, leadership, and nity life) influence on people’s ability to cope
communities being proactive in managing their with, adapt to, and recover from earthquakes.
own recovery activities. The fact that government However, the discussion also highlighted a need
agencies could help or hinder community recov- to accommodate cultural differences in the devel-
ery was also identified as an important determi- opment and delivery of earthquake recovery
nant of community recovery following a interventions.
significant seismic event. However, some cul-
tural differences in how this was achieved were
evident when the two countries were compared. Summary
In Taiwan, the nature of the social recovery
was strongly influenced by the Hakka spirit, the In conclusion, the quality of people’s earth-
essence of Hakka culture, which comprises char- quake recovery is a function of their ability to
acteristics such as frugality, diligence, self- deal with survival and planning issues (e.g.,
reliance, shared responsibility, and persistence having water and being able to develop house-
(Paton and Jang 2014). Residents highlighted hold and neighborhood plans to deal with the
the key role that Hakka spirit played in their immediate impact of an earthquake), being able
recovery following the Chi Chi earthquake. For to confront psychological and livelihood/eco-
example, it was identified as generating nomic disruption, and being able to work with
a sustained sense of purpose, self-reliance, and government agencies and NGOs. How well
perseverance (which creates a kind of culture- people and communities can deal with these
specific self-efficacy) throughout the several demands and recover effectively is a function
years of their recovery and facilitated people’s of the degree to which several personal and
efforts to rebuild both their homes and their community competencies are present. These
communities. include self-efficacy, place attachment, active
Recovery was also facilitated by the influence community participation, collective efficacy,
of Hakka beliefs regarding the importance of empowerment, trust, community leadership,
accepting disasters as part of life experience, inclusivity, and sense of community.
446 Complex Fractional Moments and Their Use in Earthquake Engineering
• In the Fourier and in Laplace domains, they with power law kernel. The fractional integral is
exactly behave like the derivatives and inte- the generalization of Eq. 1 to an arbitrary order b
grals of integer order. in the form
ðt
For all these reasons, it will be more appropri- 1
I baþ f ðtÞ ¼ ðt tÞb1 f ðtÞdt; b>0
ate to call such operators as generalized deriva- GðbÞ a
tives and integrals, but for historical reasons all the (2) C
people continue to call them as fractional ones.
In the last half century, fractional calculus where G() is the Euler-Gamma function that
becomes more and more popular also among the interpolates the factorials (G(n 1) = n!).
communities of physicists and engineers. This is
I baþ f ðtÞ is referred as Riemann-Liouville (RL)
due to its versatility to describe dynamic proper-
ties of the systems in many fields such as electro- fractional integral. The suffix a+ stands for the
chemistry, thermal engineering, electricity, initial time at which the operator applies and the
apex + means that t in Eq. 2 is greater than a. For
stochastic dynamics, and mechanics and espe-
cially to describe viscoelasticity. this reason I baþ f ðtÞ is called left RL fractional
In this entry some relevant examples on the
integral. The value a may be
also –1
and in this
use of fractional calculus to earthquake ground
case in literature is written I b1þ f ðtÞ or simply
motion modeled as a stationary normal colored
noise are presented. Applications of fractional I bþ f ðtÞ . The right RL fractional integral still
calculus for the description of mono- and multi- exists and the integration is performed in t
a,
variate earthquake accelerations and exact filter a > t. The right RL is usually written as
b
equation are obtained in an integral form involv- I f ðtÞ , and if a ! 1 then is written as
ing Riesz fractional operator in zero. The latter ab
I 1 f ðtÞ or simply I b f ðtÞ.
are related to complex spectral moments by As the derivative of order n may be viewed as
Mellin transform operator. Other relevant appli- the derivative of order m + n of the m th prim-
cation in probability may be found in Cottone itive, the Riemann-Liouville fractional derivative
et al. (2010) and Di Paola and Pinnola (2012). is defined as
ðt
1 dn
Some Basic Elements of Fractional Dbaþ f ðtÞ ¼ ðt tÞnb1 f ðtÞdt;
Gðn bÞ dtn
Calculus a
b>0
Hereinafter some few preliminary definitions of (3)
fractional calculus are introduced for clarity
sake’s as well as for introducing appropriate sym- where n ¼ ½b þ 1 ℕ and [b] means integer part
bologies. Let us start with the definition of the of b. The definition of left and right RL fractional
Cauchy formula for multiple integrals derivative is like for the RL fractional integrals,
a ðtÞ if t < a,or if a ! 1,
b
then we denote
D f
ðt ð t1 ð tn1 we denote Db1 f ðtÞ or simply Db f ðtÞ and
I naþ f ðtÞ ¼ dt1 ... f ðtÞdt similarly for the right RL derivative. Other nota-
a a a
ðt tions are used by different authors, and here the
1
¼ ðt tÞn1 f ðtÞdt; nℕ notation of Samko et al. (1993) has been adopted.
ðn 1Þ! a RL fractional integrals and derivatives remain
(1) meaningful also for complex order g ¼ r þ i.
In the following the fractional operators are
Equation 1 shows that the n – th primitive of always defined for complex order. All the rules of
f(t) may be obtained as a convolution integral classical derivatives and integrals of integer order
448 Complex Fractional Moments and Their Use in Earthquake Engineering
q refer to the asymptotic behavior of f(t) at t ! 0 overlapped by summation (12) including the
and t ! 1, respectively: trend at t = 1.
lim xðtÞ ¼ Oðtp Þ; lim xðtÞ ¼ Oðtq Þ (11) For more informations on fractional calculus,
t!0 t!1
integration of fractional differential equation,
where OðÞ means order of the term in parenthe- readers are referred to the books of Miller and
sis. Such an example if f(t) = 1/(1 + 1), Ross (1993), Samko et al. (1993), Gorenflo and C
limt!0 f(t) = 1, then p = 0, limt!0 f (t) = t1 Mainardi (1997), and Podlubny (1999).
then q = 1. It follows that f (t) = 1/(1 + t) is
Mellin transformable as the real part of g is
Power Spectral Density, Correlations,
selected in the range 0 < r < 1. The
and Spectral Moments
strip –p
q is called fundamental strip
(FS) of the Mellin transform. In discretized
The simplest model of earthquake acceleration
form Eq. 10 may be also rewritten as
Z€g ðtÞ is that it is represented as a stochastic
stationary zero mean colored normal process. In
D X m
f ðtÞ’ Fm ðgk Þtgk this case the power spectral density (PSD) func-
2p k¼m
tion, or its inverse Fourier transform, namely, the
1 r X
m h
p
p i
¼ t Fm ðgk Þ cos k x isin k x autocorrelation function, fully describes the pro-
2b k¼m b b cess Z€g ðtÞ in probabilistic setting. In the follow-
(12) ing, for simplicity of notation, Z€g ðtÞ will be
indicated as Y(t). More realistic models may be
where gk = r + ikD, b = p/D, and x = ln t. found in literature giving a variety of
The number m of the summation in Eq. 12 gives a nonstationary models. Among them the
cutoff value along the imaginary axis and is so-called uniformly modulated (Shinozuka and
selected in such a way that terms of order m + n Sato 1967) have been proposed. These processes
do not produce sensible variations on f(t). It is to are constructed as the product of a stationary
be stressed that Eq. 12 is independent of the value process, with assigned PSD, and a deterministic
of r selected, provided it belongs to the FS of the modulating function. Modulation in frequency
Mellin transform. This is due to the fact that the and time has been proposed by Solomos and
Mellin transform Fm(g) is holomorph into the FS. Spanos (1984), Shinozuka and Deodatis (1988),
From Eq. 12 some considerations may be and Conte and Peng (1997).
withdrawn: For simplicity’s sake, in this entry, we confine
ourselves to the stationary model. In order to
• Inverse Mellin transform returns f(t)tr in the define the stationary seismic motion, two strate-
domain x and looks as a Fourier series, and gies are commonly assumed:
Fm(gk)/2b are the (complex) coefficients of the
series. • The seismic ground acceleration is thought as
b
e p
e , the trigonometric func-
b
• In the range a filtered normal white noise.
tions exp ik b lnt are orthogonal, and as • The PSD is modeled in such a way that the
m ! 1 they form a basis in such an interval. corresponding response spectrum of
It follows that the orthogonality condition in x accelerograms generated by the PSD matches
domain is in b
b. the target one.
• The factor tr in Eq. 12 ensures us that f (t)
tends to zero as t ! 1, while in zero Eq. 12, In the former case, the equation of filter, usu-
unless f(0) = 0, gives an infinite value due to ally a system of differential equation of integer
the factor tr. Apart from this pathological order, is already known, but the selection of the
behavior in t = 0, all the function f(t) is various parameters to model the effective
450 Complex Fractional Moments and Their Use in Earthquake Engineering
behavior of the ground deposit requires a level of neither the PSD nor the correlation of the process
knowledge that is seldom available. Y(t) itself because they may be divergent quanti-
On the other hand, the PSD consistent with the ties starting from an assigned order. As an exam-
target response spectrum gives a shape of PSD ple for the classical Tajimi-Kanai filter, the
(Falsone and Neri 1999; Cacciola et al. 2004) in spectral moments diverge for j 4.
frequency domain, whose parameters of the filter An extension of the definition given in Eq. 14
equation may not be easily obtained by using is obtained in the form lg1
Y ðg ℂÞ in the follow-
classical differential equations. In this regard dis- ing denoted as Lg(g 1) defined as
crete autoregressive (AR) or moving-average
ð1
(MA) or their combination ARMA models have
L Y ð g 1Þ ¼ GY ðoÞog1 do g ¼ r þ i
been proposed in literature (Spanos 1983, 1986) 0
to generate sample functions with assigned PSD (15)
or correlation function.
In section “Filter Equation,” the issue of the that is a complex function representing the Mellin
definition of the filter equations for any shape of transform of the one-sided PSD GY(o). The func-
PSD will be addressed. tion Lg(g 1) will be called fractional spectral
moment (FSM) function. On the other hand, the
Fractional Spectral Moments inverse Mellin transform returns GY(o) and SY(o)
In the analysis of stochastic processes, an impor- in the form
tant role is played by the so-called spectral
moments (SM), introduced by Vanmarcke ð rþi1
1
(1972), important for the definition of some char- GY ðoÞ ¼ LY ðg 1Þog dg
2pi ri1
acteristics of stochastic processes and in reliabil-
1 X m
ity analysis (Michaelov et al. (1999)). These ’ LY ðgk 1Þogk ; o
quantities are defined as the moments of the 2b k¼m
one-sided PSD with respect to the frequency ori-
>0 (16a)
gin. Let SY (o) be the PSD of the earthquake
acceleration with SY(o) = SY(o). Let GY(o)
1 X m
be the one-sided PSD defined SY ð o Þ ’ LY ðgk 1Þjojgk (16b)
4b k¼m
Now we define the correlation function of the generalized Taylor series. The difference with
ground acceleration as RY ðtÞ ¼ E½Y ðtÞY ðt þ tÞ, the classical Taylor series is that in the latter
where E[] denotes ensemble average. By making series the coefficients are the derivatives of
the Fourier transform of the Riesz integral, the given function in zero and in Eqs. 20a
according to Eq. 8a, we obtain and 20b are the fractional derivative in zero.
Since in the FS the various FSMs never
F I 1g RY ðtÞ, o ¼ jojg1 SY ðoÞ; 0<r<1 diverge and |o|gk remains finite (provided r C
(17) belongs to the FS), the whole PSD is restituted
by Eq. 16b or 20b.
It follows that (11g RY)(t) is the inverse Fou-
rier transform of |o|g1SY(o), that is, Fractional Spectral Moments of the
ð1 Correlation function
1
I 1g RY ðtÞ ¼ eiot jojg1 SY ðoÞdo It is to be emphasized that by using integer
2p 1 order derivatives of the correlation function,
(18) we get
and then the Mellin transform of CY(t) is Mellin transform is evaluated for a certain value
defined as of gk , then we may evaluate directly the Mellin
transform of the one-sided correlation function in
ð1
gk (which is for a value of r translated into the
MfCY ðtÞ; gg ¼ tg1 CY ðtÞdt ¼ YY ðg 1Þ
0 fundamental strip by using the technique
(25) described in appendix).
As a concluding remark of this section, we
that are the spectral moments of the correlation may assert that:
function. Inverse Mellin transform of Eq. 25
returns CY(t) in the form • The FSMs are able to restitute both power
spectral density (see Eq. 16) and the correla-
ð rþi1 tion function (see Eq. 27) in the whole domain
1
C Y ð tÞ YY ðg 1Þtg dg with exclusion of o = 0 or t = 0, respec-
2pi ri1
tively. This pathological behavior, due to the
D X m
discretization in axis, may be overcome by
ffi YY ðgk 1Þtgk ; t>0 (26)
2p k¼m assuming the value in zero as the value in
o = exp(b).
and as usual gk ¼ r þ ikD. Because of the sym- • The FSMs never diverge provided r lies in the
metry of RY(t), the latter may be found in the fundamental strip of the Mellin transform,
form then the cutoff value along the imaginary
axis may be selected also for a finite value of
D X m m, and the truncation of the complex series
RY ð t Þ ffi YY ðgk 1Þjtjgk ; t>0 does not produce divergences as o ! 1 or
4p k¼m
t ! 1.
(27) • Since LY ðgk 1Þ ¼ LY ðgk 1Þ and
gk gk
o ¼ ðo Þ , then the summation may
Fourier transform of Eq. 27 restitutes SY(o) in the
run from zero up to m with a very simple
form
manipulation.
• For any Mellin transformable function
D X m
gp
SY ðoÞ ¼ YY ðgk 1ÞGð1gk Þsin k jojgk 1 limn!
1 LY ðg 1Þ ! 0 , this happens since
2p k¼m 2
the LY ðg 1Þ is holomorph into the funda-
(28) mental strip. This means that the truncation of
the summations in Eq. 16 or in Eq. 27 may be
By letting gk 1 ¼ gk, Eq. 28 may be rewritten properly performed.
as • A very important property that allows us to
solve fractional differential equations,
1 X m
SY ð o Þ ¼ YY ðgk Þnc ðgk Þjojgk (29) Fokker-Planck equation, and many other
2b k¼m problem is that once the Mellin transform is
evaluated for a fixed value of r, say r1, then it
Comparison with the expression given in Eq. 16b may be easily evaluated for any other value of
reveals that r, say r2, provided r1 and r2 belong to the FS
(see “Appendix”).
LY ðgk 1Þ ¼ 2YY ðgk Þnc ðgk Þ (30) • Since Lg(g 1) is able to reconstruct both
RY (t) and SY(o), then it may be stated that
From this equation it may be recognized that the Lg(g 1) is another representation that fully
FSM may be also evaluated directly from describes the process Y(t). It follows that we
the correlation function simply by evaluating may now assert that, for a stationary normal
the Mellin transform of the correlation. If the process, we have three equivalent
Complex Fractional Moments and Their Use in Earthquake Engineering 453
a b
c 0.007 d 0.006
0.006
0.005
0.005
0.004
0.004
RY ( )
RY ( )
0.003
0.003
0.002
0.002
0.001
0.001
0.000 0.000
−20 −10 0 10 20 1 10 100 1000
Log
Complex Fractional Moments and Their Use in PSD (Eq. 31) continuous line, approximated (Eq. 16) dot-
Earthquake Engineering, Fig. 1 FSM, PSD, and cor- ted line; (c) correlation function (Eq. 32) continuous line,
relation: (a) FSM along the imaginary line r = 0.5; (b) approximated (Eq. 27) dotted line; (d) Fig. 1c in log plot
2p
eb is plotted, that for the present case
The FSM function is (D = 0.5) leads to 1.87 103
1.87 103.
454 Complex Fractional Moments and Their Use in Earthquake Engineering
a b
1.5
SY ( )
1.0
0.5
0.0
−40 −20 0 20 40
Complex Fractional Moments and Their Use in (r = 0.5); (b) PSD evaluated by Eq. 34 continuous line,
Earthquake Engineering, Fig. 2 FSM and PSD of the PSD evaluated with Eq. 16 dotted line
Tajimi-Kanai ground acceleration: (a) FSM versus
From this figure it appears that into the interval plotted (continuous line) and contrasted with the
ep=D
ep=D Eq. 16b shows a divergence PSD given in Eq. 16 (dotted line) for D = 0.5
phenomenon, but outside the two curves totally and m = 20.
overlaps each another. However, the aforemen- From the previous examples, it is evident that
tioned problem may be easily overcome by the level of accuracy reached by the use of FSM is
assuming in o = 0 (or in t = 0) the value in quite good including the trend at infinity. It is to
ep=D so avoiding the pathological behavior be emphasized that smaller values of D do not
in zero. Moreover, the divergence phenomenon produce sensible variations on both correlation
in eb
eb is self-similar in the sense that in and PSD evaluated by FSM. By contrast the cut-
the range e3b
e3b , e5b
e5b . . . the off value mD ¼ on the summation in Eq. 16 or
shape of the curve remains the same and then in Eq. 27 may produce variations and then has to
into eb
eb we have a fractal structure. be appropriately selected. Finally the number of
However, for technical purposes this fact is not FSM that is necessary to describe the Tajimi-
influent and is produced by the fact that D is Kanai PSD is very limited; with these complex
finite. This phenomenon disappears if D ! 0 quantities, all the PSD is restored (with the excep-
and m ! 1 that corresponds to the integral tion of the value in zero).
form of the inverse Mellin transform. Other examples may be found in Cottone and
As another example the Tajimi-Kanai (1960) Di Paola (2010) and Di Paola and Pinnola (2012).
PSD given as
from a propagation of the seismic waves in the representation of the cross-PSD with the FSM
medium, they are correlated with each another. requests the definition of the real and imaginary
Then in order to define the two processes in part of Eq. 37:
probabilistic setting, due to the hypothesis of
Gaussianity and stationarity, we need of the cor- ox
ℜ½SY 1 Y 2 ðoÞ ¼ SY ðoÞrð1, 2Þ ðo, jxjÞ cos
relation function matrix v
(38a) C
RY ðtÞ ¼ E YðtÞYT ðt þ tÞ
ox
RY ð t Þ RY 1 Y 2 ðtÞ ℑ½SY 1 Y 2 ðoÞ ¼ SY ðoÞrð1, 2Þ ðo, jxjÞ sin
¼ 1
(35) v
RY 1 Y 2 ðtÞ RY 2 ðtÞ
(38b)
where YT ðtÞ ¼ ½Y 1 ðtÞY 2 ðtÞ, RY j ðtÞ is the correla- where ℜ½ and ℑ½ are the real and imaginary part
tion of Yj( j = 1, 2), and RY 1 Y 2 is the cross- of the term in parentheses. The Mellin transform
correlation between Y1(t) and Y2(t). Since borns for functions that are defined in the range
RY i Y j ðtÞ ℝ and RY i Y j ðtÞ ¼ RY j Y i ðtÞ , then the 0
1. For SY j ðoÞ Eq. 16b restores the PSD in
Fourier transform of the correlation matrix, terms of FSM. For the cross-PSD, we may con-
namely, the PSD matrix, is given as sider the real and the imaginary part separately.
By taking into account Eq. 16b for the real part
S Y ðo Þ SY 1 Y 2 ðoÞ
SY ðoÞ ¼ 1 (36) given in Eq. 38a, because it is an even function
SY 1 Y 2 ðoÞ S Y 2 ðo Þ like SY j ðoÞ, we may write
from two different points. In any case, whatever LℑY 1 Y 2 ðg 1Þ ¼ 2 ℑ½SY 1 Y 2 ðoÞog1 do (42)
the coherence function is selected, the 0
456 Complex Fractional Moments and Their Use in Earthquake Engineering
It follows that the cross-PSD is given as then discrete inverse Mellin transform gives
m h
1 X 1 X m
SY 1 Y 2 ðoÞ ¼ Lℜ ðg 1Þ Y 1 Y 2 ð tÞ ¼
Cev Yev ðg 1Þjtjgk ;
4b k¼m Y 1 Y 2 k 4pb k¼m Y 1 Y 2 k
i
þ isgnðoÞLℑY 1 Y 2 ðgk 1Þ jojgk (49a)
(43) sgnðtÞ X
m
Y 1 Y 2 ð tÞ ¼
Cod Yod ðg 1Þjtjgk
4pb k¼m Y 1 Y 2 k
Inverse Fourier transform gives the cross-
correlation function in the form (49b)
where ns ð1 gk Þ ¼ Gð1 gk Þ sin ½ð1 gk Þp=2. Y 1 Y 2 ðgk 1Þ ¼ 2YY 1 Y 2 ðgk Þnc ðgk Þ;
Lev ev
(50)
As in the case of autocorrelation also for the Y 1 Y 2 ðgk 1Þ ¼ 2YY 1 Y 2 ðgk Þns ðgk Þ
Lod od
With this choice the corresponding impulse that is the discrete filter equation. An exact rep-
response function of N ½ is symmetric in time resentation is given in the form
domain, and then it violates the causality condi-
ð
rþi1
tion. Even though the causality condition is vio- 1
lated, in the frame of stationary processes, this Y ðt Þ ¼ LH ð1 gÞðI g W ÞðtÞdg (59)
4pi
does not produce any problem since the output of ri1
a 20 b
1.2
15
1.0
0.8
10
H( )
h(t)
0.6
5 0.4
0.2
0
0.0
−1.0 −0.5 0.0 0.5 1.0 −40 −20 0 20 40
t
Complex Fractional Moments and Their Use in evaluated with Eq. 57 (dotted line); (b) H(o) evaluated
Earthquake Engineering, Fig. 3 H(o) and h(t) of the with Eq. 54 (continuous line), H(o) evaluated with Eq. 56
Tajimi-Kanai ground acceleration: (a) h(t) evaluated as (dotted line)
inverse Fourier transform of Eq. 54 (continuous line), h(t)
For more details on the filter equations and gen- and time domain. By using FSM also the filter
eration of sample function, see Cottone et al. (2010). equations have been found. It is shown that both
impulse response function and its Fourier trans-
form, namely, the transfer function of the filter,
Summary may be found by evaluating the FSM of the
squared root of the target PSD.
Fractional calculus opens new perspectives for By using the same concepts, a very large num-
solving problems that with the classical calculus ber of other problems may be solved. Such an
may not be solved. The initial effort to understand example the probability density function of a ran-
the rules is fully rewarded from the elegance and dom variable may be obtained with the same tech-
the unbelievable accuracy obtained. nique here used for representing cross-correlations
In this entry only two problems have been in terms of FSMs. It follows that Fokker-Planck
presented: representation of the power spectral equation, Kolmogorov-Feller equation, Einstein-
density (PSD) and correlation by means of com- Smoluchowski equation, and path integral solu-
plex fractional moments and filter equations. tion (Cottone et al. 2008) may be solved in terms
Regarding the first problem, it has been shown of FSM. Moreover, wavelet transform and classi-
that both PSD and correlation may be represented cal or fractional differential equations may be
by the so-called fractional spectral moments easily solved by using fractional calculus and
(FSMs). The latter are the extension of the clas- Mellin transform in complex domain.
sical spectral moments introduced by Vanmarcke
(1972) to complex order. The appealing in using Acknowledgments The author is indebted to the Ph.D.
these FSM is that they are able to reconstruct both student Gioacchino Alotta for the help in preparing the
numerical applications.
PSD and correlation. As a result, it may be stated
that FSM function is another equivalent represen-
tation of PSD and correlation. Moreover it has Appendix
been shown that with a limited number of infor-
mations (few FSM), the whole PSD and correla- In this Appendix it is shown that once the Mellin
tion may be restored, including the trend at transform of a given function is evaluated for a
infinity. Extension to multivariate seismic pro- certain value r1 (into the fundamental strip), the
cess has been also provided in both frequency Mellin transform of the function may be
Complex Fractional Moments and Their Use in Earthquake Engineering 459
evaluated for any other value r2 (into the FS) as a Since Eqs. 60 and 61 remain valid also for
linear combination of the fractional spectral r = r2, we may write
moments of the function evaluated in r1. This
fact is very important in order to solve a lot of D X m
ð 1Þ
ð1Þ
problems like solution of fractional differential f M gk 1 tgk
2p k¼m
equation, Fokker-Planck equation, and Einstein-
D X m
ð 2Þ
Smoluchowski equation and also to evaluate ’
ð2Þ
f M gk 1 tgk (62)
C
FSM for different value of r; see Eq. 26. Results 2p k¼m
of this appendix have been found by the author
(Di Paola 2014; Di Matteo et al. 2014) for solving where gð2Þ ¼ r2 þ i. By multiplying both sides
Fokker-Planck equation and Kolmogorov-Feller of Eq. 62 for t1/2 and remembering that
equation. gðjÞ ¼ rðjÞ þ i ( j = 1, 2), we obtain
We denote as f M gð1Þ 1 the Mellin trans-
form of the function f(t) which evaluates for X
m
ð1Þ
t1=2 f M gk 1 eikblnt ’ tðDrþ1=2Þ
p
r = r1, that is,
k¼m
X
m
ð
1 ð2Þ
f M gk 1 eikblnt ;
p
ð 1Þ x>0
f M gð1Þ 1 ¼ Mff ðtÞ; gg ¼ f ðtÞtg 1
dt; k¼m
0 (63)
ð1Þ
g ¼ r1 þ i
where Dr = r2 r1. It is to be remarked
(60)
that Eq. 63 strictly holds for t > 0 because
in zero singularities appear. Now we want
Inverse Mellin transform operator returns f(t) in
ð2Þ
the form to evaluate coefficients f M gk 1 from the
ð1Þ
knowledge of f M gk 1 . Because Eq. 61 is an
D X m
ð 1Þ
ð1Þ approximation, we require that Eq. 63 holds true
f ðtÞ ’ f M gk 1 tgk (61)
2p k¼m in a weak sense in the interval t1 > 0, t2 0:
ðt2 (" Xm
p Xm
# )
1 ð1Þ isblnt ðDrÞ ð2Þ ikpblnt
f M gs 1 e t f M gk 1 e ½C:C: dt
t s¼m k¼m
t1
ð2Þ
¼ min f M gk 1 (64)
where C.C. stands for complex conjugate. It is choice two goals are achieved. First the interval
convenient to make a change of variables eb
eb is very wide for the usual values of
D(D ¼ 0:25
0:5). As in fact in the worst case
dt
x ¼ lnt, dx ¼ ; x ¼ lntj , j ¼ 1, 2 (65) ( D ¼ 0:5 ) eb ¼ 1:87 103 , eb ¼ 5:35 102 ;
t j second in the x domain the extremes of the inte-
Moreover, a good choice of t1 and t2 in Eq. 64 is gral (64) become b
b and in this interval
t1 ¼ eb ¼ eD ,
p p
t2 ¼ eb ¼ eD . With this exp i pb sx are orthogonal.
460 Complex Fractional Moments and Their Use in Earthquake Engineering
ðb (" X
m
p X
m
# )
ð1Þ isbx Drx ð2Þ ikpbx
f M gs 1 e e f M gk 1 e ½C:C: dx
s¼m k¼m
b
¼ min f M gðs1Þ 1 (66)
Samko SG, Kilbas AA, Marichev OI (1993) Fractional periods of vibration, Ti, conditioned on spectral
calculus integrals and derivatives. Gordon and Breach acceleration at a conditioning period, T*, of inter-
Science Publishers, Amsterdam
Shinozuka M, Deodatis G (1988) Stochastic process est. The conditional spectrum utilizes the corre-
models for earthquake ground motion. Probab Eng lations between spectral accelerations at different
Mech 3(3):114–123 periods (e.g., Baker and Jayaram 2008) to com-
Shinozuka M, Sato Y (1967) Simulation of nonstationary pute the expected (logarithmic mean) response
random process. J Eng Mech ASCE 93:11–40
Solomos GP, Spanos PD (1984) Oscillator response to spectrum (Baker and Cornell 2006; Baker 2011) C
non-stationary excitation. Am Soc Mech Eng Appl and additionally account for the variability
Mech Div AMD 65:159–170 (variance) of the response spectra (Jayaram
Spanos PD (1983) ARMA algorithms for ocean wave model- et al. 2011; Lin et al. 2013a). Assuming the dis-
ing. Trans ASME J Energy Resour Technol 105:300–309
Spanos PD (1986) Filter approach to wave kinematics tribution of logarithmic spectral accelerations is
approximations. Appl Ocean Res 8:2–7 multivariate normal (Jayaram and Baker 2008),
Spanos PD, Miller SM (1994) Hilbert transform general- then the first two moments (i.e., the means, stan-
ization of a classical random vibration integral. J Appl dard deviations, and correlations) fully describe
Mech 61:575–581
Tajimi H (1960) A standard method of determining the the conditional spectrum. The CS provides the
maximum response of a building structure during an link between seismic hazard and structural
earthquake. In: Proceedings of 2nd world conference response, the first two elements of performance-
on Earthquake Engineering, vol 2, Tokyo based earthquake engineering (PBEE). To main-
Vanmarcke EH (1972) Properties of spectral moments
with applications to random vibration. J Eng Mech tain consistency with probabilistic seismic hazard
ASCE 98:425–446 analysis (PSHA), the computation of the CS is
West JW, Bologna M, Grigolini P (2003) Physics of refined by incorporating multiple causal earth-
fractal operators. Springer, New York quakes and ground motion prediction models
Zerva A (1991) Effect of spatial variability and propagation
of seismic ground motions on the response of multiply (GMPMs, also known as ground motion predic-
supported structures. Probab Eng Mech 6:212–221 tion equations, attention relationships) (Lin et al.
2013a). The CS can be used as a target response
spectrum for site- and structure-specific ground
motion selection (e.g., Lin et al. 2013b, c).
Conditional Spectra Baker and Cornell (2006) and Baker (2011)
described the “conditional spectrum” as the “con-
Ting Lin1 and Jack Baker2 ditional mean spectrum” (CMS) with an empha-
1
Department of Civil, Construction and sis on the mean spectrum and its associated
Environmental Engineering, Marquette ground motion selection where the best ground
University, Milwaukee, WI, USA motion candidates have response spectra as close
2
Stanford University, Stanford, CA, USA to the CMS as possible. Bradley (2010) extended
the concept of the CMS to ground motion param-
eters other than response spectra and referred the
Synonyms resulting distribution as a “generalized conditional
intensity measure” (GCIM). Bradley (2012) also
Conditional mean spectrum; Hazard-consistent developed an algorithm to select ground motions
ground motion selection; Response spectrum; based on the GCIM. Abrahamson and Al Atik
Scenario spectrum; Spectral shape (2010) coined the term “conditional spectrum”,
but represented the CS distribution directly by
realizations of the CS rather than an analytical
Introduction distribution. Wang (2011) developed a ground
motion selection algorithm that captures response
The conditional spectrum (CS) is a response spectrum characteristics and variability of sce-
spectrum that specifies the probability distribu- nario earthquakes. Gulerce and Abrahamson
tion of spectral accelerations, Sa, over a range of (2011) extended the CS concept to vertical ground
462 Conditional Spectra
Conditional Spectra, Fig. 1 Response spectra of hazard spectrum as target spectra for Sa(2.6 s) associated
ground motions selected and scaled with (a) conditional with 2 % in 50 years probability of exceedance for an
spectrum, (b) conditional mean spectrum, and (c) uniform illustrative structure in Palo Alto, California, USA
motions. Jayaram et al. (2011) developed a com- analysis) in risk-based and intensity-based
putationally efficient algorithm to select ground assessments respectively. Intensity-based assess-
motions that match the conditional mean and var- ment estimates structural response given ground
iance of the response spectrum. motions with a specified intensity level. Risk-
Lin et al. (2013a) for the first time used the based assessment (also known as PEER integral,
term “conditional spectrum” which was used to drift hazard, time-based assessment, and proba-
represent an analytical distribution of response bilistic seismic demand analysis) estimates the
spectra in the manner described here. The CS mean annual rate of exceeding a specified struc-
builds upon the CMS (which focuses on the tural response amplitude and can be obtained by
mean) and includes a measure of the variance in integrating intensity-based assessment results at
addition to the mean. Figure 1a depicts the ana- various intensity levels with the associated seis-
lytical distribution of the conditional spectrum mic hazard curve. As illustrated in Fig. 1a, the
with conditional mean spectrum as the thick thin lines represent response spectra from indi-
solid line as well as conditional mean +/ condi- vidual ground motion records selected to match
tional standard deviation as the thick dashed the mean and variance of the CS (thick solid and
lines. Lin et al. (2013b, c) utilized the conditional dashed lines) for a specified intensity level, i.e.,
spectrum to select ground motions for nonlinear spectral acceleration at the period of 2.6 s
dynamic analysis (also known as response history (“pinched” conditioning period) associated with
Conditional Spectra 463
Conditional Spectra, Fig. 2 (a) Conditional mean spec- spectra at various conditioning periods (compared to uni-
tra at T* = 2.6 s at various intensity levels (from 50 % in form hazard spectrum) at the 2 % in 50-year intensity level
30 years to 1 % in 200 years) and (b) conditional mean
hazard spectrum, which assumes equal probabil- dispersion leads to higher and lower extremes
ities of exceedance of spectral accelerations at all of structural responses and the associated dam-
periods. However, no single ground motion is age states and losses. The increased dispersion
likely to produce a response spectrum as high as can also lead to a larger probability of struc-
that of the UHS over a wide range of periods (e.g., tural collapse. Incorporating the variability in
Naeim and Lew 1995). The conditional mean the CS has significant implications for appli-
spectrum utilizes the correlations of Sa across cations where the dispersion of the responses
periods to more realistically compute the is an important consideration.
expected Sa values at all periods given a target
Sa value at a period of interest (e.g., Baker and
Cornell 2006). The conditional spectrum quan- Basic Computation
tifies the conditional distribution (mean and var-
iance) of Sa at all periods given Sa at a period of The basic computation of the conditional spec-
interest (e.g., Lin et al. 2013a). The spectral trum requires (1) input earthquake parameters
shapes of conditional mean spectra (mean of con- such as magnitude and distance (similar to those
ditional spectra) change with amplitudes required by a GMPM), (2) specification of a con-
(Fig. 2a). Uniform hazard spectrum is the enve- ditioning period, (3) a ground motion prediction
lope of conditional mean spectra at various model to estimate mean and standard deviation of
periods (Fig. 2b). logarithmic spectral acceleration, and (4) a cor-
The target response spectrum used when relation model of spectral accelerations across
selecting ground motions has an impact on periods. The computation procedure is outlined
the distribution of structural responses from as follows:
resulting nonlinear dynamic analysis (e.g.,
Jayaram et al. 2011; Lin et al. 2013c). (a) Obtain input earthquake parameters: Obtain
Matching target response variance (in the CS the target spectral acceleration at the condi-
compared to the CMS) increases the dispersion tioning period T*, Sa(T*), from PSHA, and its
of structural responses, thereby affecting the associated mean causal earthquake magni-
distribution of structural responses and conse- tude (M), distance (R), and other parameters
quently the damage state and loss estimation (if available) such as rupture mechanism and
computations in performance-based earth- site condition (y), from deaggregation (e.g.,
quake engineering. The increase in the McGuire 1995).
Conditional Spectra 465
(b) Compute mean and standard deviation of conditional mean value mlnSa(Ti)|lnSa(T*), spec-
lnSa(Ti): Use a GMPM (e.g., Abrahamson ifies a complete distribution of logarithmic
et al. 2008) to compute the logarithmic spectral acceleration values at all periods
mean and standard deviation of Sa at all (where the distribution at a given period is
periods Ti, denoted as mlnSa(M, R, y, Ti) and Gaussian, as justified by Jayaram and Baker
slnSa(M, y, Ti). 2008). The resulting spectrum distribution is
(c) Compute epsilon, e(Ti): For any Sa(Ti), com- termed the “conditional spectrum”, to be dis- C
pute e(Ti), the number of standard deviations tinguished from the CMS that does not con-
by which lnSa(Ti) differs from the mean spec- sider the variability.
tral ordinate predicted by a given GMPM,
mlnSa(M, R, y, Ti), at Ti
Deaggregation of Ground Motion
lnSaðT i Þ mlnSa ðM, R, y, T i Þ Prediction Models
ϵðT i Þ ¼ (1)
slnSa ðM, y, T i Þ
Probabilistic seismic hazard deaggregation of
*
The target e(T ) can also be computed using GMPMs links the computation of a target spec-
this equation. Spectral shape is characterized trum (for ground motion selection) to the total
by e. hazard prediction (from PSHA). Computation of
(d) Compute conditional mean mlnSa(Ti)|lnSa(T*): a target spectrum requires deaggregation (via
Using a correlation model (e.g., Baker and Bayes’ rule) to identify the causal ground motion
Jayaram 2008) that specifies the correlation parameters, along with the predictions from mul-
coefficient between pairs of epsilon values at tiple GMPMs. Hazard-consistent ground motion
two periods, rðϵT i Þ, ϵðT Þ (hereinafter selection incorporates the aleatory uncertainties
referred to as r(Ti, T*)), the conditional from earthquake scenarios and the epistemic
mean logarithmic spectral acceleration at uncertainties from multiple GMPMs, through
other periods Ti, mlnSa(Ti)|lnSa(T*) is computed deaggregation of GMPMs (Lin and Baker 2011).
This GMPM deaggregation is consistent with
mlnSaðT i ÞjlnSaðT Þ ¼ mlnSa ðM, R, y, T i Þ the probabilistic treatment of the magnitude and
distance random variables in traditional PSHA.
þ rðT i , T ÞslnSa ðM, y, T i Þ The deaggregation of GMPMs provides additional
(2) insights into which GMPM contributes most to
prediction of Sa values of interest. To match the
The spectrum defined by mlnSa(Ti)|lnSa(T*) is contribution of each GMPM to its associated
termed the “conditional mean spectrum”, as ground motion parameters, separate deaggregation
it specifies the mean values of lnSa(Ti) condi- of earthquake parameters for each GMPM is also
tional on the value of lnSa(T*). Assuming Sa is performed. First, seismic hazard is estimated using
lognormally distributed, the exponentials of PSHA that incorporates multiple GMPMs. Next,
mlnSa are equivalent to the median values of Sa. the relative contributions of events and GMPMs to
(e) Compute conditional standard deviation the hazard prediction are identified using the
slnSa(Ti)|lnSa(T*): The standard deviation of refined deaggregation procedures.
logarithmic spectral acceleration at period Ti The deaggregation of GMPMs is performed as
conditioned on the value of lnSa(T*) is an intermediate step to develop a refined mean
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi and, more importantly, a refined standard devia-
slnSaðT i ÞjlnSaðT Þ ¼ slnSa ðM, y, T i Þ 1 r2 ðT i , T Þ tion of the CS. This can also be used to develop,
(3) in general, any new target response spectrum
that is consistent with PSHA. The side product
The conditional standard deviation of this computation is the deaggregation
slnSa(Ti)|lnSa(T*), when combined with the (posterior) weights of GMPMs, instead of the
466 Conditional Spectra
Conditional Spectra, Fig. 3 (a) Conditional means and (b) conditional standard deviations of the CS computed using
Methods 2–4 for Sa(0.2 s) with 10 % probability of exceedance in 50 years at Bissell
logic-tree (prior) weights. This methodology for Method 4, accounts for the contribution that
deaggregation of prediction models can also be each GMPM and each earthquake M/R makes to
immediately applicable to other procedures the seismic hazard (Fig. 3). Subscript k denotes
which require multiple prediction models in an that the calculations are based on a specific
earlier stage of total prediction and a later stage of GMPM indexed by k, whereas subscript
new target computation. j denotes that the parameters such as M/R are
associated with a specific earthquake indexed by j.
For each GMPM k and causal earthquake com-
Refined Computational Methods bination Mj/Rj, mean and standard deviation of
the CS can be obtained (using Eqs. 12 and 13,
Approximate and exact implementations of CS respectively) that are then combined with
computations are outlined and used for example corresponding deaggregation weights, pj,kd
calculations at Stanford, Bissell, and Seattle (Lin (Eqs. 14 and 15). The weight for model k is
et al. 2013a). Figure 3 illustrates CS mean and denoted as pkl where the superscript l refers to
standard deviation computational methods using logic-tree; similarly, the superscript d in pkd
multiple GMPMs at Bissell. Exact CS mean and and pj,kd refers to deaggregation.
standard deviation calculations (Method 4) can
incorporate multiple GMPMs and causal earth- Method 1: Approximate CS using mean M/R and
quake M/R combinations. Varying levels of a single GMPM (basic computation)
approximations are also considered, that replace
multiple M/R combinations with the mean M/R mlnSa, kðT i ÞjlnSaðT Þ mlnSa, k M, R; y; T i
from deaggregation (Methods 1–3), and either (4)
þ rðT i , T ÞslnSa, k M, y; T i ϵðT Þ
consider only a single GMPM (Method 1) or
perform an approximate weighting of several pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
slnSa, kðT i ÞjlnSaðT Þ slnSa, k M, y; T i 1 r2 ðT i , T Þ
GMPMs (Methods 2–3). Methods 1–2 utilize
the overall mean M/R, whereas Method 2 con- (5)
siders multiple GMPMs with logic-tree weights.
The deaggregation of GMPMs provides GMPM- Method 2: Approximate CS using mean M/R
specific mean M/R and GMPM deaggregation and GMPMs with logic-tree weights
weights, both of which are incorporated in X
Method 3. The most accurate computation, mlnSaðT i ÞjlnSaðT Þ plk mlnSa, kðT i ÞjlnSaðT Þ (6)
k
Conditional Spectra 467
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
2
slnSaðT i ÞjlnSaðT Þ plk s2lnSa, kðT i ÞjlnSaðT Þ þ mlnSa, kðT i ÞjlnSaðT Þ mlnSaðT i ÞjlnSaðT Þ (7)
k
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Method 3: Approximate CS using GMPM- slnSa, kðT i ÞjlnSaðT Þ slnSa, k Mk , yk , T i 1 r2 ðT i , T Þ
C
specific mean M/R and GMPMs with (9)
deaggregation weights
X
mlnSa, kðT i ÞjlnSaðT Þ mlnSa, k Mk , Rk , yk , T i mlnSaðT i ÞjlnSaðT Þ pdk mlnSa, kðT i ÞjlnSaðT Þ (10)
(8)
þ rðT i , T ÞslnSa, k Mk , yk , T i ϵk ðT Þ k
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
X
2
slnSaðT i ÞjlnSaðT Þ pk slnSa, kðT i ÞjlnSaðT Þ þ mlnSa, kðT i ÞjlnSaðT Þ mlnSa, kðT i ÞjlnSaðT Þ
d 2
(11)
k
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Method 4: “Exact” CS using multiple causal slnSa, j, kðT i ÞjlnSaðT Þ ¼ slnSa, k Mj , yj ,T i 1 r2 ðT i , T Þ
earthquake M/R and GMPMs with (13)
deaggregation weights
XX
mlnSaðT i ÞjlnSaðT Þ ¼ pdj, k mlnSa, j, kðT i ÞjlnSaðT Þ
mlnSa, j, kðT i ÞjlnSaðT Þ ¼ mlnSa, k Mj , Rj , yj , T i k j
(12)
þ rðT , T i Þϵj ðT ÞslnSa, k Mj , yj , T i (14)
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
XX
2
slnSaðT i ÞjlnSaðT Þ ¼ pj, k slnSa, j, kðT i ÞjlnSaðT Þ þ mlnSa, j, kðT i ÞjlnSaðT Þ mlnSaðT i ÞjlnSaðT Þ
d 2
k j
(15)
approximate treatment of those weights works selection algorithm (1) probabilistically gener-
less effectively. The exact conditional standard ates multiple response spectra from a target dis-
deviation is always higher than approximate tribution, and then (2) selects recorded ground
results (Fig. 3b) because of the additional contri- motions whose response spectra individually
bution from the variance in mean logarithmic match the simulated response spectra. (3) A
spectral accelerations (Fig. 3a) due to variation greedy optimization technique further improves
in causal earthquakes and GMPMs. The impact the match between the target and the sample
of CS approximations on engineering decisions means and variances. This algorithm has been
depends upon the difference between approxi- adopted by the global earthquake model (http://
mate and exact spectra: as the difference www.globalquakemodel.org/) for ground motion
increases, more refined target spectra will have selection. Alternative algorithms that can be used
increasing benefits for ground motion selection for CS-based ground motion selection include
and structural response assessment (Lin those by, e.g., Wang (2011) and Bradley (2012).
et al. 2013a). If the uniform hazard spectrum or the condi-
The deaggregation of ground motion predic- tional mean spectrum, instead of the conditional
tion models and the exact conditional mean spec- spectrum, is used as the target spectrum to select
trum (incorporating multiple causal earthquakes (and optionally scale) ground motions, the PEER
and multiple GMPMs) have been implemented in ground motion database (http://peer.berkeley.
the United States Geological Survey hazard map- edu/products/strong_ground_motion_db.html) is
ping tools (http://geohazards.usgs.gov/deaggint/ an online tool appropriate for such applications.
2008/). Carlton and Abrahamson (2014) pro- In addition to CS-based ground motion selection
posed an alternative Method 2.5 for a simplified (as shown in Fig. 1a), Jayaram et al. (2011) algo-
implementation of the CS, which gives similar rithm can also be modified to accommodate
results to Method 3 in Lin et al. (2013a). CMS- and UHS-based ground motion selection,
Deaggregation weights, instead of logic-tree as illustrated in Fig. 1b, c respectively.
weights, of GMPMs were recommended when
multiple GMPMs were used to compute the CS
(e.g., Lin et al. 2013a; Carlton and Abrahamson Hazard Consistency in Risk-Based
2014). Gulerce and Abrahamson (2011) extended Assessments
the CS concept to vertical ground motions. Brad-
ley (2010) developed a generalized conditional Risk-based assessment estimates the mean
intensity measure approach to include intensity annual rate of exceeding a specified structural
measures other than Sa. response amplitude. The conditional spectrum
computation requires specification of a condi-
tioning period. The sensitivity of risk-based
Ground Motion Selection Algorithms assessment results to the choice of conditioning
period using CS-based ground motion selection is
The conditional spectrum can be used as the investigated in Lin et al. (2013b). This study
target spectrum to select “representative” ground focuses on risk-based assessments, with a specific
motions from a ground motion database for emphasis on the rates of exceeding various levels
performance-based earthquake engineering. of peak story drift ratio (i.e., drift hazard calcula-
A computationally efficient algorithm (Jayaram tions) in the structure. Some additional engineer-
et al. 2011) suitable for such CS-based ground ing demand parameters (EDPs) are also
motion selection (http://www.stanford.edu/ considered, such as the peak floor acceleration
~bakerjw/gm_selection.html) computes a target over the full building heights, a single-story
spectrum, and then selects and scales ground story drift ratio, and a single-story floor acceler-
motions from the PEER NGA database to match ation. The primary structure considered is a
the target spectrum mean and variance. This 20-story reinforced concrete frame structure
Conditional Spectra 469
Conditional Spectra, Fig. 4 Risk-based assessments of from ground motions selected to match the CS, the CMS,
(a) peak story drift ratio and (b) peak floor acceleration of and the UHS
a 20-story reinforced concrete frame structure obtained
assumed to be located in Palo Alto, California, parameters of interest. Here “proper representa-
using a structural model (Haselton and Deierlein tion” refers to consistency with the site ground
2007) with strength and stiffness deterioration motion hazard curves at all relevant periods, and
that is believed to reasonably capture the this is achieved by using the CS approach to
responses up to the point of collapse due to determine target response spectra for the selected
dynamic instability. ground motions. The reproducibility of the risk-
The risk-based assessments are performed based assessment results, for varying condition-
several times, using ground motions selected ing periods, then results from the fact that the
and scaled to match conditional spectra, where ground motion intensity measure used to link
the conditioning period used for these calcula- the ground motion hazard and the structural
tions is varied from 0.45 to 5.0 s (i.e., the build- response is not an inherent physical part of the
ing’s third-mode structural period up to seismic reliability problem considered; it is only
approximately twice the first-mode period). For a useful link to decouple the hazard and structural
each case, the risk-based assessment results are analysis. If this link is maintained carefully then
found to be similar (Fig. 4). The similarity of the one should obtain a consistent prediction of the
results stems from the fact that the careful record risk-based assessment in every case.
selection ensures that the distributions of
response spectra at all periods are nominally
comparable, so the distribution of resulting struc- Intensity-Based Assessments and
tural responses should also be comparable (to the Evaluation of Alternative Target Spectra
extent that response spectra describe the relation-
ship between the ground motions and structural Intensity-based assessment estimates structural
responses). response given ground motions with a specified
From these results, it is observed that if the intensity level. A study on the sensitivity of
analysis goal is to perform a risk-based assess- intensity-based assessment results to the choice
ment obtaining the mean annual frequency of of conditioning period when the CS is used as a
exceeding an EDP, then one should be able to target for ground motion selection and scaling is
obtain an accurate result using any conditioning conducted by Lin et al. (2013c). This study also
period, provided that the ground motions are investigates the sensitivity of both risk-based and
selected carefully to ensure proper representation intensity-based assessments to the choice of tar-
of spectral values and other ground motion get spectrum, including evaluation of the uniform
470 Conditional Spectra
Conditional Spectra, Fig. 5 Intensity-based assess- 20-story reinforced concrete frame structure obtained
ments of peak story drift ratio (median and logarithmic from ground motions selected to match the CS, the
standard deviation in (a) and (b) respectively) of a CMS, and the UHS
hazard spectrum and the conditional mean exceedance (e.g., for a building-code-type
spectrum. check). In this extremely common case, the
The study shows the different effects of con- response prediction will always change
ditioning periods on intensity-based and risk- depending upon the choice of conditioning
based assessments. In contrast to risk-based period. This comes from the fact that the choice
assessments, which are relatively insensitive to of conditioning period is an inherent part of the
the choice of conditioning period in the CS, the problem statement, and so in this case changing
choice of conditioning period for the CS can the conditioning period changes the question that
substantially impact structural response estimates is being asked. For example, computing the
for an intensity-based assessment (NIST 2011). median drift response for a building subjected to
It is therefore critical to specify the structural a 2 % in 50 year exceedance Sa(1 s) is not the
analysis objective clearly. same as computing the median drift response for
The study also demonstrates the importance of a building subjected to a 2 % in to 50 year exceed-
target spectrum. For intensity-based assessments, ance Sa(2 s); these are two different questions.
use of the CMS, instead of the CS, does not Resolution of this issue likely lies in identifying a
significantly affect the median response estimates conditioning period and performance check that,
(Fig. 5a) but does decrease both the dispersion of when passed, confirms satisfactory reliability of
the response (Fig. 5b) and the probability of col- the structural system.
lapse, while use of the UHS typically results in
higher median response (Fig. 5a). For risk-based
assessments, use of the CMS, instead of the CS, Summary
results in underestimation of structural response
hazard due to the omission of spectral variability The conditional spectrum (CS) is a target
(Fig. 4), while use of the UHS results in response spectrum (with mean and variance) for
overestimation in the structural response hazard ground motion selection to perform nonlinear
(Fig. 4). dynamic analysis. The computation of the CS
An important issue regarding conditioning requires (1) input earthquake parameters of mag-
period arises when an intensity-based assessment nitude, distance, relevant seismological and site
is being used and the purpose is to compute the characteristics (e.g., fault type and soil type);
mean or median response associated with an (2) specification of a conditioning period;
Sa(T*) having a specified probability of (3) ground motion prediction model(s) (GMPMs)
Conditional Spectra 471
to estimate mean and standard deviation of loga- motions. Progress has been made towards the
rithmic spectral accelerations; and (4) correlation effect of conditioning period in intensity- and
model(s) to relate spectral accelerations between risk-based assessments (Lin et al. 2013b, c;
different periods of vibration. Refined computa- Carlton and Abrahamson 2014), the performance
tion of the CS incorporates multiple causal goal and design check via design spectrum cali-
earthquakes and multiple ground motion predic- bration (Loth and Baker 2013), and a new method
tion models. The level of complexity in computa- termed “adaptive incremental dynamic analysis” C
tion increases from Methods 1 (single earthquake (AIDA) (Lin and Baker 2013) that allows overlap
and single ground motion prediction model) to of ground motions across intensity levels.
4 (multiple earthquakes and multiple ground Research is ongoing to provide insights on the
motion prediction models with corresponding current engineering practice of ground motion
deaggregation weights). Factors that affect the scaling and the seismological counterpart in
accuracy of the approximations include (1) the direct ground motion simulation.
input causal earthquake parameters, (2) the
GMPMs used, and (3) the GMPM deaggregation
weights. Hence, exact calculation methods may be
needed for locations with hazard contributions Cross-References
from multiple earthquake sources especially with
multiple source types, and/or for sites with larger ▶ Physics-Based Ground-Motion Simulation
variation in predictions from various GMPMs (Lin ▶ Probabilistic Seismic Hazard Models
et al. 2013a). An intermediate method, Method ▶ Selection of Ground Motions for Response
2.5, was proposed by Carlton and Abrahamson History Analysis
(2014) as a simpler practical alternative to Method ▶ Stochastic Ground Motion Simulation
3 in Lin et al. (2013a). Probabilistic seismic hazard
deaggregation of GMPMs (Lin and Baker 2011),
which computes the relative contribution of References
GMPMs to a given intensity level, facilitates the
refined CS computation. Deaggregation weights, Abrahamson NA, Al Atik L (2010) Scenario spectra for
instead of logic-tree weights, of GMPMs design ground motions and risk calculation. In: 9th US
National and 10th Canadian conference on earthquake
were recommended when multiple GMPMs were
engineering, Toronto, 12 p
used to compute the CS (e.g., Lin et al. 2013a; Abrahamson NA, Bozorgnia Y, Boore D, Atkinson G,
Carlton and Abrahamson 2014). Notable exten- Campbell K, Silva W, Chiou B, Idriss IM, Youngs
sions of the CS concept include the CS for vertical R (2008) Comparisons of the NGA ground-motion
relations. Earthq Spectra 24(1):45–66
motions (Gulerce and Abrahamson 2011) and
ASCE (2010) Minimum design loads for buildings and
the generalized conditional intensity measure other structures. ASCE 7–10. American Society of
(Bradley 2010). Civil Engineers/Structural Engineering Institute,
The computation and use of the CS advances Reston
Baker JW (2011) Conditional mean spectrum: tool for
hazard-consistent ground motion selection for
ground-motion selection. J Struct Eng 137(3):322–331
performance-based earthquake engineering. The Baker JW, Cornell CA (2006) Spectral shape, epsilon and
automation via USGS hazard mapping interface record selection. Earthq Eng Struct Dyn
and ground motion selection web-based algo- 35(9):1077–1095
Baker JW, Jayaram N (2008) Correlation of spectral accel-
rithms enhances practical applications of the eration values from NGA ground motion models.
CS. Challenges in the CS remain in (1) the choice Earthq Spectra 24(1):299–317
of the conditioning period, (2) the appropriate Bradley BA (2010) A generalized conditional intensity
question to ask in design-orientated intensity- measure approach and holistic ground-motion selec-
tion. Earthq Eng Struct Dyn 39(12):1321–1342
based assessment, (3) a larger number of distinct
Bradley BA (2012) A ground motion selection algorithm
ground motions that are used at various intensity based on the generalized conditional intensity measure
levels, and (4) issues related to scaling ground approach. Soil Dyn Earthq Eng 40:48–61
472 Conditional Spectra
Carlton B, Abrahamson N (2014) Issues and approaches international conference on structural safety and reli-
for implementing conditional mean spectra in practice. ability. CRC Press, New York, 8 p
Bull Seismol Soc Am 104(1):503–512 Lin T, Harmsen SC, Baker JW, Luco N (2013a) Condi-
Cornell CA (1968) Engineering seismic risk analysis. Bull tional Spectrum computation incorporating multiple
Seismol Soc Am 58(5):1583–1606 causal earthquakes and ground‐motion prediction
Cornell CA, Krawinkler H (2000) Progress and challenges models. Bull Seismol Soc Am 103(2A):1103–1116
in seismic performance assessment. PEER Center Lin T, Haselton CB, Baker JW (2013b) Conditional spec-
News 3(2):1–3 trum‐based ground motion selection. Part I: hazard
Gulerce Z, Abrahamson NA (2011) Site-specific design consistency for risk‐based assessments. Earthq Eng
spectra for vertical ground motion. Earthq Spectra Struct Dyn 42(12):1847–1865
27(4):1023–1047 Lin T, Haselton CB, Baker JW (2013c) Conditional spec-
Haselton CB, Deierlein GG (2007) Assessing seismic trum‐based ground motion selection. Part II: intensity‐
collapse safety of modern reinforced concrete frame based assessments and evaluation of alternative target
buildings. Technical report, 2007/08. Pacific Earth- spectra. Earthq Eng Struct Dyn 42(12):1867–1884
quake Engineering Research Center, University of Loth C, Baker JW (2013) Reliability-based calibration of
California, Berkeley design seismic response spectra and structural accep-
Jayaram N, Baker JW (2008) Statistical tests of the joint tance criteria. In: 11th international conference on
distribution of spectral acceleration values. Bull structural safety and reliability. CRC Press, New
Seismol Soc Am 98(5):2213–2243 York, 8 p
Jayaram N, Lin T, Baker JW (2011) A computationally McGuire RK (1995) Probabilistic seismic hazard analysis
efficient ground-motion selection algorithm for and design earthquakes: closing the loop. Bull Seismol
matching a target response spectrum mean and vari- Soc Am 85(5):1275–1284
ance. Earthq Spectra 27(3):797–815 Naeim F, Lew M (1995) On the use of design spectrum
Katsanos EI, Sextos AG, Manolis GD (2010) Selection of compatible time histories. Earthq Spectra 11(1):111–127
earthquake ground motion records: a state-of-the-art NIST (2011) Selecting and scaling earthquake ground
review from a structural engineering perspective. Soil motions for performing response-history analyses.
Dyn Earthq Eng 30(4):157–169 NIST GCR 11-917-15 (ATC 82). Prepared by the
Lin T, Baker JW (2011) Probabilistic seismic hazard NEHRP Consultants Joint Venture for the National
deaggregation of ground motion prediction models. Institute of Standards and Technology, Gaithersburg,
In: 5th international conference on earthquake geo- 256 p
technical engineering, Santiago, 12 p Wang G (2011) A ground motion selection and modifica-
Lin T, Baker JW (2013) Introducing adaptive incremental tion method capturing response spectrum characteris-
dynamic analysis: a new tool or linking ground motion tics and variability of scenario earthquakes. Soil Dyn
selection and structural response assessment. In: 11th Earthq Eng 31(4):611–625
D
1,500 people were injured making this event the availability during the crisis event even
deadliest earthquake to hit Italy since 1980. The though the site affected by the crisis event is
list of earthquakes and their damages that covered by clouds or if the satellite is passing
occurred globally can be continued. Indeed, during night.
according to the United States Geological Survey Since 2008 a new generation of SAR missions
(available online: http://earthquake.usgs.gov/) (i.e., TerraSAR-X, TanDEM-X, and COSMO-
records, the frequency of such catastrophic SkyMed) has regularly acquired data over Earth.
events is high: 332 earthquakes with an MMS of These systems can perform acquisitions using
7 or higher have been reported in the last two high-resolution modes called StripMap
20 years. (SM) and SpotLight (SL). SM is characterized
In order to understand and possibly mitigate by a wide coverage (around 1,500 Km2) with a
the impact of such events on the citizenry, high resolution (3–5 m), whereas SL is charac-
several governmental initiatives and research terized by a moderate coverage (around
activities have been carried out in the past 100 Km2) with a very high resolution (1 m).
(e.g., the international charter on space This means that the two SAR modes are comple-
and major disasters (available online: http:// mentary to each other and can be used for
www.disasterscharter.org/), UN-OSAT (available performing different duties during the emergency
online: http://www.unitar.org/unosat/), or the response phase. SM mode can be used for identi-
geohazard supersites and natural laboratories fying the blocks of the city more hit by the earth-
(available online: http://supersites.earthobservations. quake, whereas the SL images can be used to
org/) to tackle each of the phases of such events. identify detailed information about collapsed
In detail a catastrophic event can be partitioned in buildings. It is worth noting that, these SAR mis-
three characteristic phases: (1) before the event in sions are daily populating a huge archive of
which a system of risk assessment for early warn- multitemporal images. This archive can be
ing is essential, (2) the moment the event occurs employed to detect changes in the state of an
in which a disaster-alerting system is the main object or phenomenon by jointly observing the
priority, and (3) after the event in which an emer- data acquired at different times over the same
gency response with impact assessment has to be geographical area. This process, which is known
initiated. in the literature as change detection, has got an
Focusing on the third phase, among all the increase of interest from the scientific community
possible damages caused by natural disasters, in the last decade, and a relevant number of novel
the building damages contribute mostly to the techniques were developed to achieve this goal.
casualties (Dell’Acqua and Gamba 2012). None- Nonetheless, SAR images with a resolution rang-
theless, the identification of struck built-up areas ing from 5 m (referred in this work as high-
may be particularly complex because of both the resolution images) of the SM images up to 1 m
possible large extension of the affected region (referred in this work as very high-resolution
and the difficulties to access remote villages. In images) of the SL images are more heterogeneous
this scenario Earth observation (EO) data than old-generation sensor images, which have
acquired by remote-sensing satellites represent a resolution generally ranging from 10 to 30 m
useful tool to support decisions of civil protection (Soergel et al. 2006; Brenner and Roessing
authorities after the event. Indeed, satellites 2008). In high-resolution (HR) and very high-
equipped with optical or synthetic aperture radar resolution (VHR) images, objects that are consid-
(SAR) sensors can monitor large regions in a ered homogeneous from a semantic point of view
quick, synoptic, and uncensored manner. In par- (e.g., buildings) show a signature that is inhomo-
ticular, SAR systems present the advantage over geneous at high spatial resolution because of the
optical sensors to be independent from sun illu- scattering contributions from sub-objects (e.g.,
mination and to be relatively insensitive to atmo- facade and roof of a building). Furthermore, on
spheric weather conditions. This ensures the data the one hand the side-looking illumination
Damage Detection in Built-Up Areas Using SAR Images 475
required by SAR systems leads to phenomena analysis (Inglada and Mercier 2007) would fail
such as layover, shadow, and multipath signals to solve the problem since they are not able to
(Dong et al. 1997; Franceschetti et al. 2002), properly take into account the high geometrical
which are very marked in urban areas. On the detail of HR and VHR data. To achieve change
other hand, the appearance of a ground object detection in HR and VHR SAR images, the anal-
depends on radar system parameters (i.e., wave- ysis between the multitemporal images should
length, polarization, pulse length, incidence be performed at a higher conceptual level that
angle, look direction, etc.), surface feature prop- models the source of change from the prospec-
erties (e.g., dielectric constant), and environmen- tive of interactions with the incidence D
tal variables (e.g., groundwater content) (Xia and electromagnetic wave.
Henderson 1997). On top of these aspects, SAR Numerous studies have been recently
images are corrupted by speckle noise. presented in the literature that deal with the prob-
All these factors, within the context of the lem of recognition of changes in urban areas
multitemporal analysis, make the problem of using VHR SAR images (Matsuoka and
the detection of changes complex. Due to the Yamazaki 2004; Chini et al. 2009; Ehrlich
high resolution, a large set of possible changes et al. 2009; Brunner et al. 2010; Balz and Liao
with different semantic meanings and scales are 2010; Dekker 2011; Dell’Acqua et al. 2011;
detectable in multitemporal HR and VHR SAR Dong et al. 2011; Jin et al. 2011; Dong and
images. In general, each change may be associ- Shan 2013; Poulain et al. 2011; Tao et al. 2012;
ated with the cause of the change itself Brett and Guida 2013). In detail, Matsuoka and
(Bruzzone and Bovolo 2013) and may present Yamazaki (2004), Chini et al. (2009), Ehrlich
an extension that varies from the single pixel to a et al. (2009), Brunner et al. (2010), Balz and
relevant portion of the entire scene. For instance, Liao (2010), Dekker (2011), Dell’Acqua
it is possible to distinguish among changes due et al. (2011), Dong et al. (2011), Jin
to the anthropogenic activity, the phenological et al. (2011), and Dong and Shan (2013) are
evolution of the vegetation, and the natural focused on the detection of earthquake damages,
disaster effects. Depending on the application, Poulain et al. (2011) are focused on the building
some of these may be of interest to the end users, databases’ updating, and Tao et al. (2012) address
whereas others may not (Bovolo et al. 2013). In the detection of changes due to the urban evolu-
addition, external factors, such as different con- tion. The strategies that are exploited in these
tent of water on the ground due to different works include post-event supervised analysis
weather conditions, may affect the local back- (Ehrlich et al. 2009; Balz and Liao 2010; Dong
scattering behaviors at the two considered dates et al. 2011; Jin et al. 2011; Dong and Shan 2013),
also in absence of any change. Therefore, the joint use of optical and SAR data (Chini
same object may show different values of back- et al. 2009; Brunner et al. 2010), joint use of
scattering even though it is not affected by a LiDAR and SAR data (Tao et al. 2012),
relevant change. This invalidates the assumption unsupervised detection of damages at the level
that two SAR images acquired on the same geo- of aggregated blocks (Matsuoka and Yamazaki
graphical area at different times are similar to 2004; Dekker 2011), or GIS polygons
each other except for the presence of changes (Dell’Acqua et al. 2011). Despite the great inter-
occurred on the ground, which is often est, the only work that addresses the problem of
considered for medium-resolution SAR images. building change detection using multitemporal
Thus, the use of standard pixel-based change VHR SAR data in an unsupervised way is
detection techniques (Bazi et al. 2005) is not presented in (Brett and Guida 2013). In detail,
possible as they would be affected by a large the authors present an approach to the detection
amount of false alarms. Given the complexity of damaged structures in urban areas from VHR
of the problem, also standard context-based SAR images, which is based on the
techniques based on a local neighborhood multitemporal detection of the double-bounce
476 Damage Detection in Built-Up Areas Using SAR Images
Satellite track
La
Antenna
Le q
D
Azimuth direction
hS
Nadir
Radar footprint
Swath
Damage Detection in Built-Up Areas Using SAR Images, Fig. 1 Side-looking geometry of SAR systems (Courtesy
of Dominik Brunner)
a b
Damage Detection in Built-Up Areas Using SAR Images, Fig. 2 SAR operation modes. (a) StripMap mode. (b)
SpotLight mode (Courtesy of Dominik Brunner)
radar antenna points along a fixed direction with a resolution that varies between 1.5 and 5 m in
respect to the flight path. As the platform moves, ground range and about 3 m in azimuth
the resulting image covers a strip of the ground depending on the platform and acquisition
surface. The extension of the imaged area in parameters. In the SL mode (see Fig. 2b) the
azimuth direction is limited only by the power azimuth resolution is improved by increasing
and free memory available onboard of the satel- the length of the synthetic antenna. This is done
lite, whereas the resolution in range direction is by steering the antenna beam so that to illuminate
constrained by half of the actual antenna length. the same area on the ground during the acquisi-
Generally speaking for the new SAR generation, tion. Nonetheless, the improvement in the azi-
the standard coverage is of about 1,500 Km2 with muth resolution is at the expense of the spatial
478 Damage Detection in Built-Up Areas Using SAR Images
coverage. The standard coverage is in fact of model that does not take into account the building
about 100 Km2 (instead of 1,500 Km2) with a features such as windows, eaves, ridges, railings,
resolution that varies between 1 and 3.5 m and so on. This is equivalent to analyze the image
(instead of 1.5 to 5 m) in ground range and 1 m at the scale level comparable with the building,
(instead of 3 m) in azimuth depending on the and it allows us to derive the appearance of a
platform. building without losing generality. The acquisi-
In the rest of the section, we analyze the back- tion geometry of a SAR system (Fig. 3a) is char-
scattering mechanism with respect to the varia- acterized by two parameters: the incidence angle
tion of the acquisition geometry when isolated y (i.e., the angle defined by the incident radar
buildings with rectangular layout are sensed by beam and the normal to the intercepting surface)
HR or VHR SAR (section “Building Backscat- that is generally included between 20 and 55
tering Mechanism in Single-Detected HR and and the aspect angle 0 y < 90 (i.e., the angle
VHR SAR Images”). This analysis is used to between the azimuth direction and the orientation
interpret multitemporal HR and VHR SAR of an object in the horizontal plane). For the sake
images in order to study and model the effects of argument, let us first assume that the SAR
associated with new/demolished building sensor illuminates the section A-A (light-blue
(section “Behavior of Changed Buildings in area in Fig. 3a) of the building from a fixed
Multitemporal HR and VHR SAR Images”). position in the azimuth. Figure 3b shows in a
This study gives the base for the development of qualitative way the amplitude of the backscatter-
the proposed approach. ing projected in ground range and slant range of
A-A. Following (Brunner et al. 2010) the main
Building Backscattering Mechanism in contributions that can be identified are the return
Single-Detected HR and VHR SAR Images a from the ground, the double-bounce effect
In the literature several works have been presented b caused by the dihedral reflector formed by the
that analyzed the scattering mechanism for differ- building wall and the ground, the backscattering
ent building models in order to derive their appear- c from the front wall, the returns d from the
ance in SAR images (Bennett and Blacknell 2003; building roof, and the shadow area e (see
Thiele et al. 2007; Brunner et al. 2010). These Fig. 3b). As one can observe the contributions
analyses make use of the geometrical optics (ray from the ground, the wall and the roof are
theory) approximation in order to model the elec- summed up into an area in front of the building
tromagnetic scattering. This has the power to be (a + c + d), which appears brighter than each of
intuitive and geometrically accurate even though these contributions taken singularly because of
the electromagnetic interactions and the transmis- their superposition. This phenomenon, called lay-
sion configurations of the antenna are not taken over, occurs when SAR sensors are imaging a
into account. In geometrical optics the wave prop- surface with a slope steeper than the incidence
agation is described by rays, which are modeled as angle y, such as the wall of a building. It is worth
lines perpendicular to the wavefronts, that may be noting that the length of the layover, the roof
reflected, absorbed, or split at the interface contribution, and the shadow depends on the
between two different media. By knowing the width (w) and the height (h) of the section A-A
geometry of acquisition of the SAR system and and the incidence angle y.
by exploiting trigonometric functions, it is possible In order to derive the appearance of the whole
to simulate the appearance of a building in HR and building in a VHR SAR image, it is necessary to
VHR SAR images. perform the aforementioned analysis for all the
Let us consider an isolated flat-roof building sections parallel to A-A that form the building.
with dimensions w1 w2 h illuminated by a Nevertheless, as the width of each parallel section
SAR sensor that is moving along the azimuth changes, the length of each mentioned contribu-
direction and is illuminating the building from tion changes accordingly. Three cases can be
the left (Fig. 3a). Let us consider a building observed on the basis of the relationship between
Damage Detection in Built-Up Areas Using SAR Images 479
a azimuth direction
e
b tud
pli
am
sla
nt
a
r an
ge
A-A
a+
h
c+
b
D
d
w
d
w1 w2 d
c
e
a b a
w
a
1 e
h
azimuth direction
w
A A
w
amplitude
b
2
w
a+c+d
d
a e a
range direction ground range
c d e
a+c
a+c
azimuth direction
b d e
a+c+d
b d e
a+c+d
a+c
a+c
range direction
Damage Detection in Built-Up Areas Using SAR backscattering is reported in both slant and ground
Images, Fig. 3 Example of backscattering mechanism range). (c) The whole extension of the building (the back-
for a flat-roof building. (a) Simplified model of a flat-roof scattering is in slant-range geometry). Different gray
building oriented with a given aspect angle y. The SAR levels represent different amplitudes. Example of (d) a
sensor is moving on the left of the building. Theoretical real flat-roof building in Trento (Italy) acquired at 1 m
scattering model for a given incidence angle 6 in the case resolution by TerraSAR-X in slant range (SAR illumina-
the sensor is illuminating. (b) Only the section A-A of the tion is from the left courtesy of DLR). (e) The same
building from a fixed position in the azimuth (the building in an optical image (GeoEye – Google)
the width w of the section A-A and the limit value shows the case in which w > wb. As the width
wb defined as (Bennett and Blacknell 2003; of the section w decreases, the layover and
Soergel et al. 2003): shadow result larger, whereas the backscattering
from the roof will be absorbed in the layover.
h This behavior is valid up to the limit condition
wb ¼ (1) w = wb for which the whole roof contribution is
tan y
sensed in the layover area before the double
where h and y are the height and the incidence bounce. In the case w < wb two contributions to
angle of the building, respectively. Figure 3a layover areas can be distinguished: one due to the
480 Damage Detection in Built-Up Areas Using SAR Images
ground, the front wall, and the roof (a + c + d) The same analysis described in this section for
and one due to the backscattering from the flat-roof building can be conducted for other
ground and the front wall of the building (a + building models, e.g., gable-roof buildings
c). Figure 3c shows the building radar footprint (Brunner et al. 2010). The outcomes of such a
generated from the building of Fig. 3a. It is a study can be summarized as follows: the footprint
convex polygon made up of (i) a bright of any type of isolated buildings with a rectangu-
L-shaped region due to the layover, (ii) a dark lar base is given by a specific convex pattern
L-shaped region due to the shadow, (iii) a made up of a bright area (due to layover and
bright L-shaped line due to the double bounce, double-bounce effects) followed by a dark area
(iv) a rectangular region due to the direct return (due to the shadow effect). These features may
from the roof, and (v) two bright triangular have different thickness, shape, and internal var-
regions due to the layover of only ground and iability of backscattering on the basis of the con-
wall (a + c). It is worth noting that since in the sidered building structure and material.
acquisition phase the radar is moving, the shadow Nevertheless, they will systematically arise
casted by a ground object is moving as well when SAR systems are sensing an isolated build-
resulting in a blurring of the border of the shadow ing with an adequate resolution.
(Zhang et al. 2013; Groen et al. 2009) (see the
dotted area in Fig. 3c). By comparing Fig. 3c, d, it Behavior of Changed Buildings in
is possible to observe that the geometrical optics Multitemporal HR and VHR SAR Images
approximation can effectively describe the real The aim of this section is to analyze the behavior
behavior of scattering in VHR SAR images. of the radar backscattering in multitemporal
The same building may appear differently in images when a building changes by taking into
SAR images according to the value of the aspect account the single data analysis carried out in the
angle f. When f is approaching the limit values previous section. This analysis introduces the
f = 0 and f = 90 , the layover, the double basic concept on which the proposed approach
bounce, and the shadow change by approaching a is based. In order to properly illustrate the prob-
rectangular shape. A special attention has to be lem, let us focus on a building that fully disap-
given to the double-bounce line with respect to pears between two acquisitions. Let us assume to
the variation of f (Franceschetti et al. 2002; sample this situation acquiring one image when
Dong et al. 1997). In Ferro et al. (2011) an empir- (i) the building is standing and (ii) the building is
ical study on the relationship between the strength totally dismissed. Let us assume to use a VHR
of the double bounce and the aspect angle SAR system configured with the same geometri-
highlighted that the double-bounce contribution cal parameters (i.e., same incidence angle y and
drops off significantly if the aspect angle increases same azimuth path) for the two acquisitions. As
from f = 0 up to 10 , whereas it decays moder- in section “Building Backscattering Mechanism
ately for higher angles. y also affects the appear- in Single Detected HR and VHR SAR Images,”
ance of a building in VHR SAR images. By let us start the analysis by comparing the back-
reverting Eq. 1, it is possible to derive a limit scattering considering the illumination source
value yb that is equivalent to wb. Nonetheless, in fixed at a given point along the azimuth direction
urban areas the choice of y can be critical for two corresponding to section A-A (Fig. 1a) in the case
reasons: (i) the value of backscattering is related to the building is present. The backscattering profile
y, i.e., small value of y generates higher value of obtained for the standing building is reported in
backscattering and vice versa; and (ii) a large value Fig. 4a. As one can notice, the backscattering
of y generates long shadow areas and small lay- behavior is the same as the one obtained in the
over areas and vice versa. Thus, the backscatter previous section and reported in Fig. 3b. Whereas
has a dependence on the incidence angle, and there for the case of totally dismissed building, by
is potential for choosing optimum configurations assuming bare ground as depicted in Fig. 4b, the
for different applications. value of backscattering is approximately
Damage Detection in Built-Up Areas Using SAR Images 481
a b
c d
b a a f a
a
h e
amplitude
amplitude
b D
a+c+d
a d e a a f a
ground ground
range range
Damage Detection in Built-Up Areas Using SAR mechanism in the cases (a) a new building is built up (the
Images, Fig. 4 Building scattering mechanisms for a envelope of backscattering of bare soil is reported for
fixed illumination source corresponding to: (a) a section comparison in dotted line) and (b) a building is dismissed
of a building and (b) bare ground. The comparison of (the envelope of backscattering of the building in (a) is
Figures (a) and (b) describes the multitemporal scattering reported in dotted line)
constant. By comparing Fig. 4a, b, one can The pattern includes two regions that can be clas-
observe that the region with a high value of back- sified as: (i) area of increase of the value of back-
scattering due to the layover a + c + d, the double scattering and (ii) area of decrease in the value of
bounce b, and the roof contribution d decreases backscattering. The order of appearance of these
its value when the building disappears, whereas two regions along the near-to-far-range direction
the region hidden by the shadow e becomes vis- defines if the pattern is due to new buildings (see
ible to the line of sight of the radar and therefore Fig. 5a), i.e., the increase area (in this work
increases its value (the dashed line in Fig. 4a depicted conventionally in magenta) is closer to
represents the envelope of backscattering of the sensor than the area of decrease (in this work
bare ground). In other words, in the case of a depicted conventionally in green) or demolished
new building we are likely to observe a structured building (see Fig. 5b), i.e., the decrease area is
pattern made up of two regions having increase closer to the sensor than the increase area.
and decrease of the backscattering oriented from By taking into account the analysis done in
near-to-far range. Vice versa if we consider the section “Building Backscattering Mechanism in
case in which a building disappears between two Single Detected HR and VHR SAR Images” on
acquisitions, we expect that a structured pattern the radar signature of a general isolated building,
made up of two regions having decrease and it is possible to derive the appearance of the radar
increase in the backscattering values arises in footprint of any type of newly built-up or
near-to-far range (the dashed line in Fig. 4b rep- destroyed buildings in VHR SAR multitemporal
resents the envelope of the backscattering ampli- images. In detail, the change building radar foot-
tude of a building). print can be identified checking: (i) the presence
This specific multitemporal behavior obtained of both the regions of increase and decrease in
when the illumination source is fixed at a given backscattering, (ii) the proportion between the
position in azimuth can be used to retrieve the areas (in pixels) of the regions of increase and
appearance of new/demolished buildings in decrease in backscattering (the proportion
multitemporal VHR SAR images by repeating depends on the incidence angle of the acquisi-
the same analysis for all the sections parallel to tion), (iii) the equality between the lengths of the
A-A that form the scene. The obtained regions of increase and decrease in backscatter-
new/destroyed building radar footprint is made ing in the azimuth direction, and (iv) the align-
up of a pattern included in a convex polygon. ment of the regions of increase and decrease in
482 Damage Detection in Built-Up Areas Using SAR Images
a b
azimuth direction
azimuth direction
backscattering backscattering
decrease increase
Damage Detection in Built-Up Areas Using SAR in the value of backscattering (unchanged pixels are in
Images, Fig. 5 Building scattering mechanisms in white) obtained in the case of (a) new and (b) fully
multitemporal VHR SAR images (illumination source on demolished building
the left). Idealized map highlighting decrease and increase
backscattering with respect to the range direction, (post-event), respectively, which have the same
i.e., the barycenters of the two regions lay on the incidence angle and centered over the epicenter
line with the range (as the regions in the model of the event. The goal is to detect the areas
are not regularized, the barycenters were calcu- severely damaged by the event (i.e., hot spots)
lated by means of the geometric decomposition in order to quickly organize the rescue opera-
method). The order of appearance of the two tions. The most critical issue for the identification
regions of increase and decrease in backscatter- of the damages is related to the presence of many
ing value determines if the changed building possible kinds of changes on the ground. A single
radar footprint is due to a new or a destroyed site may include many kinds of change that show
building. significantly different characteristics in terms of
It is worth noting that the regions of change size, shape, and semantic meaning. Note that not
may have different thickness, shape, and variabil- all the changes are associated to the built-up site,
ity of backscattering change on the basis of the and thus, they are not relevant for the rescue
considered building structure and acquisition operation. The analysis of SM image allows the
geometry. Despite this variability, the pattern detection of large portions of changed areas. Here
will systematically arise when SAR systems are we refer them as hot spots. The obtained hot spots
sensing an isolated building with an adequate are then exploited to drive the selection of a
resolution (Bovolo et al. 2012). Hence, a method pre-crisis VHR SAR X11, . . ., XQ1 and the acquisi-
for the detection of the changed building based on tion of post-crisis VHR SAR images X12, . . ., XQ 1
this concept results to be robust to both the noise taken with the same incidence angle. Note that
and the uncertainty (i.e., the impossibly of model- the acquisition of the post-crisis images X12, . . .,
ing the reality precisely) of multitemporal VHR XQ2 should be scheduled on the basis of the
SAR data. detected hot spots. These images are used to
perform change detection at high spatial resolu-
tion with the goal to detect collapsed buildings
Proposed Approach to Damage with a high level of detail.
Detection Hence, the proposed approach is made up of
two main phases: (i) a coarse detection of the
Let us assume to analyze two HR amplitude areas of change (hot spots) and (ii) a fine detec-
SAR images I1 and I2 acquired in StripMap tion of the fully collapsed buildings inside each
mode before (pre-event) and after the event hot spot. Figures 6 and 7 show the block schemes
Damage Detection in Built-Up Areas Using SAR Images 483
D
Damage Detection in Built-Up Areas Using SAR Images, Fig. 6 Architecture of the proposed approach (Marin
et al. 2014)
Damage Detection in Built-Up Areas Using SAR Images, Fig. 7 Architecture of the proposed approach (Marin
et al. 2014)
From Ifil
LR it is possible to isolate the areas affected values of B results in the selection of more splits.
by a significant backscattering change and that The split-based selection allows defining a subset
present an extension equal or larger than the of pixels IS ¼ fILR jILR P s g in which the
resolution of the selected wavelet level N + 1. classes of change show a higher prior probability
Therefore, in order to detect the hot spots, the than in the whole image. Consequently, the sta-
selected level has to be comparable to the scale of tistical estimation of the parameters related to the
aggregated buildings’ or city blocks’ size. As the three probability density functions associated
analysis of SAR images acquired in SM mode with no change, increase, and decrease of back-
involves the processing of large multitemporal scattering between the two dates (i.e., xu, x+, x)
images, it is possible that the population of can be correctly derived and used to separate the
changed pixels is in sharp minority when consid- three classes. Here the thresholding method
ering Ifil
LR. This may affect the accuracy of the described in (Bazi et al. 2007) is adopted and
adopted automatic threshold selection technique. the Bayes decision rule for minimum error is
In order to address this issue, we use the applied to separate the three classes. To this end
thresholding method presented in (Bovolo and a statistical model for class distributions is
Bruzzone 2007). This method divides the consid- required together with an approach for class sta-
ered image into sub-images of a given size tistical parameter estimation. The Gaussian
(splits) and performs the threshold selection con- model and the well-known expectation-
sidering only the splits that present the highest maximization (EM) algorithm (Dempster
probability of containing changed splits. This et al. 1977) are employed to derive MSM.
selection allows the definition of a subset of splits It is worth noting that MSM does not report any
in which the class of change shows a higher prior information about the source of the changes.
probability than in the whole image. For our Indeed the changes may be associated to the crisis
purpose, ILR is split into a set of S sub-images event as well to other source of change such as
of a user-defined size. The choice of the split size phenological changes. By exploiting prior informa-
R SA is driven by the average extension of
SSM SM
tion (e.g., cadastral map) and focusing on changes
the expected changes. Once the size SSM R and SA
SM
related to built-up areas, the hot spots possibly
are defined, the splits are identified and selected related to the crisis event that are in urban areas
according to their probability to contain a signif- can be isolated. Nevertheless, among these changes
icant amount of changed pixels. The selection is some may be more critical than others. The severity
done according to the value of the variance s2s , of the change can be inferred considering the prior
s = 1, . . ., S, computed on each split. This is a information available to civil protection or
reasonable index for predicting the presence of exploiting the intensity and texture information of
changes in the log-ratio image since the residual the SAR images as proposed in (Dell’Acqua and
multiplicative noise of the ratio image becomes Gamba 2012; Dekker 2011). Once the most rele-
additive due to the log operator. The desired set vant hot spots are detected, an algorithm for solving
P s of splits with the highest probability to con- the “packing problem” of acquiring the minimum
tain changes is defined by selecting the splits that number of SL acquisitions X12, . . ., XQ
2 by covering
satisfy the following inequality: the highest number of critical hot spots can
be applied. Under the assumption the pre-event
s2s s2 þ Bss2 , s ¼ 1, . . . , S (3) X11, . . ., XQ
1 images have been already acquired,
when a crisis occurs the acquisition of X11, . . ., XQ
1
where s2 denotes the average variance of splits, can be effectively planned (if not already done) by
ss2 is the standard deviation of the variance of the outcome of this phase.
splits, and B > 0 is a constant. High values of
B result in the selection of only those splits with Detection of Fully Collapsed Buildings
high variance (i.e., the ones with the highest The aim of the second phase is to detect collapsed
probability to contain changes); vice versa, low buildings considering higher resolution data
Damage Detection in Built-Up Areas Using SAR Images 485
acquired in SpotLight mode over the hot spots is given by a pattern made up of increase and
detected in the previous phase. This is the most decrease of backscattering regions. The first
challenging part of this work and it requires stage of the proposed approach is thus devoted
advanced techniques for change detection. For to extract the areas that present both a significant
the sake of notation, and without any loss of increase or decrease in the backscattering value
validity in the following, we assume that Q = in SL (i.e., xu, x+, x). Nonetheless, at a resolu-
1 and thus that critical hot spots can be fully tion of a meter or less, the small objects that form
covered by means of a single couple of images the buildings such as windowsills or rain drains
X1 and X2 acquired with the same incidence are visible. This results in an inhomogeneous D
angle on the same geographical area at different signature of the building, which may affect also
times t1 and t2, respectively. Let O = {ou, Oc} be the appearance of new/destroyed buildings in
the set of classes of changes to be identified: ou multitemporal images rendering the regions of
represents the class of unchanged pixels, whereas increase and decrease far from being homoge-
Oc = {oc1, oc2, . . ., ocK} is a meta-class that neous. It is worth noting that working at the full
gathers all the K possible classes (kinds) of resolution of SM images (i.e., between 3 and 5 m)
change that may arise on the ground. One of the is also problematic due to the massive presence of
most critical issues dealing with this kind of sce- noise. Therefore, we propose to work at an opti-
nario is related to the presence of many kinds of mum scale level for representing buildings and
change on the ground. Nevertheless, in this work not subparts of them, derived from the VHR
we are interested to investigate an urban area with images. This can be used to identify the regions
the goal to detect collapsed buildings which usu- of increase and decrease of backscattering, which
ally represent the most critical kind of change in characterize the buildings, with the advantages of
an earthquake emergency. In detail, we consider (i) reducing the impact on the detection of small
K = 4 classes of change: (i) fully destroyed build- changes and thus reducing the false alarm rate
ings (oc1), (ii) new buildings (oc2), (iii) changes and (ii) making it possible a mitigation of the
that have a size comparable to a building but do speckle effect on the detection.
not present the typical pattern of full In order to work at the scale of a building,
new/demolished buildings (oc3), and (iv) all the we propose to build a multiscale representation
other changes that do not show a size comparable of the multitemporal information made up of
to a building and are therefore not related to N scale level. The (N 1)th resolution level,
changed buildings (oc4). In order to achieve this which represents the optimum scale level, is
classification, we introduce an approach made up selected according to the minimum size of the
of two stages: (i) identification of the areas building in the investigated scene. The multiscale
affected by changes in the backscattering at the representation can be obtained by applying sev-
scale of buildings and (ii) exploitation of the eral methods, e.g., Laplacian/Gaussian pyramid
backscattering models presented in section decomposition (Burt and Adelson 1983) or wave-
“Behavior of Changed Buildings in let transform (Mallat 1989). Here the
Multitemporal HR and VHR SAR Images” in two-dimensional discrete stationary wavelet
order to detect the classes Oc = {oc1, oc2, . . ., transform (2D-SWT) is exploited, which has the
oc4}. Each stage of the proposed method for SL advantage of avoiding the down-sampling after
images analysis is explained in detail in the fol- each convolution step (Nason and Silverman
lowing. The block scheme is reported in Fig. 7. 1995). In detail, we used the wavelet-based
As described in section “Behavior of Changed decomposition and reconstruction of the
Buildings in Multitemporal HR and VHR log-ratio image approach presented in (Bovolo
SAR Images,” under the hypothesis of using and Bruzzone 2005) for change detection in
a SAR sensor with a resolution comparable medium-resolution SAR images. The log-ratio
to the building size, the signature of isolated image XLR, which is derived as XLR = log X2/
new/demolished buildings in multitemporal data X1, efficiently points out the differences in
486 Damage Detection in Built-Up Areas Using SAR Images
8
>
> h
þ w1 sin y, if y tan 1 h
<
cos y w1
SSL ¼ (4)
R
>
> h
: w1 þ sin y þ hcosy Otherwise:
tan 1y
A ¼ w2
SSL (5) by evaluating the matching between the proprie-
ties of the building candidates with respect to the
Figure 8a reports an example of a change proprieties of new and destroyed building models
detection map Mopt SL obtained with the proposed described in section “Behavior of Changed Build-
method. As one can notice, in this example the ings in Multitemporal HR and VHR SAR
change detection map Mopt SL verifies the assump- Images.” In detail, this task is based on two
tions defined for the model presented in section steps: first, the areas of change candidate to be
“Building Backscattering Mechanism in Single associate to one of the classes oc1 (demolished
Detected HR and VHR SAR Images,” which are building), oc3 (new buildings), or oc3 (changes
(i) the detected regions of increase or decrease in that have a size comparable to a building but do
backscattering at the building scale are homoge- not present the typical pattern of full
neous and (ii) the change detection map presents a new/demolished buildings) are detected among
reduced number of changes smaller than the size all the backscattering changes highlighted in
of buildings, i.e., changes that are not related to Mopt
SL . The set of changed building candidates is
changed buildings (oc4) are filtered out. Hence, denoted by G = {g1, g2, . . ., gH}. Then the
from MoptSL it is possible to identify the building matching between the expected backscattering
radar footprints of new or destroyed buildings by behavior of changed buildings (presented in sec-
locating the pattern of increase (x+) and decrease tion “Behavior of Changed Buildings in
(x) of backscattering (building candidates) and Multitemporal HR and VHR SAR Images”) and
Damage Detection in Built-Up Areas Using SAR Images 487
a b
c
d
Damage Detection in Built-Up Areas Using SAR the regions of change d1, d2, . . ., d8 that form the candi-
Images, Fig. 8 Conceptual example of detection of a date. (c) Illustration of the parameters SI, Sd, lI, lD, and a,
destroyed building (illumination source on the left). (a) which are evaluated by the fuzzy rules described in Table 1
Mopt
SL . The pixels associated to the classes of backscatter- in order to check the matching between the candidate and
ing decrease (x) and increase (x+) are reported in green the expected pattern. (d) Identified demolished building
and magenta, respectively. The idealized dismissed build- footprint (red continuous line) compared with the actual
ing radar footprint is drawn in dashed line. (b) In gray the footprint (gray dashed line)
bounding box of changed building candidate gh containing
the characteristics of the changed building candi- The first step of the detection of individual
dates is evaluated. This is done by considering the collapsed building phase aims at detecting the
physical proprieties and the relation of the regions set of changed building candidates G = {g1, g2,
of change inside each candidate, i.e., D = {d1, d2, . . ., gH}. Ideally each changed building in Mopt is
. . ., dK}. In order to properly model the uncer- made up of regions of both increase and decrease
tainty inherent in the process (that can be due, for in backscattering, whose total extension is com-
example, to cluttered objects placed in front of parable with the expected size of buildings. One
buildings), fuzzy theory is used (Zadeh 1965). In simple and effective approach to identify the
Fig. 8 an illustrative example that depicts the changed building candidates g1, . . ., gH is to use
detection process of the proposed approach is a sliding window algorithm for each pixel in the
reported for the case of a demolished building. image. Pixels within the moving window are
This example will be used in the paper to better taken and used to compute the detector output.
illustrate the proposed approach. Figure 8b–d In this work, the window is moved in Mopt from
illustrates the process of the proposed building the left to the right by one pixel, and the number
change detection stage for the case of a of changed pixels (labeled both as increase x+ or
demolished building. In the following each step decrease x) is counted. The process is repeated
is described in detail. over five moving windows: four rectangular
488 Damage Detection in Built-Up Areas Using SAR Images
Damage Detection in
Built-Up Areas Using
SAR Images,
Fig. 9 Windows used to
derive the changed building
candidates
moving windows showing different directions values when the sliding window contains a small
and a square window having the same area of amount of changes or the windows do not match
the rectangular ones. This choice allows us to the orientation of the change. Therefore, it is
properly capture the most of the building orien- possible to detect the changed areas that can be
tations. The detector output is given by the win- associated in terms of size to the expected radar
dow that results in the maximum value. The five footprint of changed buildings by selecting the
windows are depicted in Fig. 9. The size of the areas of C exceeding a certain threshold TC, thus
windows z1 z2 is chosen according to the min- obtaining the image C as follows:
imum size of the buildings in the considered
scene in order to nminimize missed oalarms. 1, if C T C
C (7)
Hence, let Moptb ¼ Mopt jMopt Wbi, j be the 0, otherwise:
W i, j
b
set of pixels of Mopt SL included in the window Wi,j
TC is selected according to the expected min-
centered at the pixel (i, j)i = 1, . . ., I; j = 1, . . ., J,
imum size of buildings z1 z2 in the investigated
with b = 1, . . ., 5 indicating the different win-
scene. The map C represents the areas containing
dows. Let C be the image with size I J that
changes with size comparable or bigger than the
reports for each pixel Ci,j an index of the size of
minimum building size and therefore they belong
the changes. Ci,j is computed as the maximum on
to one of the classes oc1, oc2, oc3. As one can
the five windows as follows:
notice, the pixels belonging to oc4 are implicitly
identified in this operation. From the map C the
opt
Ci, j ¼ arg max M b x _ M b x þ opt
set of changed building candidates G = {g1, g2,
Wi, j Wi, j
b 1, 2, ..., 5
. . ., gH} is extracted by calculating the connected
8 i ¼ 1, . . . , I; j ¼ 1, . . . , J components considering an 8-connected neigh-
(6) borhood. This is done by means of a flood-fill
algorithm (Heckbert 1994). Each candidate
where || indicates the cardinality of a set. The gh(h = 1, . . ., H) contains a subset of regions
obtained index image C exhibits relatively high D = {d1, d2, . . ., dK} labeled as xu, x+ or x.
values when the sliding window contains a large The information associated to the kind of change
amount of changes oriented in the same direction is derived considering Mopt SL in the areas delimited
of the moving window. C exhibits relatively low by the region gh. Figure 8b shows an example of
Damage Detection in Built-Up Areas Using SAR Images 489
candidate gh that contains dk(k = 1, . . ., 8) regions presence of d+ and d inside the candidate gh.
for the case of a demolished building. In general This is done by means of a crisp membership
the number of regions, k, may vary from 1 to function mp that takes value 0 or 1 (mp(p) =
K depending on the size and proximity of the {0, 1}) depending on whether regions of increase
changes. For the sake of clarity, in the next step and decrease are simultaneously present
of the individual collapsed building detection mp(p) = 1 or not mp(p) = 0.
procedure, we assume that only two regions
belonging to increase (i.e., d+ = dk x+) or Proportionality of Areas
decrease (i.e., d = dk x) occur inside a This rule aims at verifying that the area d+ is not
D
candidate. If more than two regions are present, prevailing with respect to the area of d and vice
all the combinations among the regions of versa. The mathematical modeling of this rule
increase and decrease are automatically analyzed depends on the parameters of acquisition and the
by the proposed approach and the most reliable(s) size of the building as discussed in section
selected. “Building Backscattering Mechanism in Single
In order to properly classify each collapsed Detected HR and VHR SAR Images.” It can be
building candidate according to the classes oc1, effectively represented by a sigmoid member-
oc2, and oc3, the physical characteristics and the ship function that evaluates the attribute
spatial arrangement of the pattern formed by the rs = min{sI/sD, sD/sI}, where SI and sD indicate
regions dk(h = 1, . . ., K) have to match with the the areas of d+, i.e., SI = |d+|, and d, i.e.,
four characteristics of the models of destroyed SD = |d|, respectively. The sigmoid function is
building discussed in section “Behavior of defined as follows:
Changed Buildings in Multitemporal HR and
VHR SAR Images.” The matching is tested 1
mrs ðr s Þ ¼ : (8)
exploiting four fuzzy rules called here: 1þ ea1 ðrs b1 Þ
(i) completeness, (ii) proportionality of areas,
(iii) equivalence of lengths, and (iv) alignment. The constant a1 > 0 tunes the slope of function
These rules aim at associating a grade of mem- and the constant b1 > 0 locates the center of
bership to the changed building candidate for the function. Equation 8 returns values in [0,1].
each of specific performed test. On the basis of Figure 11 shows an example of sigmoid function.
the aggregate membership, the final classification As one can notice this is in accordance with what
decision is taken. In the following each rule is it is expected: the smaller rs is, the smaller the
presented in detail. It is worth noting that the membership grade of the candidate.
theoretical model of a new building is just alike
to the model of a destroyed building but the Equivalence of Lengths
orientation of the pattern is reversed (see section This rule aims at checking that the lengths of the
“Behavior of Changed Buildings in regions d+ and d in the azimuth direction are
Multitemporal HR and VHR SAR Images”). equivalent. This is done by identifying the
Therefore, the proposed method can detect col- extrema of the regions d+ and d. To test the
lapsed buildings as well as new constructed reliability of a candidate with respect to this cri-
buildings. Nonetheless, since the focus of this terion, the attribute rl = min{lI/lD, lD/lI} (where lI
work is on the detection of collapsed building and lD are the lengths in azimuth direction of d+
due to earthquakes, we concentrate in the follow- and d, respectively) is used in the sigmoid mem-
ing on this case. bership function defined in Eq. 8 with appropriate
parameters a2 > 0 and b2 > 0. Since the farther rl
Completeness is from 1 the smaller the membership grade of the
A pair of regions of increase and decrease is candidate to be associated to a new/destroyed
strictly needed to identify a destroyed building building, the slope of the sigmoid is steep and
(oc1). The first rule evaluates the simultaneous the center moved toward 1. Figure 8c illustrates
490 Damage Detection in Built-Up Areas Using SAR Images
a b c
Damage Detection in Built-Up Areas Using SAR respectively. (b) In gray the changed building candidate
Images, Fig. 10 Example of a candidate gh that does gh containing the regions d1 and d2, which form the can-
not respect the equivalence of lengths. (a) Mopt
SL . The pixels didate. (c) Analysis of psychical proprieties and relations
associated to backscattering decrease (x) and increase of the regions d+ and d
(x+) classes are reported in green and magenta,
Damage Detection in 1
Built-Up Areas Using
SAR Images, 0.9
Fig. 11 Example of
sigmoid function mrs ðr s Þ 0.8
defined according to Eq. 8.
0.7
The slope of the sigmoid is
tuned by the parameter a1. 0.6
b1 defines instead the center
s
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
rs
an example of correct alignment between d+ and sigmoid Eq. 8. a ranges from [p/2, p/2). As
d, whereas Fig. 10c depicts a case in which this expected, by imposing a3 < 0 the larger is |a|
rule is not satisfied. the smaller is the membership grade of the can-
didate to be associated to a new/destroyed build-
Alignment ing. Figure 8c illustrates an example of correct
This rule is devoted to test the alignment of the orientation of the pattern increase/decrease,
entire pattern increase/decrease (which is made whereas Fig. 12c depicts a case in which this
up of regions d+ and d) with respect to the range rule is not satisfied.
direction. The alignment of the centroids of the A summary reporting the description of each
regions of increase and decrease is considered, rule, the parameters, and the membership func-
which is expected to lay on the same line. In tions used to represent the rules is reported in
detail, the absolute value of the angle a defined Table 1.
as the angle included by the line that connects the The final decision on the label of the candidate
centroids of the regions d+ and d and the SAR is done according to the aggregated membership
range direction is used in the equation of the grade U, which is calculated by combining the
Damage Detection in Built-Up Areas Using SAR Images 491
Damage Detection in a b c
Built-Up Areas Using
SAR Images,
Fig. 12 Example of a
candidate gh that does not
respect the fourth rule. (a)
Mopt
SL . The pixels associated
to backscattering decrease
(x) and increase (x+)
classes are reported in
green and magenta, D
respectively. (b) In gray the
changed building candidate
gh containing the regions d1
and d2, which form the
candidate. (c) Analysis of
psychical proprieties and
relations of the regions d+
and d
Damage Detection in Built-Up Areas Using SAR Images, Table 1 Fuzzy rules used to classify gh as new building
(oc1), demolished buildings (oc2), or change comparable to the size of a building but not related to fully new/demolished
buildings (oc3)
Rule Parameter Membership function
1. Presence of both regions d+ and d (completeness) p mp(p) = {0, 1}
2. Proportionality of areas of d+ and d (proportionality of the areas) r s ¼ min ssDI , ssDI mrs ðr s Þ ¼ a11ðrs b1 Þ
1þe
3. Equivalence of the lengths of d+ and d along the azimuth direction r l ¼ min llDI , llDI mrl ðr l Þ ¼ 1
1þea2 ðrl b2 Þ
(equivalence of the lengths)
4. Alignment of the barycenters of the regions d+ and d (alignment) |a| mjaj ðjajÞ ¼ 1
1þea3 ðjajb3 Þ
four membership functions. Among the fuzzy factors in Eq. 9. Nonetheless, the value of
aggregation methods presented in the literature M obtained for each of these classes is
in this work, the Larsen product implication representing a grade of reliability of the detection
(Larsen 1980) is used, i.e., that can be used by the final users.
For each changed building candidate recog-
M ¼ mp mrs mrl ma (9) nized as demolished building, the 8-extrema
points (i.e., from top left to bottom right) of
In this way, one low value among the individ- both the regions d+ and d are calculated and
ual membership grades yields the final result to used to find the smallest convex polygon that
be low regardless of the other values. The candi- contains the two regions, i.e., the convex hull
dates that present a value of M greater than a (see Fig. 8d). The computation of the convex
user-defined threshold TM are selected as fully hull is done using the Quickhull algorithm(Bar-
demolished (oc1) if d is appearing before than ber et al. 1996). This allows a generation of an
d+ in range direction. The others are classified as envelope of a dismissed building radar footprint
oc3 (i.e., they are not fully demolished or new as much as similar to the ideal expected one.
buildings). It is worth noting that TM will be a Once the building footprint candidate is classified
small value since the algebraic product of value a demolished building, reverting the equations 4
smaller than 0 results in value smaller than the and 5, it is possible to obtain the actual size of the
492 Damage Detection in Built-Up Areas Using SAR Images
building affected by the change, which is of the parameters to give as input to the proposed
expressed in terms of w1[m] w1[m] h[m] approach requires very basic prior information
(see Fig. 3a). and is not critical.
Damage Detection in
Built-Up Areas Using
SAR Images,
Fig. 13 L’Aquila: (a) map
of the area covered by the
SM COSMO-SkyMed
images (Google). (b) RGB
multitemporal composition
of SM COSMO-SkyMed
images (red channel, 21/04/
2009, green channel, D
04/05/2009, blue channel,
21/04/2009). (c) Hot spots
map MSM. In red and blue
the most damaged regions.
The size of these regions
corresponds to the coverage
guarantee by SL images
494 Damage Detection in Built-Up Areas Using SAR Images
(red channel, 21/04/2009, green channel, 04/05/ decomposition Daubechies filter of order 4 is
2009, blue channel, 21/04/2009) in which pixels given by the following coefficient set:
with an increase in the value of backscattering
appear in magenta tone, pixels with a decrease in f0:0105974, 0:0328830, 0:0308414, 0:187035,
the value of backscattering appear in green tone, 0:0279838, 0:630881, 0:714847, 0:230378g
and unchanged pixel in gray scale. It is worth
noting that a relevant number of changes are The finite impulse response of the high-pass filter
present between the two acquisitions (see for the decomposition step can be computed by
Fig. 13b). This increases the complexity of the satisfying the properties of the quadrature mirror
detection of collapsed buildings. filters. The level (N 1)th = 3rd, which gives a
In order to apply the proposed method, the two resolution of approximately 8 m, was selected as
SM images were calibrated and coregistered the optimum one. This allows to observe radar
(with sub-pixel accuracy), and the log-ratio footprints of buildings as small as 8 by 8 m, which
image ILR calculated. ILR was filtered by is compatible with the minimum size of building
applying recursively a 2D-SWT transformation footprints in the considered scene. X3LR was
and reconstructing considering only the low-pass thresholded according to the unsupervised split-
information at the 4th level. From Ifil LR the hot based procedure described in section “Detection
spots of changes were derived by applying the of Damages (Hot-Spots).” The split size was cal-
R
split-based thresholding. The size of splits SSM culated starting from the average building size,
A = 64 64 pixels (corresponding to 160
SSM which was estimated to be w1 w2 h = 25
160 m2 on the ground) was selected according to 20 15 m3. By substituting these values in Eq. 4
average size of expected changes in the test site. SSR was obtained to be equal to 49.50 m in slant-
The obtained map is reported in Fig. 13c. From range geometry. Since the CSK images are
MSM and the cadastral map Q = 4, areas with ground projected, SSR was projected in ground
extension comparable to SL images were selected range considering the reference incidence angle,
to be further analyzed in order to detect the pres- i.e., 58 obtaining in this way SGR = 58.36 m.
ence of collapsed buildings. Among these areas From the calculation and taking into account a
we selected a representative case (red box in pixel spacing in a range of 0.5 m, SR results to be
Fig. 13c) to be used in the following in order to equal to 120 pixels. By applying Eq. 5 and con-
effectively illustrate how the proposed method sidering a pixel spacing in azimuth of 0.5 m, SA
for the detection of collapsed buildings works. was estimated to be equal to 40 pixels. It is worth
Indeed, the considered area presents a relevant noting that the split-based thresholding technique
number of collapsed buildings, differently from has a relative high tolerance to the selection of the
the other areas that present changes due to the values of SR and SA. This was proven by a sensi-
construction of refugee camps. tivity analysis in which SA was ranged from 20 to
As described in the methodological part, the 60 pixels and SR from 80 to 170 pixels. Despite the
proposed approach for the identification of col- large range (about 1/3 of the optimal selected
lapsed buildings firstly calculates the log-ratio size), the error on the estimation of the thresholds
image XLR. This was computed from the two is small. Moreover, in all the cases the generated
calibrated and coregistered CSK SL images X1 maps did not present any critical behavior and
and X2. In detail, the coregistration was resulted in a very similar final detection of
performed with sub-pixel accuracy, and the changed buildings. The obtained CD map Mopt
radar brightness calibration was applied in order highlighting the three classes xu, x+, x is
to render the two images comparable. From XLR reported in Fig. 14c. The detector described by
the set XMS = {X0LR, . . ., XnLR, . . ., XN1
LR } was Eq. 6 was then applied to Mopt in order to obtain
computed by applying sequentially the 2D-SWT the image C. With regard to the considered scene,
and the 2D-ISWT with an 8-length Daubechies it is expected that the minimum size of building
filter. The impulse response of low-pass footprint is 40 20 pixels. Therefore, the set of
Damage Detection in Built-Up Areas Using SAR Images 495
5 windows reported in Fig. 9 were set to size z1 considered area in order to limit the missed
z2 = 40 20 pixels in order to derive C (see alarms. Each of the candidate areas was analyzed
Fig. 14d). A set of changed areas G = {g1, . . ., in order to detect the changed buildings. To this
g38} compatible with building footprint changes end the four fuzzy rules were automatically eval-
were derived by thresholding C with uated for each candidate gh, (h = 1, . . ., 38).
TC = 160 (which corresponds to 20 % of maxi- Finally, for each building candidate that presents
mum value of C) and by applying a flood-fill an aggregated membership TM greater than
algorithm. It is worth noting that both the window 0.125, the building radar footprint was approxi-
size and the threshold value were selected mated by the convex hull containing the candi- D
according to minimum size of buildings in the date. It is worth noting that TM was selected in
Damage Detection in Built-Up Areas Using SAR Actual range and the azimuth directions are reported in
Images, Fig. 14 L’Aquila (Italy) dataset: (a) orthophoto white arrows. (c) Change detection map MoptSL obtained at
acquired few days after the earthquake on April 2009 the level N 1 = 3. (d) Map showing the index of the size
(Geoportale Abruzzo) and (b) RGB multitemporal com- of changes C. (e) Changed building map overlapped to the
position of SpotLight COSMO-SkyMed images (R, RGB multitemporal composition of CSK data. (f) Cadas-
21/04/2009, G, 04/05/2009, B, 21/04/2009). The images tral map of the area (destroyed buildings are reported in
are geo-projected according to the reference ellipsoid. red)
order to minimize the missed alarms. This is done Damage Detection in Built-Up Areas Using SAR
by taking into account the definition of the mem- Images, Table 2 Parameters used in the experiments
carried out on L’Aquila dataset
bership functions and the aggregation strategy
(i.e., algebraic product). In the present experi- Parameter Value
ment mrs = mrl = ma = 0.5 were selected as limit N1 3
case, which results in an aggregate membership SR 120 pixels
SA 40 pixels
function M = 0.125. Figure 14e shows the final
z1 40 pixels
result. In the specific case, all the reconstructed
z2 20 pixels
building radar footprints have an aggregate mem-
TC 160 pixels
bership function greater than 0.615. Table 2a
a1,b1 10, 0.3
reports the settings of the parameters used for a2,b2 10, 0.5
the dataset of L’Aquila. a3,b3 10, p3
By comparing the changed building map with TM > 0.125
the false color composition of the multitemporal
images (Fig. 14e), one can notice that the pro-
posed approach can effectively detect the back-
Damage Detection in Built-Up Areas Using SAR
scattering changes associated to disappeared Images, Table 3 Performance of the proposed approach
buildings as described in section “Behavior of for L’Aquila dataset in terms of number of detected
Changed Buildings in Multitemporal HR and destroyed buildings, missed alarms, and false alarms
VHR SAR Images,” which patterns present a Total Detected
decrease and increase in backscattering oriented number of destroyed Missed False
near-to-far range (see Fig. 14e). From a quantita- buildings (oc1) destroyed destroyed
tive point of view, Table 3 reports the number of 200 6 0 1
Damage Detection in Built-Up Areas Using SAR Images 497
Damage Detection in
Built-Up Areas Using
SAR Images,
Fig. 15 (a–d) Zoom of the
6 buildings destroyed and
correctly detected by the
proposed method (Bing
Maps, Microsoft
Corporation). See Fig. 14
(f) for the correspondence
of the buildings on the map. D
As one can notice the
building was totally
demolished after the
earthquake
building radar footprints of demolished building identified as totally destroyed (see Fig. 15e) even
detected by the proposed approach. Even though though it was half collapsed on itself. Figure 14f
the dataset presents a relevant number of changes shows the positions of the destroyed buildings in
in the backscattering value, none of the actual the cadastral map obtained by transforming the
demolished buildings (see Fig. 15a–d) were geometry of SAR images to the geo-referenced
misdetected, and one of the building was wrongly cadastral map. The analysis confirms the high
498 Damage Detection in Built-Up Areas Using SAR Images
accuracy of the proposed technique. It is worth and a de-noising operation based on the wavelet
noting that even though the density of buildings transformation was applied in order to obtain the
in the considered site is high, the proposed hot spots. By exploiting the available prior infor-
approach can effectively discriminate between mation about the scene, i.e., cadastral maps, it was
demolished and standing buildings. This points possible to define the regions to be further ana-
out the high robustness of the proposed technique lyzed in order to detect the collapsed buildings.
to the possible deviations from the adopted ideal The analysis at the level of single building is
model of changes. A similar analysis has been conducted in the second phase of the proposed
performed for changes that are not associated approach by taking advantage of the theoretical
with buildings, i.e., oc3 (yellow areas in behavior of collapsed building in HR and VHR
Fig. 14d). The orthophotos point out that these SAR images. Hence, from this modeling it is
changes are not associated to fully destroyed known that destroyed buildings are identified as
buildings but to the post-earthquake activities, a pattern made up of an area of increase and
e.g., emergency housing units or to partially dam- decrease of backscattering with specific physical
aged buildings. This confirms the effectiveness of proprieties and a specific alignment. In order to
the proposed method also in the detection of extract the changes associated with increase and
changes different from demolished buildings. decrease in backscattering in the SL images, the
proposed approach makes use of a multiscale
representation of the multitemporal information.
Summary This allows a detection of changes at the optimal
building scale. This means that changes smaller
In this entry an approach for the rapid and accu- than the building scale are rejected, while
rate detection of damages in built-up areas after a changes related to building size are made homo-
catastrophic event using SAR data has been geneous. This information is used to identify the
presented. SAR systems onboard of new genera- candidates to be destroyed buildings. The build-
tion SAR missions are able to acquire images ing candidates are analyzed in order to properly
with a geometrical resolution that varies from detect the collapsed building by means of four
5 m up to 1 m using two different operational fuzzy rules. The fuzzy rules are formulated by
modes called StripMap and SpotLight. By taking into account the ideal backscattering mech-
increasing the resolution the maximum ground anisms that arise when a building appears or dis-
coverage guaranteed decrease accordingly. In appears between two acquisitions. The aggregated
this entry we have proposed to jointly exploit membership resulting from the application of the
these two acquisition modes in order to: fuzzy rules makes it possible to identify the class of
(i) quickly identify the areas severely affected each building candidate (i.e., collapsed building,
by the crisis event (i.e., hot spots) by analyzing new building, or general change with size compa-
large areas using data characterized by a high rable to the building size but not related to new or
geometrical resolution (3–5 m) and destroyed building). Moreover, the membership
(ii) accurately detect collapsed buildings inside value gives an indication of the reliability of the
each hot spot by analyzing small areas using data detected class. The proposed approach requires the
with very high geometrical resolution (1 m). tuning of some parameters that depend on the
In the first phase of the proposed approach, the approximated size of the buildings in the investi-
hot spots, which are areas with both a significant gated area. This is a prior information usually
increase or decrease in the backscattering value, available and easy to include in the processing. It
are detected from the SM images. Since the goal is worth remarking that after this tuning the
during this phase is to quickly identify the areas method is completely automatic.
severely affected by the crisis event and not per- The proposed approach was tested using SAR
form detection at a fine resolution, a comparison data related to the earthquake of L’Aquila (Italy) in
between the SM images with a log-ratio operator 2009 acquired by COSMO-SkyMed constellation.
Damage Detection in Built-Up Areas Using SAR Images 499
The proposed method demonstrated to be effec- Bazi Y, Bruzzone L, Melgani F (2007) Image thresholding
tive in detecting all the demolished buildings based on the EM algorithm and the generalized Gauss-
ian distribution. Pattern Recogn 40(2):619–634
with a high accuracy. In detail, only one building Bennett A, Blacknell D (2003) Infrastructure analysis
was misclassified over 200 present in the two from high resolution SAR and InSAR imagery. In:
datasets with no missed detection. This was pos- Remote sensing and data fusion over urban areas.
sible because the high robustness of the pro- 2nd GRSS/ISPRS joint workshop on, May 2003,
pp 230–235
posed method to deviations from the ideal Bovolo F, Bruzzone L (2005) A detail-preserving scale-
model considered for changed buildings. This driven approach to change detection in multitemporal
robustness is due to both the use of adequate SAR images. IEEE Trans Geosci Remote Sens D
fuzzy rules and the intrinsic reliability obtained 43(12):2963–2972
Bovolo F, Bruzzone L (2007) A split-based approach to
when working on the comparison of two images unsupervised change detection in large-size
rather than on the modeling of building in each multitemporal images: application to tsunami-damage
image. assessment. IEEE Trans Geosci Remote Sens
As future development of this work, we plan to 45(6):1658–1670
Bovolo F, Marin C, Bruzzone L (2012) A novel approach
test the proposed approach considering SAR to building change detection in very high resolution
dataset related to other earthquakes. Moreover, SAR images. In: Proceedings of SPIE, vol 8537, Bel-
we plan to engineer the proposed approach to be lingham, WA, pp 85 370Y–85 370Y–12
compliant with the cogent requirements of real Bovolo F, Marin C, Bruzzone L (2013) A hierarchical
approach to change detection in very high resolution
emergency scenarios. SAR images for surveillance applications. IEEE Trans
Geosci Remote Sens 51(4):2042–2054
Brenner A, Roessing L (2008) Radar imaging of urban
Cross-References areas by means of very high-resolution sar and inter-
ferometric sar. IEEE Trans Geosci Remote Sens
46(10):2971–2982
▶ Building Damage from Multi-resolution,
Brett P, Guida R (2013) Earthquake damage detection in
Object-Based, Classification Techniques urban areas using curvilinear features. IEEE Trans
▶ Earthquake Damage Assessment from VHR Geosci Remote Sens 51(9):4877–4884
Data: Case Studies Brunner D, Lemoine G, Bruzzone L (2010) Earthquake
damage assessment of buildings using VHR optical
▶ Estimation of Potential Seismic Damage in
and SAR imagery. IEEE Trans Geosci Remote Sens
Urban Areas 48(5):2403–2420
▶ InSAR and A-InSAR: Theory Bruzzone L, Bovolo F (2013) A novel framework for the
▶ SAR Images, Interpretation of design of change-detection systems for very-high-
resolution remote sensing images. Proc IEEE
▶ SAR Tomography for 3D Reconstruction and
101(3):609–630
Monitoring Burt P, Adelson E (1983) The laplacian pyramid as a
▶ Urban Change Monitoring: Multi-temporal compact image code. Commun IEEE Trans
SAR Images 31(4):532–540
Chini M, Pierdicca N, Emery W (2009) Exploiting SAR
and VHR optical images to quantify damage caused by
the 2003 Bam earthquake. IEEE Trans Geosci Remote
References Sens 47(1):145–152
Dekker R (2011) High-resolution radar damage assess-
Balz T, Liao M (2010) Building-damage detection using ment after the earthquake in Haiti on 12 January
post-seismic high-resolution SAR satellite data. Int 2010. IEEE J Sel Top Appl Earth Obs Remote Sens
J Remote Sens 31(13):3369–3391 4(4):960–970
Barber CB, Dobkin DP, Huhdanpaa H (1996) The Dell’Acqua F, Gamba P (2012) Remote sensing and earth-
quickhull algorithm for convex hulls. ACM Trans quake damage assessment: experiences, limits, and
Math Softw 22(4):469–483 perspectives. Proc IEEE 100(10):2876–2890
Bazi Y, Bruzzone L, Melgani F (2005) An unsupervised Dell’Acqua F, Bignami C, Chini M, Lisini G, Polli D,
approach based on the generalized Gaussian model Stramondo S (2011) Earthquake damages rapid map-
to automatic change detection in multitemporal ping by satellite remote sensing data: L’Aquila April
SAR images. IEEE Trans Geosci Remote Sens 6th, 2009 event. IEEE J Sel Topics Appl Earth Obs
43(4):874–887 Remote Sens 4(4):935–943
500 Damage to Ancient Buildings from Earthquakes
Dempster AP, Laird NM, Rubin DB (1977) Maximum Poulain V, Inglada J, Spigai M, Tourneret J-Y, Marthon
likelihood from incomplete data via the EM algorithm. P (2011) High-resolution optical and SAR image
J R Stat Soc Ser B 39(1):1–38 fusion for building database updating. IEEE Trans
Dong L, Shan J (2013) A comprehensive review of Geosci Remote Sens 49(8):2900–2910
earthquake-induced building damage detection with Rignot E, van Zyl J (1993) Change detection techniques
remote sensing techniques. ISPRS J Photogramm for ERS-1 SAR data. IEEE Trans Geosci Remote Sens
Remote Sens 84:85–99 31(4):896–906
Dong Y, Forster B, Ticehurst C (1997) Radar backscatter Soergel U, Thönnessen U, Stilla U (2003) Iterative build-
analysis for urban environments. Int J Remote Sens ing reconstruction from multi-aspect insar data. In:
18(6):1351–1364 Maas HG, Vosselmann G, Wayne Streilein J (eds) 3-
Dong Y, Li Q, Dou A, Wang X (2011) Extracting damages D reconstruction from airborne laserscanner and
caused by the 2008 Ms 8.0 Wenchuan Earthquake InSAR data. International archives of Photogrammetry
from SAR remote sensing data. J Asian Earth Sci and Remote Sensing, vol 34. Institute of Photogram-
40(4):907–914, the 2008 Wenchuan Earthquake, metry and Remote Sensing, Dresden, Germany,
China and Active Tectonics of Asia pp 186–192
Ehrlich D, Guo H, Molch K, Ma J, Pesaresi M (2009) Soergel U, Thoennessen U, Brenner A, Stilla U (2006)
Identifying damage caused by the 2008 Wenchuan High-resolution sar data: new opportunities and chal-
earthquake from VHR remote sensing data. Int lenges for the analysis of urban areas. Radar Sonar
J Digit Earth 2(4):309–326 Navig IEE Proc 153(3):294–300
Ferro A, Brunner D, Bruzzone L, Lemoine G (2011) On Tao J, Auer S, Reinartz P (2012) Detecting changes
the relationship between double bounce and the orien- between a dsm and a high resolution sar image with
tation of buildings in VHR SAR images. IEEE Geosci the support of simulation based separation of urban
Remote Sens Lett 8(4):612–616 scenes. In: Synthetic aperture radar, 2012. EUSAR.
Franceschetti G, Iodice A, Riccio D (2002) A canonical 9th European conference on, 2012, Nuernberg, Ger-
problem in electromagnetic backscattering from many, pp 95–98
buildings. IEEE Trans Geosci Remote Sens Thiele A, Cadario E, Schulz K, Thonnessen U, Soergel U
40(8):1787–1801 (2007) Building recognition from multi-aspect high-
Groen J, Hansen R, Callow H, Sabel J, Sabo T (2009) resolution InSAR data in urban areas. IEEE Trans
Shadow enhancement in synthetic aperture sonar using Geosci Remote Sens 45(11):3583–3593
fixed focusing. Ocean Eng IEEE J 34(3):269–284 World Bank (2012) The great east Japan earthquake learn-
Heckbert P (1994) Graphics gems IV, ser. Graphics gems ing from megadisasters: knowledge notes, executive
series. AP professional summary. In EAP DRM knowledge notes
Inglada J, Mercier G (2007) A new statistical similarity Xia Z-G, Henderson F (1997) Understanding the relation-
measure for change detection in multitemporal ships between radar response patterns and the bio- and
SAR images and its extension to multiscale change geophysical parameters of urban areas. IEEE Trans
analysis. Geosci Remote Sens IEEE Trans 45(5): Geosci Remote Sens 35(1):93–101
1432–1445 Zadeh L (1965) Fuzzy sets. Inf Control 8(3):338–353
Jin D, Wang X, Dou A, Dong Y (2011) Post-earthquake Zhang Y, Chen H, Ding C, Wang H (2013) A novel
building damage assessment in Yushu using airborne approach for shadow enhancement in high-resolution
SAR imagery. Earthq Sci 24(5):463–473 SAR images using the height-variant phase compen-
Larsen PM (1980) Industrial applications of fuzzy logic sation algorithm. IEEE Geosci Remote Sens Lett
control. Int J Man–Mach Stud 12(1):3–10 10(1):189–193
Mallat S (1989) A theory for multiresolution signal
decomposition: the wavelet representation. Pattern
Anal Mach Intell IEEE Trans 11(7):674–693
Marin C, Bruzzone L, Bovolo F (2014) Rapid and accurate
damage detection in built-up areas combining
Damage to Ancient Buildings from
stripmap and spotlight SAR images. In: Geoscience Earthquakes
and remote sensing symposium (IGARSS), 2014 Qué-
bec City, Canada Miklós Kázmér
Matsuoka M, Yamazaki F (2004) Use of satellite SAR
Department of Palaeontology, Eötvös University,
intensity imagery for detecting building areas
damaged due to earthquakes. Earthq Spectra 20: Budapest, Hungary
975–994
Nason G, Silverman B (1995) The stationary wavelet
transform and some statistical applications. In:
Antoniadis A, Oppenheim G (eds) Wavelets and sta-
Synonyms
tistics, ser. Lecture notes in statistics, vol 103.
Springer, New York, pp 281–299 Archeoseismology
Damage to Ancient Buildings from Earthquakes 501
Damage to Ancient Buildings from Earthquakes, up between them. Shiva temple, Prambanan complex,
Fig. 1 Typology of earthquake-induced damages in Yogyakarta, Indonesia. Tape measure extended 20 cm
ancient buildings. (a) Cracks/fractures within a block #6498. (e) Dropped ashlars in a Roman arch (Damascus,
(second century AD Palmyra theater, Syria). Photo Syria) #2017. (f) Broken lintels in twelfth-century
Kázmér #4281. (b) Chipped edges of column and capital al-Marqab citadel, coastal Syria #4663. (g) Clockwise-
in the sixth-century Euphrasius cathedral, (Poreč, Croa- rotated blocks in twelfth-century al-Marqab citadel,
tia), damaged by the 1440 earthquake #0080. (c) Through- coastal Syria #5168. (h) Displaced drums of masonry
cutting fractures (repaired) in the eight- to ninth-century columns of the fifth-century BC Hephaisteion temple,
Brahma temple of the Prambanan complex, Yogyakarta, Athens, Greece #1132. (i) Extruded portion of house
Indonesia, caused by the 2006 earthquake #6488. (d) wall. Bosra, Syria #4183
Horizontal shift of large ashlars; vertical joints opened
and down. The hammering effect of the upper or peeling due to weathering as seismically
block on the lower one yields cracks in either or broken chips have sharp edges.
both blocks (Fig. 1a). Penetrating cracks allow a Fractures cutting through two or more blocks
chip, corner, or edge to be separated from the are often oriented close to vertical, this being the
block (Fig. 1b). The direction of the crack is weakest plane to resist bending forces (Fig. 1c).
influenced by the lithology of the rock and Fractures through individual blocks are more or
shape of the block and by direction of hammer- less connected to each other, resulting from sud-
ing. Chipping is not to be confused with rounding den seismic shaking.
Damage to Ancient Buildings from Earthquakes 503
Damage to Ancient Buildings from Earthquakes, in corners of walls (al-Marqab citadel, coastal Syria)
Fig. 2 (a) Arcuate out-of-plane collapse in the market, #4671. (f) Deformed circular dome of fourth- to
Pompeii, Italy, repaired before the 79 AD eruption of eleventh-century Samtavro cathedral, Mtskheta, Georgia
Vesuvius, which buried the city. Photo: Kázmér #9266. #1633. (g) Tilted Buddhist stupa (approximately fifteenth
(b) Warped northern side of the Valens aqueduct century) in Chiang Mai, Thailand. While a single tilted
(Istanbul, Turkey). Completed in 368 AD, several subse- building does not indicate seismic origin, this stupa is one
quent earthquakes left their traces on the massive con- of 21 sites displaying tilting in the city (Kázmér
struction #0340. (c) V-shaped damage on the SW wall of et al. 2011) #3411. (h) Buttresses support the pillars of
the twelfth-century donjon of al-Marqab citadel, coastal the Valens aqueduct in Istanbul, Turkey, completed in
Syria. The damage was inflicted by the 1202 earthquake 368 AD #0335. (i) Valens aqueduct in Istanbul, completed
(Kázmér and Major 2010) #4589. (d) Fallen masonry in 368 AD. A brick arch was built to support the weakened
fence in the Roman city of Carnuntum (Deutsch- stone arch after one of the frequent earthquakes in the
Altenburg, Austria) #5496. (e) Triangular missing parts region (Istanbul, Turkey) #0335
Damage to Ancient Buildings from Earthquakes 505
settlement, tilting, and collapse. Traces of total and illustrated and arranged in a simple, descrip-
failure are mostly removed, but minor tilting – not tive system. Deformations affecting single and
affecting the use of the building – is often left multiple blocks of dressed masonry, single
unrepaired. walls, adjacent walls, and whole buildings record
Very rarely a building – if built across an evidence of seismic damage. Their recognition,
active fault – is sheared by fault movement. The careful analysis, and distinction from other types
displaced walls indicate the sense and cumulative of damage (by aging, warfare, poor construction,
offset of the fault since construction time. Walls etc.) need thorough and systematic study but pro-
are either displaced along a single fault (i.e., vide much needed data on past earthquakes. D
Vadum Iacob crusader tower, Israel: Ellenblum
et al. 1998) or along a splay of faults, where the
building maintains integrity, but the floor plan Cross-References
gets deformed (i.e., St. Simeon monastery,
Syria: Karakhanian et al. 2008). ▶ Ancient Monuments Under Seismic Actions:
Modeling and Analysis
▶ Archeoseismology
Repairs ▶ Damage to Buildings: Modeling
▶ Geotechnical Earthquake Engineering:
Ruins of completely collapsed buildings are usu- Damage Mechanism Observed
ally removed after an earthquake. Moderate dam- ▶ Masonry Structures: Overview
ages are repaired on the spot, and additions, new ▶ Paleoseismology
walls, buttresses (Fig. 2h), reinforced arches ▶ Retrofitting and Strengthening Masonries of
(Fig. 2i), and other supporting structures can be Heritage Structures: Materials Used
identified as evidence of past earthquakes. ▶ Seismic Analysis of Masonry Buildings:
Repairs are often built of different, often inferior Numerical Modeling
material (Fig. 2i: brick arc supporting a stone ▶ Seismic Behavior of Ancient Monuments:
arch), poorly fitted to masonry patterns, From Collapse Observation to Permanent
obstructing windows – these are features of repair Monitoring
following structural damage. ▶ Seismic Vulnerability Assessment: Masonry
Study of damaged ancient buildings is a major Structures
source of evidence for past earthquakes: various
seismic parameters (e.g., date, intensity, strong
motion direction) can be assessed from a careful,
References
critical study. However, before anyone jumps
head-fast into archeoseismology, here is a cau- Ambraseys NN (2006) Earthquakes and archaeology.
tionary note: attributing seismic origin to J Archaeol Sci 33:1008–1016
observed damage needs careful consideration Ellenblum R, Marco S, Agnon A, Rockwell T, Boas
A (1998) Crusader castle torn apart by earthquake at
and exclusion of all other potential causative
dawn, 20 May 1202. Geology 26:303–306
agents (Ambraseys 2006). Galadini F, Hinzen K-G, Stiros S (2006)
Archaeoseismology: methodological issues and proce-
dure. J Seismol 10:395–414
Hinzen K-G (2009) Simulation of toppling columns in
Summary archaeoseismology. Bull Seismol Soc Am
99(5):2855–2875
Damage to ancient buildings is a major source of Hinzen K-G (2011) Archeoseismology. In: Gupta HK
information about the parameters of past earth- (ed) Encyclopedia of solid earth geophysics. Springer,
Berlin, pp 11–15
quakes (date, intensity, strong motion direction,
Hinzen K-G, Fleischer C, Reamer SK, Schreiber S,
just to name a few). Features to aid in the identi- Sch€utte S, Yerli B (2011) Quantitative methods in
fication of seismic damage are briefly described archaeoseismology. Quat Int 242:31–41
506 Damage to Buildings: Modeling
These include uncertainties in the quantification A well-established tool used to relate building
of seismic hazard, the identification of different damage to seismic hazard probabilistically is the
building types and their characteristics within the fragility function, commonly referred to as the
building stock, the relationship between seismic fragility curve when presented graphically (e.g.,
hazard and physical damage, and the relationship Rossetto and Elnashai 2003). For a given level of
between the physical damage and socioeconomic ground motion intensity, fragility functions
consequences. determine the probability that a building or
In the case of seismic hazard definitions, there a group of buildings will be in, or exceed, the
are uncertainties over the time, location, depth, ith damage state, di. The probabilistic relation- D
and magnitude of future earthquakes. One ship is commonly assumed to be a cumulative
approach to account for these is probabilistic lognormal distribution. In this case, the fragility
seismic hazard assessment (PSHA) (Bommer function can be expressed conditionally in the
2002). In PSHA, maps of ground motion intensity form of Eq. 1.
are produced for a study area in which the values
Damage to Buildings: 1
Modeling, Fig. 1 A
0.9 P [ d = d1 | PGA ]
fabricated set of fragility
curves for a fictitious 0.8
structural typology 0.7
P [ d ≥ di | PGA ]
0.6
0.5 d3 = Total damge
0.4 d2 = Moderate damage
0.3
d1 = Slight damage
0.2
0.1 P [ d = d3 | PGA ]
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Peak ground acceleration (g)
Damage to Buildings:
Modeling,
Fig. 2 Suggested steps for
assessing the earthquake-
induced damage, and
related functional impacts
and direct losses, to
buildings
This section will provide a literature review on elaborated classification systems, such as the
different building taxonomies that have been pro- one proposed within HAZUS (NIBS 2003) for
posed worldwide to classify building stocks the United States and the European
according to their expected seismic performance Macroseismic Scale EMS-98 (Gr€unthal 1998)
and occupational function. for Europe, where, further to the material, con-
Early building taxonomies grouped building sideration is given to primary parameters
typologies according to their building material influencing the building damage, namely,
only. Table 1 reports, for example, the three (1) the building structural system, (2) the seis-
building classes identified and defined by the mic design criteria (design code level), (3) the
Medvedev–Sponheuer–Karnik (MSK-64) building height (low rise, mid rise, high rise),
macroseismic scale. and (4) the nonstructural elements. Table 2
The need for a deeper diversification of reports and compares HAZUS and EMS-98 tax-
building behaviors, when subjected to an earth- onomies for unreinforced masonry and
quake event led to the definition of more reinforced concrete buildings.
510 Damage to Buildings: Modeling
Damage to Buildings: Modeling, Table 1 Classes of Damage to Buildings: Modeling, Table 2 HAZUS
buildings according to the MSK-64 macroseismic scale (NIBS 2003) and EMS-98 (Gr€
unthal 1998) classifications
for unreinforced masonry and reinforced concrete
A Buildings with dry stone or clay, adobe, or mud
buildings
walls
B Buildings with brick walls, mortar blocks, masonry EMS 98 HAZUS
and mortar, stone block, timber frame Unreinforced Rubble stone Unreinforced
C Buildings with metal structure or reinforced masonry Adobe (earth masonry bearing
concrete bricks) walls (URM)
Simple stone
Massive stone
Recently, international projects and initiatives U Masonry
worldwide have developed and defined several (old bricks)
further taxonomy systems aiming to represent U
the variety of building types around the world Masonry – RC
floors
and the continued upgrading/modification of con-
Reinforced Reinforced Concrete moment
struction techniques. Pitilakis et al. (2013) pro- concrete concrete frame frame
posed a modular and flexible taxonomy for Shear walls Concrete shear
classifying European masonry and reinforced walls
concrete buildings, where buildings are classified Concrete frame
according to some main categories, including the with U Masonry
infill walls
force-resisting mechanism (FRM), material of
FRM, plan (regular or irregular), elevation
(regular of irregular), cladding, detailing
(ductile, non-ductile, with tie-rod beam, without building on a comprehensive review of existing
tie-rod beam), floor system, roof system, height taxonomies (Charleson 2011) and on the findings
level, code level, and material. Subcategories from WHE. The effectiveness of existing taxon-
(e.g., FRM, material of FRM, cladding character- omy to represent building typologies worldwide
istics, floor system material, roof system mate- was checked against different criteria (Porter
rial, number of stories) were introduced for some 2005) aiming to define a simple, collapsible,
of the aforementioned categories, to provide and detailed taxonomy, international in scope,
a more refined classification, when more accurate user-friendly, and able to distinguish differences
data are available for the building stock under in seismic performance, but extensible to future
analysis. needs (e.g., to other natural hazards). Following
The Prompt Assessment of Global Earth- this comprehensive and critical review, the
quakes for Response (PAGER) project (Porter Syner-G taxonomy (Pitilakis et al. 2013) was
2005) developed a comprehensive building tax- selected as the benchmark to define the GEM
onomy with an international coverage including building taxonomy. The GEM taxonomy is orga-
structural types common in non-developed coun- nized as a series of expandable tables, describing
tries (i.e., adobe, clay, poor quality wood con- various attributes at increasing levels of detail
structions). The Earthquake Engineering (up to 5 depending on the specific attribute).
Research Institute (EERI) developed a World Fabricated examples of building taxonomy and
Housing Encyclopedia (WHE) (EERI 2000) building occupancy taxonomy are presented,
collecting information related to housing con- respectively, in Tables 3 and 4.
struction practices in seismically active areas Further to taxonomies for building structures,
worldwide. specific building taxonomies have been devel-
The Global Earthquake Model (GEM) devel- oped for nonstructural components and content
oped a global building taxonomy aiming to rep- (e.g., NIBS 2003; NIST 1999; Porter 2005).
resent all the building types around the globe HAZUS (NIBS 2003) provided a classification
(Brezev et al. 2013; Gallagher et al. 2013) by for architectural components (e.g., veneer and
Damage to Buildings: Modeling, Table 3 Example of building taxonomy (Adapted from GEM buildings taxonomy, Brezev et al. 2013)
Structural system Building information Exterior attributes Roof/floor system
Damage to Buildings: Modeling
Direction Material Lateral Height Date of Occupancy Building Shape of Structural Exterior Floor Foundation Roof Pitched with
of lateral load- construction position the irregularity walls system gable endsa
load- resisting or retrofit within building Clay or
resisting system a block plan concrete tile
system roof
covering
Wooden roof
Wooden
structure
with light
roof
covering
Roof-wall
diaphragm
connection
not provided
Level 1 attributes Higher-level
attributes
a
An example of higher-level attributes for the roof system
511
D
512
Damage to Buildings: Modeling, Table 4 Example of building occupancy taxonomy (Adapted from GEM taxonomy, Brezev et al. 2013)
Building occupancy class
Unknown Residential Mixed Industrial Agriculture Assembly Government Education Other Commercial Offices, professional/
occupancy use occupancy and public technical servicesa
type Hospital/medical clinic
Entertainment
Public building
Covered parking garage
Bus station
Railway station
Airport
Recreation and leisure
Level 1 attributes Higher-level attributes
a
An example of higher-level attributes for the “Commercial and Public” occupancy class
Damage to Buildings: Modeling
Damage to Buildings: Modeling 513
finishes, cantilever elements and parapets, etc.), (Tenerelli et al. 2013) include (1) ground screen-
mechanical and electrical components (e.g., pip- ing procedures (e.g., field reconnaissance, side-
ing systems, elevators, etc.), and contents (e.g., walk survey, street-front direct visual
file cabinets, bookcases etc.). NISTIR 6389 observation, interpretation of street-front photo-
(NIST 1999) provided taxonomy for shell, inte- graphs, etc.), (2) satellite-remote sensing
riors, services, equipment, furniture, and special approaches, or (3) a combination of ground
constructions. Porter (2005) proposed an expend- screening procedures and satellite-remote sens-
able and flexible taxonomy for nonstructural ele- ing approaches. Ground screening procedures
ments based on the review, combination, and collect building attributes through direct obser- D
improvement of the existing ones, including vations, usually via (a) paper forms, ad-hoc for-
NISTIR 6389 (NIST 1999). matted to quickly record relevant building
characteristics and attributes (e.g., FEMA
Building Inventories 154 1988), and (b) handheld equipment and soft-
A building inventory for seismic risk analysis ware (e.g., tablet, phone) that enables user-
purposes can be defined as an enumeration and defined electronic inventory data. For both the
characterization of the building stock under anal- approaches, a critical aspect is the definition of
ysis according to the assumed taxonomy. Prepa- the survey form that has to be coherent with the
ration of a building stock inventory is a critical adopted building taxonomy (section “Building
part of a damage and loss estimation study and Taxonomies”) and with the adopted approach to
can be both time and resource consuming. assign damage to buildings as a function of the
A compromise needs to be sought between the ground motion (section “Assigning Damage to
time and cost required for higher-quality data on Buildings Based on Ground Motions”). Further-
the building stock and the uncertainties arising more, proper training of staff those performing
from low-quality data. In order to obtain an the ground screening survey is essential since
inventory affordable both from the cost and the identifying structural types and other building
time point of view, it would be worth to collect, as attributes from the street can be challenging. Sat-
a first step, all the already available data sources, ellite and airborne remote sensing techniques
while being conscious that these data are often may provide information on building stocks,
overlapping and incomplete and that additional e.g., building footprint and height information,
work might be required to reconcile multiple in a timely and cost effective manner. However,
sources and missing data. Possible available these techniques must be closely calibrated to the
sources, in order to built-up regional building particular region under study, and the analysis
inventory, are databases belonging to state, must be supported by expert judgment, coupled
regional, and local authorities and private sector with some field reconnaissance.
sources as well as inventory information arising
from previous loss or available within seismic Building Classification
risk assessment tools or platforms (e.g., HAZUS The final aim of the building classification process
NIBS 2003). Census data, for example, can pro- is to identify and enumerate the buildings com-
vide aggregated data on the size, the number, and posing the building stock under analysis in each of
the age of residential buildings; conversely spe- the building types and subtypes considered by the
cific databases prepared, for example, for insur- assumed classification system. Although a lot of
ance purposes or for civil defense planning information sources on the building stock might
purposes can provide information on commercial be available at regional level, and further data can
and/or strategic buildings. be collected (section “Building Inventories”), they
Further to available databases, there are dif- might be insufficient for a direct classification of
ferent methods for collecting, processing, and the building in terms of the construction types and
archiving building inventory datasets. Currently, occupancy classes identified by an assumed tax-
the most common and widely used approaches onomy (section “Building Taxonomies”).
514 Damage to Buildings: Modeling
Slight (D1) Moderate (D2) Heavy (D3) Very heavy (D4) Collapse (D5)
Masonry
Reinforce
Concrete
Damage to Buildings: Modeling, Fig. 3 Damage grades for masonry and reinforced concrete buildings according to
the EMS-98 macroseismic scale (Adapted from Gr€
unthal 1998)
Therefore, inferences have to be established damage grade, the EMS-98 scale provides
between larger groups of buildings (referred to as a qualitative description of the expected physical
categories), recognized on the basis of more gen- impacts to structural and nonstructural elements.
eral information (such as land use patterns or Wailes and Horner (1933) propose a five-
building age) and building types or occupational damage state scale, where for some of the damage
classes. Inferences may take the form of “for states, further to qualitatively describing the
masonry buildings built before 1919, 40 % expected impacts to structural and nonstructural
belong to rubble stone typologies and 60 % are elements, an estimate of the damage extension is
old brick masonry buildings.” The assumed infer- provided (Table 5). ATC-13 (1985) proposes a -
ences should be verified on the basis of the six-damage grade scale, plus a “none damage”
“expert opinion” of local engineers and building (Table 5).
officials in order to verify that they reflect the In order to estimate the functional impacts and
construction features of the region. A ground consequences deriving from the earthquake-
screening procedures can be used to develop or induced physical damage to buildings, functional
check the inference rules, used to characterize the impact scales are needed and a correspondence
building stock into categories, as well to assess between physical damage grades/states and func-
the accuracy of the available information. tional levels needs to be established. Vision 2000
(SEOAC 1995) proposes a five-level perfor-
mance objective scale, where the following func-
Describing Building Damage tional level/objectives for buildings are
identified: fully operational, operational, life
Earthquake-induced damage to building struc- safety, and near collapse. Rossetto and Elnashai
tures and nonstructural elements is usually (2003) propose damage scales specific for
described in a scale of multiple increasing dam- reinforced concrete buildings (referred to as
age states, also referred to as damage grades, homogenized reinforced concrete, HCR) and
from no damage to total collapse. The damage a correspondence between physical damage
scale adopted in the HAZUS (NIBS 2003) plat- states and performance objectives, namely, ser-
form, for example, accounts for four discrete viceability, damage control, and collapse preven-
damage states further to “no damage,” namely, tion. The Applied Technology Council Report
minor damage, moderate damage, extensive (ATC-21 1988) defines a tagging system,
damage, and collapse. In the European corresponding to a three-level scale, to be used
Macroseismic Scale EMS-98 (Gr€unthal 1998), for post-earthquake safety assessment aiming to
the damage is represented in a discrete form identify buildings, which are safe to occupy
through five-damage grades Di (i = 1 5), fur- (green tag) being undamaged or with light dam-
ther to “no damage,” D0. Figure 3 represents the age, temporarily unusable building (yellow tag)
EMS-98 (Gr€ unthal 1998) five-damage grades for being affected by significant damage and requir-
masonry and reinforced building types. For each ing repairs before reoccupation, and unusable,
Damage to Buildings: Modeling 515
Damage to Buildings: Modeling, Table 5 Examples of damage and functional scales: comparison of damage grades
and functional levels (Adapted from Rossetto and Elnashai 2003)
Wailes and
HRC HAZUS Vision2000 EMS-98 ATC-13 ATC-21 Hornera
None No Fully No damage None Green tag A
damage operational (D0)
Slight Slight Slight (D1) Slight
Light Operational Light B
Moderate Moderate Life safe Moderate (D2) Moderate Yellow C D
Extensive Extensive Near collapse Heavy (D3) Heavy tag
Partial Very heavy Major Red tag D
collapse (D4)
Collapse Collapse Collapse Collapse (DG5) Destroyed E
a
A – Undamaged or minor cracking, no significant structural damage, minor veneer damage
B – Parapet failure or separation of veneer, major wall cracking, and interior damage
C – Failure of portion of exterior walls, major damage to less than 50 % of walls
D – Major damage to more than 50 % of walls
E – Non-repairable damage, demolition probably appropriate
dangerous buildings (red tag) being affected by parameters or qualitative parameters can be
heavy structural damage up to partial or total selected for the ground motion description.
collapse, deemed in some cases for demolition. Physical–mechanical parameters for the
Table 5 presents few examples of damage scales ground motion description aim to characterize
and functional scales and compares their damage the amplitude, the frequency content, and the
grades and functional levels. duration of strong ground motion; some of them
Hill and Rossetto (2008) critically reviewed describe only one of these characteristics, while
and compared building damage scales, identify- others may reflect several of them. In engineering
ing their potentialities and shortcomings for their application the most commonly used measure of
use in seismic risk analysis and for loss assess- acceleration for a particular ground motion is the
ment applications. peak horizontal acceleration, PHA, the maximum
horizontal acceleration at the ground surface is
also known as the peak horizontal ground accel-
Selecting Appropriate Earthquake eration. For most earthquakes, the horizontal
Hazard Parameters accelerations is greater than the vertical acceler-
ation, and thus the peak horizontal ground accel-
The selection of earthquake hazard parameters, eration also turns out to be the peak ground
commonly referred to as intensity measure (IM), acceleration (PGA). The PGA for a given
is a crucial step when assessing seismic risk component of motion is the largest (absolute)
for buildings. Performance, integrity, or damage value of horizontal acceleration obtained from
of structures is, in fact, influenced by the intensity, accelerogram of that component. By taking the
frequency content, and duration of the ground vector sum of two orthogonal components, the
motion (Weatherill et al. 2011a). The choice maximum resultant PGA can be obtained. Hori-
of a suitable IM should possibly account for zontal accelerations have commonly been used to
the characteristics of the building stock analyzed describe ground motions because of their natural
(section “Classifying Buildings”) and the method relationship to inertial forces; however they do
used to predict the building performance, when not provide any information about the dynamic
subjected to seismic solicitations (section behavior of a structure. This can be obtained
“Assigning Damage to Buildings Based on by the use of response spectra. Response
Ground Motions”). Physical–mechanical spectrum describes the maximum response of
516 Damage to Buildings: Modeling
earthquake (e.g., the maximum historical event adjacent predicted ground motions do not corre-
from a pertinent seismogenetic source or the late with each other and the resulting ground
maximum earthquake compatible with the motion field is random. This is fine when dealing
known tectonic framework) at a specific location with spatially clustered analysis (i.e., analysis at
(e.g., the shortest distance from the source to the a single site). Conversely, when the analysis is
studied region, when the most unfavorable situa- targeting spatially distributed building stocks
tion is considered). and the simultaneous generation of ground
It is well recognized that there is not motions at multiple sites is required, the ground
a mutually exclusive dichotomy between DSHA motion field should be spatially correlated to D
and PSHA (Bommer 2002) and different more realistically represent the earthquake sce-
approaches combining DSHA and PSHA have nario. Crowley et al. (2008) found that the spa-
been proposed in the literature. Bazzurro and tial correlation of ground motion field can
Cornell (1999) propose to select the event that is significantly influence the outputs of a seismic
most likely to contribute to the hazard for a given loss analysis for distributed portfolios, when
intensity measure at a site, from the disaggrega- simulating a single earthquake scenario.
tion of PSHA outputs. Bazzurro and Luco (2005) A number of spatial correlation models are
proposed the use of Monte Carlo simulations to available in the literature. Weatherill et al.
reproduce a series of single earthquake scenarios (2011b) present a comprehensive discussion
within PSHA. and description of different methods to produce
For PSHA, DSHA, and any combination of spatially correlated ground motion fields and to
them, it is necessary, as a first step, to identify the account for site amplifications.
potential seismogenetic sources pertinent to the
region under analysis and characterize them in
terms of locations, geometry, seismic activity, Assigning Damage to Buildings Based
and possibly released energy. As a second step, it on Ground Motions
is necessary to represent the propagation of the
ground motion by the use of a suitable ground The damage that might affect a building stock,
motion prediction equation (GMPE). GMPEs, when subjected to a seismic event, can be
also referred to as “attenuation” relationships, predicted via different approaches, including
are equations derived from observed strong (a) actual damage observation, (b) expert judg-
motion records and provide a means for ment, (c) simplified mechanical and analytical
predicting the level of ground shaking (in terms models, or a combination of the three aforemen-
of the desired intensity measure, e.g., PGA, tioned approaches. Methods based on actual
PGV) and its associated uncertainty at any damage observations are derived statistically by
given site or location, based on the earthquake processing the damage data from different earth-
magnitude, source-to-site distance, local soil quakes and by summarizing the outcomes in
conditions, fault mechanism, etc. Due to the terms of (a) damage probability matrices (DPM)
empirical nature of GMPEs, the ground motion (e.g., Whitman et al. 1973, Table 7),
at each site/location is estimated by the median, (b) vulnerability curves (e.g., Giovinazzi 2005,
or logarithmic mean, of the ground motion plus Fig. 4a, c), or fragility curves (e.g., Rossetto and
two random variables, referred to as residuals. Elnashai 2003). A DPM is a matrix that summa-
The residuals represent the aleatory variability, rizes the seismic damage that occurred in a set of
namely, the difference between the predicted buildings, represented as the statistical distribu-
and observed ground motions in the regression tion for each of the damage grades Di considered
analysis, occurring intra-event, i.e., within and for the different macroseismic intensities I. In
a single event, and inter-event, i.e., between other words, DPM expresses, in a discrete form,
separate events (Douglas 2011). For this reason, the conditional probability p[Di|T,I] to obtain
when GMPEs are applied to a large area, a damage grade Di due to an earthquake of
518 Damage to Buildings: Modeling
Damage to Buildings: Modeling, Table 7 Example of system and the earthquake demand curve, ade-
DPM, representing the probability of damage in percent quately reduced to account for the inelastic
by macroseismic earthquake intensity and damage states
(Adapted from ATC-13 1987) behavior and energy dissipation capacity of the
system (Fig. 4e). On the other hand, according to
Damage Earthquake intensity
state VI VII VIII XI X
displacement-based approaches, the periods
1 None 95.0 49.0 30.0 14 3.0 associated to the boundary of different limit
2 Slight 3.0 38.0 40.0 30 10 states can be evaluated by the intersection
3 Light 1.5 8.0 16.0 24 30 between capacity curves, represented in terms
4 Moderate 0.4 2.0 8.0 15 26 of period displacement relationship, and the dis-
5 Heavy 0.1 1.5 3.0 10 18 placement spectrum demand curves, scaled by
6 Major – 1.0 2.0 4.0 10
equivalent viscous damping factors. Other pro-
7 Destroyed – 0.5 1.0 2.0 3.0
posals for mechanical-based methods are based
on the evaluation of collapse multipliers associ-
ated to alternative collapse mechanisms (e.g.,
macroseismic intensity I and for a defined build- Cosenza et al. 2005) or on the derivation of
ing type T (Table 7). Examples of vulnerability fragility curves directly from the results of
and fragility curves obtained from the processing extensive numerical analyses (e.g., Elnashai
of damage observations are provided in section and Jeong 2005). The examples of an observed-
“Assigning Damage to Buildings Based on based approach and a mechanical-based
Macroseismic Intensities.” approach are summarized in the following sub-
Due to the inherent difficulty to retrieve com- sections (sections “Assigning Damage to Build-
prehensive damage data for all the building types ings Based on Macroseismic Intensities” and
considered within an assumed taxonomy and for “Assigning Damage to Buildings Based on
a complete range of damaging earthquake inten- Physical–Mechanical Ground Motion Parame-
sities, “hybrid” methodologies have been ter,” respectively). Despite being different for
defined, where the available observed damage derivation and conception, a quantitative com-
data have been completed with the use of differ- parison and reciprocal calibration has been pos-
ent approaches, including numerical analysis, sible for these two specific methods, thanks to
neural network system, fuzzy set theory, expert their closed-form formulations (Lagomarsino
judgment, etc. and Giovinazzi 2006). As a useful result, refine-
Expert-based methods imply the human ments of the methods based on numerical/exper-
judgment to completely replace the processing imental analysis were directly implemented
of observed data, leading to expert-defined into an equivalent macroseismic approach.
DPM (e.g., ATC-13 1987) or score assignment Dually, the reliability of assumed force-based
procedures (e.g., ATC-21 1988; FEMA capacity curves was cross-validated on the
154 1988). basis of real observed damage data (Giovinazzi
Mechanical-based models for territorial scale et al. 2006).
analysis on classes of buildings can be defined on
the basis of either traditional force-based proce- Assigning Damage to Buildings Based on
dures (e.g., capacity spectrum method Macroseismic Intensities
implemented in HAZUS, in NIBS (2003), or in Combining classical probability theory and fuzzy
RISK-UE (2004)) or according to displacement- set theory, a macroseismic approach was derived
based designed approaches (e.g., Calvi (Giovinazzi and Lagomarsino 2004; Giovinazzi
et al. 2005). According to force-based proce- 2005), based on the EMS-98 macroseismic scale
dures, the building performance is identified, definitions (Gr€unthal 1998) that summarize the
within acceleration-displacement response spec- damage observations following several European
tra (ADRS) by the intersection point between the earthquakes. Numerical and complete DPM have
capacity curve of an equivalent nonlinear SDOF been evaluated, in the range of damaging
Damage to Buildings: Modeling 519
Damage to Buildings: Modeling, probabilities pk and fragility curves for RC types for
Fig. 4 Macroseismic- and mechanic-based approaches: IEMS-98 = 9 (mD = 2.25). Mechanic: (d) elastic response
basic steps. Macroseismic: (a) medium, upper, and lower spectrum; (e) evaluation of the performance point through
vulnerability curves for RC building types; (b, c) damage capacity spectrum method; (f) lognormal fragility curves
EMS-98 intensities, IEMS-98, and damage grades the expected damage, as provided by the
Di (i = 1 5), for the EMS-98 building classes DPMEMS-98, have been in terms of a mean dam-
and types (Table 2). Fuzzy set theory was effec- age grade, mD, defined as the mean value of the
tive to consider, within the proposed DPMEMS-98 damage distributions:
macroseismic approach, the epistemic uncer-
tainties affecting the damage assessment and the X
5
building classification procedures (section “Clas- mD ¼ pi i 0 mD 5 (3)
sifying Buildings”). Upper and lower bounds of i¼1
520 Damage to Buildings: Modeling
8
where TC and TD define the onset of constant Estimating Losses and Consequences
spectrum velocity and displacement range within Due to Building Damage
the elastic response spectrum Sae(T) evaluated
from deterministic or probabilistic hazard ana- Earthquake-induced structural and nonstructural
lyses (Fig. 4d). damage to buildings is the root cause for several
Damage to Buildings: Modeling 521
Damage to Buildings: Modeling, Table 8 Example of weight factors to assess consequences from earthquake-
induced physical damage to buildings, referred to EMS-98 damage scale
Damage grade Structural (S), nonstructural (NS) wk MRDa MRDb MRDc UFUd Se
Slight (D1) S = no, NS = slight w1 0.02 0.01 0.01 0 0
Moderate (D2) S = slight, NS = moderate w2 0.1 0.1 0.1 0 0
Heavy (D3) S = moderate, NS = heavy w3 0.5 0.35 0.35 0.40 0
Very heavy (D4) S = heavy – NS = very heavy w4 1 0.75 0.75 1 0
Collapse (D5) S = very heavy w5 1 1 1 1 0.3
a
Mean damage ratio, derived from HAZUS (NIBS 2003) D
b
Mean damage ratio, derived and adapted from ATC-13 (1987)
c–e
Mean damage ratio, unfit for use buildings, severely injured from Bramerini et al. (1995)
losses and impacts. Consequences to buildings The MRD weight factors in Table 8 are
and their contents, direct economic losses (e.g., intended for the evaluation of economic losses
cost of repair/reconstruction of building struc- due to the physical damage of buildings. Similar
tural or nonstructural elements or replacement correlations can be derived for building contents
of machinery and content), and consequences to and for business interruption.
inhabitants and building occupants can be esti-
mated once the physical damage sustained by the
buildings has been assessed. Different empirical Summary
correlations, translating structural and
nonstructural damage into percentage of losses, Earthquake risk is a public safety issue that
are proposed by different authors and in the requires appropriate risk management measures
framework of different seismic risk analysis and means to protect citizens, properties, infra-
methods. These correlations apparently different structures, and the built cultural heritage. The aim
can be, actually, represented according to of a seismic risk analysis is the estimation and the
a common formulation (Giovinazzi 2005). hypothetical, quantitative description of the con-
Having the physical damage described in sequences of seismic events upon a geographical
terms of a discrete damage scale (see Table 6 area (a city, a region, a state, or a nation) in
for some examples), the probability of having a certain period of time. The effects to be
consequences pc induced by an earthquake can predicted are the physical damage to buildings
be obtained as and other facilities, the number and type of casu-
X alties, the potential economic losses due to the
pc ¼ w i pi (6) direct cost of damage along with the indirect
i economic impacts (loss of the productive capac-
ity and business interruption), the loss of function
where pi is the probability of having a damage for lifelines and critical facilities, as well as
grade di and wi is a weight factor, specifically social, organizational, and institutional impacts.
defined for a specific consequence c and The results provided by a seismic risk analysis
a damage grade i. could be regarded as helpful guidelines with
Table 8 summarizes weight factors for the respect to all the phases of risk management:
evaluation of different consequences, including during normal periods, during crisis periods, as
(a) the ratio between cost of repairing and cost of well as in the recovery and post-emergency
replacement defined as mean damage ratio, periods.
MDR; (b) the number of buildings unfit for use, This entry has summarized the basic steps
UFU; and (c) the number of severely injured needed for modeling earthquake-induced dam-
people, S. age to buildings at territorial or urban scale as
522 Damage to Buildings: Modeling
part of seismic risk analysis. It has described the ▶ Earthquakes and Their Socio-economic
importance and purpose of inventorying and Consequences
classifying buildings. It has explained the impor- ▶ Economic Impact of Seismic Events:
tance of using an appropriate damage scale for Modeling
classifying physical damage to building struc- ▶ Empirical Fragility
tural and nonstructural components or for ▶ Estimation of Potential Seismic Damage in
describing the expected performance objectives Urban Areas
for the buildings. The recommended parameters ▶ Engineering Characterization of Earthquake
for describing seismic ground motion, against Ground Motions
which damage should be evaluated, have been ▶ Insurance and Reinsurance Models for
introduced. The use of either probabilistic seis- Earthquake
mic hazard assessment or deterministic scenario ▶ Learning from Earthquake Disasters
earthquakes or a combination of the two, to ▶ Probabilistic Seismic Hazard Models
evaluate ground motions for building damage ▶ Seismic Fragility Analysis
assessment, has been discussed. A concise ▶ Seismic Actions Due to Near-Fault Ground
review of hazard-damage relationships has Motion
been provided along with the presentation ▶ Seismic Loss Assessment
of the basic steps for the two of them, a ▶ Seismic Risk Assessment, Cascading Effects
macroseismic- and a mechanic-based approach, ▶ Seismic Vulnerability Assessment: Masonry
for which a closed-form formulation assures an Structures
easy computational implementation within seis- ▶ Seismic Vulnerability Assessment: Reinforced
mic risk analysis tools and platforms. For both Concrete Structures
the methods, the classification of the building ▶ Site Response for Seismic Hazard Assessment
stock can be performed with data of different
origin and quality (e.g., from data properly sur-
veyed to statistical poor existent data), and References
a different uncertainty can be associated with
the resulting damage evaluation depending on ATC-13 (1985) Applied technology council report. Earth-
quake damage evaluation data for California. Red-
the quantity and on the consistency of dataset
wood City
used. Finally, a simple approach for a first esti- ATC-21 (1988) Applied technology council report. Rapid
mation of the expected direct economic losses visual screening of buildings for potential seismic haz-
and consequences to people, based on the esti- ards: a handbook. Palo Alto
Bazzurro P, Cornell CA (1999) Disaggregation of seismic
mation of the physical damage to buildings, has
hazard. Bull Seismol Soc Am 892:501–520
been presented. Bazzurro P, Luco N (2005) Accounting for uncertainty
and correlation in earthquake loss estimation. In:
Augusti G, Schueller GI, Ciampoli M (eds) Proceed-
ings 9th international conference structural safety and
reliability (ICOSSAR), 20th–23rd June 2005, Rome,
Cross-References Italy. Millpress, Rotterdam, pp 2687–2694
Bommer JJ (2002) Deterministic versus probabilistic seis-
▶ Building Damage from Multi-resolution, mic hazard assessment: an exaggerated and obstruc-
tive dichotomy. J Earthq Eng 6:43–73
Object-Based, Classification Techniques
Bramerini F, Di Pasquale G, Orsini A, Pugliese A,
▶ Building Disaster Resiliency Through Disaster Romeo R, Sabetta F (1995) Assessing seismic risk
Risk Management Master Planning for the Italian territory. Servizio Sismico Nazionale
▶ Earthquake Mitigation Strategies Through technical report, SSN/RT/95/01, Roma (in Italian)
Brezev S, Scawathorn C, Charleson AW, Allen L,
Insurance Greene M, Jaiswal K, Silva V (2013) GEM building
▶ Earthquake Response Spectra and Design taxonomy version 2.0. GEM technical report 2013–02,
Spectra p 180
Damage to Buildings: Modeling 523
Calvi GM, Magenes G, Pinho R (2005) Displacement Hill M, Rossetto T (2008) Comparison of building damage
based methods to predict earthquake damage at vari- scales and damage descriptions for use in earthquake
able geographical territorial scale. In: Proceedings of loss modelling in Europe. Bull Earthq Eng 6(2):
250th anniversary of the 1775 Lisbon earthquake, 335–365
Lisbon Lagomarsino S, Giovinazzi S (2006) Macroseismic and
CEN (2003) Comité Europeén de Normalisation, prEN mechanical models for the vulnerability assessment of
1998-1-Eurocode 8: design of structures for earth- current buildings. Bull Earthq Eng (Spec Iss Risk-Ue
quake resistance. Part 1: general rules, seismic actions Project) 4(4)
and rules for buildings. Pr-EN 1998–1. Draft N 6, Doc NIBS (2003) HAZUS MR4 technical manual. National
CEN/TC250/SC8/N335, Jan 2003, Brussels institute of Building Sciences, Washington, DC
Charleson A (2011) Review of existing structural NIST (1999) National institute of standards and technol- D
taxonomies. http://www.nexus.globalquakemodel. ogy, UNIFORMAT II elemental classification for
org/gem-building-taxonomy/posts/review-of-existing- building specifications, cost estimating, and cost anal-
structural-taxonomies. Accessed Feb 2014 ysis, NISTIR 6389, Washington, DC, p 93. http://
Cosenza E, Manfredi G, Polese M, Verderame GM www.bfrl.nist.gov/oae/publications/nistirs/6389.pdf
(2005) A multi-level approach to the capacity assess- Pitilakis K, Argyroudis S, Kakderi K, Gehl P,
ment of existing RC buildings. J Earthq Eng 9(1):1–22 Desramaut N, Hancilar U, Taucer F, Kaynia AM,
Crowley H, Bommer JJ, Stafford PJ (2008) Recent devel- Fardis M, Tsionis G, Crowley H, Colombi M,
opments in the treatment of ground-motion variability Broglio S, Monteiro R, Franchin P, Lupoi A (2013)
in earthquake loss models. J Earthq Eng 12(S2):71–80 D.8.8 guidelines for typology definition of European
Douglas J (2011) PEER report 2011/102: ground motion physical assets. Syner-G project: systemic seismic
prediction equations 1964–2010. Pacific Earthquake vulnerability and risk analysis for buildings, lifeline
Engineering Centre, Berkeley networks and infrastructures safety gain. http://www.
EERI (2000). World housing encyclopaedia. www.world- vce.at/SYNER-G/pdf/deliverables. Accessed 20 Jan
housing.net. Accessed Feb 2014 2014
Elnashai AS, Jeong SH (2005) Rapid assessment of struc- Porter KA (2005) A taxonomy of building components for
tural system in earthquake regions. In: Proceedings of performance-based earthquake engineering, Pacific
250th anniversary of the 1775 Lisbon earthquake, Earthquake Engineering Research Center, PEER
Lisbon (CD-Rom) 2005/03, Berkeley, p 58
FEMA 154 (1988) Rapid visual screening of buildings for RISK-UE (2004) An advanced approach to earthquake
potential seismic hazards: supporting documentation. risk scenarios with applications to different European
FEMA, Washington, DC towns. Research project, European commission, DG
Foulser-Piggott R, Spence R (2013) Extending EMS-98 XII2001-2004, CEC contract number: EVK4-CT-
for more convenient application outside Europe I: 2000-00014
review of field experience using EMS-98: develop- Rossetto T, Elnashai A (2003) Derivation of vulnerability
ment of the international macroseismic Scale. In: Pro- functions for European-type RC structures based on
ceeding of Vienna congress on recent advances in observational data. Eng Struct 25:1241–1263
earthquake engineering and structural dynamics 2013 SEOAC (1995) Vision 2000, performance based seismic
(VEESD 2013) engineering of buildings. Structural Engineers Associ-
Gallagher H, Farmer B, Mendoza C, Lee C, Dickson H, ation of California, Sacramento
Greene M (2013) GEM building taxonomy v2.0: eval- Tenerelli P, Hancilar U, Taucer F, Ehrlich D, Polli D,
uation and testing report, GEM building taxonomy Dell’Acqua F, Lisini G, Crowley H, Colombi M,
global component. GEM foundation Argyroudis S, Kakderi K, Raucoules D, Gehl P,
Giovinazzi S (2005) The vulnerability assessment and the Wenzel H, Schafer D (2013) D.8.9 development of
damage scenario in seismic risk analysis. PhD disser- inventory datasets through remote sensing and direct
tation. University of Florence (I) and Technical observation data for earthquake loss estimation,
University of Braunschweig (D) SYNER-G reference report 3 (D8.9)
Giovinazzi S, Lagomarsino S (2004) A macroseismic Wailes CD, Horner AC (1933) Earthquake damage
model for the vulnerability assessment of buildings. analysed by long beach building officials, Engineering
In: 13th world conference on earthquake engineering, Xeu’s record, 25 May 1933
Vancouver Weatherill G, Crowley H, Pinho R (2011a) Efficient inten-
Giovinazzi S, Pampanin S, Lagomarsino S (2006) Alter- sity measures for components within a number of
native retrofit strategies for pre’ 70 buildings: vulner- infrastructures, deliverable D2.12 SYNER-G project:
ability models and a damage scenarios. In: First systemic seismic vulnerability and risk analysis for
European conference on earthquake engineering and buildings, lifeline networks and infrastructures safety
seismology, Geneva gain. University of Pavia (part of SYNER-G Consor-
Gr€unthal G (1998) European macroseismic scale 1998. tium). http://www.vce.at/SYNER-G/pdf/deliverables.
European Seismological Commission, Luxembourg Accessed 20 Jan 2014
524 Damage to Infrastructure: Modeling
Weatherill G, Crowley H, Pinho R, Franchin P, components, which include facilities used for
Cavalieri F, Iervolino I, Esposito S (2011b) A review generation, storage, treatment, or redistribution
and preliminary application of methodologies for the
generation of earthquake scenarios for spatially dis- of a utility, and linear components that connect
tributed systems, D2.13. SYNER-G project: systemic these facilities acting as conduits for the
seismic vulnerability and risk analysis for buildings, distribution and transmission of the service.
lifeline networks and infrastructures safety gain. http:// Earthquake-induced physical damage to infra-
www.vce.at/SYNER-G/pdf/deliverables. Accessed
20 Feb 2014 structure generates direct losses due to the cost
Whitman RV, Reed JW, Hong ST (1973) Earthquake of repairs to these components. Furthermore
damage probability matrices. In: Proceedings of the operational failure resulting from physical dam-
5th European conference on earthquake engineering, age diminishes the capacity of the infrastructure
Rome, p 2531
system to provide services to its consumers,
resulting furthermore in indirect losses due to
business interruption and subsequent long-term
economic impacts.
Damage to Infrastructure: Modeling The modeling of earthquake-induced damage
can be either probabilistic or deterministic or,
Indranil Kongar1 and Sonia Giovinazzi2 since seismic risk assessment is a multistep pro-
1
EPICentre, Department of Civil, Environmental cess, a combination of both. The benefit of prob-
and Geomatic Engineering, University College abilistic analysis is that they can account for the
London, London, UK many uncertainties associated with seismic risk
2
Department of Civil and Natural Resources assessments of infrastructure. These include
Engineering, University of Canterbury, uncertainties in the quantification of seismic
Christchurch, New Zealand hazard, the relationship between seismic hazard
and component damage, and the relationship
between component damage and component
Synonyms functionality.
In the case of seismic hazard, there are uncer-
Economic impact; Functional impact; Infrastruc- tainties over the time, location, depth, and mag-
ture systems; Lifelines; Physical impact; Seismic nitude of future earthquakes. One approach to
fragility; Seismic vulnerability account for this is probabilistic seismic hazard
assessment (PSHA) (Bommer 2002). PSHA
maps of ground motion intensity are produced
Introduction for a study area in which the values of the
intensity measure at each location represent
Infrastructure systems, also commonly known as the ground motion that is predicted to occur on
lifelines, can be defined as the “systems or net- average once within a specified return period
works, which provide for the circulation of people, based on all possible earthquake sources. The
goods, services and information upon which earthquake consequences (e.g., damages, losses,
health, safety, comfort and economic activity or casualties) are computed for these ground
depend” (Platt 1991). Infrastructure systems motions. A feature of PSHA maps is that the
include utility networks such as energy, water, evaluated ground motions are not necessarily
telecommunications, and roads or discrete critical predicted to occur simultaneously due to the
facilities such as hospitals, ports, and airports. same earthquake and therefore are not represen-
They are crucial to the routine functioning of soci- tative of a real earthquake. An alternative
ety and play an important role in emergency relief, approach which allows for more realistic repre-
reconstruction, and recovery after an earthquake. sentation of individual events is to use scenario
Infrastructure systems are made up of differ- earthquakes. A single scenario earthquake does
ent types of components, including “point” not account for the uncertainties in seismic
Damage to Infrastructure: Modeling 525
Damage to 1
Infrastructure:
0.9 P [ d = d1 | PGA ]
Modeling, Fig. 1 A
fabricated set of fragility 0.8
curves for a fictitious 0.7
P [ d ≥ di | PGA ]
structural typology
0.6
0.5 d3 = Total damage
0.4 d2 = Moderate damage
0.3
D
d1 = Slight damage
0.2
0.1 P [ d = d3 Ω PGA ]
0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Peak ground acceleration (g)
by risk assessment platforms, probabilistic func- components that exist in the system and, sec-
tions do not necessarily have to assume ondly, identification of the typologies of that
a cumulative lognormal shape. Moreover, fragil- component.
ity functions can also be deterministic, exempli- For example, an electric power network may
fied by the repair rate functions used to evaluate comprise three components: generation plants,
damage to pipelines (Kakderi and Argyroudis substations, and conduits. Each of these serves
2014). a different function in the system. Generation
Seven steps are necessary toward the model- plants are responsible for production of electric
ing of infrastructure damage. The first step is to power, conduits transport the electricity from one
produce taxonomy of infrastructures and their location to another, and substations allow for
constituent components so that elements that voltage change and/or redistribution. Then for
are expected to behave similarly are considered each of these components there are different
uniformly. The second step is to define the appro- typologies, relating to structural or operational
priate scales for classifying damage of each com- attributes. Electrical conduits, for example, can
ponent. The third step is to determine the vary based on their position (overhead lines or
appropriate earthquake hazard parameters underground cables), their insulation material, or
against which damage should be evaluated. The their size. Substations may vary by load capacity
fourth step is to generate the appropriate ground or voltage.
motions necessary to assess damage. The fifth For infrastructure systems it is not just the
step is to identify the appropriate hazard-damage damage to the individual components that is
relationship and assign damage to each compo- important but also how the whole system conse-
nent based on the ground motion. The sixth step is quently performs, e.g., is a water supply system
to assess the performance of the whole infrastruc- able to supply water in sufficient quantity at ade-
ture system based on the damages that have been quate pressure to all consumers? While damage
assigned to each component within it. The final informs direct losses, system performance is
step is to model the restoration times for infra- needed to inform indirect losses as well as emer-
structure components, which is necessary to eval- gency management decision-making. To ensure
uate indirect losses from business interruption. that system performance can be properly evalu-
This process is summarized as a flow chart in ated, classification of a single infrastructure sys-
Fig. 2, which shows the steps involved in the tem should therefore include all constituent
risk assessment of a single infrastructure system components whose damage condition may affect
due to a single earthquake. service provision. It is essential also to under-
stand the connections between these components
and their spatial distribution relative to urban
Classifying Infrastructure populations, since some components will be
more critical than others in terms of their impact
The purpose of classifying infrastructure into on system performance. Another feature of infra-
taxonomy is to group together elements that can structure systems that does not apply to building
be expected to behave similarly following an risk assessments is the interdependency between
earthquake. While buildings are discrete struc- systems, for example, the reliance of the water
tures which can be grouped together based on supply system on electricity in order to power
certain structural characteristics (e.g., material, pumping stations. It is good practice to identify
geometry, year of construction), infrastructure and document these connections in order to fully
systems are made up of different components understand the potential risk to an infrastructure.
whose differences are based on their function as Although it is possible to model all infrastructure
well as their constructive characteristics. As such systems as a single composite system of systems,
classification of infrastructure systems is in two this is computationally intensive. There is
parts: firstly, identification of the different a growing field of research investigating
Damage to Infrastructure: Modeling 527
Damage to
Infrastructure: 1. Infrastructure system classification
Modeling, • Inventory of components
Fig. 2 Procedure for • Typologies of components
assessing the risk to a single
infrastructure system due to
a single earthquake
2. Describe infrastructure damage
• Physical or operational impacts?
• Binary or multi-part damage scale? D
modeling methods that can account for such the attribute “seismic design level” refers to
interdependencies quantitatively, but at present whether the building is seismically designed or
little of this has been adopted in risk management whether the component is secured. Sometimes
practice. components are made up of a systemic arrange-
Table 1 shows a fabricated example of infra- ment of smaller subcomponents. In this case
structure taxonomy, including some important “seismic design level” refers to whether or not
infrastructure systems, their key constituent com- these subcomponents are secured to protect them
ponents, and component attributes. The compo- from ground shaking.
nent attributes are the descriptors that could be An inventory is an enumeration of the compo-
used to group components into distinct typolo- nents and facilities in each of the typologies con-
gies. Components can sometimes be pieces of sidered by the assumed taxonomy. Preparation of
equipment housed inside a building. In this case an inventory for infrastructure can be difficult
528 Damage to Infrastructure: Modeling
since attributes cannot always be identified by a wide geographical area. If such analysis of an
simple visual inspection, and in many cases the infrastructure component is required, then the
components are not visible (e.g., buried pipe- arrangement of its constituent subcomponents
lines). The inventory should therefore be pre- may also need to be documented in the taxonomy.
pared by collecting data from available sources.
In the case of infrastructure, however, usually the
only source is the system operator. To protect Describing Infrastructure Damage
commercial interests and due to concerns over
security, this type of information is not usually The damage scales used to model infrastructure
available in the public domain to a high level of damage vary depending on the type of component
detail, and so depending on the granularity of the being assessed. All networked infrastructures can
modeling exercise, it may only be feasible to be represented as an arrangement of “point” com-
work in partnership with infrastructure owners. ponents, also known as nodes and linear compo-
Some infrastructure components may need to nents, also known as edges. The nodes are the
be assessed in more detail individually because of facilities within the infrastructure system and the
their criticality or damage consequences. An edges are the components that connect the facil-
example of this would be a nuclear power station ities and connect consumers (or end users) to the
within an electric power network, since the con- service. Figure 3 illustrates a small fictitious elec-
sequences of damage to nuclear power stations tric power transmission system represented dia-
have the potential to be extremely serious over grammatically as a network, where the
Damage to Infrastructure: Modeling 529
Damage to
Infrastructure:
Modeling, Fig. 3 A
schematic diagram of
a fictitious electric power
transmission network. The
network would continue to
divide into further branches
as power is distributed to
customers and voltage is
stepped down D
generation plant and substations are nodes and no damage to total collapse. The damage scale
the overhead lines and buried cables are edges. In adopted in the HAZUS (NIBS 2003) platform, for
a potable water system, the nodes could be wells, example, accounts for no damage, minor damage,
treatment plants, or pumping stations and the moderate damage, extensive damage, and col-
edges would be pipelines. This classification of lapse. For buildings it is desirable for damage
nodes and edges determines the way in which scales to correspond to damage limit states
damage is described. Damage to edges is usually including serviceability, damage control, and col-
described per unit length, whereas damage to lapse prevention (Rossetto and Elnashai 2003).
nodes is described for the whole component. Damage scales for equipment should relate to its
Damage to infrastructure nodes can be operational state or functionality since this is
described either in terms of physical damage or most critical for evaluating indirect losses. The
operational failure. Where components are difficulty in identifying distinct physical damage
housed inside buildings, it may be necessary to states for equipment means that damage to equip-
separately assess the operational state of the ment is often assessed only in terms of its opera-
equipment and physical damage to the building. tional state. The post-earthquake functionality
For infrastructure components housed in build- can be expressed simply in binary terms, i.e.,
ings, it is possible for the structure to be signifi- whether the equipment is operational or has
cantly damaged yet the component is fully failed, which is equivalent to a single damage
operational as none of the equipment is damaged. state, or a set of fragility curves may be used to
Conversely it is possible for the structure to be represent varying degrees of functionality.
sound yet the component does not function While damage to infrastructure nodes is usu-
because equipment inside has been damaged. ally classified qualitatively, damage to infrastruc-
The purpose of infrastructure is to provide ture edges can be assessed quantitatively. The
a service to consumers, and so while assessing only type of infrastructure edge that has been
structural damage is important for calculating studied extensively in terms of seismic risk is
direct losses, it is also important to assess the pipelines, which exist in water supply, waste
operability of the component after an earthquake, water, natural gas, and fuel delivery networks.
noting that the two do not necessarily correlate. Fragility curves are not used to assess risk to
Damage to buildings is usually described in pipelines, and instead the common parameter
a scale of multiple increasing damage states from for pipeline damage is the repair rate. The repair
530 Damage to Infrastructure: Modeling
rate is a deterministic calculation of the mean permanent ground deformation (Tromans 2004),
number of damages that the pipeline is expected and those that do exist often require detailed
to experience per unit of length, usually per kilo- geotechnical data (Bird et al. 2006). While this
meter. For the damages, no distinction is made may be feasible for single-point analysis such as
between breakages or leakages. The relationship for buildings, it is not always practical to obtain
between repair rate and earthquake hazard com- this data across a spatially distributed infrastruc-
monly follows a power law relationship although ture system. Furthermore, even if ground defor-
more complex functions do exist (Kakderi and mation can be predicted, there are very few
Argyroudis 2014). Most repair rate functions relationships that can then be used to relate this
have been derived empirically, and the typical to infrastructure damage (Bird and Bommer
form of a power law repair rate function is 2004).
shown in Eq. 2, where RR is the repair rate, IM
is the earthquake hazard parameter, and a and
b are coefficients determined using some regres- Generating Ground Motions
sion technique (ALA 2001):
For modeling building damage, ground motion
RR ¼ a IM b
(2) intensities are usually evaluated by probabilistic
seismic hazard assessment (PSHA) to account for
To account for different material properties or the uncertainty in the attributes of potential earth-
soil conditions, the repair rate function may quakes over the assessment period. However, this
include additional multiplying factors which method is not suitable for spatially distributed
vary according to attribute, or a set of functions infrastructure systems. Since the ground motion
may be proposed for different conditions. intensity map produced by PSHA is a composite
statistical representation of multiple possible
earthquakes, the ground motions do not replicate
Selecting Appropriate Earthquake those that might be produced by a single real
Hazard Parameters earthquake. However, a true earthquake repre-
sentation is important for infrastructure systems.
Ground shaking hazard is commonly represented In order to evaluate the performance of a system
by either peak ground acceleration (PGA) or peak after an earthquake, it is necessary to know the
ground velocity (PGV). It is usual for above- damage state and functionality at each compo-
ground components such as substations to be nent simultaneously and hence also the ground
assessed against PGA, while buried components motion at each site simultaneously. Because
such as pipelines are assessed against PG- PSHA ground motions are usually evaluated for
V. Various studies (e.g., Isoyama et al. 2000; long return periods, they are conservative com-
O’Rourke et al. 2001) have shown that pipeline pared to the ground motions that would be gen-
damage correlates better with PGV. This may be erated by most possible earthquakes in the region.
because PGV relates to ground strains, which are This increases the likelihood that the model will
more relevant to buried components, whereas predict component failures, leading to larger
PGA relates to inertia forces, which are more reductions in system performance and
relevant to above-ground structures (Pineda- a conservative estimation of indirect losses
Parros and Najafi 2010). accrued.
Infrastructure risk can also be assessed for Infrastructure systems require a method that
earthquake-induced ground deformation hazard can evaluate seismic hazard simultaneously over
such as liquefaction or lateral spreading, although a large area and should therefore be modeled
this is less common since the prediction of per- using scenario earthquakes, with Monte Carlo
manent ground deformation is more complex. simulation or a similar technique used to account
There are very few predictive relationships for for the distribution of potential earthquake
Damage to Infrastructure: Modeling 531
attributes. In the Monte Carlo simulation, multi- producing spatially correlated ground motion
ple earthquake events are simulated, the attri- fields is included in the European seismic risk
butes of which are assigned randomly but platform, SYNER-G, and is described by
probabilistically based on data from historic Weatherill et al. (2014).
earthquakes in the region. The result is a hybrid
between deterministic and probabilistic hazard
analysis (Bommer 2002). Damage, operational Assigning Damage to Infrastructure
performance, and losses are assessed for each Components Based on Ground Motions
event, generating a distribution of impacts. Sta- D
tistical measures can be used to make inferences The method for assigning damage to infrastruc-
from the distribution, ranging from very simple ture components for a scenario earthquake varies
measures such as median or percentile values to depending on whether the relationship between
more complex statistical techniques. An impor- hazard and component damage is defined by
tant criterion for this type of analysis is that the a probabilistic fragility function or by a repair
number of events in the simulation is sufficiently rate function. For a component whose damage is
large to represent the likely distribution of defined by a fragility function, described below
attributes. are two approaches that can be applied to ensure
For scenario earthquakes, ground motions the assignment of damages to each component
can be evaluated using ground motion prediction reflects the distribution of damages defined by the
equations (GMPEs). However, since GMPEs are fragility function.
derived empirically with resulting uncertainties, The taxonomy produced in the first step of
GMPEs contain a deterministic part, which cal- modeling process has already defined and
culates the median value of ground motion and grouped the different typologies of all compo-
added to this are one or two random variables nents by attributes. The first method for damage
representing the variability that occurs within assignment involves subdividing the typologies
a single event (intra-event) and between separate further by grouping together the occurrences of
events (inter-event) (Douglas 2011). Therefore, each typology that are exposed to the same
when GMPEs are applied to a large area, adja- ground motion. If there are a large number of
cent ground motions do not correlate with each components, it may be possible to only group
other and the resulting ground motion map is occurrences exposed to exactly the same level of
random. This is not a problem for analysis at ground motion, but if they are few in number,
a single site, but for analysis of spatially distrib- then it may be necessary to define ground motion
uted systems, the requirement for simultaneous intervals and group together the occurrences that
ground motions at multiple sites means that the fall within the same interval. With the compo-
ground motions must be spatially correlated to nents grouped by attributes and ground motion,
more realistically represent an earthquake sce- it is possible to assign damage states to all com-
nario. It is possible to generate a spatially corre- ponents that reflect the probabilistic distribution
lated ground motion map by ignoring the of the relevant fragility functions at the
random variables, hence using only median corresponding ground motion level. So, for
ground motions, or assuming a constant value example, if the fragility function defines that
for the random variables. This approach is 20 % of a particular component typology
straightforward, but the distribution of uncer- experiencing a PGA of 0.5 g should fail, then
tainties is then not represented. More complex 20 % of the components of that typology being
geostatistical methods can be used to empiri- exposed to 0.5 g should be assigned to the “fail-
cally determine the spatial correlation of ground ure” damage state. This is repeated for each
motions in a region based on historic data and damage state and then for groups of the same
apply them to simulated earthquake scenarios. typology exposed to different ground motions.
These are not discussed here but one method for The whole process can then be repeated for other
532 Damage to Infrastructure: Modeling
A more detailed analysis can be undertaken to manually identify connections between all
determine how many repairs occur on each sec- source-demand pairs, and computational tech-
tion of pipe, which is useful for evaluating flow niques may be required. A simple metric of con-
reductions. In a Poisson process the probability nectivity for infrastructure is to count the number
distribution of the interval between events can be of customers who are disconnected from an infra-
described exponentially. Therefore by repeated structure system after an earthquake. However,
sampling from an exponentially distributed ran- since infrastructure systems can be modeled as
dom variable with a rate parameter equal to the networks of nodes and edges, it is possible to
repair rate, it is possible to simulate the locations apply a field of mathematics known as graph the- D
of repairs along a section of pipe (Hwang ory to compute more complex metrics of connec-
et al. 1998). The sampling continues until the tivity that capture phenomena such as redundancy
sum of all the length intervals between repairs and path length (Duenas-Osorio et al. 2007),
exceeds the length of pipe being assessed. Alter- although these are not discussed further here.
natively the number of repairs on each pipe sec- For serviceability analysis, each infrastructure
tion can also be evaluated directly by using system will have its own unique metrics and
Poisson distribution random number generation methodologies for evaluating system perfor-
with the repair rate and pipe length as parameters. mance in terms of service flow levels or quality.
The repairs can be further classified as either In general physical flow evaluations can be used
a break or a leak by sampling from a uniformly to quantify the post-earthquake capacity of an
distributed random variable between 0 and 1 and infrastructure system. For the electric power sys-
comparing this value to the corresponding prob- tem, for example, it is possible to determine flow
abilities for breaks and leaks, which are deduced at each node by solving a set of power flow
empirically from historic data. Since the simula- equations simultaneously to find an equilibrium
tion of the number of repairs on each pipe section state across the network (Ang et al. 1996), while
is just one possible network response, the simu- hydraulic analysis can be used for flow calcula-
lation should be repeated multiple times. tions in the potable and waste water systems. The
methods for each individual infrastructure system
are not described here but they are based on
Measuring System Performance Based established physical models which can be found
on System Damage Response by consulting relevant texts.
A simple serviceability metric that can be
There are two paradigms for measuring system applied to all infrastructure systems is to compare
performance: connectivity analysis and service- capacity or supply to customer demand as
ability analysis. Connectivity analysis determines a simple ratio. Another type of metric is to deter-
whether two points in the system remain mine whether an infrastructure system is able to
connected and so can be used to determine provide service at some threshold condition, e.g.,
whether a demand node (customer) remains potable water at sufficient pressure and quality or
connected to a source node. Serviceability anal- electric power at sufficient voltage. Table 2 lists
ysis determines not just whether a customer is some examples of metrics that may be used to
connected to a service but also what the quality measure serviceability for different infrastructure
of that service is. systems, although this list is not exhaustive.
To determine the connectivity between two Irrespective of the system performance metric
points only requires knowledge of the damage selected, it should be evaluated for all system
or failure state of all the components between damage responses that exist for each scenario
those points. If there is a path from a source earthquake. The final calculations for loss esti-
node to a customer node in which all components mation should include appropriate statistical
are functional, then a connection exists. For large methods to make use of the full distribution of
networks, however, it may not be feasible to impacts that this generates.
534 Damage to Infrastructure: Modeling
Damage to Infrastructure: Modeling, Table 2 List of including the uncertainties in the process. It has
serviceability system performance metrics for infrastruc- described the purpose of infrastructure system
ture systems
classification, the types of infrastructure compo-
Infrastructure system Serviceability metrics nents that exist, and the procedure for creating an
Electric power Power supply/demand ratio, inventory of infrastructures. It has explained the
voltage reduction
different uses of probabilistic fragility curves and
Potable water Water supply/demand ratio,
hydraulic pressure reduction deterministic repair rate functions to evaluate
Waste water Capacity/demand ratio, damage for infrastructure nodes and edges,
probability of overflow respectively, as well as the differences between
Natural gas Gas supply/demand ratio descriptions of physical damage and operational
Fuel Fuel supply/demand ratio failure. Damage to infrastructure can be com-
Telecommunications Probability of call not being pared to both peak ground acceleration and peak
connected ground velocity, with the former most relevant
Highways Travel time delay, increase in
for above-ground components and the latter most
travel distances
appropriate for buried components. Although
many infrastructure components are susceptible
Restoration Functions to damage caused by ground deformation, there
are no established methods for predictive model-
The purpose of restoration functions is to evalu- ing of this risk. Although it is common to use
ate the time it might take for damaged infrastruc- probabilistic seismic hazard assessment to evalu-
ture components to be repaired. This is important ate ground motions for building damage assess-
for determining indirect losses, which are ment, for infrastructure systems, because of their
affected by the outage time of infrastructure sys- systemic and spatially distributed nature, it is
tems. A restoration function will provide infor- more valuable to use realistic scenario earth-
mation on the average repair time for quakes reinforced by simulation techniques to
a component based on its damage state or the account for uncertainties in earthquake attributes
average percentage of repairs that will take if this is required. The systemic nature of infra-
place within a specified time period (Cagnan structure means that it is important to know
et al. 2006). Such functions may be derived which, and not just how many, components are
empirically or be based on assumptions for avail- damaged. Damage can be assigned to specific
able resources and work rate. If derived empiri- components by either grouping components
cally there will be some uncertainty associated together by typology or by considering each com-
with the function, and this should be accounted ponent individually, but in both cases uniform
for in its application. Figure 4 shows a fabricated random assignment is used and so a simulation
example of a component restoration function. By technique is required to account for uncertainties.
evaluating the repair times for components, it is For pipelines it is often assumed that the distri-
possible to reevaluate system performance met- bution of damage follows a Poisson distribution.
rics at time steps after the earthquake based on Similar methods involving random assignment
improved system conditions. This in turn means and simulation techniques are used to locate dam-
that business interruption impacts and indirect ages making use of exponentially distributed
losses can be better estimated. rather than uniformly distributed variables.
Because of the uncertainties in damage assign-
ment, there should be multiple damage state sce-
Summary narios for each earthquake. System performance
can be measured either in terms of connectivity
This entry has summarized the importance of between source nodes and consumers or as
modeling earthquake-induced damage to infra- a measure of the level of service being provided.
structure systems and the steps involved Connectivity analysis is more straightforward
Damage to Infrastructure: Modeling 535
Damage to 100
Infrastructure:
Modeling, Fig. 4 An 90
example of a restoration 80
since it only requires knowledge of the network ▶ Probabilistic Seismic Hazard Models
topology and does not require physical models to ▶ Resiliency of Water, Wastewater, and
be solved. However level of service analysis can Inundation Protection Systems
generate a more informative view of the impacts ▶ Seismic Actions Due to Near-Fault Ground
of damage to the system. Finally restoration func- Motion
tions can be used to estimate the length of time it ▶ Seismic Design of Pipelines
will take to repair the components. Since this ▶ Seismic Fragility Analysis
method for modeling infrastructure damage uses ▶ Seismic Loss Assessment
simulation techniques for both earthquake hazard ▶ Seismic Vulnerability Assessment: Lifelines
and damage state assignment, it will not produce ▶ Seismic Vulnerability Assessment: Reinforced
a single outcome but rather a distribution of Concrete Structures
potential outcomes. Statistical measures and ▶ Site Response for Seismic Hazard Assessment
techniques therefore need to be applied to make
inferences from this distribution.
References
Cross-References Adachi T, Ellingwood BR (2008) Serviceability
of earthquake-damaged water systems: effects of
▶ Earthquake Risk Mitigation of Lifelines and electrical power availability and power backup sys-
Critical Facilities tems on system vulnerability. Reliab Eng Syst Saf
93:78–88
▶ Economic Impact of Seismic Events:
ALA (2001) Seismic fragility formulations for water sys-
Modeling tems: part 1 – guidelines. American Lifelines Alliance,
▶ Empirical Fragility Federal Emergency Management Agency, Washing-
▶ Estimation of Potential Seismic Damage in ton D.C.
Ang AH, Pires JA, Villeverde R (1996) A model for the
Urban Areas seismic reliability assessment of electric power trans-
▶ Insurance and Reinsurance Models for mission systems. Reliab Eng Syst Saf 51:7–22
Earthquake Bazzurro P, Luco N (2005) Accounting for uncertainty
▶ Liquefaction: Performance of Building and correlation in earthquake loss estimation. In:
Augusti G, Schueller GI, Ciampoli M (eds) Proceed-
Foundation Systems
ings 9th international conference structural safety and
▶ Post-Earthquake Diagnosis of Partially reliability (ICOSSAR), 20–23 June 2005, Rome.
Instrumented Building Structures Millpress, Rotterdam, pp 2687–2694
536 Design of Cable-Supported Bridges: Control Strategies
Even though evolution in cable materials design, which mainly depends on the specific
towards strength and weight is nowadays impres- structural configuration, and by a low damping
sive, cable-stayed bridges are limited by the stiff- of the stay cables or hanger ropes. The flexible
ness of their superstructure. Simply put, the design may lead to unwanted motion or vibra-
weight of the bridge superstructure would create tions that could compromise the bridge safety
a sag in the cable, leading to nonlinear behavior. and serviceability performance and even lead
For this reason, for longer spans, the suspension to immediate structural failure and also have
scheme, in which the deck is supported by sus- an influence on the fatigue performance. Excita-
pension cables on vertical hanger ropes, is pre- tions due to wind, earthquake, or anthropic loads D
ferred. The suspension bridge with the longest (railway, traffic, pedestrian loads, etc.) are the
central span in the world is the Akashi Kaikyo principal causes of unwanted vibrations.
Bridge in Japan, completed in 1998, with a cen- Unlike conventional bridges, cable-suspended
tral span of 1,991 m. The feasibility of longer bridge design is not covered by bridge codes,
spans has been demonstrated, with different standards, and specifications. Nowadays,
design paradigms at a conceptual or preliminary performance-based design (PBD) is mainly
stage. employed for the seismic design of cable-
Nowadays, cable-suspended bridges are suspended bridges (Bontempi 2006).
designed in different schemes (e.g., self- The dynamic wind action is the principal
anchored, multi-span), in different materials design concern (especially for longer spans),
(essentially steel and concrete), and with differ- since it affects different parts of the bridge
ent deck types (e.g., truss, plate). Several varia- (deck, pylons, main cables, and hanger ropes of
tions to the structural form have been a suspension bridges; stay cables of a cable-
successfully built as a variation of either forms stayed bridge). Thus, the principal concern is
(self-anchored suspension bridges, side-spar the alleviation of wind-induced vibration. The
cable-stayed bridges, etc.) or as a combination commonly observed aeroelastic phenomena
(cable-stayed-suspension hybrid bridges). (wake galloping, torsional and coupled flutter,
In general, cable-supported bridges, due to the and vortex-induced vibrations) have a strong
fact that they are supported by cables which con- influence on the bridge behavior and are usually
stitute elastic supports, behave very well under suppressed or controlled by appropriate design
earthquake excitation. Nevertheless, structural measures and control strategies. Furthermore,
control for seismic loads is equally a concern bridge substructures and components have been
for seismically active regions (with several shown to act as bluff bodies, resulting in self-
cases in the USA, Japan, China, Turkey, and excited vibration of the bridge.
Greece). The issue is the structural behavior to Vibrations due to anthropic (pedestrian, rail-
long-period and long-duration ground motion. way, and traffic) loads can be of issue on a case-
The principal seismic concern is the pounding by-case basis, depending on the bridge type and
between the deck and the abutments. This also use. Other additional effects can be an issue
accounts for the differential motion at the support (e.g., the rain-induced vibration that mostly
system (towers), influenced also by the spatially affects cables). Design solutions and counter-
varying soil conditions. measures for these vibration effects have been
applied, after meticulous studying during the
years, starting from notable disasters and fail-
Performance Requirements and Design ures (e.g., the wind-induced collapse of the
Issues Tacoma Narrows Bridge in 1940 caused by aero-
dynamic flutter or the synchronous lateral exci-
Both cable-stayed and suspension bridges have tation of the pedestrian London Millennium
unique characteristics compared to ordinary Footbridge in 2000). In the latter, frame-
structures and are characterized by a flexible mounted viscous dampers and tuned mass
538 Design of Cable-Supported Bridges: Control Strategies
vibration absorbers were retrofitted after the ini- There are reports of ordinary bridge failures
tial undesirable performance of the bridge. that may have been caused or aided by differen-
Regarding earthquake excitation, due to its tial support motion during earthquakes (Burdette
(typically) very long fundamental periods, seis- et al. 2008). However, it is difficult to fully attri-
mic behavior of cable-supported bridges is in bute these failures to asynchronous earthquake
general satisfactory and easily reparable with motion, even though the phenomenon contributes
extraordinary repairs (replacement of fuse to large longitudinal displacements between
restrainers, etc.). Seismic loads have in general pylons. Nevertheless, different numerical ana-
a direct influence on the foundations and towers, lyses have shown that in long-span suspended
while the deck is indirectly influenced. bridges, the spatially varying ground motion
For seismic design, performance-based earth- may result in bigger structural response than the
quake engineering (PBEE) design entitles bridge synchronous ground motion.
owners and stakeholders to decide on the behav-
ior of the suspension bridge under different
design solutions in terms of seismic vulnerabil- Structural Control
ity assessment. Usually, two seismic hazard
levels are identified, with performance criteria The mitigation of extensive vibration or motion
related to safety or serviceability: the design has been a concern in the civil engineering field
earthquake and the maximum considered earth- since the early conception of complex structural
quake. These hazard levels are characterized by systems (high-rise buildings, long-span bridges),
different probabilities of exceedance (or return characterized by nonlinear behavior. The issue is
periods) for the earthquake. All other hazards to control how input energy (from strong wind,
(e.g., high wind) are fulfilled also on the basis earthquake, etc.) is absorbed by a structure, by
of appropriate limit states (e.g., ULS, Ultimate means of different methods and techniques, as an
Limit States; SLS, Serviceability Limit States), alternative to conventional design methods based
while additional limit states (e.g., durability, on ductile response. Most commonly
fatigue, impact, etc.) are appropriately defined implemented methods include installing isolators
on the basis of the tender design document and or passive energy dissipation devices to dissipate
the bridge site. vibration energy and reduce dynamic responses.
For long-span bridges, one of the principal In addition to these, with the advent of advanced
design issues is the asynchronous earthquake computational methods (hardware and software),
excitation. While a normal structure frequently it is now possible to alter the structural configu-
has a rather limited imprint, a long-span bridge is ration for mitigating the induced energy. In long-
supported by towers placed sometimes hundreds span bridge engineering, recent bridges mostly in
of meters away. Thus, the seismic excitation from Japan (e.g., the 1993 Rainbow Bridge in Tokyo)
pylon to pylon can vary in time and intensity, and deploy such more advanced, “active” control
a uniform ground motion assumption is not real- strategies, especially for the transitory phase of
istic anymore. construction.
Although the phenomenon is complex, the
spatial variation of an earthquake mainly is Structural Control Strategies
divided for simplification in three effects and In the civil engineering field, the term “structural
conditions: the geometric incoherence effect, control” with explicit reference to “active con-
the wave passage effect, and the local soil condi- trol” was introduced in 1972 by Yao (1972).
tions. The asynchronous earthquake motion in Since then, the increase in the conception and
bridges has shown to have an effect on the deck design of complex structural systems and the
displacement during earthquake, which in cable- increased demand for structural performance
supported bridges is in part mitigated by the lead to numerous studies, design cases, and appli-
cables. cations of structural control for such structures.
Design of Cable-Supported Bridges: Control Strategies 539
a c
Computer
Sensors Sensors
Excitation Structure Response Controller
No control Control
Actuators
b PED D
Excitation Structure Response Excitation Structure Response
Control Control
Actuators Actuators
PED
PED
These were also motivated by devastating disas- Passive control is achieved by employing a
ters through the years (earthquakes, tsunamis, passive energy device (PED), e.g., a Tuned Mass
typhoons, etc.) that put in evidence the issues of Damper or mechanical energy dissipation tool,
structural safety and reliability. that responds to the motion of the structure to
Structural control systems can be grouped into dissipate energy or by isolating the structure
four broad categories (Housner et al. 1997): from the ground and horizontal loads (base isola-
tion systems). Passive control is employed exclu-
– Passive control sively by the motion of the structure, and no
– Active control external force or energy is applied to effect the
– Semi-active control control (Fig. 1b). It is nowadays a broadly
– Hybrid control accepted vibration control strategy, with many
consolidated cases worldwide, while in the recent
Figure 1 (adapted from Spencer and Soong years, innovative approaches have been applied
1999) provides an overview of the different con- in long-span bridge engineering. Typical applica-
trol strategies, together with the no-control case tions of passive control in cable-supported brid-
(Fig. 1a). ges are in the form of dampers, placed principally
540 Design of Cable-Supported Bridges: Control Strategies
on the deck, pylons, and the hangers, and base that the passive control part can function also
isolation schemes on the foundations. during a power failure, thus providing a mini-
Active control, on the other hand, is a mum degree of protection.
response-triggered control method. In other Considering the above, there are significant
words, an external force is applied to the structure advantages and disadvantages for each control
every time a displacement or vibration is mea- method. Also, optimal fields of application can
sured by means of a control actuator. The latter be significantly different: in general, consoli-
takes input from sensors that measure the external dated passive methods are preferred for their
excitation (open-loop control) and, in some cases, cost-effective solution and reliability. However,
the structural response (closed-loop control) – in the last years, alternative active, semi-active,
(Fig. 1c). Although in the last three decades or hybrid methods are implemented in case a
numerous methods have been proposed and passive method results inadequate (two cases
many researches have been carried out, practical worth mentioning are the implementation of
applications in cable-supported bridges are rather active methods as a support to passive ones dur-
limited due to cost and reliability issues. Most of ing the tower erection and the use of semi-active
the applications regard pylon control during erec- tendon control). Finally, different control
tion, as in the case of the Akashi Kaikyo Bridge. methods and devices are preferred for different
In fact, a change in the pylon height leads to a parts of the structure.
change in its natural frequency, something that
poses as a challenge to deal with using a small Structural Control Issues
number of mass dampers. Taking into account what was said above, the
In between, two other methods have been following sections overview control devices and
widely studied (and applied in a few cases world- strategies with respect to different parts of cable-
wide) that try to combine advantages and exclude supported bridges. On the basis of consolidated
shortcomings of the abovementioned control cases, Fig. 2 synthesizes the control options.
strategies. It should be noted that many of the issues can
In semi-active control, the control actuator be dealt with an improved design or structural
does not directly apply force to the structure modifications. This is the case, for example, for
(as in the case of active control), but instead it is flutter- and vortex-induced vibration suppres-
used to control the properties of a passive energy sions on bridge deck, obtained by means of
(Fig. 1d). Within the class of semi-active control, section modifications or aerodynamic counter-
the use of controllable electrorheological (ER) and measures (e.g., deflectors, fairings, and flaps).
magnetorheological (MR) fluids in dampers has Similarly, an appropriate designed cable cross
found ground in the recent years, with applications section and surface can minimize cable vibration.
mostly on the stay cables of cable-stayed bridges.
Finally, hybrid control consists in the com- Foundation Isolation
bined application of both passive and active Foundations transmit lateral loads to the entire
methods. This means that a part of the control bridge through the connection to the other parts
derives from a passive control device, something (principally deck and cables). The typical control
that alleviates the required effort in terms of strategy in long-span bridge design is base isola-
forces and energy for the active device (Fig. 1e). tion. Base isolation, that is, to decouple the bridge
Hybrid control is applied as an alternative to superstructure from its foundations, is now con-
active control during the erection of towers of sidered a mature technology. A modern interpre-
suspension bridges. Spencer and Nagarajaiah tation of this strategy has been applied in the
(2003) report several cases of hybrid control in recent years in several long-span suspension and
bridges during tower erection. cable-stayed bridges.
Compared to pure active control, semi-active Two notable cases are the 2004 Rio-Antirrio
and hybrid control strategies have the advantage multi-span cable-stayed bridge in Greece and the
Design of Cable-Supported Bridges: Control Strategies 541
Main Cables
Hanger Ropes / Stay Cables
Towers
Deck
Suspension Bridge Foundations Cable - Stayed
Bridge
D
Design of Cable-Supported Bridges: Control Strategies, Fig. 2 Control strategies for different substructures of
cable-supported bridges
TMD–3
TMD –3 HMD
TMD – 1 TMD–1
TMD–2
Erection crane
TMD – 2
Inside tower shaft Slewing frame
Elevating deck
Tower TMD–E3,E4
TMD–E1,E2
Post
Design of Cable-Supported Bridges: Control Strategies, Fig. 4 TMDs on the Akashi Kaikyo towers (Courtesy of
“Honshu-Shikoku Bridge Expressway Co., Ltd.”)
TLDs (Tuned Liquid Damper systems) dissipate (Fujino and Siringoringo 2013). The first appli-
energy using a secondary mass in the form of a cation of active mass dampers (AMD) to a bridge
body of liquid introduced into the structural sys- pylon was on the 1991 Rainbow suspension
tem and tuned in such a way to act as a dynamic bridge in Tokyo, Japan. The AMDs were placed
vibration absorber. TLDs are particularly effec- on top of the outer edge of the 126 m high pylons
tive for the suppression of lateral vibrations (e.g., to help reduce large-amplitude vortex-induced
due to wind or pedestrial loads) and were used for vibration. During erection, the Akashi Kaikyo
vibration control of the Toda Park Bridge in Bridge pylons were equipped with two sets of
Japan. Similar devices to TLDs are TSDs Hybrid Mass Dampers (HMD), a combination
(Tuned Sloshing Dampers) and TLCDs (Tuned of a combination of passive TMD and active
Liquid Column Dampers). control actuator. Figure 4 (Courtesy of “Hon-
Another important issue is the control of the shu-Shikoku Bridge Expressway Co., Ltd.”) out-
pylons during erection and also during the cable lines the position of the vibration control devices
spinning, since their frequency varies signifi- on the bridge tower.
cantly in that phase. The use of passive dampers
(mostly Tuned Liquid Dampers) proves to be Deck Control
ineffective since they have to be constantly As in the case of the pylons, the deck of a long-
tuned to the new pylon configuration. In the past span bridge is exposed to turbulent wind and
20 years, this problem has been dealt with the thus to different wind-induced phenomena.
implementation of active and hybrid control Depending on the bridge configuration, the
strategies. This is a particularly consolidated bridge deck is subjected from minor to average
practice in Japan, with several cases reported solicitation from earthquakes. The former is
Design of Cable-Supported Bridges: Control Strategies 543
particularly truth for bridges with a deck totally Shock Transmission Units (STUs), otherwise
suspended (without deck-pylon connections). known as Lock-Up Devices (LUDs), are devices
Vibrations and displacements due to anthropic that allow the slow movement of the deck when
loads (railway, highway, and pedestrian) have to the bridge structure slowly expands and contracts
be controlled as well. (under thermal loads or creep) but lock-up (under
An important issue is the pounding effect that seismic or wind). In this way, the structural sys-
usually takes place during strong earthquakes tem responds differently to slowly applied loads
between the main span and the approaching (daily temperature variation and creep) and fast
span of long suspended bridges or the main span loads (impact, earthquake, wind gusts). Thus, D
and the pylon of self-anchored suspended bridges large seismic or extreme event forces are
and may cause considerable damage of the col- restricted from the bridge towers.
liding bridge deck segments, in correspondence STUs have been used broadly in major cable-
to the position of the expansion joints. This is supported bridges. Patel (2013) reports several
important considering that the vibration periods applications both in suspension bridges (the
of the flexible and usually unrestricted main span Storebaelt in Denmark) and cable-stayed bridges
are different from the vibration period of the (the Sidney Lanier in the USA, the Second
pylon or the approaching spans. Severn Bridge in the UK, the Seohae Grand
For long spans the usual design practice now- Bridge in South Korea, and the Stonecutters
adays for seismic isolation is to have a fully or Bridge in Hong Kong). In many cases, the con-
semi-floating deck (where the deck is suspended cern was not only seismic but wind loads as well
from stay cables or hanger ropes and there is little (Fig. 5).
or no contact at the pylons through lateral control Fluid viscous dampers (FVDs) are passive
devices and bearings). control devices that dissipate energy by applying
Control devices are placed usually between a resisting force over a finite displacement. The
the deck and the pylons and the main span and damper’s output force acts in a direction opposite
the approaching span, and consolidated practice to that of the input motion. Such dampers consist
indicates the implementation of passive devices. usually of a hollow cylinder filled with a fluid
Again, design decisions drive the adoption of (typically silicon based). As the damper piston
control devices. In fact, a compromise has to be rod and piston head are stroked, fluid is forced to
made for what regards the longitudinal move- flow through orifices either around or through the
ment and the possible deck-pylon connection, piston head, something that produces a force in
considering both impact forces (earthquake, the opposite direction. The friction between fluid
wind, vehicle braking) and slow developed particles and the piston head leads to energy
stresses due to the temperature gradient. The dissipation in the form of heat. Fluid viscous
compromise lies in choosing between a fully dampers can be tuned to have either linear or
floating system, prone to an elevated movement, nonlinear behavior, depending on the specific
and a fixed deck-pylon that results to a more rigid application. A strongly nonlinear behavior
system. For very long spans either case is impos- means that the opposing force is very low for
sible, and in many cases, an intermediate solution low applied forces but can be high as the applied
is adopted, with the deck-pylon being connected loads increase.
with damping systems that provide limitations to FVDs have been implemented in many long-
the free movement of the deck. span suspended bridges in the last years, espe-
Shock Transmission Units and (usually fluid) cially in China. An example is the Xihoumen
dampers are implemented to limit longitudinal Bridge with a main span of 1,650 m (currently
deck movements, of the otherwise undamped the second longest suspension bridge in the
deck, and, in the case of suspension bridges, to world) that relies on four stroke dampers that
avoid pounding between the stiffening trusses have a range of movement of 1,100 mm
and the towers. installed at the end of the bridge deck.
544 Design of Cable-Supported Bridges: Control Strategies
Design of Cable-Supported Bridges: Control Strategies, Fig. 5 STU units under installation in the Stonecutters
Bridge (From Patel 2013)
Other innovative ideas include the case of the continuous deck is free to accommodate the
Sutong Bridge, the world’s second longest cable- movements from seismic excitation in the longi-
stayed bridge with a span of 1,088 m and a totally tudinal direction, since it is fully suspended. The
suspended deck; eight isolators are used to reduce movements in the transverse direction are con-
the axial motion of the suspended deck (Fig. 6). trolled by nonlinear fluid viscous dampers and
Each isolator consists in a combined spring- fuse restraints installed parallel to the dampers
damper element (consisting in a fluid damper, (Fig. 7). In this way, the deck is linked rigidly to
an elastomer spring, and an internally placed the substructure when subjected to lateral loads
gapping mechanism). For the initial 85 % of the not exceeding the design capacity. In major seis-
available displacement from the neutral position, mic events, the fuse restrainers are designed to
the isolator has only damping response (that fail and leave the dampers free to dissipate the
accommodates also the daily thermal expansion earthquake-induced energy acting upon the struc-
and contraction of the suspended deck) and for ture. After their failure the deck is free to swing
the rest 15 %, combined spring and damping coupled to the dampers.
response. This, since the long period of the Other passive control devices (solid viscoelas-
deck, provides intrinsic vibration control in the tic dampers, metallic dampers, friction dampers,
longitudinal direction, while additional displace- etc.) have a different hysteric behavior with sev-
ment has to be limited in the longitudinal direc- eral advantages and disadvantages also for what
tion for other events (e.g., vehicle braking forces, concerns their reliability.
typhoon loads). The control device has a maxi- In the case of the Tianxingzhou Yangtze River
mum displacement of 85 cm of which about Bridge in China (a bridge with a mail span of
10 cm is the final combined spring and damping. 504 m, the longest combined road and rail cable-
Lateral movement is also an issue, especially stayed span in the world), a combination of
for floating deck configuration. For example, in magnetorheological (MR) and hydraulic dampers
the Rio-Antirrio cable-stayed bridge, the is used in correspondence to the piers, to limit
Design of Cable-Supported Bridges: Control Strategies 545
Design of Cable-Supported Bridges: Control Strategies, Fig. 6 Sutong Bridge pylon-deck viscous damper system
(above) and device schematic (below) (From You et al. 2008)
deck vibrations. MR dampers consist of a hydrau- such as snowstorms) or, indirectly, due to the
lic cylinder containing a solution that, as their vibrations of the supported deck. To suppress
name suggests, can reversibly change, when these vibrations, commonly passive devices are
exposed to a magnetic field or electric current, used (hydraulic, oil, rubber, or viscous dampers).
from a free-flowing liquid to a semisolid with Usually these devices are attached to the cables
controllable yield strength. These devices allow near the deck anchorage, although other configu-
to actively control impact forces (due to train rations are possible.
braking and earthquakes), where the hydraulic More recently, adaptive dampers have been
damper provides a velocity-related recovery implemented on many occasions. Their principal
(damping) force that relieves slow movements advantages compared to passive dampers lie in
(e.g., due to daily temperature change). the easy tuning (by adjusting the current or volt-
age) both before and after the installation and for
Cable Control different modes of vibration.
Stay cables and hanger ropes are characterized by Duan et al. (2006) report the first successful
very low internal damping, relatively small mass, application of MR dampers for the retrofitting of
large flexibility, and considerable length. They the Dongting Lake Bridge in China. A total of
are susceptible to vibrations due to direct loading 312 dampers were installed on 156 cables of the
(caused by wind, rain, and other extreme events bridge (two for each cable), at a height of 1.8 m
546 Design of Cable-Supported Bridges: Control Strategies
Design of Cable-
Supported
Bridges: Control
Strategies,
Fig. 7 Vibration control
on the deck of the
Rio-Antirrio Bridge
(courtesy of Gefyra SA)
Cable
smart damper
st
ay
ca
Cable-stayed MR Damper
bl
es
bridge b
MR Damper
Design of Cable-Supported Bridges: Control Strate- et al. 2006) and the Shandong Binzhou Yellow River
gies, Fig. 8 Cable-stayed vibration control with MR Bridge (right, from Li et al. 2007)
dampers at the Dongting Lake Bridge (left, from Duan
from the bridge deck, and in an inclined T-shape Other recent applications of MR dampers for
support configuration (Fig. 7). A similar cable control include the Jiashao Bridge in China,
approach has been used in the retrofitting of the the world’s longest and widest multi-span cable-
Shandong Binzhou Yellow River Highway stayed bridge, and the Sutong Bridge.
Bridge (a three-tower cable-stayed bridge in
China, with two main spans of 300 m). The
bridge suffered unaccepted vibrations after its Overview of Control Methods
opening in 2004 and was retrofitted with MR The above are reassumed in Table 1. Many
dampers on 20 of its longest stay cables issues can be minimized by adopting an
(Li et al. 2007; Fig. 8). appropriate structural shape. For example, deck
Design of Cable-Supported Bridges: Control Strategies 547
Design of Cable-Supported Bridges: Control Strategies, Table 1 Overview of principal control issues and
methods for cable-suspended bridges
Substructure Principal issue(s) Control type Control device
Foundations Earthquake Passive (seismic Foundation design
isolation)
Towers Vortex-induced Passive TMD (Tuned Mass Dampers)
vibrations
Earthquake Active TLD (Tuned Liquid Dampers)
Semi-active TSD (Tuned Sloshing Dampers) D
Hybrid TLCD (Tuned Liquid Column
Dampers)
HMD (Hybrid Mass Dampers)
Stay cables and hanger Wind Passive Viscous dampers
ropes Rain Active MR (magnetorheological) dampers
Ice
Deck Vortex-induced Passive FVD (fluid viscous dampers)
vibrations
Earthquake Semi-active STU (Shock Transmission Units)
Pedestrian TMD
Vehicle TLD
Train
vortex-induced vibration can be suppressed by east of the bridge. The construction of the foun-
changing the section or by using aerodynamic dations, both towers, and main cables had been
appendages or flaps. completed before the earthquake occurred, but
neither hangers nor girders had been attached to
the main cables, while the stiffening girders were
Seismic Response of Cable-Suspended not yet in place. The bridge towers were equipped
Bridges with temporary semi-active mass dampers, tuned
for wind excitation.
In the last 30 years, many new bridges were There was no damage to bridge structures, but
constructed, and many old bridges were displacements of the four foundations were
retrofitted with instrumentation systems. Mea- caused by crustal movements. There was also an
surements from these systems during major seis- 80 cm elongation at the center span and 30 cm at
mic events are allowed to assess the seismic one of the side spans. There was also a change of
performance of the structures, the performance angle horizontally and vertically at the towers
of the control systems, and the post-earthquake and at the anchorages. The construction reprised
behavior of the bridge. A number of significant accounting the above facts in the updated design
case studies based on SHM (Structural Health (Kitagawa 2004).
Monitoring) data or field investigations are
reported below. Higashi-Kobe Bridge
The Higashi-Kobe Bridge is a cable-stayed
Akashi Kaikyo Bridge bridge with a center span of 485 m. With the
The Akashi Kaikyo Bridge, presently the suspen- aim of augmenting the fundamental period of
sion bridge with the longest span in the world, bridge, the bridge deck is supported by towers
was hit by the Hanshin Kobe earthquake while and piers in such a way that the girder is movable
under construction. The epicenter of the 6.8 mag- in the longitudinal direction independently from
nitude earthquake was located just about 2–3 km the towers; thus, the displacements are mainly
548 Design of Cable-Supported Bridges: Control Strategies
restricted by the cables. The bridge is equipped events by means of different mechanisms
with two vane-type oil dampers at the end piers, between the pylon and the soil and the deck
and the cables are arranged in a harp pattern, in and the pylons and abutments (where a dissipa-
order to prevent an excessive longitudinal dis- tion system with hydraulic dampers is installed).
placement. The vane-type dampers help in The seismic design was tested during the
preventing excessive displacement since its 2008 6.5 magnitude earthquake about 35 km
effect increases with the velocity of the girder from the bridge location, which had no perma-
movement. During the 1995 Hanshin Kobe nent effects to the bridge. The bridge suffered
earthquake, the bridge suffered significant very mild damage, principally in the lateral
yet repairable damage (Naganuma et al. 2000). restrainers (sacrificial elements) that yielded
The upper set of bolts of the transverse wind (Papanikolas et al. 2010). As a remedial, the
brace on the end pier broke, and the upper shoe deck was realigned and six fuses were replaced.
was taken off, something that put excessive
force on the vane-type dampers that were not Ji-Lu Bridge
designed to carry transversal loads. It should be The Ji-Lu Bridge is a single pylon cable-stayed
observed that the design acceleration was signif- bridge with prestressed concrete girders that suf-
icantly lower than the one monitored during the fered important damage in September 1999 dur-
event. ing the 7.7 magnitude Chi-Chi, Taiwan
earthquake (Kosa and Tasaki 2003). The
Yokohama Bay Bridge 120 + 120 m bridge, supported by a single con-
The Yokohama Bay Bridge with a total length crete pylon, was under construction at the time.
of 860 m and a central span of 460 m is a The damage was widespread, extending over the
densely instrumented cable-stayed double deck main girder, the main pylon, and the two piers,
highway bridge in Japan. The bridge opened in due to the seismic force that acted in both the
1989 and was already retrofitted in 2005 for axial and the perpendicular directions of the
additional seismic safety. Specifically, to bridge. Pounding occurred between the main
accommodate longitudinal girder displacement span and the approaching spans and between the
in case of large earthquake, LBCs (link-bearing superstructure and the towers. Also, one cable
connections) were implemented: 10 m long was put out from the anchorage section of the
end-links at the two end piers and 2 m long pylon, and many sockets were also pulled out
tower links at the two towers. The bridge is from the anchorage section.
located 180 km from the closest fault area of
the 2011 9.0 magnitude Great East Japan Shipshaw Bridge
(Tohoku) Earthquake, notably the largest earth- The Shipshaw Bridge is an asymmetrical single
quake in Japan’s modern history. The bridge pylon cable-stayed bridge in Quebec, Canada,
experienced large transverse displacements dur- with a deck 183 m long supported by a double-
ing the earthquake (55 and 62 cm in the top of leg steel tower. At a later inspection after the
the tower and the mid-span, respectively) and 1988 Saguenay 6.0 magnitude earthquake
very small longitudinal displacements (15 cm). (Filiatrault et al. 1993), it was found that one of
No damage was detected. The excessive trans- the four interior anchorage plates in one abut-
versal and longitudinal motion leads to scratch ment failed. Although the damage cannot be
marks on the wind shoes and a few crushed nuts linked directly to the earthquake, later analysis
and bolts on one of the tower links (Siringoringo point towards this, together with the fact that
et al. 2013). the anchorage plates were already subjected to
high stress concentrations. The bridge was
Rio-Antirrio Bridge immediately closed, and all four plates in corre-
The Rio-Antirrio Bridge is a multi-pylon cable- spondence to that abutment were replaced
stayed bridge designed to withstand seismic (Table 2).
Design of Cable-Supported Bridges: Control Strategies 549
Design of Cable-Supported Bridges: Control Strategies, Table 2 Notable earthquake events and bridge behavior
Length
(m) (total/ Magnitude
Bridge Type longest span) Seismic event (Mw) Damage References
Akashi Suspension 3,911/1,991 1995 Hanshin 6.8 No damage Kitagawa
Kaikyo Kobe earthquake 2004
Bridgea
Higashi- Cable stayed 885/485 1995 Hanshin 6.8 Vane-type Naganuma
Kobe Kobe earthquake dampers et al. 2000
Bridge D
Yokohama Cable stayed 860/460 2011 Great East 9.0 No damage Siringorino
Bay Bridge Japan et al. 2013
Earthquake
Rio- Cable stayed 2,252/560 2008 Achaia-Ilia 6.5 Fuse Papanikolas
Antirrio (multi- earthquake replacement et al. 2010
Bridge pylon)
Ji-Lu Cable stayed 240/120 1999 Chi-Chi, 7.7 Pounding, Kosa and
Bridge (single Taiwan cable Tasaki 2003
pylon) earthquake detachment,
Shipshaw Cable stayed 183/13,725 1988 Saguenay 6.0 Anchorage Filiatrault
Bridge (single earthquake plate failure et al. 1993
pylon)
a
During construction
Burdette NJ, Elnashai AS, Lupoi A, Sextos AG congress incorporating the PCI annual convention and
(2008) The effect of asynchronous earthquake motion bridge conference, Washington, DC, 29 May–2 June
on complex bridges. Part 1: methodology and input Patel DJ (2013) Shock transmission units in construction.
motion. J Br Eng ASCE 13(2):158–165 ICE, London
Duan YF, Ni YQ, Ko JM (2006) Cable vibration control Shen WA, Zhu S (2014) Harvesting energy via electromag-
using magnetorheological dampers. J Intell Mater Syst netic damper: application to bridge stay cables. J Intell
Struct 17(4):321–325 Mater Syst Struct. doi:10.1177/1045389X13519003,
Filiatrault A, Tinawi R, Massicotte B (1993) Damage to Published online before print 21 Jan 2014
cable-stayed bridge during 1988 Saguenay earthquake. Siringoringo D, Fujino Y, Namikawa K (2013) Seismic
I. Pseudostatic analysis. J struct Eng 119(5):1432–1449 response analyses of the Yokohama Bay cable-stayed
Fujino Y, Siringoringo D (2008) Structural health moni- bridge in the 2011 Great East Japan Earthquake. J Br
toring of bridges in Japan: an overview of the current Eng ASCE. doi:10.1061/(ASCE)BE.1943-
trend. In: Fourth international conference on FRP com- 5592.0000508, A4014006
posites in civil engineering (CICE2008), Zurich, Spencer B, Nagarajaiah S (2003) State of the art of struc-
22–24 July. tural control. J Struct Eng ASCE 129(7):845–856
Fujino Y, Siringoringo D (2013) Vibration mechanisms Spencer BF, Soong TT (1999) New application and devel-
and controls of long-span bridges: a review. Struct Eng opment of active, semi-active and Hybrid control tech-
Int 3:248–268 niques for seismic and non seismic vibration in the
Gimsing NJ, Georgakis CT (2012) Introduction, in Cable USA. In: International post-SmiRT conference semi-
Supported Bridges: Concept and Design, Third Edi- nar on seismic isolation, passive energy dissipation
tion, John Wiley & Sons, Ltd, Chichester, UK. doi: and active control of vibration of structures, Cheju,
10.1002/9781119978237.ch 23–25 Aug
Housner G, Bergman L, Caughey T, Chassiakos A, Sun L, Sun Z, Dan D, Zhang Q, Huang H (2009)
Claus R, Masri S, Skelton R, Soong T, Spencer B, Researches and implementations of structural health
Yao J (1997) Structural control: past, present, and monitoring systems for long span bridges in China.
future. J Eng Mech 123(9):897–971 Doboku Gakkai Ronbunshuu A 65(1):15–29
Jacques C (2011) The Rion-Antirion bridge – when a Wierschem NE, Spencer B, Bergman LA, Vakakis AF
dream becomes reality. Front Archit Civil Eng China (2011) Numerical study of nonlinear energy sinks for
5(4):415–426 seismic response reduction. In: ANCRISST
Kitagawa M (2004) Technology of the Akashi Kaikyo 2011 – The sixth international workshop on advanced
bridge. Struct Control Health Monit 11:75–90 smart structures and technology, Dalian
Koh H-M, Lee H-S, Kim S, Choo JF (2009) Monitoring of Yao JTP (1972) Concept of structural control. J Struct Div
bridges in Korea. In: Christian B, Fou-Kuo C, Yozo F 98(7):1567–1574
(eds), Encyclopedia of structural health monitoring. You Q, He P, Dong X, Zhang X, Wu S (2008) Sutong
Wiley. http://onlinelibrary.wiley.com/doi/10.1002/ bridge – the longest cable-stayed bridge in the world.
9780470061626.shm163/abstract Struct Eng Int 18(4):390–395
Kosa K, Tasaki K (2003) Detailed investigation of PC
Cable-stayed Bridge damaged in the 1999 Taiwan
earthquake. In: Proceedings of the 19th US – Japan
Bridge Engineering Workshop, Tsukuba Science City, Design of Mechanically Stabilized
Japan, 27–29 Oct.
Li H, Liu M, Li J, Guan X, Ou J (2007) Vibration control
Walls (MSWs)
of stay cables of the Shandong Binzhou Yellow River
Highway Bridge using magnetorheological fluid Richard J. Bathurst
dampers. J Br Eng ASCE 12(4):401–409 GeoEngineering Centre at Queen’s-RMC Civil
Naganuma T, Kitazawa M, Adachi Y, Noguchi J (2000)
Seismic design and behavior of the Higashi-Kobe
Engineering Department, Royal Military College
Bridge and restoration after the 1995 Kobe Earth- of Canada, Kingston, ON, Canada
quake. 12 WCEE, paper no 2793, Auckland
Ni YQ, Wong KY (2012) Integrating bridge structural
health monitoring and condition-based maintenance
management. Lecture 06: workshop on civil structural
Synonyms
health monitoring (CSHM-4) “SHM systems
supporting extension of the structures’ service life”, Dynamic finite difference method; Dynamic
Berlin, 6–8 Nov finite element method; Earthquakes;
Papanikolas P, Stathopoulos-Vlamis A, Panagis A,
Pecker A, Infanti S (2010) The behavior of Rion-
Geosynthetics; Metallic reinforcement; Pseudo-
Antirion Bridge during the Earthquake of “ACHAIA- static; Reinforced soil; Retaining walls; Seismic
ILIA” on June 8, 2008. In: Third international fib design; Sliding block
Design of Mechanically Stabilized Walls (MSWs) 551
Reinforcement Facing
The soil reinforcement materials that are used in MSWs must include a facing that acts as a con-
MSWs can be broadly classified into metallic and struction aid to assist with placement and com-
geosynthetic categories. Metallic reinforcement paction of the soil at the wall face and ensures that
includes steel strips, steel bar mats and ladders, the wall face will be stable at a vertical or near-
welded wire, and steel anchor plates attached to vertical orientation over its design life. The sim-
the facing using steel rods. Almost all steel rein- plest facing can be constructed by simply
forcement products are galvanized for corrosion extending the primary reinforcement layers in a
552 Design of Mechanically Stabilized Walls (MSWs)
a b
1.0 m
0.83 m
7.1 m
1.0 m
0.6 m
0.6 m
0.52 m
3 at 0.38 m
5.0 m
2.1 m 1
Surcharge Strap
0.8 m
0.8 m
1H:20V 5 to 8 m
1.0 m
6.1 m
0.8 m
0.6 m
0.6 m
0.6 m
0.6 m
0.2 m
0.6 m
4.3 m Alignment rod
Plan view of facing panel and straps
Design of Mechanically Stabilized Walls (MSWs), concrete panel. (c) Geotextile wrapped face wall. (d)
Fig. 1 Example MSW structures and components (a) Full-height concrete panel wall. (e) Modular block wall.
Steel strip reinforced soil wall with incremental concrete (f) Polyester strap reinforcement
panel. (b) Steel ladder reinforcement with incremental
Design of Mechanically Stabilized Walls (MSWs) 553
Design of Mechanically Stabilized Walls (MSWs), capacity. (d) Pullout. (e) Tensile over-stress. (f) Internal
Fig. 2 Failure modes for MSWs with modular block sliding. (g) Connection failure. (h) Column shear failure.
facing (a) Base sliding. (b) Overturning. (c) Bearing (i) Toppling
wrapped configuration at the wall face and Potential failure mechanisms are illustrated in
tucking the free end of the wrap back into the Fig. 2 for the case of geosynthetic and metallic
reinforced soil mass. A temporary moving form- reinforced soil walls constructed with a modular
work is required to support each wrap during block facing. External failure modes are related
construction. These facings are flexible and to the stability of the composite reinforced soil
cost-effective and best used for temporary struc- zone comprised of the backfill soil, reinforce-
tures when long-term durability of the facing is ment layers, and facing. The composite acts as
not a concern. More robust facings can be formed a gravity block to resist sliding at the base of the
using welded wire cages, masonry or wet-cast block and overturning at the toe in response to
concrete blocks, incremental concrete panels, or active earth forces from the retained soil located
full-height concrete panels. Historically, steel immediately behind the block (reinforced soil
soil reinforcement materials have been used in zone). In addition, the block must not generate
conjunction with incremental concrete panels a bearing capacity failure in the foundation soil
that come in square or hexagonal shapes. or excessive settlement. For walls with simple
geometry, sliding resistance typically controls
the length of the block in the direction perpen-
Design Concepts dicular to the running length of the wall face and
consequently the length of the reinforcement.
Static Load Environments Regardless of the computed minimum reinforce-
MSWs are first designed to have adequate mar- ment length, design codes typically restrict the
gins of safety against collapse and excessive minimum length of the reinforcement to be not
deformation under static loading conditions. less than 60–70 % of the height of the wall.
554 Design of Mechanically Stabilized Walls (MSWs)
Additional stability calculations for global sta- damage, chemical degradation, and creep.
bility must always be carried out, as is the case Modern geosynthetic reinforcement products
for any retaining wall structure, to ensure that have been specially formulated to minimize deg-
the structure is not part of a larger failure mech- radation. Polyester geogrids and strips are
anism that extends beyond the reinforced soil manufactured with a coating for additional sta-
zone and into the foundation. These calculations bility and for protection during installation and
can be carried out using conventional slope sta- exposure to UV light.
bility programs. The facing must also be stable over the life of
In order for the reinforced zone to maintain its the structure, and the connections between the
coherence, it is necessary to design the reinforce- facing and the reinforcement layers must have
ment layers against internal stability modes of sufficient tensile capacity. Adequate connection
failure. These include tensile overstressing of capacity is most important for walls with a hard
the reinforcement leading to excessive deforma- facing (e.g., concrete blocks and panels). For mod-
tion or even rupture and pullout of the reinforce- ular block walls, stability against toppling of the
ment layers from the anchorage zone which units at the top of the facing column and interface
extends from the location of the potential internal shear must also be ensured. The failure modes for
failure wedge to the free end of the reinforcement column shear and toppling in Fig. 2h, i are unique
layers. In some cases it may be possible for a to modular block walls and thus do not apply to
failure mechanism to propagate at an angle from incremental concrete panel, full-height concrete
the heel of the wall facing or at a higher elevation panel, and wrapped face walls.
at the back of the facing column through the Post-construction wall deformations for well-
reinforced soil zone and into the retained soil constructed MSWs should be less than 1.5 % of
zone. These failure mechanisms are called com- the wall height. Deformations for the same walls
posite failures and are best investigated using but using much stiffer steel reinforcement will be
circular or noncircular slope analysis methods less. This difference may be a factor when deciding
implemented in slope stability computer pro- which category of reinforcement to use in a partic-
grams that have been modified to include the ular project.
tensile capacity of any intersected reinforcement
layer. For reinforced soil walls with incremental Seismic Design
panel or full-height panels, the internal sliding The potential modes of failure identified for static
failure through the concrete facing is unlikely design of MSWs are equally applicable to the
(Fig. 2f). same structures subjected to additional loads
The design tensile strength of the reinforce- due to earthquake.
ment will vary depending on the reinforcement Current methods for the earthquake design of
type. For steel reinforcement the unfactored ten- MSWs fall into three categories: (1) pseudo-static
sile strength of the reinforcement is easily com- methods, (2) sliding block displacement
puted as the product of yield strength of the methods, and (3) dynamic finite element or finite
tensile member and the cross-sectional area. difference methods.
Most steel reinforcement products are galva-
nized for cathodic protection, and a sacrificial Pseudo-Static Methods
thickness of steel is included in the calculation of The calculations for earthquake loads used in
tensile load capacity to account for possible cor- internal and external stability design of MSWs
rosion. This thickness is computed based on the have historically been based on extensions of the
soil type, porewater chemistry and resistivity, classical (static) Coulomb wedge approach for
and design life. For geosynthetic MSWs, the active earth forces acting against retaining
reference tensile strength of the reinforcement walls. The advantage of the method is that in
from standard laboratory testing must be the limit of no earthquake-induced ground accel-
reduced to account for possible installation erations, the equations for active earth force
Design of Mechanically Stabilized Walls (MSWs) 555
Here, g = unit weight of the soil and H = height DKdyn ¼ ð1 kv Þ KAE KA (5)
of the wall. The total earth pressure coefficient
(KAE) is calculated using the Mononobe-Okabe Here, KA = static active earth pressure coefficient
equation which is an extension of the Coulomb and DKdyn = incremental dynamic active earth
equation: pressure coefficient. In stability calculations the
556 Design of Mechanically Stabilized Walls (MSWs)
Design of Mechanically
Stabilized Walls (MSWs),
Fig. 4 External stability
geometry and forces
forces or coefficients of earth pressure shown panel wall systems. Quantity R in Fig. 4 is the
above must be multiplied by cos(d o) when sliding resistance at the base of the reinforced soil
factor of safety expressions are for loads and zone.
resistances acting in the horizontal direction. As During an earthquake, additional dynamic
distance from the earthquake epicenter increases, transient tensile loads will be carried by the rein-
the most important ground acceleration is the forcement. In the simplest approaches to internal
horizontal component, and the vertical compo- stability seismic design, the dynamic load (Tdyn)
nent is typically ignored (i.e., kv = 0). Further- due to earthquake is most often added to the static
more, the likelihood of peak horizontal and load (Tsta) calculated for the same structure as
vertical acceleration components of an earth- illustrated in Fig. 5. For example, the tensile
quake arriving simultaneously is small. In prac- reinforcement load Tmax in a layer can be calcu-
tice, the choice of kh is typically based on local lated as:
experience or prescribed by design codes that are
applicable to the site area. In current US practice, Tmax ¼ Tsta þ Tdyn ¼ Kr sv Sv þ DKdyn sv Sv (6)
the location of the dynamic earth force compo-
nent is assumed to act at the same elevation as the Tensile load is expressed with units of force per
static component (AASHTO 2012). If the conse- unit running length of wall. Here Sv is the rein-
quences of failure are high, then the same code forcement spacing and sv is the vertical stress
recommends locating the resultant at a higher acting at the elevation of the reinforcement due to
elevation. soil self-weight and any uniformly distributed
The calculation of factor of safety against surcharge. The quantity sv is the average vertical
external base sliding is carried out by substituting stress over the height of the wall. For modular
PAE for PA in a conventional block sliding anal- block wall systems, the horizontal inertial force
ysis and adding horizontal inertial forces due to due to the facing units over the layer spacing
the mass of the reinforced zone (sliding block) as height Sv should be included in the calculation
illustrated in Fig. 4. Quantities Ws and Ww are the of Tmax. The example here is simplified by con-
weight of the reinforced soil and the facing col- servatively ignoring wall-soil friction, wall incli-
umn, respectively. Typically, the horizontal iner- nation, and weight of the facing. Parameter Kr is
tial force due to Ws is reduced by setting l = 0.5. the static earth pressure coefficient. For
The contribution of the facing column to horizon- geosynthetic reinforcement, Kr = KA. For steel
tal inertial loads is usually applied to facings reinforcement cases, Kr 1.2 KA at the top of the
constructed with modular blocks and is ignored wall (with actual values depending on steel rein-
for thinner incremental or full-height concrete forcement type) and decreases linearly with depth
Design of Mechanically Stabilized Walls (MSWs) 557
Design of Mechanically
Stabilized Walls (MSWs),
Fig. 5 Calculation of
maximum internal
reinforcement loads
to 1.2KA at a depth of 6 m. The distribution of Here, C is the coverage ratio of the reinforcement
dynamic horizontal pressure in the above exam- in the plane of the reinforcement and is equal to
ple is assumed to be constant over the height of one for continuous sheet reinforcement (e.g.,
the wall. However, assumptions regarding the geogrid, woven geotextile). Parameter f* is the
distribution of this dynamic pressure component interface shear coefficient and is a function of the
can vary based on reinforcement type reinforcement type and soil type, and for the case
(geosynthetic or steel) and between different of steel strips, it will also be a function of depth.
design codes. As noted in the previous section Parameter sv is vertical stress acting at the ele-
on static design principles, the tensile load Tmax vation of the reinforcement and Le is the anchor-
in the reinforcement must not exceed a critical age length. The anchorage length can be
value to prevent overstressing that can lead to computed knowing the geometry of the internal
rupture. For static design of geosynthetic active wedge. In Fig. 5 this wedge is shown as
MSWs, this value is largely determined by the triangular shape which is the typical assumption
constant load creep behavior of these polymeric for MSWs constructed with relatively extensible
materials. However, under transient dynamic geosynthetic reinforcement materials. For steel
loading, the residual strength of the reinforce- reinforcement, the active wedge is assumed to
ment is assumed to be available. This strength is be bilinear in shape with a constant distance
computed by ignoring the creep-induced strength from the back of the wall equal to 0.3H to a
loss that is assumed for the design life of a wall depth of H/2 below the crest of the wall and
based on static loading only. In other words, the then tracing a straight line to the heel of the wall
residual strength value is calculated as the origi- face thereafter. For MSWs constructed with
nal available strength of the reinforcement imme- select backfills, there is no evidence to suggest
diately after the end of construction. that the value of f* should be less for the case of
In order for the internal active wedge to be earthquake loading than for the static loading
stable, the reinforcement layers must have ade- condition.
quate anchorage (pullout) capacity. The pullout Rigorous analytical or numerical solution of the
capacity (Tpull) can be computed as: active Coulomb wedge problem for the case of
addition seismic inertial forces shows that the ori-
Tpull ¼ 2C f sv Le (7) entation of the internal failure wedge (aAE)
558 Design of Mechanically Stabilized Walls (MSWs)
Design of Mechanically
Stabilized Walls (MSWs),
Fig. 6 Double integration
method to calculate
displacements in sliding
block analysis
illustrated in Fig. 3 will decrease with increasing applied to the horizontal ground acceleration
magnitude of kh. This can lead to excessive rein- record at the structure, and double integration
forcement lengths. Conventional practice for is performed to calculate the cumulative dis-
geosynthetic reinforced walls is to limit this placement, as illustrated in Fig. 6. Specifically,
angle to the static loading value which is reason- the cumulative displacement is calculated by
ably approximated using the Rankine solution integrating those portions of the accelerogram
a = p/4 + f/2. Similarly, the bilinear wedge that are above the critical acceleration and those
shape for steel-reinforced MSWs is assumed to portions that are below until the relative veloc-
be applicable to both static and dynamic loading ity between the sliding mass and the sliding base
cases. Pullout may be a concern at the top of the reduces to zero.
wall where vertical stress (sv) is low. The solution
is to locally increase the length of these layers (Le) Dynamic Finite Element or Finite Difference
to meet an acceptable factor of safety against this Methods
mode of failure. MSWs are complex mechanical systems and this
complexity is compounded when the structures
Sliding Block Displacement Methods are subjected to transient dynamic loading due to
A shortcoming of pseudo-static design and anal- earthquake. Pseudo-static and displacement
ysis methods is that they cannot be used to methods may be sufficient for simple structures,
predict serviceability performance of a wall for stable ground conditions and MSWs in low
with respect to displacement. Sliding block dis- seismic risk areas, and/or for preliminary design.
placement methods have been used for many Otherwise, more sophisticated analyses may be
years to estimate the displacement of earth warranted using advanced dynamic finite element
structures during earthquake. This general model or finite difference computer programs.
approach is easily adaptable to failure limit Today commercially available computer pro-
states for MSWs where the failure surface is grams offer the user a suite of constitutive models
horizontal (see Fig. 2a, f). The first step is to for the component materials and in some cases
find the critical acceleration, kc, which is the allow the user to implement their own constitu-
value of kh that will give a factor of safety of tive models. In order to accurately model the soil
unity in the corresponding static factor of safety in a MSW, it may be necessary to use nonlinear
expression. The critical acceleration is then cyclic stress-strain models. The equivalent-linear
Design of Mechanically Stabilized Walls (MSWs) 559
method (ELM) is one example. The ELM Factors of Safety and Load and Resistance
requires strain-dependent shear stiffness degrada- Factor Design (LRFD)
tion curves and damping curves to capture the The concepts introduced in this entry for seismic
reduction of shear modulus of the soil and the design of MSWs have been related to margins of
increase in soil damping that occurs with increas- safety using classical notions of factor of safety.
ing shear strain. These parameters can be found in In this approach, factor of safety is simply the
published databases or from project-specific res- ratio of resistance capacity to load. In North
onant column testing for small shear strain ampli- American practice, the factors of safety against
tudes and cyclic triaxial tests for larger strain the modes of failure identified in Fig. 2 are D
amplitudes. Another important consideration for reduced to values that are typically 75 % of static
the correct numerical modeling of MSWs with values or as low as 1.1. The exception is the factor
polymeric reinforcement materials is the strain- of safety for overstressing (rupture) for
rate-dependent stiffness of these materials. Often geosynthetic reinforcement products which is
forgotten is that the stiffness of polymeric PP and taken as unity for seismic design.
HDPE reinforcement products increases signifi- Modern US design codes for MSWs now
cantly with rate of loading. Hence, stiffness adopt a load and resistance factor design
values deduced from conventional laboratory (LRFD) approach (AASHTO 2012; FHWA
tensile tests at low strain rates will underestimate 2009). Load factors are applied to each load
the stiffness of the reinforcement during dynamic term in a limit state function, and a single resis-
loading. Consequently, numerical-predicted rein- tance factor is applied to the resistance term in the
forcement loads will be nonconservative for same function. The factors have been computed
design (i.e., underestimated). using rigorous reliability theory to satisfy an
acceptable minimum target probability of failure
or to give design outcomes with equivalent fac-
Additional Considerations tors of safety to those using the conventional
factor of safety approach. Load and resistance
Frequency Response factors are different for seismic and static design
The seismic response of MSWs of typical heights consistent with the expectation that severe earth-
(say H < 10 m) can be dominated by the natural quakes are relatively rare events over the design
frequency of the structure. A recommended life of a typical MSW. In Europe, a partial factor
check of the response of any soil structure is the of safety approach is used in which partial factors
proximity of the natural (resonant) frequency of are applied to material properties on the resis-
the system (fo) to the predominant frequency of tance side of each limit state function. At present,
the input acceleration record (f1). An approxi- partial factors have not been updated for seismic
mate upper-bound estimate of the natural fre- design of MSW systems in the British code (BSI
quency of a MSW structure can be computed 2010). Eurocode guidelines are limited to a sum-
from linear-elastic theory as: mary of the general approaches to the design and
analysis of geotechnical soil structures.
sffiffiffiffi
G
f o ¼ Vs =4H ¼ 1=4H (8) Performance of MSWs During Earthquake
r In recent years, surveys of MSW structures have
been carried out in the Americas and in Japan
where Vs = shear-wave velocity of the soil back- after earthquakes. These surveys and project-
fill, G = shear modulus, and r = density of the specific case studies of reinforced soil walls
soil. Modifications to this formula are available in after earthquake have demonstrated generally
the literature to adjust this one-dimensional solu- good behavior, and, where direct comparisons
tion to the case of two-dimensional rectangular are available, they have behaved better than con-
volumes to match the reinforced soil zone shape. ventional gravity structures and at least as well as
560 Design of Mechanically Stabilized Walls (MSWs)
traditional reinforced concrete structures (Koseki so that static designs do not need to be aug-
et al. 2006; Yen et al. 2011). Geosynthetic MSWs mented for peak ground accelerations less than
constructed with full-height panel facings for 0.4 g. Details of the conditions that meet this
railway track embankments and bridge abut- waiver can be found in the American Associa-
ments (including bullet train lines) were virtually tion of State Highway and Transportation Offi-
undamaged and exhibited good seismic resis- cials design guidelines (AASHTO 2012). For
tance during the recent 11 March 2011 Great wall cases exceeding the 0.4-g criterion, more
East Japan Earthquake compared with conven- sophisticated analyses executed by experienced
tional unreinforced embankments (JGS 2011). and knowledgeable design engineers will be
The reason for this good performance is that required. Often this will include displacement
MSWs are generally more flexible than conven- methods and numerical modeling using com-
tional walls. They are more forgiving of the mercial computer software programs that use
uneven ground displacements that may occur dynamic finite element or finite difference
due to surface faulting during an earthquake methods.
event. Importantly, the ductility of these systems
has prevented collapse of structures during earth-
quakes so that adjacent structures, property, or Summary
persons were not put at risk even though
supporting infrastructure (e.g., railway lines) Mechanically stabilized walls (MSWs) are a
was no longer in service. Where there have been well-accepted technology to perform the earth
detectable performance problems, these are typi- retaining wall function in civil engineering earth-
cally related to inadequate details at the wall work applications. The design of these structures
corners, the joints between full-height panels, for earthquakes is most often based on conven-
and the wall top due to inadequate coping details tional extensions of static wall design that are
or toppling of masonry concrete blocks. well known and accepted by geotechnical engi-
In addition to field surveys, there is now a neers. The performance of MSWs during earth-
large body of results from centrifuge and 1-g quake loading has been excellent in most cases
gravity tests on model walls subjected to simu- due to their relative flexibility compared to con-
lated earthquake loading using laboratory shak- ventional gravity wall structures and their reserve
ing tables. Overturning about the toe of a MSW load capacity due to conservatism in the static
structure is rare and can be safely omitted for the design of these structures.
external stability design of MSW structures.
Similarly, bearing capacity failure is unlikely
and this mode of failure is not expected for Cross-References
walls built on competent foundations. Field sur-
veys and laboratory shaking table tests have ▶ Analysis and Design Issues of Geotechnical
demonstrated that MSWs designed to satisfy Systems: Rigid Walls
current factors of safety for static design are ▶ Engineering Characterization of Earthquake
typically stable at peak ground accelerations up Ground Motions
to 0.4 g if the structures are seated on competent ▶ Seismic Analysis of Masonry Buildings:
level ground below and in front of the walls and Numerical Modeling
the walls support horizontal backslopes (Mikola ▶ Seismic Design of Earth-Retaining Structures
and Sitar 2013). Consequently, for these scenar-
ios additional earthquake-induced loads need
not be considered in design. The explanation
References
for this recommendation is that the static design AASHTO (2012) AASHTO LRFD bridge design specifi-
of these structures is very conservative. Hence, cations, 6th edn. American Association of State High-
these walls have sufficient excess load capacity way and Transportation Officials, Washington, DC
Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings 561
BSI (2010) Code of practice for strengthened/reinforced semi-active. Among them, passive control sys-
soil and other fills (BS8006). British Standards Insti- tems are less expensive and have the advantage
tution, Milton Keynes
FHWA, Berg RR, Christopher BR, Samtani NC of not requiring an external source of power. The
(2009) Design of mechanically stabilized earth walls passive control systems consist basically on
and reinforced soil slopes. vol 1. FHWA GEC installing energy-dissipating devices (dampers)
011-VolI. National Highway Institute – Federal High- in the main structure, in order to increase its
way Administration, U.S. Department of Transporta-
tion, Washington, DC energy dissipation capacity and to prevent or
JGS (2011) Lessons and recommendations from the 2011 limit the damage in the main structure (plastic
Great East Japan earthquake. In: 2011 committee for deformations) in case of severe earthquakes. The D
Geo-hazards during earthquakes and mitigation mea- use of seismic energy-dissipative devices for pas-
sures. The Japanese Geotechnical Society, Tokyo, 91 p
Koseki J, Bathurst RJ, G€ uler E, Kuwano J, Maugeri sive control is increasing exponentially in recent
M (2006) Seismic stability of reinforced soil walls, years for both new and existing buildings
Invited keynote paper. In: Proceeding of the 8th inter- (Martelli 2006). The challenge now is to produce
national conference on geosynthetics, Yokohama, low-cost and effective dampers feasible for mas-
Japan, 18–22 Sept 2006, 30 p
Mikola RG, Sitar N (2013) Seismic earth pressures on sive use in developing countries.
retaining structures in cohesionless soils. Report No, Several mechanisms have been used for passive
UCB GR 13-01. Department of Civil and Environmen- energy dissipation: yielding of metals, phase trans-
tal Engineering, University of California, Berkeley formation of metals, friction sliding, fluid orificing,
Yen W-HP, Chen G, Buckle I, Allen T, Alzamora D,
Ger J, Arias JG (2011) Postearthquake reconnaissance and deformation of viscoelastic solids or liquids.
report on transportation infrastructure impact of the The former is one of the most popular and there are
February 27, 2010, Offshore Maule Earthquake in a number of ways to induce the yielding of the
Chile. Publication No. FHWA-HRT-11-030. Office metal. In the ADAS damper (Berman and Goel
of Infrastructure Research and Development, Federal
Highway Administration, McLean, VA 1987), yielding is attained by subjecting metallic
plates to out-of-plane bending. In the honeycomb
damper (Kobori et al. 1992) or the slit damper
(Benavent-Climent et al. 1998) a metallic plate
Design, Testing, and Evaluation of with openings is subjected to in-plane shear defor-
the Web Plastifying Damper for the mations. Another popular damper is the so-called
Aseismic Protection of Buildings Buckling Restrained Brace (Chan and Albermani
2008), which consists on yielding a metal core
Amadeo Benavent-Climent subjected to axial deformations, preventing buck-
Department of Structural Mechanics and ling under compressive loads.
Industrial Constructions, Polytechnic University A new type of hysteretic damper, called Web
of Madrid, Madrid, Spain Plastifying Damper (WPD), was recently pro-
posed by Benavent-Climent et al. (2011) that is
based on subjecting the web of standard wide-
Synonyms flange or I shape sections to out-of-plane bend-
ing, to dissipate energy in seismic applications.
Earthquake resistant structure; Energy dissipa- Yielding the web under out-of-plane bending is a
tion; Hysteretic damper; Metallic damper; potential source of energy dissipation because:
Passive control (i) the transition from the web-ends to the flanges
has a smooth curve shape (preventing stress con-
centrations), and (ii) the union between flange
Introduction and web is weld-free, which avoids uncertainties
and imperfections associated with welding. This
Over the last three decades, structural control has energy-dissipating device requires a very simple
achieved significant progress. Structural control manufacturing process (cutting a wide-flange or
includes three categories: passive, active, and I-section in short segments and drilling openings
562 Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings
SECTION A A
auxiliary element (U-section)
Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings,
Fig. 1 Details of a typical WPD
fy h2 faux L 1 h
as auxiliary elements. dTy ¼ þ ¼ 0:5fy þ faux L
2tE E E t
The relative displacement dy between flanges
(6)
at yielding, the yielding load Qy, and maximum
load QB of a single segment of wide-flange or
From QTy and dTy , the axial stiffness KT of the
I-shaped section of length l, clear height of the
brace damper is given by:
web h, and web thickness t are given by:
fy h2 QTy nlt3 fy E
dy ¼ (1) KT ¼ ¼ 3 (7)
2tE dTy fy h þ 2thLfaux
h
Mpu,w
region D
intended to
My,f My,f l yield
tf
displacement, QTy and dTy, respectively, the fol- wide-flange section of 140 73 mm2 (total
lowing paths must be followed: depth flange width) – referred to as the
IPE-140 section hereafter – to different histories
(i) Choose the type of steel and the stress level of imposed out of plane bending, until failure.
in the auxiliary elements faux Figure 2 shows the test specimen.
(ii) Using Eq. 6 determine the required h2/t The mechanical characteristics of each test
(iii) Choose a commercial wide-flange or specimen, in terms of length l, the clear height
I-shape section that fits the required h2/t of the web h, and the web thickness t are:
(iv) Determine the value of the product nl to l = 90 mm, h = 112 mm, and t = 5.4 mm. The
attain the target strength QTy using Eq. 4 specimens were made of mild steel S275, with
(v) Determine an appropriate combination of fy = 340 N/mm2, fB = 441 N/mm2 and
n and l, compatible with architectural con- E = 2.1 105 N/mm2. Applying above equa-
straints of the buildings tions, the lateral displacement dy, the load Qy at
(vi) Ensure that the following design require- yielding, and the apparent maximum lateral load
ments are fulfilled: QB of a single IPE-140 section damping device
are dy = 1.88 mm, Qy = 3.98kN and
pffiffiffi
h 3 3 QB = 5.17 kN.
> (8) The only variable considered in the tests was
t 4
the loading history. Patterns of cyclic displace-
t 2 3fB ments were applied to the specimens until they
f
> (9)
t 4fy failed. Failure was assumed to occur at the onset
of strength degradation under increasing forced
Above requirements ensure that the WPD fails displacements. The cyclic rate applied by the
under out-of-plane flexural yielding of the web of actuator was 1/80Hz. In the first, second, and
the wide-flange or I-shape section, near the third tests cycles of incremental amplitude were
round-shaped transition zone from the web to applied to the specimens, with the increment of
the flanges (see Fig. 2). amplitude in each consecutive cycle normalized
by the yielding displacement dy being f = 2,
f = 5 and f = 10, respectively. The specimens
Testing the Web Plastifying Damper were subjected to cycles of constant amplitude in
the fourth and fifth tests; the amplitude normal-
Six component tests were conducted to investi- ized by the yielding displacement dy was D = 20
gate the energy dissipation capacity and to char- and D = 30, respectively. The loading protocol
acterize the hysteretic behavior of the WPD. Each ATC-24 developed by the Applied Technology
test consisted in subjecting 90 mm length speci- Council (Applied Technology Council 1992) was
mens cut from a single segment of a structural applied to the sixth specimen. This protocol
564 Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings
a Q 17 18
b Q Kp2 18
Q6 12 12,17
11 6 QB Kpl 6,11
5 1
− Qy 1,5
Sd u Ke Ke Ke Ke Ke
− − −
16 10 4 0 2 7 13 19 d Sd 3 Sd 2 S d 1 0,2
d −max Sd
d +max dy Sd2+ Sd3
+
Sd4
+
Ke Ke Ke + d
+
Sd u
Sd1 S B
3 8 3,8
9 9,14 Trilinear approximation
15 14
15 of the skeleton curve
c Q 17
11
5
− − Ke Ke Ke
Bd2 13 Bd1 4
10 Bd
7 + 16 +
+ Bd2 Bd3
Ke Ke Bd1
8
14
Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings,
Fig. 3 Decomposition of a Q-d curve: (a) Q-d curve; (b) skeleton part; (c) Bauschinger part
consists of several sets of three cycles; the The scheme proposed by Kato et al. (1973)
amplitude of the cycles is constant within was applied for decomposing the Q-d hysteretic
each set but it is increased every consecutive curves obtained during the cyclic tests, as illus-
set of cycles, following the sequence trated in Fig. 3a, b, and c. In these figures, the
0.5dy ,0.75dy ,1.0dy ,2dy ,3dy ,4dy and so on. segments 0–1, 5–6, 11–12, 17–18 in the positive,
and 2–3, 8–9, 14–15 in the negative domain are
Hysteretic Behavior the paths that exceed the load level attained by
Figure 3a shows a typical load–displacement Q-d the preceding cycle in the same domain of load-
hysteretic curve obtained from the test. It can be ing. Connecting these paths sequentially, the
seen that the WPD has stable hysteretic behavior two curves called “skeleton part” (Fig. 3b) are
and high energy dissipation capacity. obtained. These curves can be approximated by
a three-segment curve that is shown with dash
Ultimate Energy Dissipation Capacity and dot lines in Fig. 3b. The trilinear approxi-
Following the procedure formerly proposed by mation is defined by the yielding load, Qy , the
Kato et al. (1973) a Q-d hysteretic curve such yielding displacement, dy , the first and second
that is shown in Fig. 3a can be decomposed into: plastic stiffnesses Kp1 and Kp2 (Kp1 Kp2), and
(i) the segments controlled by the Bauschinger the load QB corresponding to the transition point
effect, so-called “Bauschinger part”; (ii) the seg- from Kp1 to Kp2. The unloading paths are the
ments controlled by the strain-hardening phe- segments 1–2, 6–7, 12–13, 18–19, 3–4, 9–10,
nomena, so-called “skeleton part”; (iii) and the and 15–16. The slope of the unlading paths coin-
unloading paths, so-called “unloading segments”. cides with the initial elastic stiffness Ke (=Qy/
It has been shown (Kato et al. 1973) that the curve dy). In Fig. 3b, S dþ
u and S du are the plastic
obtained by connecting subsequently the seg- deformation accumulated in each skeleton
ments of the skeleton part coincides with the curve when the WPD device fails. On the other
relation obtained from a monotonic test. hand, SdB is the plastic deformation accumulated
Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings 565
Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings,
Table 1 Ultimate energy deformation capacity of the damping devices
Load pattern Qy kN dy mm ep S B B/S
f¼2 3.98 1.88 44.32 67.19 532.79 599.98 7.93
f¼5 3.98 1.88 60.93 97.62 355.60 453.53 3.64
f ¼ 10 3.98 1.88 75.12 124.54 269.45 393.99 2.16
D ¼ 20 3.98 1.88 53.27 79.25 555.08 634.32 7.00
D ¼ 30 3.98 1.88 67.86 107.63 338.05 445.68 3.14
ATC-40 3.98 1.88 20.77 29.31 923.11 952.41 31.50 D
in the approximate trilinear skeleton curve at preceding cycles in the same loading domain;
Q = QB. The area enveloped in Fig. 3b they constitute the so-called “Bauschinger part”
for each domain of loading by the skeleton (Fig. 3c). The sum of the area enveloped by the
curve is called hereafter S W þ
u and S W u . The Bauschinger part for each domain of loading will
segments 4–5, 10–11, and 16–17 in the positive be referred to as B W þ u and B W u .
þ
domain and 7–8 and 13–14 in the negative In addition, S W u , S W u , B W u , B W
þ þ
u , S du and S du
domain of loading begin at Q = 0 and terminate can be expressed in a nondimensional by,
at the maximum load level previously attained in
þ Wþ W þ Wþ W þ dþ d
S ¼ ; S ¼ ; B ¼ ; B ¼ ; ep ¼S ; ep ¼S ;
S u S u B u B u u u
(10)
Q y dy Q y dy Qy dy Qy dy dy dy
Next, the following ratios are defined: Bauschinger parts (i.e., B/S) depend on the
loading pattern. If the WPD is subjected to cycles
S ¼ S þ þ S ; B of small constant amplitude (i.e., the loading pat-
tern with D = 20) or to cycles whose amplitude
¼ B þ þ B ; ep grows in small increments (i.e., the loading pat-
¼ ep þ þ ep ; ¼ S þ B (11) terns ATC-40 or with amplitude f = 2), and
B/S are larger than when applying cycles of
large constant amplitude (i.e., the loading pattern
ep denotes the apparent ultimate cumulative
with D = 30) or than cycles whose amplitude
plastic deformation ratio on the skeleton part; S
enlarges in large increments (i.e., the loading
is the ultimate cumulative plastic deformation
pattern with f = 10).
ratio on the skeleton part; B is the ultimate
It can be seen that the relation between the
cumulative plastic deformation ratio on the
values ep and B shown in Table 1 is almost
Bauschinger part; is the total ultimate cumula-
linear and can be formally expressed by:
tive plastic deformation ratio. represents in
nondimensional form the total energy dissipated
by the damper until failure, i.e. B ¼ aep þ b: (12)
Wu ¼ SWþ u þ W
S u
þ W
B u
þ
þ W
B u
Þ .
Table 1 shows the values of the parameters The value of parameters a and b for the WPDs
defined above obtained from each test. It can be tested are a = 12 and b = 1140. It has been
seen in the table that the total energy dissipated shown in previous research that while the relation
by the WPD as expressed by , and the distribu- between B and ep is linear irrespective of the
tion of this energy between the skeleton and the steel type, the values of a and b depend on the
566 Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings
material properties and must be determined for tB = QB/Qy . Using these parameters, the energy
each type of steel. a and b can be obtained by dissipated on the skeleton part expressed in
testing only two specimens under different pat- nondimensional form by S can be related to ep
terns of cyclic loading. assuming that the skeleton curves in the positive
As for the skeleton part, the approximated and negative domains of loading are equal and
trilinear curve shown in Fig. 3b can be normal- that |ep þ |=|ep |= ep ¼ 0:5ep :
ized by dividing the abscissa by dy given by Eq. 1,
and the ordinate by Qy given by Eq. 2. The
resulting values of the normalized plastic stiff- 2ðtB 1Þ 1 kp1
ness kp1 = Kp1/(Qy/dy), and kp2 = Kp2/(Qy/dy) for ep :
kp1
for the WPD tested, are kp1 = 1/5 and kp2 = 1/ (13)
kp1
S ¼ 0:25ep þ ep
2
30. Now, a new parameter tB is defined to char- 1 kp1
acterize the normalized trilinear skeleton curve as
2ðtB 1Þ 1 kp1
for : ep > :
kp1
2
"
#
tB 1 1 kp1 ep ðtB 1Þ 1 kp1 kp2 ep ðtB 1Þ 1 kp1
S ¼ þ 2tB þ
kp1 2 kp1 1 kp2 2 kp1
(14)
By adding Eqs. 12, 13, and 14, the ultimate 2ðtB 1Þ 1 kp1
for ep :
energy dissipation capacity in terms kp1
(15)
of the nondimensional ratio is obtained as kp1
¼ 0:25ep 2 þ ep ð1 þ aÞ þ b
follows: 1 kp1
2
2ðtB 1Þ 1 kp1 t 1 1 kp1
for : ep > : ¼ B
k kp1
p1
"
# (16)
ep ðtB 1Þ 1 kp1 kp2 ep ðtB 1Þ 1 kp1
þ 2tB þ
þ aep þ b
2 kp1 1 kp2 2 kp1
Numerical Model for Characterizing the For characterizing the shape of the skeleton
Hysteretic Behavior of the WPD part, the trilinear idealization proposed above is
adopted. The shape of the Bauschinger parts is
The load–displacement relationship of the WPD studied next. Focusing on Fig. 3a, consider for
under an arbitrary cyclic loading pattern can be example that under a given loading history, the
predicted with a simple hysteretic model, and a WPD has covered the path 0-1-2-3-4-5-6-7-8-9-
procedure for predicting its failure. In this hys- 10, and an approximation for the Bauschinger
teretic model, the shape of the skeleton part and segment 10–11 located in the positive loading
the shape of each segment of the Bauschinger are domain is sought. The ordinate at the origin of
determined as follows. the segment (i.e., point 10) is Q = 0, and the
Design, Testing, and Evaluation of the Web Plastifying Damper for the Aseismic Protection of Buildings 567
ordinate at the end (i.e., point 11) is known is Q1. If the damper is unloaded at point 1, the Q-d
because it coincides with the maximum force curve will follow the unloading line of stiffness
Qm attained in the skeleton part in previous cycles Ke (=Qy/dy) until point 2. If the WPD is now
of deformation in the same domain of loading loaded in the opposite domain of loading, the
(i.e., the ordinate at point 6 which is referred to Q-d curve follows the skeleton curve, from
as Qm = Q6 in Fig. 3a). The Bauschinger seg- point 2 to 3. SS d ¼ jS dþ
1 j þ jS d1 j is the deforma-
ment 10–11 can be further approximated with a tion accumulated on the skeleton part at point
bilinear curve determined by three parameters: 3. If the damper is unloaded at point 3 it will
(i) the stiffness of the first line; (ii) the displace- again follow a line of stiffness Ke until point D
ment Bd of the segment (i.e., B dþ 2 in Fig. 3c for the 4. Reloading the damper in the positive domain,
example segment 10–11); and (iii) the ordinate of the Q-d curve will follow a Bauschinger
þ path
the intersection point of the two lines, which will whose amplitude Bd is B dþ ¼ b S d j þ jS d .
1 1 1
be expressed as a fraction a of the maximum Within this Bauschinger path, initially the Q-d
force attained in the skeleton part in previous curve follows a line of stiffness Ke until the point
cycles of deformation in the same domain of of ordinate aQ1; from here, the Q-d curve follows
loading aQm (i.e., aQ6 for the example segment a line oriented to a point determined by B dþ 1 and
10–11). For the WPD, and based on the results of Q1. Following this rule, the complete hysteretic
the test, these parameters can be set as curve under arbitrary loading can be constructed
follows: (i) the first stiffness is taken equal to until failure.
the initial elastic stiffness Ke; (ii) Bd is taken as For the prediction of the failure of the WPD,
a fraction b of the displacement
X accumulated deformation accumulated on the skeleton curve
in the skeleton part S d up to the start of in each step i of the loading process, SSdi, and the
the Bauschinger segment under consideration total energy Wi dissipated by the WPD up to this
(i.e., for the example segment 10–11, step, must be monitored. The corresponding
þ
þ þ ratios epi and i at
step i are: ep i ¼ SS di =dy
B d2 ¼ b S d1 j þ jS d1 j þ jS d2 j þ jS d2 ); and
(iii) the value of a is determined so that the area and i ¼ W i = Qy dy . Failure is assumed to
(i.e., the dissipated energy) under the occur when this point reaches the “failure line”
Bauschinger segment obtained from the test and defined by Eqs. 15 and 16 (Benavent-Climent
the area under the approximated bilinear curve 2007).
are equal. Based on the test results, for the WPD
the following values can be adopted: a = 0.66
and b = 0.61. Summary
Once the shape of the skeleton part and the
Bauschinger part are determined by the parame- A new type of seismic damper based on
ters Qy, QB, dy, kp1, kp2, a and b, the hysteretic plastifying the web of short length segments of
curve of the WPD under an arbitrary history of wide-flange or I-shaped steel sections under
displacements can be constructed as follows. The out-of-plane bending is presented. The damper
first time the damper is loaded in a given domain can be installed in a frame like a conventional
(positive or negative), the proposed model con- brace. The damper device was tested to charac-
siders that the Q-d relationship follows the skel- terize its hysteretic behavior and ultimate energy
eton curve. Referring to Fig. 3a, consider dissipation capacity. The results of the tests indi-
for example that, first – starting from point cate that the seismic damper has stable restoring
0 – the damper is deformed an amount S dþ 1 in force characteristics and high energy dissipation
the positive domain; it will follow the skeleton capacity. Further, the load–displacement curve of
curve moving from point 0 to 1. At point 1 the the damper under arbitrarily applied cyclic load
deformation accumulated on the skeleton part is can be easily predicted with a simple numerical
SS d ¼ jS dþ1 j and the corresponding force attained model. Also, its ultimate energy dissipation
568 Downhole Seismometers
Downhole
Seismometers,
Fig. 1 Posthole and vault electrical
seismometers cable
hoist cable
(optional)
connector eye-bolt
sensor D
feet (optional)
levelling
bubble sensor
azimuth connector
indicator
electrical
adjustable cable
N
feet
pressure tubes and are watertight to a depth of typically include a holelock assembly as part of
50–1,000 m. Vault seismometers are normally the instrument. A holelock is used to mechani-
not designed for continuous submersion in cally support and connect the seismometer to the
water. The diameter of a downhole seismometer borehole casing. It includes an actuator or a pas-
is minimized to enable the seismometer to fit sive spring that pushes the entire seismometer
down narrow boreholes. The diameter of a vault against the side of the borehole with a substantial
seismometer is not as critical as a downhole seis- force. This force holds the seismometer rigidly in
mometer, and vault seismometers are typically place and prevents the seismometer sliding down
shorter and wider for stability (see Fig. 1). the hole.
Vault seismometers are usually leveled by A posthole seismometer is a downhole seis-
hand using a reference bubble level and adjusting mometer that does not include a holelock and is
three leveling feet. Downhole seismometers are designed for direct burial or installation in
normally not accessible for manual leveling and shallower, uncased holes, known as postholes.
have automatic leveling mechanisms where A posthole seismometer is designed to be
leveling is required. installed in sand at the bottom of a posthole.
Downhole seismometers can be further broken The sensor will typically also have three fixed
down into two different instrument types, feet, which can be useful in side-by-side testing
namely, borehole seismometers and posthole on a pier, or placement at the bottom of a cased
seismometers. hole, as long as that surface is reasonably hard
Borehole seismometers are downhole seis- and level. These seismometers are designed for
mometers that are designed for deployment in holes up to 50 m deep and may not have the
steel-cased deep boreholes. These seismometers underwater pressure rating of a borehole
570 Downhole Seismometers
seismometer. Furthermore, a posthole seismom- due to lower tilt tolerances. This entry will focus
eter’s drivers and power supplies may not support on broadband downhole seismometers and their
the cable lengths required for a borehole installation. However, many of the installation
seismometer. techniques and challenges presented in this
A downhole seismometer can be a single- entry apply to short-period seismometers, accel-
component seismometer, meaning that the seis- erometers, and geophones in varying degrees.
mometer only measures vertical motion or that it
can be a three-component seismometer which Leveling
measures vertical, north–south, and east–west In order for a seismometer to measure true verti-
motion. Three-component seismometers are also cal and north–south, east–west horizontal
known as triaxial seismometers. motion, the verticality error due to the as-drilled
An important parameter in selecting a borehole must be compensated for along with the
downhole seismometer is the self-noise level of north–south orientation. The methods of
the instrument. The self-noise of a seismometer north–south orientation will be discussed in a
defines the limit of the instrument’s ability to later section (see section “Water Egress”), with
resolve ground motion. The self-noise of a seis- this section focusing on leveling. Since a
mometer is often graphed with respect to the downhole seismometer cannot be leveled manu-
minimum earth noise as defined by the new ally because it is down a borehole or posthole,
low-noise model (NLNM) (Peterson 1993). downhole seismometers have automatic leveling
Although a self-noise graph provides a compre- mechanisms that can adjust the sensing elements
hensive view of seismometer performance over or axes to measure true vertical and horizontal
frequency, the long-period performance of an motion.
instrument is typically the limiting parameter of The leveling range of a borehole seismometer
the self-noise. This is because all seismometers is typically 5 to cover the range of verticality
have self-noise lines that steadily curve upward at error due to the drilling. Some downhole seis-
long periods. mometers have wider leveling ranges, but this is
Seismometers can be further categorized by only needed in special situations.
the lower corner frequency (3 dB point) of the Leveling mechanisms in seismometers are
amplitude frequency response of the instrument. challenging to design due to the space constraints
Geophones have lower corner frequencies from inside the seismometer, the need to have high
1 to 40 Hz. Short-period seismometers have fidelity coupling to ground motion, and the need
lower corner frequencies from 1 to 4 Hz, and to maintain a design that is not subject to sponta-
broadband seismometers have lower corner fre- neous micro-mechanical movements. Different
quencies from 0,027 to 1 Hz. Broadband seis- instrument designers have taken different
mometers typically have lower noise floors over approaches to designing leveling mechanisms
wider bandwidths than short-period seismome- for seismometers.
ters, and short-period seismometers are typically There are four main methods for leveling sens-
quieter than geophones. Accelerometers are ing axes within a seismometer, namely, gimbal
approaching the performance levels of some geo- leveling, horizontal axis leveling, mainspring
phones and short-period seismometers and can be adjustment, and electronic centering. Each of
considered for downhole applications too. these methods has benefits and drawbacks that a
Nevertheless, in most downhole applications user should be aware of in selecting a downhole
today, broadband seismometers are typically seismometer. Some of the drawbacks relate to the
used since they can measure all three components axis topology within the seismometer. More
of motion, are quieter instruments, and can be details will be given below.
realized in packages that fit downhole. Broad- Within a triaxial seismometer, there are three
band seismometers are more challenging to sensing components or axes which measure
install downhole than short-period seismometers motion in three orthogonal directions. These
Downhole Seismometers 571
The Y-component is a more complicated expres- Eqs. 7, 8, and 9, the angles y and a are zero after
sion (Eq. 8), but that is the outcome of the order of leveling, and the three equations simplify to:
the rotations. Changing the order of the rotations,
namely, x ¼ x0
z ¼ z0
results in the following expression for the
Y-component:
Horizontal Axis Leveling
The horizontal axis leveling method is a variation
y ¼ y0 cos y z0 sin y (11) of gimbal leveling where the vertical axis is fixed
and not leveled, and the two horizontal axes are
Notice this equation has the same form as the leveled in one direction only. This method can
X-component in Eq. 7, so for conceptualizing, only be used on XYZ topology seismometers. It
this form can be used for both the X-component cannot be used in symmetric triaxial seismome-
and Y-component. ters because it would affect the mixing ratios
Equations 7, 8, and 9 are important because when converting from UVW to XYZ. Referring
some leveling methods do not actually level the to Eqs. 7, 8, and 9 again, the vertical axis is fixed
seismometer axes. This results in the seismom- in position, so Eq. 9 describes the Z-component
eter not outputting a true XYZ signal, but rather signal. The east–west axis or X-axis is leveled
an XYZ as defined by Eqs. 7, 8, and 9 where an about the Y-axis, so Eq. 7 with a equal to zero
axis output is mixed with signals from the other holds. Lastly, the north–south axis or Y-axis is
two axes. These equations will be referred to as leveled about the X-axis, so Eq. 8 with y equal to
each leveling method is discussed in detail. zero describes the Y-axis signal:
The seismometer axes are fixed in orthogonal There are a number of advantages of this
positions within the seismometer, so there is no method. There are no leveling actuators or main-
loss of orthogonality, but the XYZ components spring adjustment actuators, making the design
do not measure true vertical and horizontal more reliable and lower in cost. Depending on the
signals. lower corner frequency of the seismometer (see
This method of leveling is less costly to “▶ Principles of Broadband Seismometry”), the
implement, but it is limited to higher-noise seismometer can have a wide leveling range of up
seismometers. to 10 .
Downhole Seismometers, Table 2 A sample of downhole broadband seismometers available on the market in 2013
Lower corner Axis Leveling
Seismometer name Manufacturer frequency topology Leveling method range
Trillium Nanometrics 120 s Symmetric Gimbal 5
120 Borehole triaxial
Trillium Posthole Nanometrics 120 s Symmetric Gimbal 5 standard
triaxial 10 optional
Trillium Compact Nanometrics 120 s Symmetric Electronic 2.5
Posthole triaxial centering 10 D
KS-54000 Geotech 360 s XYZ Unknown 10
KS-2000BH Geotech 120 s (standard) XYZ Unknown 15
CMG-3TB Guralp 120 s XYZ Horizontal axis 3 standard
leveling 10 optional
CMG-40T-B Guralp 60 s (standard) XYZ Electronic 8
centering
CMG-Flute Guralp 60 s (standard) XYZ Mainspring 12
adjustment
STS-5A Streckeisen 120 s Symmetric Gimbal 5
triaxial
gimbal leveling as it does not compromise the seismometers due to factors in the design and
quality of the data while maintaining orthogonal- the quality of the manufacturing. The consistency
ity. The next most suitable method is horizontal of the seismometer refers to the spread in the self-
axis leveling. Although the data will be neither as noise floor over time and from instrument to
accurate nor as orthogonal, there are cost benefits instrument. This can best be assessed by huddle
to such an approach. Mainspring adjustment and testing at least three instruments over a period of
electronic centering are best reserved for higher- time (Sleeman et al. 2006). There are a number of
noise, lower-cost seismometers. published papers on self-noise floor of seismom-
For the leveling range, the user needs to ensure eters (Ringler and Hutt 2010).
that the seismometer has sufficient leveling range Table 2 documents some commercially avail-
to cover their deployment conditions. Insisting on able downhole broadband seismometers to be
a larger leveling range than needed unnecessarily considered in a selection process.
limits the choice of seismometer. Most downhole See company websites for further information.
applications do not require large leveling ranges.
The type of seismometer, namely, whether it is
a borehole or posthole type, will be determined Downhole Seismic Noise Environment
by the type of installation (see section “Types of
Installations”). Borehole seismometers tend to One of the goals of a seismic station installation is
cost more than posthole seismometers but can to minimize seismic noise and maximize the level
be deployed in deeper holes. of a seismic signal. In other words, the goal is to
The user needs to ensure that the seismometer maximize the signal-to-noise ratio (SNR) that can
will fit into the posthole or borehole that has been be achieved at a seismic station. A higher SNR
drilled. Deeper holes tend to have smaller diam- ratio will enable a station to detect smaller-
eters to reduce the cost per meter of drilling; magnitude seismic events. This goal applies
hence a narrower diameter instrument may be equally to surface vaults and downhole seismic
required. installations, but the means to accomplish this
Lastly, a user needs to consider the reliability goal are different for vaults and downhole instal-
and consistency of a seismometer. Some seis- lations due to the characteristics of the seismic
mometers are more reliable than other noise, the seismic signal, and the local geology.
576 Downhole Seismometers
Downhole
Seismometers,
Fig. 3 USGS graphic of
seismic noise types
Seismic noise can be broken down into differ- appears as an apparent horizontal acceleration.
ent types or sources of noise. These include wind- The resulting accelerations due to tilt are indis-
induced seismic noise, cultural seismic noise tinguishable from accelerations due to seismic
(noise generated by human activities), microseis- motion (Rodgers 1968; Graizer 2005). This effect
mic noise, moving-water-induced seismic noise, is relatively minor on the vertical axis of a seis-
and tilt-induced noise (Fig. 3). Each of these mometer as long as this axis is substantially ver-
noise sources will be briefly described. tical. On the horizontal axes, the effect is
significant and is described as
Microseismic Noise
Microseismic noise is caused by wave actions on H ¼ g0 sin y
large bodies of water such oceans, seas, and
lakes. As waves enter the continental margins Since the angles are small, the approximation
and crash onto shorelines, energy is transferred sin y = y can be used:
from the water into the earth, generating Rayleigh
waves. Microseismic noise ranges from 1 Hz to H ¼ g0 y
30 s, with a peak at 4–8 s known as the micro-
seismic peak. Microseismic noise is a global phe- where
nomenon, but has higher amplitudes closer to
coastlines and shorelines. • g0 = 9.8 m/s2 is the acceleration due to
gravity,
Tilt Noise • H is the horizontal acceleration in m/s2, and
Inertial sensors, such as seismometers, measure • y is the tilt in radians.
accelerations due to the motion of a seismic mass.
However, inertial sensors are also sensitive to Ground tilts can be caused by human activi-
static accelerations such as gravity. Normally, ties, wind, or changes in local barometric pres-
accelerations due to gravity are counterbalanced sure, temperature, or soil moisture content. The
by the mainspring in the suspension of a seis- magnitude of ground tilts is influenced by the
mometer, but when a seismometer is tilted stiffness of the ground. Hard-rock sites will be
(rotated around a horizontal axis), gravity much less susceptible to tilt as compared to a site
Downhole Seismometers 577
on clay. Tilt noise causes horizontal components energy into the ground. Regions with steeper
to be typically 20–50 dB noisier than the vertical gradients, such as waterfalls and rapids, generate
components at long periods (Bormann 2002). more seismic noise. Moving-water seismic noise
Tilt noise tends to decrease with depth as is continuous but varies in amplitude depending
many sources of tilt noise are surface effects, on water levels. Moving water induces surface
namely, wind, air pressure, soil moisture, and waves similar to wind noise.
human activity. Tilt noise can also be caused by
air currents circulating around a seismometer. Cultural Noise
This effect can be significant in borehole instal- Seismic noise generated by human activities is D
lations if strong vertical convection air currents referred to as cultural noise. The principal
are allowed to flow. Precautions must be taken to sources of cultural noise are related to transpor-
prevent the air moving. This problem can be tation, construction, military, and mineral extrac-
mitigated with a sand installation, which elimi- tion activities. However, any human activity can
nates air spaces around a seismometer, or with generate cultural noise, even a person walking
insulation methods, such as foam plugs, to limit down a path. Cities have higher levels of cultural
the airflow below and above a seismometer. noise than rural areas. Cultural noise can be sta-
tionary or nonstationary, periodic or random, and
Wind-Induced Noise generated by sources distributed or localized.
Wind is another major source of seismic noise. Cultural noise does tend to be diurnal, meaning
Wind transfers energy into the ground as it decel- it varies over the day with quietest times in the
erates from colliding with the terrain, water bod- middle of the night when there is the least human
ies, buildings, vegetation, and other objects activity. Like wind-induced noise, cultural seis-
coupled to the ground. The chaotic nature of mic noise is predominantly composed of high-
wind often excites resonances in objects it strikes. frequency surface waves but also includes a body
Trees and buildings sway in the wind, waves are wave component.
generated on water, and the ground is com-
pressed. Although the wind blows predominantly Seismic Noise Mitigation with Depth
in a horizontal direction, interactions with the The common factor in wind-induced, cultural,
surface translate the motion into the vertical com- moving-water, and microseismic noise is that
ponent. For example, as trees sway back and forth the noise is predominately generating surface
in the wind, alternating pressure is placed on waves. At short periods, the amplitude of surface
opposite sides of the root bundle imparting verti- waves decays exponentially with depth due to
cal motion. Wind-induced seismic noise is short amplitude drop from the surface and due to the
period in nature, meaning that is above 1 Hz. upper layers of the earth being highly attenuating
Wind noise is dependent on the wind velocity. or “lossy.” It is well known that high-frequency
Higher velocities generate more noise. For refer- seismic noise decreases with depth making bore-
ence, a wind speed below 3 m/s (11 km/h) does hole sites typically quieter than surface vault sites
not raise the background seismic noise at the from 1 to 100 Hz (Carter et al. 1991). At long
surface for grassland sites on sedimentary rock periods, the seismic noise also drops with depth,
in rolling hills (Withers et al. 1996). Wind particularly on the horizontal channels due to a
induces predominantly surface waves (Love and reduction in the tilt noise (Sorrells 1971).
Rayleigh waves) along with a small component While seismic noise attenuates with depth,
of body waves onto the ground. seismic signal amplitudes are dependent on a
number of factors relating to the geology of the
Moving-Water-Induced Noise site. As a seismic signal propagates toward the
Seismic noise is generated in watercourses from surface, it typically travels through progressively
small streams to large rivers. As water flows lower Q layers and loses energy. The layers at the
downhill, turbulence is created which imparts surface have the lowest Qs and are responsible for
578 Downhole Seismometers
−100
Clay
−140
Bedrock (at 15m)
−150
−160
30 m deep cased hole installation
of posthole seismometer
−170
−180
Downhole Seismometers, Fig. 4 Vault installation and posthole installations at various depths as measured in
Ottawa, On, Canada (Nanometrics Inc 2013)
a significant portion of the loss in energy (Aster seismometer be placed to result in an improve-
and Shearer 1991). However, as the velocity of ment in the signal-to-noise ratio? The answer is
the signal slows, the amplitude increases, and within 3–10 m of the surface depending on the
there is also surface amplification effect. In sum- geology due to the attenuation of temperature, tilt
mary, it is difficult to estimate the amplitude of noise, and surface noise.
the signal with depth. Consider an extreme example of 15 m clay
Nevertheless, the effect of seismic noise atten- above quartzite bedrock in an urban environment
uation with depth results in improved signal-to- shown in Fig. 4. This was measured in Ottawa,
noise ratios especially when the final layers of Ontario, Canada, using Trillium120 Posthole
earth are highly lossy (Young et al. 1996). Ide- seismometers (Nanometrics Inc 2013). Clay has a
ally, a downhole seismometer should be placed in very low Q, and the quartzite has a much higher Q.
hard rock below a highly lossy overburden. Unsurprisingly, the surface vault has the highest
The question that remains is, how deep should noise at low and high frequencies. The 1.2 m direct
the downhole seismometer be placed? The burial posthole seismometer installation shows
answer is as deep as economically and practically improved low-frequency vertical and horizontal
possible as signal-to-noise ratios typically noise and improved high-frequency noise. At 7 m,
improve with depth in the first few hundred the posthole seismometer installation shows further
meters. However, the question can be turned improvements in low-frequency horizontal noise
around to: how shallow can a downhole and improved high-frequency noise. At 30 m from
Downhole Seismometers 579
the surface in bedrock, the horizontal cased in steel or pvc and cemented in. Shallow
low-frequency noise has dropped 30 dB, the holes drilled into hard rock do not need to be
low-frequency vertical noise has improved 10 dB, cased. Seismometers are typically deployed into
and the high-frequency noise on the vertical and shallow holes using a sand installation method.
horizontal channels has dropped 40 dB as com- Shallow holes that are used for temporary
pared to the surface vault. This example shows deployments are typically not cased. This mini-
that improvements in noise levels can be achieved mizes the cost of the installation. A seismometer
in relatively shallow downhole installations. deployed in an uncased hole is backfilled with
Deeper installations could possibly yield further sand or the excavated materials from the hole. D
improvements, but in this example an improvement This type of deployment is also known as a direct
was achieved in the first 15 m. It should be noted burial installation. A direct burial installation is
that these results are specific for this site. In general, quicker and easier to install than an installation
noise levels will be site specific. using a vault seismometer. It is often challenging
In summary, even shallow posthole installa- leveling a vault seismometer in a field deploy-
tions can provide SNR improvements over sur- ment since it requires a hard surface, access to the
face vaults. leveling feet, and bubble level. With a posthole
seismometer, the procedure is to dig a small hole,
place the seismometer in the hole, backfill with
Downhole Seismometer Installations earth or sand, and start the autoleveling in the
seismometer. There is no further work required
Types of Installations to insulate or weatherproof the installation.
Downhole seismometers can be deployed into
boreholes or shallow holes. Seismometer Emplacement
Boreholes are steel-cased holes that are Once a seismometer is lowered into a borehole, it
cemented into competent rock. Typical seismic must be secured rigidly in the hole to ensure good
boreholes are 50–150 m deep, but there are seismic seismic coupling to the borehole casing and
boreholes 2,000 m deep and beyond (Young the ground. There are two methods for
et al. 1994). Although a cemented steel casing accomplishing this. The first is with a mechanical
extends to the surface, a borehole poorly propa- borehole lock and the second is with a sand
gates surface noise down the borehole. There are installation.
two reasons for this. First, the coupling of the A borehole lock is a device attached to the
casing to the upper softer layers of material is not seismometer which clamps the seismometer to
as good as the coupling to stiffer materials lower the side of the borehole. Typically, a borehole
down. Second, the borehole is relatively compliant lock has three fixed pins on one side of the
horizontally over the length of the borehole. holelock which create a three-point contact with
These holes can be drilled using water-well or the borehole and one moving pin on the opposite
oil-well drilling equipment. The steel casings are side which clamps the holelock to the side of the
cemented into the borehole to stabilize the casing, borehole. The moving pin pushes the holelock
provide good coupling to the ground, and main- three-point contact pins against the borehole
tain the integrity of underground aquifers. with a force sufficient to hold the seismometer
Shallow holes are typically holes 0.5–5 m in place. The seismometer is then supported by
deep that are dug by shovel or lightweight drilling the four pins. This is a kinematic coupling. The
equipment such as augers, small drill rigs, screw- seismometer must not be touching the bottom of
pile equipment, and the like (Fig. 5). These holes the borehole; otherwise the coupling will be over-
are typically larger in diameter as compared to constrained which can result in undesired
boreholes to enable a wider range of seismome- motions. With this method of locking, the seis-
ters to be deployed and to improve access to the mometer takes on the vertical angle of the bore-
seismometer. At permanent sites, the hole may be hole at that point. The moving pin on the holelock
580 Downhole Seismometers
Downhole Cellular
Seismometers, Antenna
Fig. 5 Components of a
typical posthole station
GPS
with the sensor 1–6 m Antenna
below the surface
Solar Panel
Inside:
- Digitizer
- Battery
- Solar Regulator
- Cellular Modem
Cable
Mounting Conduit
Pole
PVC Tube,
Capped
Sand
Posthole
Seismometer
Earth (consistency may vary)
Bedrock
can be actuated with an electric motor or a material that is prone to spontaneous movements
mechanical system controlled from the surface. that can be detected by a seismometer during a
The sand installation method involves lower- settling period. Another issue with sand installa-
ing the seismometer to the bottom of a borehole tions is that after a period of time, moisture and
or shallow hole and pouring dry sand into the hole dissolved minerals in the sand can solidify the
sufficient to cover the seismometer in sand. The sand making it challenging to remove the
sand flows around the entire seismometer filling seismometer.
the entire gap between the seismometer and the
hole. This method of installation has a few bene- Economics of Installations
fits and drawbacks. A sand installation is simpler One of the primary considerations in deploying a
and less costly than a holelock installation. It downhole seismic station is the cost of the instal-
eliminates convection air currents around the lation, especially in comparison to vault installa-
seismometer and effectively couples the seis- tions. In the past, downhole installations have
mometer to the casing. However, sand is a loose been substantially more costly than vault
Downhole Seismometers 581
installations because downhole installations were used. Uncased shallow holes are the least expen-
predominately 100–200 m-deep borehole instal- sive and can be dug with hand or powered augers.
lations with high-cost, top-performance borehole Screw-pile installations have costs between bore-
seismometers. A typical borehole was drilled holes and shallow holes. Screw piles are installed
using oil-well drilling equipment which costs with skid-steers and have the added benefit of
approximately US$100,000 per borehole. being removable at the end of a deployment.
A borehole station of this kind would have had Local construction contractors can provide cur-
a total installed cost of US$150,000–300,000 rent prices for comparison.
depending on the location and ancillary D
equipment included. This limited borehole instal- Security of Installations
lations to those who had budgets large enough Boreholes and shallow holes are in general more
to cover cost of a borehole network and had secure than vault installations for the seismome-
a requirement for excellent signal-to-noise ratios. ters. The overall footprint of a downhole instal-
Recently, newer instruments and installation lation is much smaller and less visible than a
methods have dramatically lowered the cost of a typical permanent vault installation and can be
downhole installation making downhole stations camouflaged more easily. If desired, the entire
cost competitive with vault installations. borehole wellhead can be buried. Buried under-
One of the main innovations is switching away ground conduits can be routed from the borehole
from using oil-well drilling equipment and mate- or shallow hole to the communications and power
rials. Oil-well drilling equipment is designed and hub. The curious would assume that a visible
sized to drill deep wells (up to many kilometers) wellhead is that of a water well which is
in complex geologies. This equipment typically uninteresting. Often winches and tripods
includes steerable drill bits to enable horizontal (or masts) are required to extract seismometers
drilling. For a typical vertical seismic borehole of from the hole providing a level of physical secu-
50–150 m deep, this equipment is expensive to rity that is hard to overwhelm even for the most
use and in most geologies unnecessary. determined thief.
Water-well drilling rigs are more suitable and Nevertheless, a downhole seismometer instal-
substantially cheaper to operate as compared to lation only protects half the equipment at the
oil-well drilling equipment. Water-well drilling station. The communications and power equip-
equipment is designed for drilling only vertical ment is still highly visible and attractive to a thief.
boreholes 10–300 m deep reducing rig size, com- Solar panels and backup batteries are popular and
plexity, and cost. Water-well drilling services are useful everywhere in the world and are routinely
typically available in most locations reducing stolen from less-secure sites (see Fig. 6). A solu-
deployment costs and are competitively priced tion to station security must cover all the equip-
due to the demand for water wells. A typical ment at a site (Fig. 7).
60 m water well can be drilled in North America Typically, there is a trade-off to be made
for $US 5,000–25,000. between a rural site which has low cultural seismic
A seismic borehole includes costs for casing, noise but is less secure due to its isolation and a
cementing, and pressure testing, in addition to the secure urban site, such as police or fire station,
drilling costs. Casing and cementing costs will which has higher cultural seismic noise. Ideally,
vary with depth, casing type, and diameter. Local an operator wants a low cultural noise site that is
drilling contractors who are often very familiar also secure. With downhole installations, a secure
with the geology in the region can readily provide urban site enables the reduction of seismic noise by
up-to-date prices and guidance on drilling. providing vertical separation from seismic noise
For posthole installations, cased boreholes, which originates at the surface. The degree of
screw piles, and uncased shallow holes can be attenuation of seismic noise with depth will depend
582 Downhole Seismometers
Downhole
Seismometers,
Fig. 6 Rural posthole
installation with cellular
communications in Alberta,
Canada
Downhole
Seismometers,
Fig. 7 Rural vault
installation with satellite
communications in Fiji
on the local geology and nature of the cultural can be as little as the area of the hole. Typically, a
noise. An urban borehole or posthole may not be borehole or shallow-hole installation will require
as quiet as a good rural site, but it is secure and can a 0.25–1 m2 of space (Fig. 9). Seismic vaults
be quickly serviced if there is a problem. typically use much more space, particularly if
A network operator will have to trade security the vault is a walk-in type of vault.
and ease of access for uptime and data quality The smaller footprint of a downhole site offers
based on the local geology, cultural seismic more flexibility in station placement, especially
noise, economic conditions, cultural norms in an urban environment where available land is
regarding property, and cost. limited and costly and local building codes
restrictive.
Station Footprint
Downhole installations typically have a smaller Site Environmental Conditions
footprint as compared to vault installations. The The site environmental conditions, namely,
land area required for a borehole or shallow hole weather, water, and wildlife, have a large impact
Downhole Seismometers 583
Downhole
Seismometers,
Fig. 8 Ground
temperature variations with
depth for Ottawa
reproduced from Williams
and Gold (1976)
on the performance of seismic stations. While penetration depth and depends on the composi-
great care must be taken with the design of vault tion of the soils and the moisture content of the
installations to mitigate the effects of tempera- soils. For daily temperature variations, the pene-
ture, wind, air pressure, water egress and tration depth is less than 1.2 m for all soil types.
flooding, and wildlife, downhole installations The yearly penetration depths vary from 5 to
are much less susceptible to these effects and 20 m depending on the soil conditions. For most
require less design effort. downhole seismometer installations, an annual
temperature variation of 3 is acceptable and
Temperature can be achieved in as little as 3–6 m below-
Seismometers are sensitive to temperature ground. See Fig. 8: ground temperature variations
changes, and ideally, a seismometer should be with depth for Ottawa, reproduced from Williams
exposed to as little temperature variation as is and Gold (1976).
economic. At the surface, the temperature can
range from 20 C in the winter to +35 C in Water Egress
the summer in a northern continental climate, Seismic vaults are usually located at or just below
with daily fluctuations between 10 C and 20 - the surface on undisturbed soils and are subject to
C. Other climates can be more or less extreme. water egress and flooding from time to time.
A vault installation uses thermal insulation and While vaults are often designed to be waterproof,
structures with high thermal mass to reduce the there is usually some weakness in the implemen-
temperature variation of the seismometer. This tation that enables water to enter the vault over
requires design work and cost. time or during flood situations. Some vaults are
A downhole installation relies on the heat equipped with pumps, but over time, pumps fail.
capacity and thermal conductivity properties of This would not be an issue except that vault
the ground to reduce the temperature fluctuations seismometers and ancillary equipment are not
to a minimum. The depth where the temperature designed for continuous submersion. Flooding
fluctuation is 1/100 of the surface temperature causes the temporary shutdown of many vault
variation for a given time period is known as the seismic stations.
584 Downhole Seismometers
Deployment Considerations
Borehole Verticality
Boreholes are typically not vertical holes but
have small tilt angles of up to 10 from vertical.
This is due to the accuracy of the driller’s equip-
ment and technique and the nature of the mate-
rial being drilled. Drilling practices that affect
the verticality of a borehole include the weight
on the drill bit, the speed of drilling, the degree
of hole cleaning, drill bit selection, whether a
steerable drill bit is used, and the use of vertical-
ity monitoring equipment while drilling. Expe-
rienced drillers understand the trade-offs in
maintaining a vertical borehole. In maintaining
a vertical borehole, a driller will want to under-
stand the geology of the area prior to drilling. It
is harder for a driller to maintain verticality with
Downhole Seismometers, Fig. 9 Rural borehole seis- complex borehole profiles. For example, it is
mometer deployment in Alberta, Canada
more challenging to maintain verticality when
drilling soft sediments such as clay or sand or in
sediments with materials with differing hardness
Downhole seismometers are designed for con- such as glacial till or limestone with cavities.
tinuous submersion and are unaffected by peri- With glacial till, the drill bit can be deflected
odic flooding. Generally, boreholes and postholes by rocks and boulders in the sand, and with
are designed to be dry, but it is not strictly nec- limestone, the drill bit can drift in the cavities.
essary for the equipment. Shallow hole can be dry Other challenging borehole profiles include
or wet, or even below the water table. Water may highly fractured rock and sedimentary rocks
degrade the seismic performance due to move- with steep dip angles.
ment of the water, but it will not cause the total In most situations, drillers can maintain the
loss of data from a station. tilt to less than 3 off vertical. Typically, drilling
specifications stipulate that the borehole shall be
Wildlife within 3 off vertical or less. USGS suggests a
Animals from time to time cause problems drilling specification of 2 or less (Albuquerque
at seismic stations. This can range from large Seismological Laboratory 2001). Even with
animals walking in the vicinity of the station or a vertical error of 3 , the vertical offset or
scratching themselves against the station down drift at the bottom of the 100 m borehole is
to mice building nests in the insulation sur- just over 5 m. The drift at the bottom of a
rounding a seismometer or chewing cables borehole may be a concern in some installations
and every critter and its activities in between. that have underground features that must be
Being at the surface, vault installations are avoided.
simply more accessible and susceptible to Another issue related to borehole verticality
these invasions. Downhole Installations are is the tilt of the seismometer in the borehole. It is
away from the surface and difficult to access not possible to maintain the verticality of a
Downhole Seismometers 585
seismometer as it is lowered; a seismometer will correlated with the surface data set to determine
always be resting against the side of the borehole the offset from north.
as it goes down and will tend to assume the angle • Rifle scope alignment involves mounting a
of the borehole when it reaches the bottom. For rifle scope vertically pointing down the bore-
example, if the seismometer is lowered to 50 m hole. The rifle scope cross hair is focused on
in the borehole, and the borehole at 50 m is 2 off the north–south marking on the seismometer
vertical, the seismometer will tend to be 2 off downhole. A reading of the rifle scope rotation
vertical when it comes to rest. See section relative to north is taken.
“Levelling” for how to compensate for the ver- • North-finder alignment uses a north-finder D
tical error. instrument which measures earth’s rotation
vector to determine north. The north finder
can be mounted on the seismometer or
Aligning to North lowered down the borehole to engage the
Seismometers are typically aligned to north. seismometer.
Vault seismometers are aligned to north manually
using a compass. A compass bearing is taken, a The second step is to rotate the seismometer to
north–south line is scribed on the vault floor, and be north aligned. This can be done digitally,
the seismometer is manually rotated to align with mechanically with a “bishop’s hat” mechanism,
the north–south line. There are many variations or manually with an alignment rod.
of this method, but overall the process is
straightforward. • Digital alignment involves mathematically
Aligning downhole seismometers to north is rotating the digital data set to align to north.
more challenging. The user has only limited This can be done with no loss of precision. The
access to the instrument downhole, and a com- user enters a rotation parameter into the seis-
pass cannot be used downhole due to the mag- mometer or digitizer configuration, and the
netic shielding by the steel borehole casing. instrument subsequently rotates all time-
Nevertheless, methods have been developed to series data.
align downhole seismometers. • Bishop’s hat mechanical alignment is a
The downhole alignment process can be bro- multistep process. A holelock with a bishop’s
ken down into two steps. The first step is to hat mechanism is lowered into the borehole
determine the seismometer orientation relative and clamped in place. A north-measuring
to north. This can be accomplished using a num- device is lowered onto the bishop’s hat mech-
ber of different methods including downhole anism to determine the orientation of the
gyroscopic alignment, surface seismometer holelock. The north-measuring device is
referencing, rifle scope alignment, and north- removed, and the seismometer bishop’s hat
finder alignment. See below: mate is rotated and locked at the surface. The
seismometer is lowered down onto the
• With gyroscopic alignment, a gyroscope is bishop’s hat and is now aligned to north.
activated and aligned to north at the surface. A bishop’s hat mechanism is a device that
It is then lowered down the borehole and forces the mating assembly (the seismometer)
engages with a north-aligned feature on the to a particular orientation. It is called a
seismometer or holelock. A reading is taken bishop’s hat because it is shaped like a
from the gyroscope to determine the rotational bishop’s hat.
offset from north. • Lastly, a seismometer can be aligned with an
• Surface seismometer alignment involves alignment rod. The alignment rod is attached
collecting a seismic data set from the downhole to the seismometer and the seismometer is
seismometer and a surface seismometer located lowered into the hole. From the surface, the
at the wellhead. The downhole data set is user rotates the alignment rod and the
586 Downhole Seismometers
amplification and resonance (mostly around 2 s) of In the MLIT (2010a) document, a total of
the lakebed areas of Mexico City (Anderson 9 areas in Japan were specified (4 in Tokyo, 3 in
et al. 1986; Çelebi et al. 1987). To the best knowl- Nagoya, and 2 in Osaka). For each one of these
edge of the authors, however, there are no publicly areas, simulated acceleration and velocity time-
available records of the responses of tall structures histories were specified. Takewaki and others
from these past earthquakes. Therefore, records (2011) report that velocity response spectra of
obtained from numerous instrumented tall build- all the simulated ground motions have significant
ings during the Great East Japan earthquake of large peaks within 2–8 s range. It is important to
March 11, 2011 offer a rare opportunity to study state that Takewaki and others (2011) report that
and understand how structures characterized by in Shinjuku area of Tokyo, where tall buildings
predominantly long-period responses behave dur- are discussed later in this entry, the KNET
ing medium to large events originating at long Shinjuku station (TKY007) motions exhibit pre-
distances. Such effects have consequences for dominant peaks between 3 and 8 s range.
large metropolitan areas in Japan but also in other In this entry, we present acceleration and veloc-
parts of the world, including the United States ity response spectra of ground motions recorded
(e.g., Los Angeles area from Southern California during the Tohoku earthquake from two stations.
earthquakes, Chicago from NMSZ, and the Seattle As an example, for Osaka Bay, spectra of motions
(WA) area from large Cascadia subduction zone from KIKNET station OSKH02 are provided in
earthquakes). For example, the recent M = 5.8 Fig. 1 and display the long periods both at surface
Virginia earthquake of August 23, 2011, was felt and downhole. This station is closest to the exam-
in 21 states of the Eastern and Central United ple building in Osaka Bay discussed also later in
States that include large cities such as New York the entry. In Fig. 2, we present spectra of motions
and Chicago (http://earthquake.usgs.gov/earth- from KNET stations TKY007 (representative of
quakes/eqinthenews/ 2011/se082311a/#summary, motions in the Shinjuku area of Tokyo, as
last accessed September 11, 2014). During this discussed above). These spectra clearly display
event, occupants of a 28-story building at MIT the long periods capable of affecting responses of
Campus in Cambridge, MA, experienced the shak- the tall buildings. TKY007 was also examined by
ing (Toksoz, 2011, “personal communication”). Takewaki and others (2012).
Drift Issues of Tall Buildings in 2011 Tohoku Earth- event at KIKNET OSKH02 station in Osaka Bay. Velocity
quake, Fig. 1 (Left) NS and EW acceleration response response spectra exhibit the large amplitudes between
spectra and (right) velocity response spectra of ground 5 and 7 s
motions recorded at surface and downhole during Tohoku
Drift Issues of Tall Buildings in 2011 Tohoku Earth- TKY007 station in Shinjuku area of Tokyo where tall
quake, Fig. 2 (Left) NS and EW acceleration response buildings are concentrated. Velocity response spectra
spectra and (right) velocity response spectra of ground exhibit the large amplitudes particularly for periods >2 s
motions recorded during Tohoku event at KNET
Figure 3 shows vertical sections and locations of The parameters used in computing the site trans-
triaxial accelerometers. Figure 4 shows plan fer functions are the (depth vs. Vs) Profiles A, B,
views, orientation of the building, and location and C shown in Fig. 3 (left). Profile A is an
of accelerometers at the 52nd floor. approximation based on the geotechnical data
The building was subjected to an input motion for free-field KIK-NET station OSKH02 that is
at ground level with peak accelerations of
3 % g near (
2.5 km) the building. In this profile, the
during the Tohoku earthquake. Furthermore, the upper and softer layers have been ignored. By
building was founded at a site with a site fre- way of comparison with the transfer functions
quency (period) of 0.13–0.17 Hz (5.88–7.69 s) computed for Profiles B and C, which underlie
as seen in Fig. 5 (left) (Çelebi et al. 2012). the building, it is concluded that the upper layers
590 Drift Issues of Tall Buildings in 2011 Tohoku Earthquake
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake, Table 1 Upper limits of drift ratios (Japan, Chile, the
United States, Turkey, New Zealand, and Eurocode 8)
Upper limit
Code of drift ratio (%) Comment Reference
Chile 0.2 Results in elastic design NCh433.Of96 (1996)
Japan 1.0 Max for buildings taller than 60 m. For Building Center of Japan
collapse prevention (level 2) motions 2001a, b
The United States 2.0 No collapse state ASCE7-10 (2007)
Turkey 2.0 Can be increased by 50 % in case of some DBYYHY (2007):
steel frames Section 2.10.1.3, Eqn:
2.19 in Turkish Code
(2007)
New Zealand 2.5 NZS1170.5(2004)
Eurocode 8 1.0 (max) For buildings having nonstructural elements Section 4.4.3.2, Eqn: 4.33
fixed in a way so as not to interfere with Eurocode 8 (2008)
structural deformations or without
nonstructural elements
do not significantly alter the computed funda- or taller (The Building Center of Japan 2001a,
mental frequency of the site of this building. b)). However, average drift ratios are much
Q values used in calculating the transfer functions larger than expected for an input motion with
range between 25 and 60 for shear wave veloci- a small peak acceleration in the order of only
ties between 200 and 600 m/s – having been 3 % g. In the United States, the comparative
approximately interpolated to vary linearly maximum drift ratio for tall buildings for
within these bounds. As seen in Fig. 5 (left), the Risk Category 1 or 2 is 2 % (Table 12.12,
site fundamental frequency is computed to be in ASCE7-10 2007).
the range of 0.13–0.17 Hz due to the dominant Detailed studies of the Osaka basin are
characteristics of layers three and four (typically presented in Yamada and Horike (2007),
of the area of the site of the building and KIKNET Sekiguchi et al. (2007), and Iwaki and Iwata
OSKH02 strong-motion station as described in (2008), but the results of these studies likely do
Fig. 5 [left]). not accurately characterize the local site transfer
The building exhibited fundamental frequen- function. Instead, we computed the site transfer
cies (periods) of 0.152Hz (6.58 s) in the functions using software developed by
X-direction and 0.145 Hz (6.90) in the C. Mueller, 1997 (“personal communication”),
Y-direction as seen in Fig. 5 (right). Such close which is based on Haskell’s shear wave propaga-
frequencies are clear confirmation of resonance tion method (Haskell 1953, 1960). In this
that led to prolonged responses (
1,000s method, the transfer function is computed using
recorded) and long-duration high-amplitude linear propagation of vertically incident SH
shaking as exhibited in Fig. 4 which shows aver- waves and has, as input, data related to the lay-
age drift ratios. ered media (number of layers, depth of each
The average drift ratios (Fig. 6) computed layer, corresponding Vs, damping, and density),
from relative displacements between many desired depth of computation of transfer function,
floors indicate that maximum average drift ratios computation frequency (df), half space substra-
experienced during the mainshock were between tum shear wave velocity, and density. Damping
0.5 % and 1.0 % for the X-direction and (x) in the software is introduced via the quality
0.2–0.4 % for the Y-direction. These average factor (Q), a term used by geophysicists that is
drift ratios are less than the maximum 1 % limit related to damping by x = 1/(2Q).
usually used in Japan for collapse protection In summary, during the Tohoku event, the
level motions (level 2 used for buildings 60 m building was subjected to ground level input
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake 591
Drift Issues of Tall Buildings in 2011 Tohoku Earth- on the 52nd, 38th, 18th, and ground level (First floor).
quake, Fig. 3 Vertical sections of the building showing X and Y denote principal axes of the building (see Fig. 4)
major dimensions and locations of triaxial accelerometers (From Çelebi et al. 2012)
acceleration of 3 % of g. But the deformations Furthermore, the actual drift ratios computed
were large enough (to realize sizeable average from relative displacements divided by story
drift ratios) that the building lost its functionality heights between some of the pairs of two consec-
for several weeks. Elevator cables were utive floors are certainly to be larger than the
entangled. This case, along with others, deserves average drift ratios computed from differential
further studies as to how tall buildings would displacements between several floors, due to
respond if the ground input accelerations were sparse deployments of instruments, as in
20–30 % of g with similar low-frequency content this case.
that could be caused by a closer earthquake (e.g.,
100 km or less) [e.g., similar to 1923 Kanto and Building B: 55-Story Shinjuku Center Building
1994 Kobe earthquakes]. It is important to note in Shinjuku, Tokyo
that same elevator cable tangling problems also According to EERI Special Earthquake Report
occurred in several tall buildings in the Shinjuku (EERI Newsletter 2012), the 54-story Shinjuku
area of Tokyo, Japan (Hisada et al. 2012a). Center Building was constructed in 1979.
592 Drift Issues of Tall Buildings in 2011 Tohoku Earthquake
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake, Fig. 4 (Left) Typical plan view (the figure shows 52nd
floor) and (right) orientations (From Çelebi et al. 2012)
Drift Issues of Tall Buildings in 2011 Tohoku Earth- peaks in the transfer function, particularly for the funda-
quake, Fig. 5 (Left) Transfer functions computed for mental mode of the site. (Right) Spectral ratios of ampli-
Profile A (near the OSKH02 strong-motion site; www. tude spectra at 52nd floor, 38th floor, and 18th floor with
kik.bosai.go.jp/, last accessed 09/16/2011) and Profiles respect to that at first floor. Note that the third mode in
B and C below the building. The depth of the softer X-direction is identified from the ratios. Different colors
upper two layers (to about 1,500 m depth) below the (red, black, and blue) are used only to distinguish lines
building does not significantly change the position of the corresponding to different floors in descending order
The report states: “The structure’s height is Figure 7 shows a picture of the building and the
223 m, and the first natural period of the structure oil dampers installed in 2009.
is 5.2 and 6.2 s in two perpendicular directions. In Figure 8 shows recorded acceleration time
2009, the building was retrofitted from the 15th to history at first floor and computed displacements
39th floor with 288 oil dampers that were config- at the roof during the Tohoku earthquake at
ured to exhibit a form of deformation depen-
375 km from the epicenter. Without the actual
dency, in addition to velocity dependency. In data being available, as shown in Fig. 8, the
the March quake, the dampers were calculated maximum roof displacement is in the transverse
to have reduced the maximum accelerations (Y) direction and reported as 54.2 cm
by 30 % and roof displacements by 22 %.” (Figure adopted from Moehle, 2012, “personal
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake 593
D
Drift Issues of Tall Buildings in 2011 Tohoku Earth- frame, the numbers in denominators are distances (in cm)
quake, Fig. 6 Average drift ratios computed from dis- between the designated floors
placements between 52nd, 38th, and 1st floors. In each
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake, Fig. 7 (Left) Picture of Shinjuku Center Building and
(right) dampers installed in 2009 (From EERI Newsletter 2012)
Drift Issues of Tall Buildings in 2011 Tohoku Earth- (max
54 cm) during Tohoku earthquake (Figure
quake, Fig. 8 Recorded first-story accelerations adopted from J. Moehle and Y. Sinozaki)
(max
14 % of g) and computed roof displacements
roof displacement divided by the height of the Drift Issues of Tall Buildings in 2011 Tohoku Earth-
building, due to unavailable detailed data, as quake, Table 2 Maximum observed responses (From
Hosozawa et al. 2012 and Sinozaki, 2012, “personal
done in this case. communication”)
Max. Accel.
Buildings C and D: Two Tall Buildings at (cm/s/s) Max displ. (cm)
Kogakuin University Campus, Shinjuku, Long. Trans. Long. Trans.
Tokyo (X) (Y) (X) (Y)
Without mentioning drift ratios, detailed studies Roof 236.0 161.3 49.4 54.2
of two buildings on the Kogakuin University 28th 112.7 171.3 26.3 33.3
Shinjuku, Tokyo, campus are provided by Hisada floor
et al. (2012a, b). Figure 9 shows the vertical 1st floor 94.3 142.1
sections of the two subject buildings (Fig. 10).
Drift Issues of Tall Buildings in 2011 Tohoku Earth- storeys STEC Office Building in Shinjuku area of Tokyo,
quake, Fig. 9 Vertical sections showing instrumented Japan (Figure adopted from Hisada et al. (2012a and b))
floors of 30 storeys Shinjuku Campus Building and 28
buildings for Risk Category 1 or 2 (Table 12.12, ratios under the low peak input motions expe-
ASCE7-10 2007). Even better, drift ratios closer rienced during and following the Tohoku
to the 0.2 % imposed by the Chilean Code for earthquake of 2011.
Earthquake Resistant Design of Buildings 3. Prolonged cyclic behavior of the buildings,
(NCh433.Of96 1996) should be considered. And even at moderate to small amplitudes with
(c) finally, further applications of unique relatively acceptable elastic drift ratios, can
response modification features are feasible to be a cause for low-cycle fatigue. This issue is
reduce the drift ratios. significant, and unless structural dynamics
Finally, the following points are made: characteristic of the buildings is modified,
low-cycle fatigue can be an important concern
1. Behavior and performances of these particular for the health of the building in the future.
tall buildings far away from the strong shaking
source of the M9.0 Tohoku earthquake of
2011 and large-magnitude aftershocks should Summary
serve as a reminder that, in the United States as
well as in many other countries, risk to such One of the most significant effects of the M9.0
built environments from distant sources must Tohoku, Japan, earthquake of March 11, 2011, is
always be considered. the now well-known long-duration (>10 min)
2. The potential risk from closer large- shaking of buildings in Japan – particularly
magnitude earthquakes that could subject the those in Tokyo (
350–375 km from the epicen-
buildings to larger peak input motions should ter) and in places as far as Osaka (
770 km from
be assessed in light of the substantial drift the epicenter). Although none collapsed, the
596 Drift Issues of Tall Buildings in 2011 Tohoku Earthquake
Drift Issues of Tall Buildings in 2011 Tohoku Earth- in Shinjuku area of Tokyo, Japan. Highest drift ratio
quake, Fig. 10 Combinations of average drift ratios (
0.5%) is computed from displacements in the NS direc-
computed from recorded data of 30 storeys Shinjuku tion between 24th and 22nd floors (bottom left frame)
Campus Building and 28 storeys STEC Office Building
strong shaking caused many tall buildings not to The other three examples are from Tokyo. For
be functional for days and weeks. example, based on records obtained from
The purpose of this entry is to discuss the a 54-story building retrofitted with 288 oil
behavior and performance of four tall buildings, dampers on 24 floors, computations show that
considered to be representative of most tall build- average drift ratio may have reached
0.3 %
ings in Tokyo and other locations and from which and maximum drift ratio likely was > .3 %. The
shaking data were retrieved. Of particular interest maximum allowed by Japanese practice for
is a building in Osaka that almost reached an buildings taller than 60 m and for collapse pro-
average of 0.5 % drift ratio – this, with a ground tection (level 2) motions is 1 % (The Building
level input motion of
3 % g is significant. What Center of Japan, 2001a, b).
might have happened during an event with input Performances of tall buildings in many seis-
level motions with similar low-frequency content mically active regions of the world (e.g., Chile,
and in 10–20 % g range is a legitimate question Turkey) or those tall buildings affected by long-
that must be pondered. The particular building distance long-period effects by sources at
had serious site effects and was in resonance. a distance (e.g., Abu Dhabi, Dubai) are of
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake 597
Drift Issues of Tall Buildings in 2011 Tohoku Earthquake, Table 3 Epicentral distances, peak accelerations and
displacements, and computed average drift ratios of the buildings supplemented by peak accelerations of ground
motions recorded at stations in proximity to the buildings
Rel. H
Epicentral Peak accel. at Peak accel. at top No of displ Height (Ave)
Bldg distance (km) ground level (gals) floor (gals) floors (cm) (m) DR (%)
A 767 30 130 54
130 218
.5
Ba
375 142 236 54 54.2 216
.2
Cb
375 92 340 31 33.4 119
.5
D
Db
375 92 302 29 34.9 90
.4
Peak accelerations (gals) at Free-Field Stations in the vicinity of the buildings
Free-Field Station NS EW
KIKNET:OSKH02 13 15
\surface
OSKH02\downhole 1 1
KNET: TKY007 198 165
H height of building used to computed average drift ratios, DR drift ratio
a
From Sinozaki, 2012, (“personal communication”)
b
From Hisada and others (2012a, b)
interest to the earthquake engineering commu- Çelebi M, Prince J, Dietel C, Onate M, Chavez G (1987)
nity. Chile imposes 0.2 % drift limit that results The culprit in Mexico City – amplifications of
motions. Earthq Spectra, J EERI 3:2. 315–328
in elastic design. The United States and Turkey Çelebi M, Sanli A, Sinclair M, Gallant S, Radulescu
impose 2 % drift limit. Such wide variations D (2004) Real-time seismic monitoring needs of
of drift limits in design practices deserve discus- a building owner and the solution – a cooperative
sion in light of functionality and performance effort, journal of EERI. Earthq Spectra 20(2):333–346
Çelebi M, Okawa I, Kashima T, Koyama S, Iiba M (2014)
of tall buildings during the 2011 Tohoku event. Response of a tall building far from the epicenter of the
March 11, 2011 M = 9.0 Great East Japan earthquake
and its aftershocks. J Design Tall Build Spec Struct
(a John Wiley Journal) 23:427–441
Cross-References DBYYHY (2007) Deprem bolgelerinde yapilacak binalar
hakkinda yonetmelik (Code for buildings to be built in
earthquake prone regions of Turkey)
▶ Building Monitoring Using a Ground-Based EERI Newsletter (2012) EERI special earthquake report:
Radar Learning from earthquakes Performance of Engineer-
▶ System and Damage Identification of Civil edStructures in the Mw 9.0 Tohoku, Earthquake of 11
Structures Mar 2011in EERI Newsletter (January 2012)
Eurocode 8 (2008) Design of structures for earthquake
resistance: Part1: general rules, seismic actions and
rules for buildings, European standard, prEN1998-1
Haskell NA (1953) The dispersion of surface waves on
References multi-layered media. Bull Seismol Soc Am 43
(1):17–34
Anderson JP, Bodin P, Brune JN, Prince J, Singh SK, Haskell NA (1960) Crustal reflection of plane SH waves. J
Quass R, Onate M (1986) Strong ground motion from Geophys Res 65(12):4147–4150
the Michoacan, Mexico, earthquake. Science Hisada Y, Yamashita T, Murakami M, Kubo T, Shindo J,
233:1043–1049 Aizawa K, Arata T (2012a) Seismic response and
ASCE/SEI Standard 7–10 (2007) Minimum design loads damage of high-rise buildings in Tokyo, Japan, during
for buildings and other structures. American Society of the Great East Japan earthquake. In: Proceedings of the
Civil Engineers, 1801 Alexander Bell Drive, Reston, International symposium on engineering lessons
Virginia 20191, USA learned from the 2011 Great East Japan Earthquake,
Çelebi M (2008) Real-time monitoring of drift for occu- Tokyo, 1–4 March 2012, pp 1110–1119
pancy resumption. In: Proceedings of the 14WCEE Hisada Y, Yamashita T, Murakami M, Kubo T, Arata T,
(CD-ROM), Beijing, 13–17 Oct 2008 Shindo J, Aizawa K (2012b) Seismic response and
598 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
The content discussed in this chapter relates to properties, while in situ tests reflect the actual
several other entries in the Encyclopedia of combined effects of both material (particles size
Earthquake Engineering. While this chapter dis- and shape distribution, mineralogy, etc.) and state
cusses methods of obtaining dynamic soil prop- (confining stresses, void ratio, etc.) properties.
erties from penetration tests, closely related entry A drilling rig, shown in Fig. 1, is used to
include ▶ Dynamic Soil Properties: In Situ Char- advance boreholes. Depending on soil and
acterization Using Penetration Tests. Because groundwater conditions, hollow stem augers
dynamic soil properties are used in seismic anal- (Fig. 2), casing, or bentonite slurry (drilling
ysis and design, the entries covering geotechnical mud) may be used to maintain borehole stability. D
seismic design and analysis may also be of inter- The SPT can also be conducted using light-
est. These entries include ▶ Seismic Analysis of weight portable equipment, such as shown in
Masonry Buildings: Numerical Modeling, Fig. 3. Stable, near-surface soils can be tested in
▶ Seismic Design of Dams, ▶ Seismic Design shallow boreholes advanced by hand augers or
of Earth-Retaining Structures, and ▶ Seismic post-hole diggers.
Design of Tunnels. When a test depth is reached, a split-spoon
sampler (Fig. 4) is connected to the end of a
string of drill rods and lowered into the hole.
Overview of Penetration Tests The split-spoon sampler is cylindrical with a
1.5 in. inner diameter and 2.0 in. outer diameter.
This section describes the SPT, CPT, and DMT Its length is between 1.5 and 2.5 ft. As shown in
equipment and test procedures. Also discussed Fig. 4, the sampler is made of two symmetrical
are the readings taken during each test and the halves, splitting the cylinder lengthwise. This
engineering parameters deduced from them. allows for visual inspection and removal of the
soil after sampling. Most split-spoon samplers
Standard Penetration Test can accommodate a liner, creating a uniform
The SPT was developed in the 1920s and is the inner diameter. When the liner is not used, the
most common in situ penetration test. ASTM inner diameter is slightly larger, reducing the
D1586 details the test’s equipment and procedure. friction between the incoming soil and sampler.
Due to its maturity and popularity, there is vast The top of the rod assembly is connected to a
experience in interpreting SPT results and an drive-weight assembly consisting of a 140 lb
extensive database of correlations between the hammer, anvil, and a lift-drop system. After low-
SPT and soil properties. An advantage of the SPT ering the split-spoon sampler to the bottom of the
by comparison to other penetration tests is that the borehole, the drill rod string is attached to the
SPT is conducted in boreholes where, in addition to drive-weight assembly.
the disturbed soil samples obtained by the SPT In the early years of the SPT, equipment and
split-spoon sampler, “undisturbed” soil samples procedural differences made it impossible to
may also be collected. While no sampling method compare test results. However, the procedures
provides a truly undisturbed specimen, some described in ASTM D1583 and recommenda-
methods and devices such as piston samplers with tions summarized by Seed et al. (1985) have
thin-walled tubes are less disruptive than others. In largely standardized the test, hence its name.
terms of relating SPT results to dynamic soil prop- The test requires the 140 lb hammer to be raised
erties, this advantage means that in situ SPT results 30 in. and dropped, driving the sampler into the
can be compared to laboratory test results from soil. This process is repeated until the sampler has
samples collected from the same borehole. There penetrated 18 in. into the soil at the bottom of the
is, however, a caveat: it is virtually impossible to borehole. The number of blows required to
collect undisturbed samples of cohesionless soil advance the sampler each of three 6 in. incre-
without expensive ground freezing and coring. As ments are recorded. After lifting the split-spoon
such, laboratory tests only bracket the soil sampler from the borehole and removing the soil
600 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests, Fig. 3 Typical light-weight SPT
equipment, including a portable tripod SPT system with safety hammer and gas powered auger
against the blade by lateral earth pressure, an as the “A” reading. When the center of the dia-
electric circuit is completed, and an audible tone phragm expands to 1.1 mm from the blade,
sounds from the dilatometer control unit. When another circuit is completed, and the buzzer
the blade reaches a test depth, its advance is begins to sound again. The corresponding pres-
halted, and pressure is added until the diaphragm sure is recorded as the “B” reading. Lastly, the
is unseated from the blade. This breaks a circuit, diaphragm is slowly deflated until contact with
and the buzzer in the control unit ceases to sound. the blade is restored and the buzzer sounds again.
The gas pressure at which this occurs is recorded This pressure, taken as the “C” reading,
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests 603
corresponds to pore water pressure in free respectively. The corrected pressures are used to
draining materials as water rapidly fills the gap compute three DMT indices: the lateral stress
between the deflating membrane and course index, KD; the dilatometer modulus, ED; and the
grained soil. After recording the “C” reading, all material index, ID, as defined by Eqs. 2, 3, and 4:
pressure is released, and the blade is advanced to
the next test depth. p0 u 0
KD ¼ (2)
Following data collection, the DMT readings s0v
are corrected for the stiffness of the steel dia-
phragm and gauge offset at atmospheric pressure. ED ¼ 34:7ðp1 p0 Þ (3) D
With these corrections, the gage readings A, B,
p1 p0
and C become corrected pressures p0, p1, and p2, ID ¼ (4)
p0 u0
where:
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests, Fig. 6 Disassembled cone
penetrometer
604 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests, Fig. 7 Types of CPT rigs: (a) trailer
mounted, (b) small light-weight CPT rig, (c) large commercial rig, (d) tracked CPT rig
While G0 is commonly used in dynamic soil smaller, are small strain moduli. The SPT is a large
analysis, other small strain soil moduli related to strain test, and, therefore, correlations between
G0 may be of interest. These relationships are N and Vs should be used with caution. Correlations
given by Eqs. 6 through 8 for the Young’s mod- developed in one geologic setting, and particular
ulus, E; bulk modulus, B; and constrained modu- soil type should not be expected to accurately pre-
lus, M, respectively: dict Vs in other situations.
Many such correlations have been developed,
E ¼ 2Gð1 þ uÞ (6) and some are presented in Table 1. The following
review summarizes the conditions under which
2Gð1 þ uÞ these relationships were developed and is
B¼ (7)
3ð1 2uÞ intended to aid the practitioner in choosing an
appropriate relationship if no other means of
2Gð1 uÞ determining Vs exist. As shown in Figs. 10, 11,
M¼ (8) and 12, for sand, clay, and all soils, respectively,
1 2u
the different Vs versus N correlations do not show
where u = Poisson’s ratio. very good agreement. The authors therefore rec-
Due to their dependence on strain, it is important ommend using the correlations that were devel-
to remember that all of the soil moduli determined oped under conditions most similar to those of the
from Vs, which is measured at strains of ~104 % or users’ site.
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests 605
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests, Table 1 Summary of correlations
between SPT data and shear wave velocity
Shear wave velocity, Vs (m/s)
Citation Sand Clay All soil
Kanai et al. 1966 V S ¼ 19N 0:6
Imai and Yoshimura 1970 V S ¼ 76N 0:33
Ohba and Toriuma 1970 V S ¼ 84N 0:31
Shibata 1970 V S ¼ 32N 0:5
Fujiwara 1972 V S ¼ 92:1N 0:337
Ohta et al. 1972 V S ¼ 87N 0:36
Seed et al. (1983) adapted Ohaski and Iwasaki’s correlation on Vs down-hole measurements at
(1973) relationship based on data collected fol- 84 sites across the Los Angeles basin. Jinan
lowing liquefaction events. Sykora and Stokoe (1987) proposed a correlation based on approxi-
(1983) measured Vs using the cross-hole method mately 100 measurements in Shanghai.
and used 229 data pairs to develop their correla- Lee (1990) conducted 15 SPTs and used
tion. Fumal and Tinsley (1985) based their 88 data pairs to construct his relationship for
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests 607
N-Vs in Taipei. Athanasopoulos (1995) devel- EURO-SEISTEST database gathered from north-
oped his correlation based on 221 data pairs ern Greece near Thessaloniki.
from Greek soils. Kayabali (1996) proposed a Wride et al. (2000) proposed a correlation
relationship based on 30 alluvial silty sand based on results from experiments conducted as
deposits from an unspecified location. Pitilakis part of the Canadian Liquefaction Experiment
et al.’s (1999) correlation was developed (CANLEX) at six sand sites in Western Canada.
using 176 pairs of N and Vs data points from the The range of data shown in their equation
608 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
represents the minimum and maximum site- Carolina, six sites in Canada, and eight sites in
specific values. Jafari et al. (2002) used 57 spec- Japan. Hasancebi and Ulusay (2007) correlated
tral analyses of surface waves (SASW) tests, 97 pairs of N and Vs measurements from
9 down-hole tests, and 26 SPT boreholes to create Yenisehir, Turkey. Using Vs measurements from
a correlation for south Tehran. The database used multichannel analysis of surface waves (MASW)
in Andrus et al.’s (2004) correlation came from testing points adjacent to 38 SPT borehole loca-
published work performed in 21 soil layers from tions, 162 data pairs were used by Anbazhagan
eight sites in California, one site in South and Sitharam (2008) to develop an N-Vs
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests 609
relationship for the Bangalore Region. An N-Vs accurately determine the geologic age of soil
relationship for Delhi was proposed by deposits. Therefore, while this factor influences
Hanumantharao and Ramana (2008) based on a the correlation, it is difficult to apply in practice.
subset of SASW-SPT data compiled by the city When determining which relationship to use in
government from public and private entities. practice, the authors recommend matching the
Uma Maheshwari et al. (2010) used MASW to geologic age of the deposit of interest to the geo-
determine Vs and compared the results to 200 SPTs logic age of deposit used to develop the correlation.
in order to create their correlation for the Chennai In summary, while the SPT is a large strain
region. Anbazhagan et al. (2013) correlated results test, correlations have been developed for over D
from 11 SPTs to MASW results in Lucknow City, four decades relating corrected and uncorrected
India. A correlation for Korean soils was devel- blow counts to small strain soil properties, such
oped by Sun et al. (2013) using cross-hole, down- as Vs and G0. Local experience has demonstrated
hole, and up-hole tests from 26 sites. Some refer- the utility of some of the relationships. However,
ences listed in Table 1 were published in languages caution should be used in applying these correla-
other than English (Imai and Yoshimura 1970; tions outside of the conditions for which they
Shibata 1970; Ohta et al. 1972; Imai et al. 1975; were developed.
Sisman 1995; Iyisan 1996; Jafari et al. 1997) and
are, therefore, not explained in detail.
Dynamic Soil Properties from CPT
Other Factors Influencing the SPT
Correlations The relationships between Vs, G0, E0, and M0
As shown in Table 1, the majority of the empir- were presented in the SPT section of this chapter.
ical SPT-Vs correlations are in the form of a The research summarized here compares CPT
power fit: results to various soil moduli, but other properties
V s ¼ XN Y (9) of interest can be determined from the relation-
ships above. In the section describing the CPT
This is logical since both Vs and N60 depend on equipment, procedures, and results, qc, fs, and
confining stress. However, recent research has u were introduced. Some relationships described
shown that s0 v may need to be added explicitly in this section use dimensionless, normalized
to the empirical N-Vs relationship (Schnaid cone parameters. The normalized cone tip resis-
et al. 2004; Brandenberg et al. 2010). Schnaid tance, Qt; normalized friction ratio, FN; and nor-
et al. (2004) incorporated s0 v into the following malized pore pressure, Bq, were proposed by
relationships between G0 and N60: Robertson (1990) and are given by
1= qc sv0
G0 ¼ 1200 N 60 s0v0 p2a 3 Qt ¼ (11)
s0v0
upper bound : cemented sands
1=
G0 ¼ 450 N 60 s0v0 p2a 3 fs
FN ¼ 100% (12)
lower bound : cemented sands and upper bound : qt sv0
uncemented
sands
1=
G0 ¼ 200 N 60 s0v0 p2a 3 Du
Bq ¼ (13)
lower bound : uncemented sands qt sv0
(10)
where:
where pa = atmospheric pressure.
Additionally, geologic age has been shown to sv0 = total stress.
influence the N-Vs relationship (Ohsaki and Du = the difference between measured pore pres-
Iwasaki 1973; Lee 1991). It is often difficult to sure and hydrostatic pore pressure.
610 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
Additionally, some of the CPT-based correla- Wride et al.’s (2000) relationship was based on
tions depend on the soil behavior type index, Ic, tests conducted as part of CANLEX at six sites in
(Robertson and Wride 1998): Western Canada with loose sand deposits. The
denominator in the equation is listed as a range
h i0:5
I c ¼ ð3:47 logQt Þ2 þ ðlogFN þ 1:22Þ2 of values because Wride et al. (2000) presented a
different value for each of the six sites. The
(14) database used in Andrus et al.’s (2004) correla-
tion came from published work from eight sites
Finally, Robertson et al. (1992) introduced an
in California, seven South Carolina, six from
overburden stress-corrected shear wave velocity,
Canada, and four from Japan. Schnaid
Vs1:
et al. (2004), similar to their N-G0 relationship,
0:25 presented a qc-G0 relationship based on data
pa
V s1 ¼ Vs (15) from sites in South America, Europe, and the
s0v0
United States:
Just as with the SPT, the following correlations
between CPT parameters and small strain
1=
G0 ¼ 800 qc s0v0 pa 3
dynamic properties are best utilized under the upper bound : cemented sands
same soil conditions under which they were
1=
developed. Relationships between CPT and G0 ¼ 280 qc s0v0 pa 3
lower bound : cemented sands and upper bound
dynamic soil properties are presented in Table 2.
: uncemented
sands
1=
Baldi et al. (1989) proposed a Vs-G0 relation-
G0 ¼ 110 qc s0v0 pa 3
ship developed with Po River Sand and is meant
lower bound : uncemented sands
for uncemented, silica sand. Rix and Stokoe
(1992) updated the Baldi et al. (1989) relation- (18)
ship between G0 and Vs by including laboratory
cone penetration tests in calibration chambers: Hegazy and Mayne (2006) added 12 new sites to
databases assembled by Mayne and Rix (1993)
!0:75
and Hegazy and Mayne (1995) in order to
G0 q
¼ 1634 0 c0:5 (16) develop their correlation. Andrus et al. (2007)
qc sv0 used 229 data pairs from sites in California,
South Carolina, and Japan.
Hegazy and Mayne (1995) used the general form
Robertson (2009) utilized a database of more
of equations developed by Mayne and Rix (1993)
than 100 SCPTs and over 1,000 data pairs in his
and replaced initial void ratio, e0, and s0 v and with
correlation. Stuedlein’s (2010) CPT-Vs correla-
fs. This correlation was based on Vs from cross-
tion is based on 12 SCPTs, yielding 254 data
hole, down-hole, SASW, and the seismic cone
pairs, in the Puyallup River delta near the port
penetration tests (SCPT) from 30 sites in the
of Tacoma, Washington. SCPTs from 17 sites in
United States, Canada, Norway, and Italy.
five areas along the western coast of South Korea
Mayne and Rix (1995) complied data from
were used to develop Sun et al.’s (2013)
cross-hole, down-hole, SASW, and SCPTs
correlation.
conducted at 31 clay sites in Europe and North
McGann et al. (2014) compared predicted Vs
America and proposed a Vs-qc relationship.
from Andrus et al. (2007), Hegazy and Mayne
Simonini and Cola (2000) reported SCPT
(2006), and Robertson (2009) to Vs measured
results from the Venice Lagoon. After investigat-
using the SCPT following the 2010–2011 Can-
ing several relationships, the best fit relationship
terbury earthquakes. They found that the relation-
included the Bq pore pressure parameter:
ships over-predicted Vs and recommend future
4:59 research to develop a Canterbury-specific CPT-
G0 ¼ 21:5q0:79
c 1 þ Bq (17) Vs relationship. McGann et al. (2014) note that
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests, Table 2 Summary of correlations between CPT data and shear wave velocity
Shear wave velocity, Vs (m/s)
Stress parameters have units of kPa unless noted
Citation Sand Clay All soil
0:27
Baldi V s ¼ 277ðqc Þ0:13 s0v0 (stress units in MPa)
et al. 1989
0:3
Hegazy and V s ¼ 12:02q0:319
c f s0:047 V s ¼ 3:18q0:549
c f 0:025
s fs
V s ¼ ð10:1logqc 11:4Þ1:67 qt 100
Mayne 1995
Mayne and V s ¼ 1:75ðqc Þ0:627
Rix 1995
Wride 0:25
V s1 ¼ ð95:6 110:8Þqc1
et al. 2000
Andrus V S1cs ¼ 62:6q0:231
c1N ,cs
et al. 2004
0 0:25
Hegazy and s
V S ¼ 0:0831qc1N pv0 e1:786Ic
Mayne 2006 a
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
Robertson 0:5
V s ¼ 100:55Ic þ1:68 ðqt sv Þpa
2009
0:25
Stuedlein
V S ¼ 0:033Qt sp0a e2:11Ic
2010 v0
h
0:158
0:169 i h
0:246
0:126 i h 0
0:232
0:181 i
Sun V S ¼ 32:159qt0:053 f s0:131 s0v0 1 þ Bq V S ¼ 14:716q0:168 1 þ Bq V S ¼ 24:215q0:091 f 0:037 1 þ Bq
t f s0:018 s0v0 t s sv0
et al. 2013
611
D
612 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
g
G q correlations through the basic CPT readings.
¼1f (20) An accelerometer placed in the CPT probe is
G0 qult
used to record the arrival times of shear waves
where: generated at the ground surface as shown in
Fig. 14. A wired sledge hammer completes an
f and g = empirical constants depending on soil electrical circuit when it strikes a support pad of
type and stress history. the CPT rig. This triggers the beginning of the
q = applied load causing modulus reduction. data acquisition system’s recording and sends
qult = failure load. the seismic signal into the ground. The arrival of
the shear wave at the accelerometer is marked
Mayne (2007) suggested that f equals 1 and by the onset of a large pulse in the signal trace.
g is approximately 0.3 0.1. Mayne (2010) dem- The probe is then advanced to the next test
onstrated the utility of Eq. 20 by using data from depth, and the process is repeated. Common
laboratory tests and finding good agreement to practice is to conduct seismic tests at CPT rod
results from case histories given proper selection changes (i.e., 1.0 m intervals). However, the
of f and g. Combining Vs data from in situ tests authors find that use of 1.0 m intervals does
with laboratory tests results provides an estimate not define soil stratigraphy very well and there-
of G/G0. However, there is currently no method fore should not be used for detailed site charac-
for obtaining this relationship entirely from terization or liquefaction hazard assessment.
CPT data. However, a 1.0 m interval does provide a rea-
sonable estimate of average dynamic properties
Seismic Cone Penetration Test and is therefore very reliable for earthquake site
The seismic cone penetration test (SCPT) response analysis.
allows for direct measurement of dynamic soil An example of SCPT data is presented in
properties rather than through empirical Fig. 15. The first step in interpreting the data is
614 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
to determine the arrival time of the shear wave. 3.3 m. It could also mean that the soil between
Campenella and Stewart (1992) and Brignoli 2.8 and 3.3 m is different from the soils above
et al. (1996) explain methods of picking shear and below. This illustrates the need for smaller
wave arrival times. The authors use the first test intervals if better stratigraphic resolution is
intense energy excursion, as described by desired.
Brignoli et al. (1996). The determined shear Determining dynamic soil properties directly
wave arrival times are plotted versus acceler- from measured shear wave velocities such as by
ometer depth as shown in Fig. 16. When more the SCPT is obviously more accurate than esti-
than two data points line up, there is reasonable mates of Vs based on correlations with CPT resis-
certainty that the elevation increment can be tances. Additionally, research has shown
represented by constant Vs. However, this does excellent agreement between Vs measured using
not necessarily mean that the soil over this incre- the SCPT with other methods of measuring Vs,
ment is the same; it merely shows that the shear such as down-hole tests, cross-hole tests, SASW,
wave velocity is constant. By contrast, a break in and laboratory measurements (Robertson
the Vs profile generally means a change in soil et al. 1992; Lunne et al. 1997; Schneider
conditions. Figure 16 shows some uncertainly in et al. 1999, 2001; Mayne 2000; Saftner
the layer interface between a depth of 2.8 and et al. 2011; etc.).
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests 615
Dynamic Soil Properties from DMT density, Dr, played a significant role in relating
ED to G0, suggesting that previous relationships
The first correlations between DMT results and were too simple.
dynamic soil properties used ED. Baldi et al. Hryciw (1990) presented a method of
(1986) conducted resonant column tests on Ticino predicting G0 by the DMT based on results from
sand and compared the results to DMT results nine sites in Europe and North America.
from a calibration chamber to develop Eq. 21. Hryciw’s (1990) relationship included the confin-
ing stresses and soil unit weights determined
G0 from DMT indices:
¼ 2:72 0:59 (21)
ED
g
0 0:25 D 1
Bellotti et al. (1986) used cross-hole testing in Po s gW 0
0:5
G0 ¼ 530 v g K 0:25
0 sv pa
River sand to determine the correlation presented pa 2:7 D
in Eq. 22. gW
(23)
G0
¼ 2:2 0:7 (22) where:
ED
Jamiolkowski et al. (1988) conducted experi- gD = unit weight as determined by DMT indices.
ments using Ticino sand and found that relative gw = unit weight of water.
616 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
Monaco et al. (2009) collected data from of determining G/G0 degradation with strain
34 sites in Europe and the United States. The amplitude (g) by the DMT. Marchetti
following equations provided estimates of G0: et al. (2008) suggest a path for future research,
using two points, G (or GDMT) as determined
G0 from MDMT and G0 as determined from Vs to
Clay ¼ 26:177K 1:0066
D
MDMT construct the G/G0 curve. Suggestions for
G0
Silt ¼ 15:686K 0:921
D (24) methods of accomplishing this include using ID
MDMT to determine soil type and comparing GDMT to G-
G0
Sand ¼ 4:5613K 0:7967
D
g curves recommended for the given soil type, as
MDMT well as by comparing settlement predictions
based on MDMT to predictions from G-g curves
where MDMT = constrained dilatometer modulus.
with variable decay rates to determine the best
Marchetti (1980) suggested that MDMT can be
match. Of the three in situ penetration tests
obtained through the DMT modulus, ED by
discussed in this chapter, the DMT appears to
have the most potential for developing G/G0 – g
MDMT ¼ RM ED (25)
relationships from in situ penetration tests.
In turn, RM is a function of the DMT material
index, ID, and horizontal stress index, KD:
Summary
if I D 0:6 RM ¼ 0:14 þ 2:36logK D
if I D 3 RM ¼ 0:5 þ 2logK D Numerous empirical correlations exist between
if 0:3 < I D < 0:6 RM ¼ RM0 þ ð2:5 RM0 ÞlogK D geotechnical in situ penetration test results and
with RM0 ¼ 0:14 þ 0:15ðI D 0:6Þ dynamic soil properties. These relationships have
if K D > 10 RM ¼ 0:32 þ 2:18logK D
been developed in various locations around the
if RM < 0:85 RM ¼ 0:85
world and are best applied to sites geologically
(26) similar to those at which the correlations were
Numerous case histories (i.e., Lacasse 1986; developed. In situ penetration tests that measure
Schmertmann 1986; Schnaid et al. 2000; etc.) Vs directly, such as the SCPT and SDMT, allow
have demonstrated that the DMT gives a reliable determination of small strain dynamic soil proper-
estimate of M. ties more accurately than empirical relationships
based on the basic SPT-N, CPT- qc, or DMT
Seismic Dilatometer Test indices.
The seismic dilatometer test (SDMT) is similar to
the SCPT. The descriptions of equipment, test
procedure, and data reduction presented for the Cross-References
SCPT are essentially the same for the SDMT. The
advantage of performing SDMT is that one can ▶ Seismic Analysis of Masonry Buildings:
obtain dynamic soil properties, in this case, Vs, Numerical Modeling
directly, without the uncertainty of correlations. ▶ Seismic Design of Dams
Monaco et al. (2009) emphasized that the avail- ▶ Seismic Design of Earth-Retaining Structures
ability of SDMT makes direct measurement of Vs ▶ Seismic Design of Tunnels
the best method for determining dynamic prop-
erties from the DMT.
Whereas the SPT and CPT are large strain References
tests (~10 %), the DMT is a medium strain test
Anbazhagan P, Sitharam TG (2008) Mapping of average
(0.05–0.1 %) as shown in Fig. 13. Therefore, shear wave velocity for Bangalore Region: a case
there is also an opportunity to develop a method history. J Environ Eng Geophys 13(2):69–84
Dynamic Soil Properties: In Situ Characterization Using Penetration Tests 617
Anbazhagan P, Kumar A, Sitharam TG (2013) Seismic Fumal TE, Tinsley JC (1985) Mapping shear wave
site classification and correlation between standard velocities of near surface geological materials.
penetration test N value and shear wave velocity for Predicting areal limits of earthquake induced
Lucknow City in Indo-Gangetic Basin. Pure Appl landsliding. In: Evaluation of earthquake hazards
Geophys 170:299–318 in the Los Angeles region, USGS aper 1360,
Andrus RD, Piratheepan P, Ellis B, Zhang J, Juang H pp 127–150
(2004) Comparing liquefaction evaluation methods Hanumantharao C, Ramana GV (2008) Dynamic soil
using penetration-Vs relationships. Soil Dyn Earthq properties for microzonation of Delhi, India. J Earth
Eng 24:713–721 Syst Sci 117(S2):719–730
Andrus RD, Mohanan NP, Piratheepan P, Ellis BS, Holzer Hasancebi N, Ulusay R (2007) Empirical correlation
TL (2007) Predicting shear wave velocity from cone between shear wave velocity and penetration resis- D
penetration resistance. In: Proceedings of the 4th inter- tance for ground shaking assessment. Bull Eng Geol
national conference on earthquake geotechnical engi- Environ 66:203–213
neering, The ssaloniki, Paper no 1454 Hegazy YA, Mayne PW (1995) Statistical correlations
Andrus RD, Hayati H, Mohanan NP (2009) Correcting between Vs and cone penetration data for different
liquefaction resistance for aged soils using measured soil types. In: Proceedings, CPT ’95, vol 2. Swedish
to estimated velocity ratio. J Geotech Geoenviron Eng Geotechnical Society, Linköping, pp 173–178
135(6):735–744 Hegazy YA, Mayne PW (2006) A global statistical corre-
Athanasopoulos GA (1995) Empirical correlation Vs-N for lation between shear wave velocity and cone penetra-
soils of Greece: a comparative study of reliability. In: tion data. In: Proceedings of the GeoShanghai, site
Proceedings of the 7th international conference on soil and geomaterial characterization (GSP 149), ASCE,
dynamics & earthquake engineering, Crete, pp 19–36 Reston, pp 243–248
Baldi G et al (1986) Flat dilatometer tests in calibration Hryciw RD (1990) Small strain shear modulus of soil by
chambers. In: Proceedings of the use of in situ tests in dilatometer. J Geotech Eng 116(11):1700–1716
geotechnical engineering, Blacksburg, pp 47–73 Imai T (1977) P and S wave velocities of the ground in
Baldi G, Bellotti R, Ghionna V, Jamiolkowski M, Lo Japan. In: Proceedings of the 9th international confer-
Presit DCF (1989) Modulus of sands from CPTs and ence on soil mechanics & foundation engineering,
DMTs. In: Proceedings of the XII ICSMFE, vol 1, Rio vol 2, Toyko, pp 127–132
de Janeiro, pp 165–170 Imai T, Tonouchi K (1982) Correlation of N value with
Bellotti R, Ghionna V, Jamiolkowski M, Lancellotta R, S wave velocity and shear modulus. In: Proceedings of
Manfredini G (1986) Deformation characteristics of the 2nd European symposium on penetration testing,
cohesionless soils from in situ tests. In: In situ ’86 Amsterdam, pp 57–72
specialty conference in situ. ASCE, Virginia Tech, Imai T, Yoshimura Y (1970) Elastic wave velocity and
Blacksburg soil properties in soft soil. Tsuchito-Kiso 18(1):17–22.
Brandenberg SJ, Bellana N, Shantz T (2010) Shear wave As cited in Hasancebi N, Ulusay R (2007) Empirical
velocity as function of standard penetration test resis- correlation between shear wave velocity and penetra-
tance and vertical effective stress at California bridge tion resistance for ground shaking assessment. Bull
sites. Soil Dyn Earthq Eng 30:1026–1035 Eng Geol Environ 66:203–213
Brignoli EG, Gotti M, Stokoe KH (1996) Measurement Imai T, Fumoto H, Yokota K (1975) The relation of
of shear waves in laboratory specimens by means of mechanical properties of soil to P and S wave veloci-
piezo-electric transducers. Geotech Test J 19(4): ties in Japan. In: Proceedings of the 4th Japan earth-
384–397 quake engineering symposium, Toyko, pp 89–96. As
Campenella RG, Stewart WP (1992) Seismic cone analy- cited in Uma Maheshwari R, Boominathan A,
sis using digital signal processing for dynamic site Dodagoudar GR (2010) Use of surface waves in sta-
characterization. Can Geotech J 29(3):477–486 tistical correlations of shear wave velocity and pene-
Campenella RG et al (1985) Recent developments in in tration resistance of Chennai soils. Geotech Geol Eng
situ testing of soils. In: Proceedings of the XI interna- 28:119–137
tional conference on soil mechanics and foundation Iyisan R (1996) Correlation between shear wave velocity
engineering, San Francisco, pp 849–854 and in situ penetration test results. Tech J Chamb Civ
Chein LK, Lin MC, Oh YN (2000) Shear wave velocity Eng Turk 7:1187–1199. As cited in Anbazhagan P,
and SPT-N values of in situ reclaimed soil in west Kumar A, Sitharam TG (2013) Seismic site classifica-
Taiwan. Geotech Eng 31:63–77 tion and correlation between standard penetration
Fahey M, Carter JP (1993) A finite element study of the test N value and shear wave velocity for Lucknow
pressuremeter test in sand using a nonlinear elastic City in Indo-Gangetic Basin. Pure Appl Geophys
plastic model. Can Geotech J 30(2):348–362 170:299–318
Fujiwara T (1972) Estimation of ground movement in Jafari MK, Asghari A, Rahmani I (1997) Empirical corre-
actual destructive earthquakes. In: Proceedings of the lation between shear wave velocity and SPT-N values
4th European symposium on earthquake engineering, for south of Tehran soils. In: Proceedings of the 4th
London, pp 125–132 international conference on civil engineering, Tehran.
618 Dynamic Soil Properties: In Situ Characterization Using Penetration Tests
Schneider JA, Hoyos L, Mayne PW, Macari EJ, Rix GJ University, Geophysical Engineering Department. As
(1999) Field and laboratory measurement of dynamic cited in Hasancebi N, Ulusay R (2007) Empirical cor-
shear modulus of piedmont residual soils. In: Behav- relation between shear wave velocity and penetration
ioral characteristics of residual soils. Geotechnical resistance for ground shaking assessment. Bull Eng
special publication 92, ASCE, Reston, pp 12–25 Geol Environ 66:203–213
Schneider JA, Mayne PW, Rix GJ (2001) Geotechnical Skempton AW (1986) Standard penetration test proce-
site characterization in the greater Memphis area using dures and effects in sands of overburden pressure,
cone penetration tests. Eng Geol 62:169–184 relative density, particle size, aging and over consoli-
Seed HB, Idriss IM (1981) Evaluation of liquefaction dation. Geotechnique 36(3):425–447
potential sand deposits based on observation of perfor- Studelein AW (2010) Shear wave velocity correlation for
mance in previous earthquakes. In situ testing to Puyallup River Alluvium. J Geotech Geoenviron Eng D
evaluate liquefaction susceptibility, pp 81–544 ASCE 136(9):1298–1304
Seed HB, Idriss IM, Arango I (1983) Evaluation of lique- Sun CG, Cho CS, Minkyung S, Shin JS (2013) Correla-
faction potential using field performance data. tions between shear wave velocity and in situ penetra-
J Geotech Eng 109:458–482 tion test results for Korean Soil Deposits. Pure Appl
Seed HB, Tokumatsu K, Harder LF, Chung R (1985) Geophys 170:271–281
Influence of SPT procedures in soil liquefaction Sykora DW, Stokoe KH (1983) Correlations of in situ
resistance evaluations. J Geotech Eng 111(12): measurements in sands with shear wave velocity. Geo-
1425–1445 technical Engineering Report GR83-33, University of
Shibata T (1970) Analysis of liquefaction of saturated Texas at Austin, Austin. As cited in Sykora DW
sand during cyclic loading, disaster prevention. Res (1987) Examination of existing shear wave velocity
Inst Bull 13:563–570. As cited in Uma and shear modulus correlations in soils. Miscellaneous
Maheshwari R, Boominathan A, Dodagoudar GR Paper GL-87-26, US Army Corps of Engineers, WES,
(2010) Use of surface waves in statistical correlations Vicksburg
of shear wave velocity and penetration resistance of Uma Maheshwari R, Boominathan A, Dodagoudar GR
Chennai soils. Geotech Geol Eng 28:119–137 (2010) Use of surface waves in statistical correlations
Simonini P, Cola S (2000) Use of piezocone to predict of shear wave velocity and penetration resistance of
maximum stiffness of Venetian soils. J Geotech Chennai soils. Geotech Geol Eng 28:119–137
Geoenviron Eng 126(4):378–381 Wride CE, Roertson PK, Biggar KW, Campanella RG,
Sisman H (1995) An investigation on relationships Hofmann BA, Hughes JMO, Kupper A, Woeller DJ
between shear wave velocity and SPT and (2000) Interpretation of in situ test results from the
pressuremeter test results. MS Thesis, Ankara Canlex sites. Can Geotech J 37(3):505–529
E
Early Earthquake Warning (EEW) elevators, and enabling people to quickly protect
System: Overview themselves in various environments. Operational
EEW systems are currently running in Mexico,
Tsuneo Katayama1 and Shigeki Horiuchi2 Taiwan, and Japan. For well over one decade the
1
Professor emeritus of the University of Tokyo, Japan Meteorological Agency (JMA) had been
Bunkyo-ku, Tokyo, Japan developing early warning system to generate
2
COE, Home Seismometer Corporation, immediate alerts after earthquakes, and on Octo-
Shirakawa, Fukushima, Japan ber 1, 2007, JMA officially launched EEW ser-
vice to the residents in Japan through a number of
media outlets such as TV and radio. JMA’s EEW
Synonyms provides advance announcement of the estimated
seismic intensities and expected arrival time of
Earthquake alarm system (ElarmS); Earthquake principal motion. The system presently uses
alert system; Earthquake early alarm system; about 1,000 seismometers. EEW essentially
Nakamura’s method; Real-time earthquake makes automatic calculation of the (1) focus
warning; Urgent earthquake detection and alarm and (2) magnitude of an earthquake, and esti-
system (UrEDAS) mates, before the arrival of strong motions,
(3) the distribution of seismic intensity in the
regions where strong tremors are expected.
Introduction Quickness and robustness are the two key factors.
Estimation starts when a few seconds’ data from a
Earthquake early warning (EEW or Real-time small number of recording stations are obtained,
earthquake warning) is a system to provide warn- and the results are updated with time as the num-
ing of potentially significant shaking in neighbor- ber of data increases. The JMA’s EEW is divided
ing regions of a strong earthquake within a few into two grades: “forecast” and “warning.” In the
seconds upon detection of a severe earthquake. It case of the March 11, 2011, off the Pacific coast
differs from earthquake prediction which is of Tohoku Earthquake (Mw 9.0), a total of 15 fore-
defined as the specification of the time, location, casts were issued, and the fourth forecast at 8.6 s
and magnitude of a future earthquake. EEW is after the first trigger became a warning to the
aimed at mitigating earthquake-related damage general public. When EEW was launched in
by allowing countermeasures such as promptly 2007 in Japan, JMA notified the public the limi-
slowing down running trains, controlling tations associated with the present EEW,
Early Earthquake Warning (EEW) System: Overview, Fig. 1 Flow of JMA’s EEW (JMA 2007)
when the first station triggers, the event is determined from estimated values of hypocenter
temporarily located beneath that station; after a parameters, magnitude, and empirical attenuation
second station trigger the location moves to a equations. There, however, are difficulties in the
point between the two stations, based on the quick determination of the magnitude for a large
timing of the arrivals; with three or more trig- earthquake. Generally speaking, M8-class earth-
gered arrivals, the event location and origin time quakes have faults with size of about 100 km. It
is estimated using trilateration and a grid search takes about 30 s until all fault areas are ruptured.
algorithm. Numerical simulation indicates that the move-
A group of Japanese seismologists (Horiuchi ment of faulting continues, for nearly the same
et al. 2005) successfully developed an earthquake duration of time as its fault rupture time, even
early alarm system that determines earthquake after the stoppage of the rupture propagation.
parameters of large events within a few seconds This means that it takes almost 1 min to observe
from the arrival of P-waves at the closest stations a complete waveform for an M8-class earthquake
by using the real-time data from the Hi-net and making quick determination of the magnitude for
Kanto-Tokai seismic networks that employs large events difficult.
about 750 high dynamic range seismic stations To overcome this difficulty, a parameter has
covering the whole of the Japan Islands. These been proposed that quickly and accurately esti-
networks are maintained by the National Institute mates shaking intensity (Yamamoto et al. 2008).
for Earth Science and Disaster Prevention The shaking intensity is defined from the ampli-
(NIED). The alarm system employs a novel tude of the filtered acceleration. Usually, dis-
method to locate hypocenter using not only the placement defines magnitude from which
P-wave arrival times but also the fact that intensity is estimated. The new principle is
P-waves have not yet arrived at other stations in based on the assumption that the intensity could
the network. be better estimated from the observed accelera-
The final output of EEW is the shaking inten- tion data than from the displacement data. The
sity at locations where customers live and/or parameter indicating the earthquake size is
work. In the standard procedures, intensity is named Shaking Intensity Magnitude MI which is
Early Earthquake Warning (EEW) System: Overview 625
intended to represent the amplitude of short forecast issued at 8.6 s after the first trigger esti-
period seismic waves. Station correction factors mated magnitude to be 7.2 and the expected
are incorporated, because short period waves are intensity increased to Lower-5 for central Miyagi
greatly amplified in the sedimentary layer just prefecture. The fourth forecast, therefore,
beneath stations. It has been shown that MI became a warning to the general public in the
grows much faster to its final value compared Tohoku district, and the warning was automati-
with the moment magnitude by conventional cally broadcast nationwide through TV, radios
methods. Test result for the 2003 Off Miyagi and cellular phone mails. The EEW arrived the
earthquake (M 7.0) indicated that MI grows to coastal areas where people live at least 15 s ear-
the 95 % of the final size by only 3 s, whereas it lier than the strong phase of ground motion
E
required 18 s for MJMA to grow to the final value. (intensity Lower-5 or greater on the JMA scale).
Similar results were obtained for several different Figure 2 shows how the hypocenter and mag-
earthquakes. In the case of 2003 Off Tokachi nitude were updated (Hoshiba et al. 2011). Mag-
earthquake of M8.0, MI reached about 95 % of nitudes were estimated from maximum
its final value at 10 s, and MJMA at about 31 s. displacement amplitude at four stations.
Although the JMA’s standard procedure uses
An Example: March 11 Tsunami Earthquake records from five stations, in the case of March
of 2011 11 Tsunami earthquake, the instrument at one
The present JMA intensity scale has 10 from station was found not properly working and the
intensity 1 that people can barely detect to Inten- record from that station was discarded. Color
sity 7 the upper limit. Intensity 5 and 6 are lines represent the different stations, and the
divided into two degrees, namely Lower-5, black line the median value which was used for
Upper-5, Lower-6, and Upper-6. The JMA’s EEW. The horizontal axis shows the elapsed time
EEW is divided into two grades: “forecast” and from the first trigger, and offset axis at top the
“warning.” The “forecast” is issued to advanced elapsed time from the public warning, i.e., the
users when events are estimated to be M 3.5 or fourth forecast in this particular case. The solid
larger, or when the expected seismic intensity is line indicates the magnitude determined by the
3 or stronger. In the “forecast,” the regions are P-wave, the broken line that by the S-wave, and
specified where seismic intensity 4 or greater is the dotted line shows the period during which the
expected. When estimated intensity becomes magnitude is kept unchanged around the S-wave
Lower-5 or greater at any observation station of arrival. White stars show epicenters determined
the seismic intensity networks covering all over by EEW, one for the first to the seventh forecast
the country, “warning” is issued to the general (5.4–11.0 s after the first trigger) and another for
public in regions where intensity 4 or greater is the next eight (15.9–116.8 s). Focal depth was
expected. The “warning” (i.e., forecasts and estimated to be 10 km for all 15 announcements.
warnings) is updated as available data increases. Red star indicates the epicenter location from the
In the case of the March 11, 2011, off the unified JMA catalog (focal depth is 24 km). The
Pacific coast of Tohoku Earthquake (Mw 9.0), resolution of the JMA’s EEW is 0.1 for latitude
which will be hereafter abbreviated as the and longitude, and 10 km for focal depth. The
March 11 Tsunami earthquake caused strong fifteenth forecast at 116.8 s after the first trigger
ground motion in the north-eastern part of the estimated the magnitude as 8.1 which is almost
Honshu Island. JMA’s EEW was first triggered the upper limit of JMA displacement magnitude
when station OURI detected the initial P-wave at because of amplitude-magnitude saturation.
14:46:40.2, Japan Standard Time, March Figure 3 shows regions where forecasts and
11, 2011 (Hoshiba et al. 2011). JMA issued a public warnings were issued (Hoshiba
total of 15 EEW forecasts, but the first forecast et al. 2011), and the distribution of estimated
issued at 5.4 s after the first trigger considerably seismic intensities at three time points, i.e., at
underestimated the magnitude as 4.3. The fourth 14:46:48.8 when the “warning” was issued, at
626 Early Earthquake Warning (EEW) System: Overview
Early Earthquake 40
Warning (EEW) System:
Overview,
Fig. 2 Sequence of
epicenter and magnitude 39
M in EEW
determinations of the
March 11 Tsunami
earthquake by JMA’s EEW OURI
(Hoshiba et al. 2011) ICHINM 38
MIYKNA
KAWAUC
37
139 149 141 142 143
0 20 40 60 80 100
10
M 7
6 M (P wave)
M (Fixed)
5
M (S wave)
4
0 50 100
3/11 14:46:40.2(Trigger) sec
14:47:04.0 when intensity Lower-5 or greater confused, and did not always determine the loca-
first observed, and at 14:47:46.0 when Tokyo tion and magnitude correctly.
region was first specified as the “forecast” area. In the 19 days from the main shock to March
Pink areas indicate the region where the “warn- 29, 2011, JMA appropriately issued 15 warnings
ing” was issued, and the yellow areas are those for the 22 events for which seismic intensity
specified as the target of the “forecast.” The dis- Lower-5 or greater was actually observed. On
tribution of the observed intensity is shown in the other hand, during the same period of time,
Fig. 4. For the Tokyo region, JMA’s EEW esti- 45 forecasts were issued, but actual observed
mated intensity 4, and warning was not issued. intensities did not exceed 2 at any observation
This was an underestimation, and actual observa- station in 11 of the 45 events.
tions reached Upper-5, which corresponds to the
level of the warning issuance. After the Problems and Prospects
mainshock, the system did not properly operated When EEW was launched in 2007 in Japan, JMA
for several hours because of high background notified the public the limitations associated with
noise from the coda waves and active after- the present EEW, including (1) Timing, (2) False
shocks, and because of power failure and wiring alarms, (3) Magnitude estimation, and (4) Seismic
disconnections. For several days, when earth- intensity estimation (JMA 2007). These still are
quakes occurred sometimes simultaneously over difficult problems to solve in the present techni-
the wide source region, the system became cal state.
Early Earthquake Warning (EEW) System: Overview 627
40°
38°
36°
138° 140° 142° 144°138° 140° 142° 144° 138° 140° 142° 144°
1 2 3 4 5 6 7
Seis, Intensity(JMA scale)
Early Earthquake Warning (EEW) System: Overview, Fig. 3 Regions of EEW “warning” & “forecast,” and
distribution of estimated seismic intensities (Hoshiba et al. 2011)
The window of time from the announcement number of stations increases, i.e., if the spacing
of an EEW until the arrival of the main tremors is between stations becomes much smaller than that
very short, i.e., a matter of seconds (or between of today, EEW need not rely upon the single
several seconds and a few tens of seconds). In station estimation from the very early stage of
areas that are close to the focus of the earthquake, warning.
the warning may not be transmitted before strong A conventional receiving/alarm unit of an
tremors hit. EEW system is equipped with a CPU and mem-
Because of limitations imposed by station ory and is on-line via the Internet. By adding an
spacing and the determination and transmission inexpensive MEMS seismometer and A/D con-
of earthquake source parameters in real-time, verter, this unit is transformed into a real-time
JMA system cannot issue an EEW to areas within seismic observatory, which may be named a
about 30 km of the earthquake’s hypocenter. It is home seismometer (Horiuchi et al. 2008). If the
estimated that about 10 times the number of sta- home seismometer is incorporated in the stan-
tions would be needed to issue an effective EEW dard receiving unit of EEW, the number of
near focal areas. observing stations will dramatically increase.
JMA’s EEW adopts the single station estimate In order to properly cancel the influence of back-
for the first warning (see Fig. 1), and false EEW’s ground noise caused by human activity, spe-
may be transmitted as a result of noise from cially developed software is put in the system.
accidents, lightning, or device failure. If the When an event occurs, its predominant
628 Early Earthquake Warning (EEW) System: Overview
Earthquake Damage Assessment from VHR Data: are visible in the satellite image; white arrow in the satel-
Case Studies, Fig. 1 Collapsed single building: ground lite image shows the look angle of the photo (From Saito
photo (a) and satellite image (b); the ruins of the building et al. 2004)
Earthquake Damage Assessment from VHR Data: circled in white in the satellite image (b); the low-story
Case Studies, Fig. 2 A building with collapsed ground collapse is not visible in the IKONOS-2 acquisition (From
and first floors: ground photo (a) and the same building Saito et al. 2004)
was compared with a map derived by a ground The higher values of differences (2 and
survey. Such a comparison highlights that the 3 levels) can be ascribed mainly to the resampling
map obtained from remote sensing data under- of the ground survey map, from its original for-
estimates, on average, the level of damage by mat (i.e., aggregated at street block scale) to the
one level, and the standard deviation of the dif- square block type of satellite map. Indeed, in this
ference is 1.58. Despite this discrepancy the step a reclassification of damage level was com-
overall pattern of the two maps is very similar puted, according to the levels defined in the
(Fig. 3). IKONOS-2 derived assessment.
632 Earthquake Damage Assessment from VHR Data: Case Studies
Earthquake Damage
Assessment from VHR
Data: Case Studies,
Fig. 4 Damage mapping
results from pre- and post-
seismic data visual
comparison. The numbers
of identified damaged
buildings are 1,597 (grades
1 or 2: blue points), 3,815
(grade 3: green), 1,700
(grade 4: yellow), and 4,951
(grade 5: red) (From E
Yamazaki et al. 2005)
values of the four parameters Ev, Ed, Ta, and is categorized by the percentage of collapsed
Te correspond to damaged buildings. structures: 80–100 %, 50–80 %, 20–50 %,
The method was applied to IKONOS-2 and and <20 %. The observed the accuracy rate is
QuickBird post-event data. The vegetated areas high, demonstrating the potential of this damage
were masked out by thresholding the Normalized detection algorithm.
Difference Vegetation Index (NDVI). The results Generally, the use of a pixel by pixel approach
were compared to the damage map obtained by with VHR optical data has some limitations
visual inspection, and it was found that the distri- because of the presence of some restrictive factors
bution of detected damage from IKONOS-2 was such as different viewing angles of the pre- and
consistent with these results, while less damage post-images (and the related geometric distor-
information was detected from the QuickBird tions), different sun illumination, and shadows
image. due to the different seasons of the satellite acqui-
In Huyck et al. 2005, a similar method based sitions. These are the reason why the visual inter-
on textural analysis is performed. On both pre- pretation is considered the most reliable approach,
and post-seismic images from QuickBird, the but such problems can be reduced by adopting
edge detection is performed, and from the specific methods. One of the proposed in literature
resulting images, the dissimilarity textual param- is the object-based technique. This technique ana-
eter is calculated. At this stage, the change detec- lyzes each building (the object in the satellite
tion of dissimilarity is done. The damage in the scene) instead of considering individual pixels.
post-event data gives a high dissimilarity of In Gusella et al. 2005, an object-oriented technique
edges, while intact buildings yield a low dissim- was adopted for Bam case study. The object selec-
ilarity of edges. The image textural difference is tion is done on the pre-earthquake image, where
then classified, on the base of standard deviation intact buildings are identified by using, first,
values, and finally, the damage map is extracted a segmentation procedure with a bottom-up region
by considering the mean value of the class within merging strategy and then a supervised classifica-
200 200 pixel block. The textural analysis was tion method by using eCognition software.
also applied in a multisensor scenario, where the The aim is to separate building objects from
pre-seismic image is the QuickBird acquisition nonbuilding objects (e.g., vegetation, cars, and
and the post-event one is the IKONOS-2 data, shadows). Then the classified buildings in the
begin the latter resampled at the same resolution pre-event image are superimposed on the post-
of QuickBird image, i.e., from 1 to 0.6 m. event one, and to distinguish between collapsed
The two damage maps were compared with and non-collapsed structures, the post-earthquake
damage observations obtained from an indepen- buildings are classified, with a supervised
dent source compiled by US-AID where damage approach, for a second time using training set of
634 Earthquake Damage Assessment from VHR Data: Case Studies
Earthquake Damage
Assessment from VHR
Data: Case Studies,
Fig. 5 Building damage
map for the city of Bam,
based on object-oriented
analysis of QuickBird
imagery captured before
and after the event.
Collapsed structures are
shown in red and
non-collapsed in blue
(From Gusella et al. 2005)
Earthquake Damage Assessment from VHR Data: (c) damaged pixels, in red, superimposed to the difference
Case Studies, Fig. 6 Building damage map: (a) detail between pre- and post-seismic images (From Chini
of panchromatic image taken on September 30, 2003; (b) et al. 2009)
detail of panchromatic image taken on January 3, 2004;
damaged and non-damaged objects from the build- image (Chini et al. 2009). This results in
ing class. Note that these automated methods a feature vector which is used as input dataset of
require a preliminary co-registration of the two an unsupervised classifier. The output of the clas-
satellite images, to match the geometry of one sifier is therefore the mask of the buildings (i.e.,
image on the other (Richards 2006). the objects) of Bam city. At this stage, the pixel by
The resulting damage map is shown in Fig. 5. pixel normalized difference (ND), i.e., a simple
The accuracy assessment, performed on change detection, is computed only on the mask-
a random set of 136 buildings, reaches an overall selected objects, and by setting a threshold on ND
value of 70.5 %. value, the damaged buildings are detected. The
A similar way is proposed in the work of result of this methodology is presented in Fig. 6.
Hutchinson and Chen 2005. Here the object Such approach allowed improving roughly of
selection is performed through a hierarchical seg- 15 % the accuracy with respect to the change
mentation procedure. detection computed over the entire scene.
The objects selection can also be obtained by
following different methods. An alternative solu-
tion is the exploitations of the morphological The January 12, 2010, Port-au-Prince,
operators (Soille 2003). It is accomplished by Haiti, Earthquake
applying the closing and the opening filters,
with different size, to extract additional features On January 12, 2010, a devastating earthquake of
from panchromatic band of VHR pre-seismic magnitude 7.2 hit the city of Port-au-Prince and
Earthquake Damage Assessment from VHR Data: Case Studies 635
Earthquake Damage Assessment from VHR Data: (b) average normalized difference map of Cosmo-
Case Studies, Fig. 7 (a) Cosmo-SkyMed color compos- SkyMed pair (From Dekker 2011)
ite of Port-au-Prince (R = January 13, G = B = 26 April);
the surrounding villages in Haiti. In this very passages, allowing better performances in terms of
destructive event, more than 300,000 people time latency between satellite tasking and image
were killed and over one million were displaced. availability that reflects a prompt response for dam-
Over 310,000 houses were destroyed or dam- age map delivery. It is worth noticing that usually
aged, and medical facilities, educational build- the visual interpretation of VHR SAR images is
ings, and basic infrastructure such as roads, a very hard task to be done; therefore, most of the
bridges, ports, and airports were severely dam- approaches rely on automated methods, mainly
aged (Duda and Jones 2011). based on change detection techniques.
As for the case of Bam and Gujarat earth- For the Haiti case, pre-seismic and post-seismic
quakes, the VHR optical images were exploited images from both TerraSAR-X and COSMO-
adopting some of the methodologies (or similar) SkyMed satellites were available. The TerraSAR-
already presented (e.g., Corbane et al. 2011; X images are dated October 13, 2009, and January
Tiede et al. 2011; Voigt et al. 2011). For such 20, 2010, while the COSMO-SkyMed pair was
reason, this section will focus on the methods and acquired on April 26, 2009, and January 15, 2010.
results carried out by using VHR images captured The two datasets are on descending orbit, in strip-
from SAR sensors. map mode at 3 m resolution (Dekker 2011).
During the emergency phase the area of Port- One of the proposed method relies on a change
au-Prince was captured by the German SAR sen- detection approach which is based on the normal-
sor of TerraSAR-X mission (Pitz and Miller ized difference (ND) between the pre- and post-
2010) and by the Italian constellation COSMO- earthquake images. The damage estimation was
SkyMed (Covello et al. 2010). Both satellites can computed at grid level, by considering the aver-
acquire VHR images, at X band, reaching reso- age value of the ND within a grid cell or area.
lution of about 1 m on ground, allowing urban A cell size of 200 200 m was adopted. An
area analysis, not only at city block scale but also alternative damage index proposed for the Haiti
at single building scale. case study was estimating the de-correlation
SAR data overcome the typical limitations of coefficient between pre-seismic and post-seismic
optical images that cannot provide information in images (Yonezawa and Takeuchi 2001;
case of cloud coverage, and only daylight satellite Stramondo et al. 2006). It can be obtained by
passes are suitable for image collection. Actually, correlating the intensity pixels within a certain
SAR instruments, being active sensors, can operate grid cells, which was again of 200 200 m for
in all weather conditions and day or night orbital this analysis. In Fig. 7 an example of the damage
636 Earthquake Damage Assessment from VHR Data: Case Studies
index derived from the ND average evaluation compact, and it result in less open double-
obtained from COSMO-SkyMed data and the bounce scattering mechanism.
original SAR image are shown.
The resulting maps were compared with
a damage assessment carried out in the frame- Summary
work of the Post Disaster Needs Assessment
and Recovery Framework (PDNA) at request of This entry describes the main achievements
the Government of Haiti (http://oneresponse. about the very high-resolution (VHR) remote
info/Disasters/Haiti/Early%20Recovery/). The sensing data in the fieldwork of earthquake dam-
PDNA map was derived from VHR satellite and age estimation, from their first application. Both
airborne optical images, on a visual interpretation optical and SAR imagery exploitation is
base. Such map is at building level and on the reported, with the description of the main steps
classification scheme of the EMS98 scale. In that were adopted to obtain damage detection
order to compare the results, the PDNA map and assessment maps. Some relevant case
was also averaged with the same grid cells of studies that can be considered the basis of the
the SAR-derived maps. The average PDNA dam- VHR remote sensing methodologies are here
age map shows the same damage distribution presented.
pattern as in the average damage maps calculated
from TerraSAR-X and Cosmo-SkyMed, with
some discrepancies in the eastern area. In addi- Cross-References
tion to this visual comparison, the regression
analysis between PDNA map and SAR maps ▶ Building Damage from Multi-resolution,
was performed for both ND and grid-cell Object-Based, Classification Techniques
de-correlation maps. The regression analysis ▶ Building Disaster Resiliency Through Disaster
highlighted that, for both satellites, the ND out- Risk Management Master Planning
performs the de-correlation estimation, but the ▶ Building Monitoring Using a Ground-Based
two indices are not well correlated with the Radar
PDNA damage map. Probably not all the damage ▶ Damage to Ancient Buildings from
reflects in a strong enough change of radar back- Earthquakes
scattering. It can be explained by the different ▶ Damage to Buildings: Modeling
scattering mechanisms and radar orientation ▶ Damage to Infrastructure: Modeling
(Dekker 2011). ▶ Earthquake Disaster Recovery: Leadership and
Besides the grid-based approach, a detailed Governance
SAR damage assessment was also carried out. ▶ Earthquake Location
This experiment was done on TerraSAR-X data, ▶ Earthquake Magnitude Estimation
adopting the logarithmic ratio between the pre- ▶ Estimation of Potential Seismic Damage in
and post-event images as a damage index. The Urban Areas
logarithmic ratio is also filtered to remove the ▶ Features Extraction from Satellite Data
well-known speckle effect, and two thresholds ▶ Hyperspectral Data in Urban Areas
for the increase and decrease of radar backscat- ▶ Remote Sensing in Seismology: An Overview
ter are set (6.0 dB and -6.0 dB, respectively). ▶ Resilience to Earthquake Disasters
Comparing the resulting map with the PDNA ▶ SAR Images, Interpretation of
damage assessment, a similar trend in the center ▶ Seismic Collapse Assessment
of Port-au-Prince was observed, but TerraSAR- ▶ Seismic Loss Assessment
X damage map seems weakly correlated at ▶ System and Damage Identification of Civil
building level, especially in the shanty area of Structures
the city. Reason for this latter phenomenon is ▶ Urban Change Monitoring: Multi-temporal
that shanty areas are less regular and more SAR Images
Earthquake Disaster Recovery: Leadership and Governance 637
events such as earthquakes, volcanic eruptions, leadership as the key elements of successful
floods, or hurricanes. In the last few decades, an recovery. The sections provide a discussion on
ever-increasing vulnerability and complexity of recovery and earthquake recovery, its principles,
risks brought to prominence integrated and com- and its relation to other elements of disaster man-
prehensive, all-hazards approaches, supported by agement (see section “Recovery: Definition and
frameworks and structures based upon a shared Principles”); an identification of specific charac-
system of governance and policy making. These teristics of governance and leadership in post-
approaches presume democratic environment earthquake recovery (see section “Recovery:
with decentralized and deliberative planning Governance”); a discussion on governance trans-
and decision-making processes that integrate formation and leadership in recovery (see section
risk management into broader sustainable devel- “Governance Transformation and Leadership in
opment strategies. The system requires extensive Recovery”); considerations of strengths and
coordination and cooperation among all levels of weaknesses of different approaches to recovery
government, as well as with the many other stake- governance, as well as strategies for successful
holders, such as private sector, NGOs, and com- recovery governance and leadership (see section
munity groups involved in disaster management. “Strengths and Weaknesses of Different
The supporting legislation, policies, guidelines, Approaches to Earthquake Recovery Governance
and plans are commonly based upon sustainable and Leadership”); and a discussion on recovery
development principles; they also introduce the governance and social capital and its implications
concept of shared responsibility and empower- for governance and leadership for resilience and
ment of local communities, reflecting the recog- future trends in disaster recovery (see section
nition that it is essential to develop society’s “Recovery Governance, Leadership, Social Cap-
overall “adaptive capacity” or resilience ital (or Social Capacities), and Resilience”).
(Tierney 2012).
The all-hazards, comprehensive approach to
risk and disaster management means that the The Role of Governance and Leadership
practice has evolved from its historical sphere in Disaster Recovery
of natural hazard events and now incorporates
an array of risks, including major environmental Recovery: Definition and Principles
and technological, as well as terrorist, events. Within a broader comprehensive, integrated
Furthermore, the approach comprises risk pre- framework for disaster management, recovery is
vention or risk reduction, preparedness or readi- generally defined as a coordinated process that
ness, and response to and recovery from events, involves supporting affected communities in
internationally recognized as PPRR of risk pre- reconstruction of the built environment and the
vention, preparedness, response, and recovery, or restoration of emotional, social, cultural, eco-
4Rs of risk reduction, readiness, response, and nomic, and natural environment well-being fol-
recovery. These elements are not sequential or lowing an emergency. It is recognized that
mutually exclusive, but each element contains recovery is a complex social and developmental
components of the other three and may, at least process that involves much more than simply the
in part, be simultaneously operational. Each ele- replacement of what has been destroyed and the
ment should be fully integrated through broader rehabilitation of those affected. Effective recov-
planning programs and management processes ery presents an opportunity to reduce future risks,
for disaster management (MCDEM 2005; “build back better,” and foster a more resilient
FEMA 2011). community (MCDEM 2005; FEMA 2011; Aus-
The subsequent sections of this article focus tralian Government 2011; UK Government
on one of the elements of the integrated, compre- 2013). Current scholarship recognizes that the
hensive disaster management – disaster recovery actual recovery processes vary with each disaster
framework, with the accent on governance and and that the so-called stages of recovery overlap
Earthquake Disaster Recovery: Leadership and Governance 639
with one another, with different groups and sub- ecosystems, organizations and institutions, eco-
regions within disaster-affected areas recovering nomic activity, and culture and, hence, cannot be
at different times and in different ways, rendering solely or closely equated with post-disaster
recovery nonlinear and often iterative (Tierney reconstruction of the physical environment. It is,
and Oliver-Smith 2012). Furthermore, recovery rather, a holistic, intricate process requiring com-
is affected not only by local circumstances but by plex, integrated, and adaptive governance struc-
macro-societal and even global factors and, tures to succeed.
importantly, by their interconnection (ibid.).
Recovery is not only a set of activities that Recovery: Governance
occur after a disaster but a continuum that Disaster governance is often defined as a system
E
involves both pre-disaster and post-disaster plan- that enables collaborative activities that bring
ning, as well as risk-reduction and resilience- together multiple organizations (government
building activities that link recovery to other and nongovernment) to solve problems that
aspects of a comprehensive risk management extend beyond the purview of any single organi-
framework (Smith 2011). zation (Tierney 2012). More specifically, Tierney
It is now widely recognized that recovery is (2012) defines disaster governance as interrelated
best achieved when the affected communities are sets of norms (such as laws and regulations,
actively involved and able to exercise a high frameworks, standards), organizational and insti-
degree of self-determination or even take the tutional actors, and practices – across the 4Rs of
leadership in developing recovery priorities and risk reduction, readiness, response, and
activities (MCDEM 2005; FEMA 2011; Austra- recovery – designed to reduce the impacts asso-
lian Government 2011). ciated with disasters.
Earthquake recovery epitomizes all the above Commonly, governance structures include
characteristics of recovery. In particular, due to public sector organizations (federal or national,
the sudden and devastating nature of earthquakes, state or regional, and local governments),
especially when they impact urban environments, nongovernmental organizations (community
earthquake recovery presents communities and development corporations, homeowners’ associ-
governments with an enormous and complex ations, special service districts, regional planning
long-term process that often challenges organizations, professional associations, and col-
preexisting concepts and frameworks. leges and universities), nonprofit relief organiza-
Historically, scholars postulated that the tions (nonprofits, community-based
amount of physical damage from a disaster best organizations, and foundations), private sector
correlates with recovery speed, arguing that areas organizations (businesses and corporations,
which suffered a higher degree of damage to the financial and lending institutions, insurance, and
built environment and higher injuries and casual- media), international relief organizations and
ties will recover more slowly, requiring more nations, and emergent groups and individuals,
time for reconstruction, with housing in shorter who together provide three types of resources:
supply (Aldrich 2011). This is particularly rele- financing, policies, and technical assistance
vant for disasters caused by earthquakes. How- (Smith and Birkland 2012). The degree of coor-
ever, there are theories that the scale of damage dination and integration, underlying values and
and loss does not necessarily correlate with processes, timing, capacity, and capability of
recovery rates, and factors such as population those actors are some of the key factors that
density, pace of rebuilding, and socioeconomic determine the effectiveness of recovery processes
status play equally important role (Aldrich 2011). (Smith and Birkland 2012).
Tierney and Oliver-Smith (2012) define recovery Many countries have developed national
as a socially configured process that is inextrica- frameworks, governance models, and administra-
bly linked to the recovery of not only the built tive guidelines for recovery management and
environment and infrastructure but also governance (MCDEM 2005; FEMA 2011;
640 Earthquake Disaster Recovery: Leadership and Governance
Social
Territorial
Environment
Local
Authority Economic
Recovery Environment Natural
Manager Environment Built
Environment
Social
Environment Economic
Regional
CDEM Group
Environment Natural
Recovery Recovery
Manager Environment Built
Coordinator
Environment
Only if
appointed by
Minister under
s29 CDEM Act MCDEM
Emergency Social Economic Natural Built
Management Environment Environment Environment Environment
Advisers
National Recovery
Lead Agency: MCDEM
Reports
Recovery Office
National
Cabinet
Minister of
Civil Defence
Earthquake Disaster Recovery: Leadership and Governance, Fig. 1 Multilevel recovery management structure
from the New Zealand CDEM group recovery guidelines (MCDEM 2005)
Australian Government 2011; UK Government countries to formalize the approach in its national
2013). In general, all these models reflect the recovery framework, as illustrated in Fig. 1.
holistic, comprehensive, and integrated risk man- The recovery framework diagram in Fig. 1
agement approaches based on the principles for depicts the national arrangements in New
sustainable development. They are characterized Zealand, showing the national government as
by decentralized, “bottom-up” structures, with supporting recovery activities lead by the local
local governments having primary responsibility government in partnership with an array of stake-
for supplying disaster-related resources and holders, including private and nongovernment
regional, subnational, national, and international organizations. Depending on the scale of event,
agencies providing support as requested. the lead agency changes from local to national.
A similar approach is advocated by the United The holistic nature of recovery is achieved
Nations as well (Fraser-Molekati 2012). For through integration across different “environ-
example, New Zealand was one of the first ments”: social, economic, natural, and built.
Earthquake Disaster Recovery: Leadership and Governance 641
Establishment of the special recovery coordinate between the central and local govern-
authority in Victoria – Victorian Bushfire Recon- ment (Suzuki and Kaneko 2013).
struction and Recovery Authority (VBRRA) – Following the devastation caused by Hurri-
followed Black Saturday, within the preexisting cane Sandy in October 2012, on January
legislative framework (Smart 2012). In the case 29, 2013, President Obama signed into law the
of the Queensland floods and the Canterbury Sandy Recovery Improvement Act 2013. The
earthquakes, however, it was perceived that the legislation introduces significant changes to the
magnitude and duration of events exhausted the way the Federal Emergency Management
capacity of preexisting institutions to cope, and Agency (FEMA) delivers disaster assistance and
recovery authorities with greater powers were empowers FEMA to develop processes and pro-
deemed necessary. In Queensland the Queens- cedures for newly established authorities and
land Reconstruction Authority (QldRA) was their implementation. Prior to the introduction
established (Smart 2012). The QldRA built on of the act, in December 2012, President Obama
the experience by VBRRA, taking onboard rec- issued an executive order that established the
ommendations to align with the highest level of Hurricane Sandy Rebuilding Task Force, atypi-
government, acquire statutory powers to expe- cally chaired by the Secretary of Housing and
dite decision-making and have clearly defined Urban Development, thus allocating the lead
terms of reference (Smith and Birkland 2012). role to the Department of Housing and Urban
Both Queensland and New Zealand Development, the first for this agency (White
circumvented existing regulations, New Zealand House 2012).
effectively emulating the Queensland experi- It is important not to underestimate the influ-
ence into its arrangements for the Canterbury ence Hurricane Katrina had on disaster manage-
Earthquake Recovery Authority (CERA). Both ment in the USA and internationally. The
agencies were given similar powers; both could perceived poor recovery following Hurricane
undertake works, enter and acquire land, decide Katrina was investigated by the House Select
the priorities for reconstruction, and override Committee who identified failures at all levels
local government plans and policies; and both of government to support otherwise exemplary
had legislated sunset clauses (Smart 2012). The efforts by nongovernment organizations, groups,
key differences were in their status – CERA was and individuals. The most important failure for
a government department reporting directly to the government was lack of leadership, expressed
the newly appointed Recovery Minister, while as lack of imagination, lack of initiative, and lack
the QldRA was created as a statutory body with a of adaptive capacity (Waughn and Streib 2006).
governing board. Effectively, the New Zealand The examples from the USA, Australia, New
arrangement allowed the government to exert a Zealand, and Japan all reflect an acute awareness
more direct influence over the recovery process of the political and fiscal risks and the recognition
(Smart 2012). on the part of respective governments that there is
Similar to the governance changes in Austral- an expectation on governments to demonstrate
asia, the Great East Japan earthquake of March leadership and deliver on recovery needs. The
11, 2011, triggered a number of major initiatives changes, or transformation of governance that
to transform the current system of disaster miti- occurred following major disasters, testify to
gation governance in Japan. New government governments taking a leadership role. The
organizations were established: the Reconstruc- newly introduced arrangements were designed
tion Agency to deal with recovery and recon- to build and sustain institutional capacity for
struction (in February 2012) and the Nuclear recovery by engaging across a broad sector of
Regulatory Authority to deal with nuclear safety government and nongovernment organizations,
(September 2012). The role of the Reconstruc- to allow for formation and growth of partnerships
tion Agency is to integrate across different sec- through enabling processes, and to facilitate fuel-
tors of the national government, as well as to ing of the recovery process with money and
Earthquake Disaster Recovery: Leadership and Governance 643
information through coordination between differ- historical and, especially, more recent events,
ent levels and stakeholders. such as the Victoria bushfires in 2009, Queens-
A closer look at the transformation of gover- land floods in 2011, and Hurricane Katrina in
nance and arrangements for recovery from Can- 2005. The Committee recommended that a new,
terbury, New Zealand, earthquakes and the role central government public service department be
of the Canterbury Earthquake Recovery Author- created to ensure leadership and coordination and
ity (CERA) provides further insights into both to balance the central government’s contribution
requirements for and potential issues with recov- to funding while carefully managing its fiscal
ery governance and leadership. situation. It was also recommended that normal
parliamentary procedures and the normal time-
E
Strengths and Weaknesses of Different line for establishing a new national department be
Approaches to Earthquake Recovery waived given the emergency situation, allowing
Governance and Leadership for the Canterbury Earthquake Recovery Author-
Scholarship has identified that supportive gover- ity (CERA) to be provisionally established the
nance is necessary to ensure coping capacities in following day (Johnson and Mamula-Seadon
societies. Governance determines the ways in 2014). At the same time, the CERR Act 2010
which a variety of actors at all levels engage was repealed, including the dissolution of the
and coordinate their efforts in recovery. For Canterbury Earthquake Recovery Commission
engagement and coordination to lead to effective (CERC), and a new legislation was introduced,
recovery, governance has to ensure broad partic- effectively upholding the authorities granted
ipation, transparency, accountability, efficiency, under the CERR Act, extending the expiration
and responsiveness (Chen et al. 2013). date of those authorities where appropriate and
Two days after the September 4, 2010, earth- conferring on the Minister for CER and/or CERA
quake in Canterbury, Prime Minister John Key a series of powers that would enable the timely
announced the appointment to the new Cabinet and coordinated recovery of greater Christ-
position of the Minister for Canterbury Earth- church. The Parliament agreed to an abbreviated
quake Recovery (CER Minister) and a newly legislative process and passed the Canterbury
appointed ad hoc Cabinet Committee on Canter- Earthquake Recovery (CER) Act 2011 with a
bury Earthquake Recovery. Following this, on near unanimous vote on April 14, 2011
September 14, 2010, the Canterbury Earthquake (Johnson and Mamula-Seadon 2014).
Response and Recovery Bill was passed by the One of the first tasks for the newly established
Parliament, with urgency and unanimously, and CERA was to develop the Canterbury Recovery
became an Act (CERR Act 2010). The Bill was Strategy (CERA 2012a) that sets up the vision for
exempt from the normal parliamentary examina- the region’s recovery and identifies key goals for
tion procedures that precede the introduction of the recovery. The strategy was developed
new legislation. Its purpose was to facilitate bet- through a consultative process with other govern-
ter coordination between impacted local authori- ment agencies, local government, and local com-
ties and the central government. The Act also munities. The key goals of the recovery process
established the Canterbury Earthquake Recovery are community centered and expressed through
Commission, comprised of the mayors of the the five components of recovery, pertaining to the
three impacted local authorities and four govern- social, cultural, natural, built, and economic envi-
ment appointees (NZ Parliament 2010). Follow- ronments, as shown in Fig. 2.
ing the devastating earthquake on February To effect delivery of the complex recovery
22, 2011, the ad hoc Cabinet Committee on Can- tasks, the New Zealand government has proposed
terbury Earthquake Recovery began considering the following governance structure for CERA
new national governance arrangements to man- (Fig. 3).
age the recovery effort. The rationale for the As can be seen from Fig. 3, CERA, as a central
proposed change was given as experience from government agency, and CERA’s Chief Executive
644 Earthquake Disaster Recovery: Leadership and Governance
ECONOMIC
Revitalise greater
Christchurch as the
heart of a prosperous
region for business,
work, education, and
increased investment in
new activities.
SOCIAL BUILT
Strengthen Develop resilient, cost
community resilience, effective, accessible and
safety and wellbeing, integrated infrastructure,
and enhance quality buildings, housing and
of life for residents transport networks.
and visitors.
COMMUNITY
CULTURAL NATURAL
Renew greater Restore the natural
Christchurch’s unique environment to support
identity and its vitality biodiversity and
expressed through economic prosperity
sport, recreation, art, and to reconnect people
history, heritage and to the rivers, wetlands
traditions. and Port Hills.
Earthquake Disaster Recovery: Leadership and Governance, Fig. 2 Key goals for the six components of recovery
following Canterbury earthquakes in 2011 (CERA 2012a)
both report directly to the Minister for Canterbury To a degree, the proposed recovery gover-
Earthquake Recovery. A four-person Canterbury nance structure in Canterbury reflects recent soci-
Earthquake Recovery Review Panel, appointed by etal and political changes, such as the emergence
the Minister, was tasked with reviewing and of public-private partnerships, outsourcing, and
recommending Orders of Council, a mechanism contracting, involving not only governmental
allowing the government to make decisions requir- institutions but also private sector and civil soci-
ing legal force by bypassing the standard legisla- ety entities, as discussed in research literature
tive procedure. There is also a cross party (Tierney 2012).
parliamentary forum (CPPF), comprised of mem- It also reflects the intent to deliver on findings
bers of the Parliament from the Greater Christ- of the World Bank that identified good recovery
church area, established to provide the Minister governance demonstrated during Queensland
with information and advice about the operation recovery effort as:
of the CER Act, and a community forum (CF) of at
least 20 members to provide information and • Introduction of specific agencies to deal with
advice to the Minister that he must consider in recovery
carrying out his duties (CERA 2012b). • Commitment to community engagement
Earthquake Disaster Recovery: Leadership and Governance 645
Cabinet
Construction industry
Waimakariri and Selwyn
District Councils
Key
Contracting and procuring eg Infrastructure rebuild Key input and delivery
Recovery Plans
land remediation Key input and delivery
Community development Advisory and delivery
Monitor implementation of
and engagement
Recovery Strategy and Recovery
Plans by central and local
government and private sector
Earthquake Disaster Recovery: Leadership and Governance, Fig. 3 The proposed governance structure and roles
for CERA (2012b)
• Government agencies working with the local decision-making authority and operational respon-
government in recovery planning sibility for recovery activities. The findings dem-
• Incorporation of risk mitigation strategies into onstrate how post-disaster “time compression” can
recovery planning spotlight the strengths and weaknesses of recovery
• Provision of advice to individuals management processes and governance frame-
• Learning and improving preexisting arrange- works or cause distortions. The authors conclude
ments (World Bank 2011) that the effectiveness of vertical coordination of
multiple levels of government, capacity building
When analyzing recovery governance arrange- at the local and regional level, and public consul-
ments pre-earthquake and post-earthquake, John- tation and deliberation of key decisions are some
son and Mamula-Seadon (2014) identify that the critical areas of concern (Johnson and Mamula-
governance structures for managing recovery Seadon 2014). The research revealed that over the
following the Canterbury earthquakes have course of the first 2 years following the February
transformed significantly at both the national and 22, 2011, earthquake, the transformation of gover-
local levels, in comparison to preexisting arrange- nance structures for managing recovery has mainly
ments. At the national level, recovery governance focused on strengthening national-level centraliza-
has transformed from a national service delivery tion. The driving force behind this transformation
coordination and support role for locally was the need to make timely and effective deci-
affected areas to one of increasing national-level sions, add capacity and capabilities for the
646 Earthquake Disaster Recovery: Leadership and Governance
recovery, and ensure accountability for the consid- public good. The mode of operation presupposes
erable public expenditure, among others (Johnson consensus-oriented and deliberative processes,
and Mamula-Seadon 2014). The research indicates with shared decision-making directed toward
that there are signs that the centralization may have shaping public policy (Kapucu 2012).
helped to strengthen coordination among national Thus, one of the themes when considering
agencies and expedite policy and decision-making; governance structures, and particularly so in the
but the effectiveness of coordination among multi- recovery context, is the necessity to understand
ple levels of government, capacity building at the and address the state-civil society relationships.
local and regional levels, and public engagement
and deliberation of key decisions are some areas Recovery Governance, Leadership, Social
where the transformations may not have been as Capital (or Social Capacities), and Resilience
effective. At the local level, recovery governance Different forms of governance including authority,
transformed from a locally led and collaborative power, influence, degree of decentralization, and
approach to a supportive role with, in some cases, accountability affect disaster recovery. It is impor-
little to no involvement in recovery decision- tant to understand, for instance, how policy initia-
making or implementation. The authors conclude tives with local communities can be influenced by
that long-term outcomes of the CERA model and the way local and national governments work
the Canterbury recovery governance transforma- together and how larger trends in national policies
tions require better understanding. They also pro- affect the capacities of communities to engage with
vide some cautionary notes for future disaster their governmental agencies. Defining the forms of
recovery policy making and the development of governance that take the energy and commitment
effective and sustainable integrated recovery gov- individuals and communities show during an emer-
ernance models for large-scale disaster recovery gency and sustain it through appropriately aligned
(Johnson and Mamula-Seadon 2014). government efforts, especially in response and
Reflecting on governance and governance recovery situations, is one of the key challenges
transformation that often occurs in recovery, it is of successful recovery (MNRPG 2013). Further-
important to bear in mind that the last decade or so more, if successful recovery requires informed,
has seen a dramatic change in the way public coordinated, and sustained leadership throughout
service is delivered. Kapucu (2012) observes that all levels of the government and community and
the most important reform has occurred not only in phases of the recovery process (FEMA 2011), it is
the range of agencies involved but in the tools and paramount to design governance structures that
forms of service delivery. Government agencies foster meaningful engagement between the state
are not anymore the sole providers of services that and civil society (MNRPG 2013).
traditionally were considered “public.” In the cur- Understanding social capital and the power of
rent environment, often nonprofit and for-profit social groups to organize, generate, and deploy
agencies as well as ordinary citizens have become individuals within a community to take decisive
the stakeholders taking on the roles and responsi- action when facing hardships is one of the critical
bilities of service provision at all stages of the points in recovery governance. Leadership at all
process (Kapucu 2012). The term governance levels, including spontaneous, emergent leader-
depicts the complex web of interactions and inter- ship typical of disasters, requires governments to
dependencies among a multiplicity of actors that be able to anticipate and adapt. In nearly every
deliver specific services to citizens (Kapucu national experience, policy leaders have encoun-
2012). The term is also synonymous with collab- tered situations in which community members
orative governance and collaborative public man- have spontaneously organized and begun to
agement, formed to deliver on complex problems respond and recover to a disaster well before gov-
and describe network relationships across differ- ernment agencies and programs have mobilized
ent sectors and different levels of the government, (MNRPG 2013). It is important to understand how
all brought together for the purpose of delivering various forms of governance align with these
Earthquake Disaster Recovery: Leadership and Governance 647
mobilizations that occur largely outside the con- engage community initiatives in the early stages
trol and direction of particular agencies. To the of recovery were seen as lacking (Mamula-
extent that community leaders are the potential Seadon et al. 2012).
allies and partners for government authorities, Following on the discussion of social capaci-
governments need to learn to identify and work ties (capital), it can be argued that the strength of
with them – and even help create effective leaders local relationships is expressed as the degree of
(MNRPG 2013). prevalent, strong, and sustained-over-time
State-civil society interactions are complex and involvement of local groups and individuals in
take many forms – some nations have strong civil the recovery decision-making process and is a
society sectors, while in other nations civil society measure of horizontal integration within the gov-
E
institutions are nascent, weak, or unable to func- ernance framework (Smith and Birkland 2012).
tion openly (Tierney and Oliver-Smith 2012). Participants may include local government offi-
Recent scholarship clearly links the effectiveness cials, nongovernment groups, citizen groups,
of the recovery process to levels of trust and social business owners, local financial institutions, the
capital, including social networks or ties, the media, and individual community members. The
degree of political prowess and organization, and empowering process usually involves delibera-
the degree to which social networks are embedded tive planning and policy-making strategies at
into recovery processes (Aldrich 2011). There are the community level. Recovery governance with
many examples of successful state-civil society poor horizontal integration lacks stakeholder
partnership in earthquake recovery, in Haiti, Tai- involvement (Smith and Birkland 2012). Vertical
wan, and China, to mention a few (Tierney and integration is defined by strong linkages to state
Oliver-Smith 2012; Chen et al. 2013). and national organizations, including national
When, following the earthquake on February foundations, corporations, national lending insti-
22, 2011, the citizens of Christchurch were asked tutions, and insurance companies (Smith and
to identify what makes them resilient, they sin- Birkland 2012). Integrating social capital, ensur-
gled out the ability to take action and engage with ing meaningful deliberation and engagement
others in determining their future as one of the with affected communities, achieving effective
key ingredients for resilience and successful vertical coordination with multiple levels of gov-
recovery. They felt that the emergent, spontane- ernment and other stakeholders, and capacity
ous community leaders were the catalyst for the building at the local and regional level are
process. When asked to summarize what among key factors for effective recovery gover-
transformed community members into commu- nance and leadership.
nity leaders and made them so readily accepted International policy makers identified that, to
by other members of the community, the partic- a large extent, the challenge of building disaster
ipants responded that these leaders were found to resilience is a matter of democratic governance. It
have local knowledge and connections and have involves partnering with communities, building
the ability to listen to the community and had a mutual support within communities and across
good grasp of community dynamics and needs; jurisdictional boundaries, and sustaining involve-
they acted on their own initiative and had little ment. In general, the work of building resilience,
need for formal structure in order to be effective; both pre- and post-disaster, demands cooperation
those leaders were described as able to function among citizens and between subnational and
with little resource, to be flexible, and able to national levels of the government and integration
adapt quickly. They were skilled, open minded, of both the public and private sectors and thus
selfless, and focused on the well-being of the closely resembles the collaborative nature of
community. Whereas those leaders were recog- community recovery. Furthermore, research and
nized (and subsequently awarded) by the com- policy making have been advanced by views of
munity, the governance mechanisms that had resilience as a construct formed through the inter-
flexibility and adaptability needed to actively dependencies that evolve from established
648 Earthquake Disaster Recovery: Leadership and Governance
societal patterns and not as a replica of institu- bring together multiple organizations, including
tional, group, or program arrangements (MNRPG the public sector, nongovernmental organiza-
2013). Governance for recovery requires leader- tions, nonprofit relief organizations, the private
ship capable of transforming historical hierarchi- sector, and the community. The degree of coor-
cal government models into adaptable, dination and integration, underlying values and
integrated, networked, multi-stakeholder envi- processes, timing, capacity, and capability of
ronments with deliberative and collaborative those actors are some of the key factors that
models of state-civil society – in other words determine effectiveness of recovery processes.
effective governance for recovery is synonymous The immediate post-disaster environment cre-
with effective governance for resilience. ates enormous pressures on governments and
societies that require complex rebuilding and
societal activities to happen in a compressed
Summary period of time. Governments are acutely aware
of political and fiscal risks and the expectation to
The shift in disaster management practice toward demonstrate leadership and deliver on recovery
more integrated and comprehensive, all-hazards needs. They rally to provide and unify communi-
approaches based on principles of sustainable ties around a shared vision for recovery. Faced
development has seen an introduction of frame- with insufficient capacity to solve problems,
works and processes with a shared system of make timely and effective decisions, add capacity
governance and policy making. Common and and capabilities for the recovery, ensure account-
overlapping responsibilities are apportioned ability for the considerable public expenditure,
among layers of government and other deliver on excess demands for information and
nongovernment entities in democratic gover- stakeholder deliberation, greater needs for orga-
nance systems with decentralized and delibera- nizational integration and coordination, and
tive planning and decision-making processes. immediate demands for large amounts of money
National frameworks, governance models, and with existing distribution systems challenged to
administrative guidelines on recovery manage- meet the needs, governments often struggle. As a
ment reflect the holistic, comprehensive, inte- result, this may lead to significant transformation
grated risk management approach based on the in governance. The new arrangements are
principles for sustainable development and designed to build and sustain institutional capac-
decentralized, “bottom-up” structures, with local ity for recovery by engaging across a broad sector
governments having primary responsibility for of government and nongovernment organiza-
supplying disaster-related resources and regional, tions, to allow for formation and growth of part-
subnational, and national agencies providing sup- nerships through enabling processes, and to
port as requested. Recovery is considered a con- facilitate fueling of the recovery process with
tinuum that involves both pre-disaster and post- money and information through coordination
disaster activities. Recovery preplanning is as between different levels and stakeholders. Sup-
crucial as risk-reduction and resilience-building portive governance embracing of stakeholder
activities that link recovery to other aspects of the engagement is necessary to ensure coping capac-
comprehensive risk management framework. To ities in societies. For engagement and coordina-
a varying degree recovery governance structures tion to lead to effective recovery, governance has
also reflect recent societal and political changes, to ensure broad participation, transparency,
such as the emergence of public-private partner- accountability, efficiency, and responsiveness
ships, outsourcing, and contracting, involving not Understanding social capital and the power of
only governmental institutions but also private social groups to organize, generate, and deploy
sector and civil society entities. individuals within a community to take decisive
Disaster governance is often defined as a sys- action when facing hardships is one of the criti-
tem that enables collaborative activities that cal points in the future of recovery governance
Earthquake Disaster Recovery: Leadership and Governance 649
and leadership. Integrating social capital, ensur- Federal Emergency Management Agency. http://
ing meaningful deliberation and engagement www.fema.gov/pdf/recoveryframework/ndrf.pdf
Fraser-Molekati GJ (2012) Democratic governance at
with affected communities, achieving effective times of crisis: rebuilding our communities and build-
vertical coordination with multiple levels of ing on our citizens. Int Rev Adm Sci 78(2):191–208,
government and other stakeholders, and capac- Sage
ity building at the local and regional level GeoNet (2013) Aftershocks – our news – GeoNet.
GeoNet: EQC and GNS Science. http://info.geonet.
are among the key factors for effective recovery org.nz/display/home/Aftershocks
governance and leadership. An adaptable, UK Government, Cabinet Office (2013) Emergency
integrated, networked, multi-stakeholder envi- response and recovery guideline. https://www.gov.uk/
ronment with deliberative and collaborative government/uploads/system/uploads/attachment_data/
file/253488/Emergency_Response_and_Recovery_5th_ E
models of state-civil society governance for edition_October_2013.pdf
recovery is synonymous with effective gover- Johnson L, Mamula-Seadon L (2014) Transforming gov-
nance for resilience. ernance: organizing for and managing disaster recov-
ery following the September 4, 2010 and February
22, 2011 Canterbury earthquakes. Earthq Spectra
30(1):577–605, special edition Canterbury
Cross-References Earthquakes
Johnson LA, Olshansky TB (2013) The road to recovery:
▶ “Build Back Better” Principles for governing post-disaster reconstruction. Land Lines,
Reconstruction Lincoln Institute of Land Policy, Cambridge, MA,
pp 14–21
▶ Community Recovery Following Earthquake Kapucu N (2012) Disaster and emergency management
Disasters systems in urban areas. Cities 29:S41–S49, Elsevier
▶ Learning from Earthquake Disasters Mamula-Seadon L, Paton D, Selway K (2012) Exploring
▶ Legislation Changes Following Earthquake resilience: learning from Christchurch communities, in
Community resilience: the Christchurch experience.
Disasters MCDEM Wellington, Tephra. http://www.civildefence.
govt.nz/memwebsite.nsf/Files/Tephra-vol23/$file/Tephra
2012web.pdf
References MCDEM, Ministry of Civil Defence & Emergency Man-
agement (2005) Focus on recovery: a holistic frame-
Aldrich DP (2011) The power of people: social capital’s work for recovery in New Zealand. Director’s
role in recovery from the 1995 Kobe earthquake. Nat Guidelines, Ministry of Civil Defence & Emergency
Hazards 56:595–611, Springer Science Management, Wellington. http://www.civildefence.
Australian Government Attorney-General’s Department govt.nz/memwebsite.nsf/Files/FocusOnRecovery/$file/
(2011) Community recovery handbook 2. The Austra- FocusOnRecovery.pdf
lian Emergency Management Institute, part of the MNPRG The Multi-national Resilience Policy Group
Commonwealth Attorney-General’s Department, (2013) The New Zealand discussion, notes from the
Canberra meeting. Wellington/Christchurch
CERA, Canterbury Earthquake Recovery Authority New Zealand Parliament (2010) Canterbury Earthquake
(2012a) Recovery strategy for greater Christchurch: Response and Recovery Act 2010. Available at http://
Mahere Haumanutanga o Waitaha. CERA. http://cera. www.legislation.govt.nz/act/public/2010/0114/latest/
govt.nz/recovery-strategy/overview/read-the-recovery- DLM3233004.html
strategy Olshansky RB, Hopkins LD, Johnson LA (2012) Disaster
CERA, Canterbury Earthquake Recovery Authority and recovery: processes compressed in time. Nat Haz-
(2012b) Proposed governance arrangements ards Rev 13(3):173–178
(cabinet paper). http://cera.govt.nz/sites/cera.govt.nz/ Smart J (2012) The role of post-disaster institutions in
files/common/cabinet-paper-1-proposed-governance- recovery and resilience: a comparative study of three
arrangements-annex-2-march-2011.pdf recent disasters – Victorian Bushfires (2009), Queens-
Chen J, Chen THY, Vertinsky I, Yumagulova L, Chansoo land Floods (2010–2011), and Canterbury Earth-
P (2013) Public-private partnership for the develop- quakes (2010–2012). Working paper 12/02. Institute
ment of disaster resilient communities. J Conting for Governance and Policy Studies, Victoria Univer-
Disaster Manag 21(3):130–143, Wiley sity Wellington
FEMA, Federal Emergency Management Agency Smith G (2011) Planning for post-disaster recovery: a
(2011) National disaster recovery framework: review of the United States disaster assistance frame-
strengthening disaster recovery for the nation. work. Public Entity Risk Institute, Fairfax
650 Earthquake Early Warning System in Taiwan
Smith G, Birkland T (2012) Building a theory of recovery: Some of the destructive earthquakes caused
institutional dimensions. Int J Mass Emerg Disasters large casualties, such as the 1906, ML = 7.1
30(2):147–170
Suzuki I, Kaneko Y (2013) Japan’s disaster governance Meishan earthquake (1,258 death); the 1935,
how was the 3.11 crisis managed? Springer, New ML = 7.1 Hsinchu-Taichung earthquake (3,276
York/Heidelberg/Dordrecht/London death); and the 1999, ML = 7.3 Chi-Chi earth-
Tierney K (2012) Disaster governance: social, political, quake (2,455 death). Earthquake is one of the
and economic dimensions. Annu Rev Environ Resour
37:341–363 most serious disasters in Taiwan. Therefore, it is
Tierney K, Oliver-Smith A (2012) Social dimensions of important to seek scientific resolutions to alle-
disaster recovery. Int J Mass Emerg Disasters viate earthquake hazards.
30(2):123–146 The earthquake early warning (EEW) system
Waughn WL, Streib G (2006) Collaboration and leader-
ship for effective emergency management. Public (Wenzel and Zschau 2013; “▶ Early Earthquake
Adm Rev 66(Spec Issue):131–140 Warning (EEW) System: Overview”) is consid-
White House (2012) Executive order establishing Hurri- ered as a useful tool for real-time seismic hazard
cane Sandy Rebuilding Task Force. http://www. mitigation. An EEW system provides a few sec-
whitehouse.gov/the-press-office/2012/12/07/executive-
order-establishing-hurricane-sandy-rebuilding-task-force onds to tens of seconds of advanced warning time
World Bank (2011) Queensland recovery and reconstruc- for the damage shakings. It allows some mitiga-
tion in the aftermath of the 2010/2011 flood events and tion reactions to be taken in a short time. For
cyclone Yasi, report. The International Bank for instance, high-speed trains can slow down to
Reconstruction and Development, Washington, DC,
USA and Queensland Reconstruction Authority, resist strong ground shakings, gas pipelines can
Australia be turned off automatically, elevators can be
moved to nearby floor and open the doors, sur-
geries in hospitals can be suspended in time,
etc. The early warning time may be too short for
Earthquake Early Warning System people to evacuate from buildings. However,
in Taiwan through careful planning, training, and drills,
an educated general public can still take
Yih-Min Wu1, Nai-Chi Hsiao2, Tai-Lin Chin3, necessary response in time to avoid loss of lives
Da-Yi Chen2, Ya-Ting Chan1 and from large earthquakes. Two types of EEW sys-
Kai-Shyr Wang4 tems are operated in the world. One is a front-
1
Department of Geosciences, National Taiwan detection EEW system in which seismometers
University, Taipei, Taiwan installed in the earthquake source area give
2
Central Weather Bureau, Taipei, Taiwan early warnings for more distant urban areas.
3
Department of Computer Science and The other is an on-site EEW system which deter-
Information Engineering, National Taiwan mines the earthquake parameters from the initial
University of Science and Technology, Taipei, portion of the P waves and predicts the more
Taiwan severe ground shakings of the following S wave
4
Ministry of Science and Technology, Taipei, trains (Kanamori 2005; Wu and Kanamori
Taiwan 2005a).
Motivation for Taiwan to develop the EEW is
the painful experience of the 15 November 1986
Introduction Hualien, Taiwan, earthquake of ML 6.8 (or MW
7.8). Although the epicenter of that earthquake was
Taiwan is located on the plate boundary where located near Hualien, the most severe damage
Philippine Sea Plate collides with Eurasian occurred in the Taipei metropolitan area. The dis-
Plate with a very high plate convergence rate tance between Hualien and Taipei is about 120 km.
of 8.0 cm/year (Yu et al. 1999; Fig. 1). Nearly The shear waves traveling over this distance
18,000 seismic events occur around Taiwan should take more than 30 s. Thus, if a seismic
and the vicinity every year (Wu et al. 2008). monitoring system can reliably determine the
Earthquake Early Warning System in Taiwan 651
Earthquake Early
Warning System
in Taiwan,
Fig. 1 Distribution of the
real-time strong-motion
stations (squares) of the
Central Weather Bureau
(CWB) and sample virtual
subnetwork (VSN) with
60-km radius from the
earthquake early warning
(EEW) system. Stars show
epicenters of the four major E
damage earthquakes
location and magnitude within 20 s for a large of modern technologies for earthquake early
earthquake that could threaten the metropolitan warning purposes by the CWB in 1993. This
area, then more than 10 s will be available for system consisted of 10 three-component force
emergency response. Consequently, the Central balanced accelerometer stations (distributed in
Weather Bureau (CWB), Taiwan, started to an aperture of about 20 km) which were
develop an experimental system to explore the telemetered digitally via 9600-baud telephone
technology of EEW for the highly seismic area of lines to the CWB Hualien Station. At the Hualien
Hualien in 1993 (Wu et al. 1999). After 20 years Station, the incoming digital signals are
efforts, the EEW system is becoming one of the processed in real time and the results are
important tools for seismic hazard mitigation in telemetered to the CWB Headquarters in Taipei
Taiwan, and Taiwan is one of the leading countries (Fig. 2). A computer in Hualien handles the com-
in developing EEW system. Progress of the EEW munication with Taipei via a dedicated telephone
systems in Taiwan will be reported in this entry. line (19.2 k baud). A parallel computer system in
Taipei is used for displaying, processing, and
analyzing the received data. It also allows opera-
The Hualien Experimental EEW System tors in Taipei to access, manage, and control the
system in Hualien. A commercial company
The Hualien experimental EEW system is a implemented both the hardware and software
front-detection type prototype, which was under the CWB specifications and directions.
implemented in Hualien area to explore the use The goal of this system was to detect, locate,
652 Earthquake Early Warning System in Taiwan
Earthquake Early Warning System in Taiwan, telephone lines to the CWB Hualien Station. At the
Fig. 2 A prototype earthquake early warning system in Hualien Station, the incoming digital signals are
Hualien. This system consists of 10 three-component processed in real time and the results are telemetered to
accelerometers telemetered digitally via 9600-baud the CWB Headquarters in Taipei
demonstrated that earthquake reporting time can presented at the VSN system, they will be imme-
be significantly reduced by using a smaller net- diately processed to give simulated Wood-
work (Wu et al. 1999). This leads to the design Anderson seismograms for magnitude determina-
and configuration of a VSN approach. The VSN, tions as ML10. The earthquake magnitude ML
automatically configured by the monitoring sys- cannot be determined in the same time frame
tem, is event-dependent and its configuration due to the incomplete recording of shear waves
varies with time. In hypocenter and magnitude at some stations. However, the magnitude based
determinations, only stations close to the epicen- on the first 10 s of the signals, i.e., ML10, can be
ter (less than 60 km) contribute to the crucial found to correlate with ML as follows:
information. Within the framework of the real-
E
time strong-motion network, only signals from a ML ¼ 1:28 ML10 0:85 0:13 (1)
subset of stations are chosen to process and form
a VSN network surrounding an event. As soon as By applying Eq. 1 to magnitude determina-
the system is triggered by an event, the system tion, the VSN system can determine hypocenter
automatically extracts signals from a subset of and magnitude with tolerable uncertainty in
stations and configures a VSN with a 60-km about 20 s after the occurrence of an earthquake
radius centered on that event. Figure 1 also and early warning is thus possible in Taiwan.
gives a number of possible VSN configurations; During the period from 2006 to 2012, a total of
each normally consists of about a dozen stations. 207 earthquakes with magnitude ML ranging
The extracted data stream for this event forms the from 4.5 to 6.9 occurred on land or within
basic VSN input data for the subsequent EEW 50 km from the coastlines of the Taiwan Island
work. By working with the VSN, the reporting (Fig. 3). A total of 196 earthquakes were
time could be substantially reduced, which makes detected, processed, and reported in real time.
it possible to cover the entire Taiwan region with Detecting successful rate is about 95 %. Only
effective earthquake early warning. 5 % offshore and magnitude smaller than 5.0
Two major requirements for an EEW system events could not be detected automatically. If
are near real-time estimation of the earthquake the off-line manual measurements give correct
location and magnitude. The first requirement on values, the VSN results give average errors of
rapid location can be achieved readily in a 10-s 6.5 7.9 km and 5.3 9.2 km in epicenter and
time window immediately following the first focal depth, respectively. The error of magnitude
arrival of P waves. On the other hand, the second is 0.3 0.2 units. The reporting time is 18.6
requirement of rapid determination of the earth- 3.7 s (Fig. 3). Large uncertainties in the estimated
quake magnitude would be more difficult because magnitude using the initial portion of the
the shear-wave trains may not be recorded seismograms are inevitable. Similar results are
completely within this time window. Therefore, also presented in many other studies (e.g.,
a method for quick magnitude determination for Nakamura 1988; Grecksch and Kumpel 1997;
large events needs to be developed. The ML10 Allen and Kanamori 2003; Kanamori 2005; Wu
(Wu et al. 1998) is one of the approaches to and Kanamori 2005a, 2008; Wu et al. 2007). For
determine ML using first 10 s of the signals. larger offshore earthquakes, the VSN method
Signals collected by the stations within a 60-km may underestimate the magnitudes due to the
radius from an event are grouped and extracted limited intervals of the waveform signals used.
from the VSN input, which will then be processed To avoid this problem, the magnitude Mt
in parallel through the VSN software in a dedi- obtained from the average period tc (Kanamori
cated computer. A series of experiments have 2005) and the dominant period tmax p (Allen and
been conducted to determine the optimum Kanamori 2003) of the initial P waves may offer a
recording time for a 60-km radius network. satisfactory solution (Hsiao et al. 2009). How-
Results show that 10 s is about the optimum ever, the VSN system has achieved a short earth-
timing. As soon as the 10-s waveforms are quake reporting time about 20 s. This can offer
654
Earthquake Early Warning System in Taiwan, Fig. 3 Comparison of the real-time automatic event locations and magnitudes of the EEW system (subnet) and the manual
off-line event locations of the CWB catalog. Right-bottom figure shows the EEW reporting times versus magnitudes
Earthquake Early Warning System in Taiwan
Earthquake Early Warning System in Taiwan 655
Earthquake Early Warning System in Taiwan, Fig. 4 Schematic diagram of the new prototype system for
earthquake early warning in the CWB
earthquake early warning for metropolitan areas prototype EEW system. Real-time seismic signals
located 70 km from the epicenter and beyond. are packaged and transmitted to the headquarters
In order to provide warning for regions within via various IP-based networks, such as Frame
70 km from the epicenter, a new prototype EEW Relay, ADSL, GPRS, or satellite telemetry. Dif-
system using a P-wave method is developed using ferent telemetered networks can be arranged as a
a real-time seismographic network currently in secure environment for seismic data transmission.
operation in the CWB of Taiwan. This system is A cluster of computers running the Earthworm
based on the Earthworm environment which is system is installed at the central station in Taipei.
capable to carry out integrated analysis of real- Instead of ML10, the “Pd magnitude,” MPd (Wu and
time broadband, strong-motion, and short-period Zhao 2006; Hsiao et al. 2011), is adopted as mag-
signals. The Earthworm system distributed by the nitude indicator in the new system in order to
United States Geological Survey (USGS) is devel- shorten reporting time. Pd is defined as the peak
oped as a multifunctional seismic data processing amplitude of the initial P-wave displacement
system. Figure 4 shows the configuration of (Wu and Kanamori 2005b). Incoming signals are
656 Earthquake Early Warning System in Taiwan
processed in real time. When a large earthquake the signal resolution is 16 bits with a 2 g to
occurs, P-wave arrival times and Pd will be esti- +2 g range. The sampling rate is 100 samples per
mated for location and magnitude determinations second. Once an earthquake is detected by the
for EEW purpose. In a test of 54 felt earthquakes, trigger algorithm stored in a Palert, a
this system can report earthquake information in corresponding earthquake watch or warning
18.8 4.1 s after the earthquake occurrence with alarm is sent by the device. The major trigger
an average difference in epicenter locations of 6.3 algorithms include procedures for continuous
5.7 km and an average difference in depths of monitoring peak ground acceleration (PGA) and
7.9 6.6 km from catalogs. The magnitudes the peak amplitude of the filtered (high pass at
approach a 1:1 relationship to the reported magni- corner 0.075 Hz) vertical displacement of the
tudes with a standard deviation of 0.5 (Hsiao P wave (e.g., Pd). Palert also has networking
et al. 2011). capabilities for streaming real-time acceleration
VSN system is the current system operated by signals to the data collection servers and synchro-
CWB. When the new system performs as well as nizing clocks by the Network Time Protocol
the VSN system, it will become one of the sys- (NTP). Therefore, it is possible to connect Palert
tems operating together with the VSN. Currently, devices to build a front-detection EEW system.
the new system is capable of offering earthquake Since 2010, NTU began to build up the Palert
warning information in less than 10 s after the real-time strong-motion network. A total of
occurrence of events inland. Therefore, the new 400 stations were deployed and configured in
system will become the major EEW system of the 2013. Figure 5 provides the distribution of the
CWB in the future. Palert stations. Most of the stations are located
in elementary schools where power and Internet
connections are available. Figure 6 provides the
A Hybrid of Both On-Site and Front- configuration of the Palert network. At the field
Detection EEW System Using Low Cost site, real-time signals are processed by the Palert
Accelerometers devices. Three-component acceleration signals
are processed on-site for detecting P-wave arrival
The EEW system of the CWB can provide earth- and continuously double integrated to displace-
quake information within 20 s following an earth- ment signals for calculating Pd. The real-time
quake occurrence. Although a 20- s reporting signals are sent to the central stations every sec-
time is already very short, the reporting time ond simultaneously via the Internet. When an
can still be reduced if the density of seismic earthquake occurs, the Palert automatically
stations in the network can be increased. How- detects P-wave arrival, and once Pd or PGA are
ever, to build such a high-density seismic net- larger than 0.35 cm or 80 gal, respectively, the
work by traditional force-balance seismometers, Palert device begins its alert with a warning
the cost is extremely high. In contrast, Micro sound on site. Tasks of the central stations
Electro Mechanical Systems (MEMS) acceler- include data clustering and processing. For real-
ometers introduced in seismic applications time seismic data processing, real-time signals
(Holland 2003) are ideal and cost-saving devices are processed and stored in the Earthworm sys-
for recording strong ground motions. The EEW tem. In the Earthworm system, signals are again
research group at National Taiwan University processed in order to detect the arrival of the
(NTU) worked with a technology corporation to P wave (Hsiao et al. 2011; Chen et al. 2012).
develop a P-wave alert device, named “Palert,” P-wave information detected by the Earthworm
that uses MEMS accelerometers for on-site and platform is also sent to a shared memory for later
front-detection EEW purposes. The cost of a fusion process. During the final stage, once eight
Palert device is less than 1/10 of the cost for a Palert stations are triggered, an event is declared.
traditional accelerometer. The Palert device can The TcPd.c program fetches the event parameters
record real-time, three-component signals, and stored within the shared memory and computes
Earthquake Early Warning System in Taiwan 657
Earthquake Early
Warning System
in Taiwan,
Fig. 5 Distribution of the
Palert stations. Star shows
the epicenter of the 2 June
2013 Nantou ML 6.3
earthquake
Earthquake Early
Warning System
in Taiwan,
Fig. 6 Configuration of
the Palert network
658 Earthquake Early Warning System in Taiwan
Earthquake Early Warning System in Taiwan, magnitudes given by the real-time Palert network and
Fig. 7 A comparison of earthquake information, as fol- the CWB published earthquake catalogs. (d) The
lows: (a) epicenters, (b) focal depths, and (c) the reporting times of the Palert network
the earthquake early warning information. The the Palert network versus those from the CWB
hypocentral location is determined using a tradi- catalogue are plotted in Fig. 7. The hypocenters
tional earthquake location algorithm with a half- determined by the two systems are generally in
space linear increasing velocity model (Wu and agreement. Only a few offshore events have
Lin 2013). The magnitude (Mp) determination is larger differences. The average difference in epi-
based on the relationship of peak displacement center location is 14.1 km with a standard devia-
and velocity amplitudes of initial three P waves tion of 15.1 km. The average difference in focal
and hypocentral distance (Wu and Zhao 2006; depth is 1.6 km with a standard deviation of
Wu et al. 2013). 18.5 km. The magnitudes determined by the
During the period from June 2012 to August Palert network and the corresponding ML values
2013, a total of 33 earthquakes with ML magni- from the CWB earthquake catalog are listed. Fig-
tudes ranging from 4.6 to 6.3 have been detected. ure 7 also provides a plot of MP values against
Front-detection EEW performance of the Palert those of the CWB earthquake catalog. In general,
system is summarized in Figure 7. Results of the magnitude uncertainty is in the order of 0.3,
Palert network were compared with those which is an acceptable value for EEW applica-
reported by the CWB. Hypocenters provided by tion. More than 1 year operation, the Palert
Earthquake Early Warning System in Taiwan 659
Earthquake Early Warning System in Taiwan, ML 6.3 earthquake. The leading time is defined as the
Fig. 8 The early warning leading times for Palert stations interval between the time of on-site sound alert and the
close to the epicentral region of the 2 June 2013 Nantou time of PGA
network achieves an early warning reporting On-site approach may possibly offer early warn-
time, the time between earthquake occurrence ing for regions close to the epicenter. Palert has
and the time of the system providing earthquake an on-site alert function that triggers a warning
information, from 9 to 28 s, with an average of sound once Pd exceeds 0.35 cm or PGA is larger
15.5 s and a standard deviation of 4.3 s (Fig. 7). than 80 gal (Wu et al. 2011). The major advantage
The results indicate that for the majority of events of the on-site Pd approach is that it can provide a
the Palert EEWS can issue an early warning valuable early warning for epicentral regions.
report within approximately 20 s after an earth- Figure 8 shows the resulting early warning lead-
quake occurrence. It shows that it is feasible to ing times near the epicentral regions for the 2 June
use Palert devices to build a practical front- 2013 Nantou ML 6.3 earthquake. So far, this
detection EEW system. The system could offer earthquake is the largest event that has been
EEW messages for the regions at distance more detected by the Palert network. The early warning
than 60 km from the earthquake epicenter. leading time is defined as the interval between the
Front-detection EEW system could not offer time when the filtered vertical displacement
early warning messages to the epicentral regions. exceeds 0.35 cm or acceleration large than
660 Earthquake Early Warning System in Taiwan
80 gal and the time of the PGA. Only one station Chen DY, Lin TL, Wu YM, Hsiao NC (2012) Testing a
(W093) has a very short leading time. However, P-wave earthquake early warning system by simulating
the 1999 Chi-Chi, Taiwan, Mw 7.6 earthquake. Seismol
the majority of stations which have PGA larger Res Lett 83:103–108. doi:10.1785/gssrl.83.1.103
than 80 gal have a 1–9 s leading time. It is worth Grecksch G, Kumpel HJ (1997) Statistical analysis of
noting that those stations which are very close to strong-motion accelerogram and its application to
the epicenter (W07F, W072, W07C, W078, earthquake early-warning systems. Geophys J Int
129:113–123
W08E, and W07A) and recorded of PGA larger Holland A (2003) Earthquake data recorded by the MEMS
than 400 gal had leading times from 3 to 6 s. Such accelerometer: field testing in Idaho. Seismol Res Lett
times are long enough for automatic systems to 74:20–26. doi:10.1785/gssrl.74.1.20
make an emergency safety response. Hsiao NC, Wu YM, Shin TC, Zhao L, Teng TL
(2009) Development of earthquake early warning sys-
tem in Taiwan. Geophys Res Lett 36:L00B02.
doi:10.1029/2008GL036596
Summary Hsiao NC, Wu YM, Zhao L, Chen DY, Huang WT, Kuo
KH, Shin TC, Leu PL (2011) A new prototype system
for earthquake early warning in Taiwan. Soil Dyn
In this paper, the progress on EEW in Taiwan is Earthq Eng 31:201–208. doi:10.1016/j.
introduced in particular for the systems run by solidyn.2010.01.008
CWB and NTU. Both systems are able to reduce Kanamori H (2005) Real-time seismology and earthquake
the EEW reporting within 20 s. This represents a damage mitigation. Annu Rev Earth Planet Sci
33:195–214. doi:10.1146/annurev.earth.33.092203.
significant step toward the capability of providing 122626
realistic EEW in Taiwan. The CWB is the agency Lee WHK, Shin TC, Teng TL (1996) Design and imple-
responsible agency for earthquake reporting and mentation of earthquake early warning system in Tai-
NTU is a research and education institute. So, the wan. Paper no. 2133. In: 11th world conference of
earthquake engineering. Acapulco
front-detection EEW message is officially Nakamura Y (1988) On the urgent earthquake detection
announced by the CWB. However, the technique and alarm system (UrEDAS). In: Proceedings of the
developed by NTU will be transferred to the CWB 9th world conference on earthquake engineering,
for routine operation. Currently, front-detection Tokyo-Kyoto
Teng TL, Wu YM, Shin TC, Tsai YB, Lee WHK
EEW systems in Taiwan still cannot offer EEW (1997) One minute after: strong-motion map, effective
messages for regions within 50 km from an earth- epicenter, and effective magnitude. Bull Seismol Soc
quake epicenter. However, an on-site method can Am 87:1209–1219
provide 3–6 s leading time before the largest shak- Wenzel F, Zschau J (2013) Early warning for geological
disasters – scientific methods and current practice.
ing based on the results from the NTU EEW sys- Springer, Berlin/Heidelberg/New York. ISBN 978-3-
tem. Thus, it is envisioned that a hybrid approach 642-12232-3
such as the NTU EEW system will be one of the Wu YM, Chen CC, Shin TC, Tsai YB, Lee WHK, Teng
future systems for real operation. TL (1997) Taiwan rapid earthquake information
release system. Seismol Res Lett 68:931–943
Wu YM, Shin TC, Tsai YB (1998) Quick and reliable
determination of magnitude for seismic early warning.
Cross-References Bull Seismol Soc Am 88:1254–1259
Wu YM, Chung JK, Shin TC, Hsiao NC, Tsai YB, Lee
WHK, Teng TL (1999) Development of an integrated
▶ Early Earthquake Warning (EEW) System:
seismic early warning system in Taiwan – case for the
Overview Hualien area earthquakes. Terr Atmos Ocean Sci
▶ Seismic Accelerometers 10:719–736
Wu YM, Teng TL (2002) A virtual sub-network approach
to earthquake early warning. Bull Seismol Soc Am
92:2008–2018
References Wu YM, Kanamori H (2005a) Experiment on an onsite
early warning method for the Taiwan early warning
Allen R, Kanamori H (2003) The potential for earthquake system. Bull Seismol Soc Am 95:347–353
early warning in South California. Science 300: Wu YM, Kanamori H (2005b) Rapid assessment of
786–789 damaging potential of earthquakes in Taiwan from
Earthquake Location 661
for the later-arriving S phase can be large, phase, depth, and distance triplet. The predicted
resulting in a correspondingly large error in the travel times may be adjusted by various correc-
estimate of epicentral distance. tions, to account for the ellipticity of the Earth
These early graphical methods evolved into and topographical effects, as well as path correc-
a method of earthquake location by triangulation, tions to account for three-dimensional velocity
drawing arcs from station locations with heterogeneities. Thus, the time residual for
a compass on a large globe using a tape measure a phase at the ith station is defined as
calibrated for the travel time. For many years the
pred o
US Coast and Geodetic Survey (USC&GS) used di ¼ tobs
i ti ¼¼ tobs
i T þ ti
model
þ tcorr
i
this method to produce the Preliminary Determi- (3) E
nation of Epicenters (PDE) bulletin.
where d is the time residual; tobs and tpred are the
observed and predicted arrival times, respec-
The Basic Location Problem tively; T o is the origin time; tmodel is the travel-
time table value; and tcorr is the sum of applicable
The location estimate is determined by finding travel-time corrections. The predicted travel time
the location and origin time (hypocenter) that for a phase arrival is a function of the station and
minimizes the difference between observations source coordinates as well as the velocity model.
and predictions of phase arrival times as mea- For local earthquake location problems, where
sured at a network of seismographic stations. the epicentral distances do not exceed a few hun-
Methods to solve the location problem typically dred kilometers, a horizontally layered, or “flat”
assume that both the phase associations (i.e., all Earth, model can be used without imparting sig-
associated phases belong to the event) and the nificant computational error. Representing the
phase identifications (or at least whether a phase Earth as a series of horizontal layers with con-
is P or S type) are correct. The arrival-time pre- stant or slowly varying velocities dramatically
diction for each observation is the event origin simplifies the calculation of travel-time predic-
time plus a calculation of travel time between the tions as they can be expressed in closed form in
event and the observing station for the designated a Cartesian coordinate system. In either case,
phase. Historically, travel time was read from an travel time is a nonlinear function of event loca-
empirical travel-time table. Modern methods tion, requiring the use of nonlinear inversion
integrate time along ray paths for each phase methods.
within a model of wave speed for the Earth’s
interior. To ease the potentially heavy computa-
tional burden of computing ray paths and travel Geiger’s Method
times in a heterogeneous Earth model, most loca- One of the earliest numerical methods for earth-
tion algorithms use radially symmetric velocity quake location by the ISS consisted of least-
models, for which velocity is only a function of squares estimations using a desk calculator.
depth. For local velocity models, this translates to With the advent of computer processing, the
a horizontally layered structure. method of Geiger (1910) was used to develop
Computationally efficient methods have been programs for routine earthquake location.
developed to compute travel times in radially Among the first of these programs were those
symmetric Earth models (Buland and Chapman developed by Bolt (1960) for the ISS, by Engdahl
1983). To further increase computational speed, and Gunst (1966) for the USC&GS, and by
precomputed travel times for various phases are Edouard Arnold for the ISC.
often collected in travel-time tables for a set of Geiger’s insight was that if the initial source
depth and distance values. These travel-time coordinates (i.e., the user’s initial guess for
tables are then used as lookup tables and interpo- the starting position for the hypocenter, [x0, y0,
lated to get travel-time predictions for a specific z0, T o0]T) are sufficiently close to the true
664 Earthquake Location
Earthquake Location, Fig. 2 Neighborhood algorithm origin time and depth). The square represents the best
search results for the Mw = 6.7, January 17, 1929, solution from the neighborhood search; the star shows
11:45:44 earthquake in Venezuela. The misfit surface the final location
exhibits multiple minima (also note the trade-off between
hypocenter, the residuals can be expanded in iteration k, the model parameters are adjusted
a Taylor series with the higher-order terms con- by the current estimate of the adjustment vector,
sidered to be negligible. mk+1 = mk + mest, where mest is the least-squares
solution of Eq. 5. Textbook linear algebra
@ti @ti @ti informs us that the solution to Eq. 5 is
di ¼ Dx þ Dy þ Dz þ DT o (4)
@x @y @z
1
mest ¼ GT G GT d (6)
where d is the travel-time residual, z is event
depth, y and x are local event coordinates in the
direction of latitude and longitude, T o is event Practically all linearized location algorithms
origin time, and t is phase travel time. This yields build on Geiger’s method. Note that Geiger’s
a linear system of N equations (N = number of linearization does not make the location problem
arrival-time observations) with M 4 model linear; it assumes that the solution to the
parameters written in matrix form as nonlinear problem can be found by iteratively
solving a linear approximation to the problem,
Gm ¼ d (5) which enables application of standard numerical
methods.
where G is the (N M) design matrix containing The assumption that the initial source coordi-
the partial derivatives for N data with respect nates are sufficiently close to the true hypocenter
to M model parameters, m is the (M 1) model makes linearized location algorithms very sensi-
adjustment vector, [Dx, Dy, Dz, DT o]T, and d is tive to the initial guess for the starting hypocen-
the (N 1) vector of time residuals. Geiger ter. Figure 2 shows the misfit surface (the root
proposed to solve the system of equations with mean square of time residuals) obtained with the
an iterative least-squares method by minimizing neighborhood algorithm (Sambridge and Kennett
the root mean square of residuals. After each 2001), a nonlinear grid search algorithm (a more
Earthquake Location 665
detailed description is given later in the section For the flat Earth approximation, widely used
Nonlinear Methods) for the Mw = 6.7, January in local earthquake location algorithms, many of
17, 1929, event in Venezuela. The station cover- the above corrections are unnecessary. The
age for this event was quite poor as most of the travel-time predictions can also, in many cases,
recording stations were in Europe and North be expressed in closed forms. For instance, for
America. The misfit surface exhibits multiple a two-layer velocity model with a source at the
minima, and a linearized location algorithm surface, the travel times for the direct, reflected,
could easily be trapped in any of them, depending and refracted waves are given by
on the starting hypocenter. In this case the neigh- pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
borhood algorithm provided a good starting loca- x x2 þ 4h2 E
tion (square) for the linearized location algorithm tdirect ¼ ; treflected ¼ ;
v1 v1 (7)
that further refined the initial guess (star). The x
trefracted ¼ t0 þ
right panel clearly demonstrates the trade-off v2
between depth and origin time. It should be
noted that the problem of multiple local minima where x is distance, h is the depth of the layer
in the misfit surface is a particularly serious prob- boundary, v1 and v2 are the velocities above and
lem for local event locations in areas with strong below the layer boundary, # = asin(v1/v2) is the
3D velocity heterogeneities. critical angle, and t0 = 2h cos(#)/v1. Figure 3
shows the travel-time curves and various ray
Velocity Models/Travel-Time Tables paths. Figure 3 also illustrates one of the major
Since knowledge of Earth structure is derived pri- difficulties of the local earthquake location
marily from earthquake data, the earliest Earth problem – despite the fact that the travel times
models were at best rudimentary, often inaccurate are easily computed, one can generate a plethora
and incomplete. Travel-time predictions deter- of crustal phases, all arriving within a few seconds
mined using early Earth models were valid only from each other. This can make phase identifica-
for shallow-depth earthquakes. For many years the tion difficult, especially around the crossover dis-
standard travel-time tables used by the ISC and the tances. It should be noted that ray tracing to obtain
NEIC were the Jeffreys and Bullen tables travel-time predictions could be still necessary if
published in 1940. Although the limitations of the velocity model becomes complicated, or the
these tables had been known for some time, until source is not in the uppermost layer of the model.
only the 1990s no other tables could provide such Furthermore, the crust is the most heterogeneous
a complete representation of travel times for P, S, part of Earth’s interior, so 1D velocity models tend
and other later-arriving phases. to do a poor job of adequately describing the
In 1987 the International Association of Seis- local velocity structure. The result is increased
mology and Physics of the Earth’s Interior nonlinearity, and it is not uncommon to have misfit
(IASPEI) initiated a major international effort to surfaces with multiple local minima.
construct new global travel-time tables for earth-
quake location and phase identification. A model Depth Resolution
resulting from this effort was iasp91 (Kennett and Until only recently the ISC and NEIC relied
Engdahl 1991). Subsequently, Kennett almost entirely on the times of first-arriving
et al. (1995) constructed an improved model for P phases to determine the locations of earth-
the average P and S radial velocity profile of the quakes that are observed globally, which for
Earth (ak135). The ak135 model has proven very P rays bottoming in the lower mantle
suitable for predicting the arrival times of a wide (teleseismic) the ak135 travel times are particu-
variety of seismic phases for use in event location larly accurate. These agencies now use all later-
and phase identification procedures, and the arriving phases for earthquake location. In the
ak135 model is presently used by both the NEIC absence of near-station phase data, direct
and ISC. teleseismic phases alone provide little depth
666 Earthquake Location
Earthquake Location,
Fig. 3 Ray paths and
travel-time curves for
a surface source in
a two-layer velocity model.
The refracted wave takes
over the direct wave at
Xcross, the crossover
distance. The travel-time
curves of the direct and
refracted waves are straight
lines with a slope of the
slowness (inverse velocity)
of the first and the second
layer, respectively
resolution, as the derivative of travel time with makes picking the depth phases extremely diffi-
respect to depth at teleseismic distances is only cult, and therefore they are hardly ever used.
a small, slowly varying function. P-waves and However, if there are stations where both the
S-waves that initially travel upward and reflect direct (Pg) and refracted (Pb or Pn, traveling
off of the Earth’s surface before traveling to the along the Conrad or the Moho discontinuities in
recording station provide important constraints the crust, respectively) phases are picked, they
on event depth because their travel-time deriva- could provide sufficient resolution to constrain
tives with respect to depth are opposite in sign to the depth as their travel-time derivatives with
those of direct teleseismic phases, as illustrated in respect to depth have opposite signs. Figure 4b
Fig. 4a. The surface-reflected phases are com- shows that at close distances the travel-time
monly referred to as “depth” phases. The trade- derivative of the direct phase changes rapidly,
off between event depth and origin time is also thus providing depth resolution. As a rule of
avoided by the inclusion of depth phases. The thumb, stations closer than twice the event depth
arrival times of depth phases can either be used are needed to reliably constrain event depth.
directly as an independent observation or the
depth can be estimated based on differential Error Budget
pP-P times. Alas, due to interference with the The error budget in seismic location problems is
direct phase, as well as potentially strong rever- traditionally described as the combination of
berations in the Earth’s crust, a distinct depth “measurement” and “model” errors. The effect
phase is only reliably observed for events below of errors due to ignoring the higher-order terms
approximately 50 km. in the Taylor expansion in Geiger’s method,
For local problems depth phases are little help except for some highly degenerate network
as there is not enough separation between the geometries, is typically second order compared
direct P arrival and the depth phase, which to the model and measurement errors.
Earthquake Location 667
Earthquake Location, Fig. 4 (a) Travel-time deriva- direct teleseismic P phase. (b) Travel-time derivatives
tives with respect to depth as a function of epicentral with respect to depth at local distances for a source
distance for a source depth of 600 km for direct depth of 15 km for the direct Pg and the refracted Pb and
teleseismic P and the teleseismic depth phases pP and Pn from the Conrad and Moho discontinuities. At close
sP. The direct teleseismic P phase cannot resolve the distances the depth derivatives of the travel times of the
depth in itself as the depth derivative is a slowly varying direct phase are rapidly changing which provides depth
function, but when depth phase picks are available, resolution. Further away the direct and refracted phases
together they can resolve the depth. The inset shows that together provide depth resolution owing to the opposite
the depth phases pP and sP first reflected from the Earth signs of their derivatives
surface and then travel along a similar ray path as the
Earthquake Location, Fig. 5 Waveforms of nuclear two larger explosions the first-arriving Pn can be easily
explosions carried out at the Nevada test site recorded at picked, for the two smaller explosions the first-arriving Pn
Elko, Nevada, about 400 km distance. The explosions is completely masked by noise, and the more energetic,
were detonated within 15 km of one another. The wave- later-arriving Pg could be erroneously picked as the first
forms are band-pass filtered between 1 and 3 Hz and arrival
shown with increasing event magnitude. While for the
model will generate correlated travel-time pre- slowness and back-azimuth measurements from
diction errors along similar ray paths. Thus, the arrays (Bratt and Bache 1988; Lienert and
assumption of independent errors is violated Havskov 1995; Schweitzer 2001). While early
when rays sample the same velocity anomaly. locators allowed only for first-arriving P and
Ignoring the correlation structure of the travel- S phases, recent algorithms use most of the
time predictions results in both location bias and phases allowed for by the ak135 model (e.g.,
increasingly underestimated location uncertainty Bratt and Bache 1988; Engdahl et al. 1998;
estimates. Bondár and Storchak 2011). Several popular
algorithms focus on the local earthquake location
problem and use travel-time predictions from
Linearized Inversion Methods
E
a local velocity model (Lee and Lahr 1972;
Lienert and Havskov 1995; Schweitzer 2001).
All linearized location algorithms originate from Almost all linearized location algorithms rely on
Geiger’s method (1910). The differences are typ- one-dimensional velocity models or travel-time
ically in the details, such as how the inverse of the tables to obtain phase arrival-time predictions. To
G matrix is obtained, what kind of weighting account for large-scale three-dimensional effects,
scheme is applied, and how the formal the EHB algorithm (Engdahl et al. 1998) applies
uncertainties are calculated. Expressed in patch corrections, Bratt and Bache (1988) allows
a probabilistic framework and assuming indepen- for local and regional source-station-phase-
dent, normally distributed data, linearized loca- specific corrections as well as slowness-azimuth
tion algorithms maximize the likelihood function corrections, and the ISC locator (Bondár and
Storchak 2011) is capable taking travel-time pre-
1 dictions directly from a 3D velocity model
ℒðmÞ ¼ exp ðd GmÞT C1
d ðd GmÞ
2 (Myers et al. 2010).
(8)
Uniform Reduction
where Cd is the data covariance matrix describing One of the most successful location algorithms is
the uncertainties in the data (picking and model Jeffreys’ uniform reduction method (Jeffreys
errors). Maximizing ℒ(m) is equivalent to solv- 1932). It was in use with minor modifications
ing the equation (Bolt 1960; Buland 1986) up until the introduc-
tion of a new location algorithm at the ISC
Gw m ¼ WGm ¼ Wd ¼ dw (9) (Bondár and Storchak 2011). Uniform reduction
also relies on Geiger’s method but solves
where W = C d
1/2
is the diagonal (N N) weight the linear system of equations in an iteratively
matrix. Some early location algorithms even reweighted least-squares fashion. This means
make the quite simplistic assumption that all that each arrival time will get a weight
data variances are the same and collapse the depending on its residual. The weighting func-
data covariance matrix into a single scalar value. tion is defined as
Popular location algorithms solve Eq. 9 in
various ways, such as standard least squares or " ( )#1
calculating the general inverse of the G matrix. ðd mÞ2
wðdÞ ¼ 1 þ mexp (10)
More recent algorithms tend to obtain the general 2s2
inverse by singular value decomposition. Seismic
arrays offer reliable estimates of back-azimuth where m, m, and s are parameters estimated from
(station-to-event azimuth) and apparent velocity the data (Jeffreys 1932). The weights are
(or more precisely, its inverse, the horizontal recalculated after each iteration. Hence, outliers
slowness vector) as a plane wave sweeps through get progressively smaller weights and their
the array. Several algorithms make use of the contribution to the final hypocenter will be
670 Earthquake Location
predictions not only produces more reliable hypocenter parameters (M 4) determines the
formal uncertainty estimates, but also reduces order of the chi-squared function (w2), and m ^ is
location bias. the global minimum of the misfit function.
The chi-squared function is divided by the
degrees of freedom, which is the number of
Nonlinear Methods observations (N) minus the number of free
parameters (M). The shape of the probability
Poor network configuration and/or the use of region determined in this way is nonparametric,
a model with discontinuous travel-time deriva- which can provide an accurate assessment of
tives can result in multiple local minima in the location uncertainty. However, the nonparamet-
E
arrival-time misfit function, which is problematic ric form can be difficult to summarize into
for algorithms based on Geiger’s method. a compact form that is convenient for distribu-
Numerous strategies have been developed to mit- tion in a bulletin format.
igate the tendency of Geiger’s method to get Obviously, a grid search on hypocenter
“stuck” in a local minimum. Most commonly, parameters is the most straightforward approach
a number of widely varying starting locations to map the misfit function. However, a global grid
are iterated to convergence, and the solution search can have computational limitations, even
with the lowest misfit is selected. However, in the case of single-event location. For example,
there is no guarantee that the global minimum is a grid with 1 km spacing on Earth’s surface
found and the formal uncertainty at the chosen extending to a depth of 35 km (ak135 crustal
hypocenter is not an accurate representation of thickness) and sampling 0.1 s for 1 min would
the non-Gaussian misfit function. result in a grid of over 1013 points. Even if travel
A number of nonlinear inversion procedures times were computed in a microsecond,
have been developed with the goal of reliably a location based on 100 arrival times would
finding the global minimum in the misfit function require years to compute on a single processor.
(i.e., the best location) and better representing Clearly, strategies to reduce the grid dimensions
location uncertainty. The vast majority of are needed in order to make a grid search on all
nonlinear methods use a strategy of explicitly four hypocenter parameters practical.
sampling the misfit function on a discrete grid Sambridge and Kennett (1986) limit the
of points (grid search). If the sample points pro- search domain by determining hypocenter
vide a high-fidelity representation of the misfit bounds based on the station locations at which
function, then it is straightforward to find the P-waves arrive first, S-wave minus P-wave
hypocenter, which is the global minimum in the arrival time, and plausible ranges in wave
misfit function. speed. In this case the domain of the grid search
Using grid search, it is only slightly more is determined by the data set, so computational
difficult to determine probability contours that demand can vary greatly. However, if the search
define hypocenter uncertainty. Sambridge and domain were limited to an epicenter domain of
Kennett (1986) provide a succinct description of 100 km by 100 km by 35 km by 60 s on a km
how misfit may be mapped to probability. spatial grid and a 0.1 s time grid, then computa-
Assuming Gaussian data errors, the likelihood at tional time can be a few minutes on a single
hypocenter grid points, ℒ(m), that satisfy the processor.
inequality Rodi (2006) describes an adaptive grid search
method that further reduces computational time
^ Þ
½ℒðmÞ ℒðm w2 ðpÞ by starting with a course grid and progressively
4 (14)
^Þ
ℒ ðm ðN M Þ densifying the grid in the neighborhood of points
where the misfit function is lowest. Adaptive
are within the p probability region, where gridding has the benefit of enabling an arbitrarily
p ranges between 0 and 1. The number of fine grid in the neighborhood of the event, which
672 Earthquake Location
removes limitations on hypocenter accuracy that proposed hypocenter and the current state. If the
is inherent for a fixed grid. procedure of proposing a new hypocenter adheres
Sambridge and Kennett (2001) use a natural to the critical balance condition, which requires
neighbor approach to further refine the sampling reciprocity of probability in the proposal pro-
strategy. The natural neighbor method starts with cesses, then the series of accepted states comprise
a random sample in hypocenter space. More ran- a Markov chain, for which each state is a random
dom samples are then generated in the neighbor- sample of the multidimensional hypocenter prob-
hood of hypocenter samples with the lowest ability function. Therefore, high sample density
misfit values. This process is iterated until equates to high probability. Many Markov chains
a global minimum is found. Similar to the grid with widely varying starting locations can be
search method, the misfit function can be used to ensure robustness (Myers et al. 2007).
contoured to assess location uncertainty. As will The process of proposing, accepting, and
be discussed further, the misfit may also be rejecting new hypocenters typically finds the
used to guide numerical integration of neighborhood of the global minimum in a few
multidimensional probability of the hypocenter, hundred iterations. The process is iterated thou-
which provides a high-resolution estimate of the sands of more times to adequately characterize
hypocenter volume. hypocenter probability, but the number of for-
It should be noted that features in the misfit ward calculations is still orders of magnitude
function, including the global minimum, can be fewer than in a full grid search. Sample density
missed if the spacing of the sample points is too may be contoured to characterize the nonpara-
large. For example, if a grid with 10 km spacing is metric probability volume over hypocenter
used, then a feature in the misfit function with space. Alternatively, summary statistics (e.g.,
spatial dimension of a few km might not be sam- probability ellipses) may be computed by assum-
pled. This is particularly problematic for adaptive ing that the samples are drawn from an analytical
grid approaches, for which computational gains distribution (e.g., Gaussian).
are derived from a coarse starting grid. The safest
practice is to initialize sampling with spacing
smaller than the expected dimension of location Event Location with 3D Velocity Models
uncertainty. That way the edges of the depression
in the misfit function will be reliably sampled. As discussed in the section on model errors,
Estimating the probability density function of travel-time prediction errors are particularly
the hypocenter via stochastic sampling can more problematic because they are repeatable, not ran-
accurately characterize the probability volume dom. This property of model error can result in
and can, in some instances, improve computa- a similar vector shift (bias) in hypocenter param-
tional efficiency. Lomax et al. (2000) introduce eters and unrepresentative location uncertainty
the Metropolis-Gibbs method to sample the estimates (Bondár and McLaughlin 2009).
a posteriori location probability density function. Accounting for data covariance can mitigate the
The Metropolis-Gibbs method begins with effects of model error, but a more accurate loca-
a starting hypocenter (state), which is commonly tion and smaller uncertainty bounds can be
selected to be the location of the station at which achieved by improving travel-time prediction
the first P-wave arrived. A new hypocenter is accuracy with improved models of seismic veloc-
proposed based on a random perturbation of the ity. Practical application of 3D models depends
current state. If the probability of predicted data greatly on the distance domain of the network:
(weighted data fit) is higher at the proposed hypo- teleseismic, regional, or local.
center, then the proposed hypocenter becomes
the new “state.” If the probability is lower at the Teleseismic
proposed hypocenter, then it is accepted at a rate Teleseismic travel times can be accurately com-
proportional to the ratio of the probabilities at the puted by integrating slowness along a ray that is
Earthquake Location 673
computed using a 1D model. In other words, the computed in about one-half second on current
derivative of teleseismic travel time with respect computers, real-time location may require
to observed variations in lower mantle velocity precomputation. Precomputation approaches at
is approximately linear. As a result, ray tem- regional distance are the same as those described
plates could be used to compute travel times on for teleseismic phases, but the grid spacing must
the fly. However, the most common approach is be smaller in order to capture strong gradients
to precompute differences between travel times and first-order discontinuities in the travel-time
for a 3D and a 1D model at a fixed event depth. corrections. Myers et al. (2010) achieve milli-
The travel time for the 3D model may then be second regional travel-time computation for Pn
approximated by using the 3D-1D model differ- and Sn phases by parameterizing upper mantle
E
ence as a correction to the computationally velocity as a linear gradient, which enables flex-
efficient travel-time lookup for the 1D model. ible, on-the-fly implementation in a real-time
For example, a zero-depth correction surface location procedure. However, this approach is
may be applied to locate events throughout the not applicable for rays interacting with velocity
crust. Travel-time corrections are attractive at discontinuities at approximately 410 and 660 km
teleseismic distance because deviations from depth.
1D travel times vary slowly and interpolation Because crust and upper mantle velocity
of corrections is accurate even when points of anomalies can be large compared to anomalies
precomputation are on the order of 100–200 km in the lower mantle, regional travel-time errors
apart. Studies of improvement in epicenter accu- for 1D models can be many times larger than
racy generally find that average error can be those at teleseismic distances. Epicenter error
reduced from approximately 15 km to approxi- can be reduced from approximately 20 km on
mately 8 km (e.g., Yang et al. 2004). average for a 1D model to approximately
6–8 km for a 3D model (e.g., Ritzwoller
et al. 2003; Yang et al. 2004; Flanagan
Regional et al. 2007; Myers et al. 2010).
Due to strong velocity anomalies, minimum-
time rays turning in the upper mantle can differ
significantly for 3D and 1D models. Ray shoot- Local
ing techniques have been used to compute Velocity anomalies are largest in the crust, and
regional travel times. Shooting methods iterate important velocity discontinuities can be unique
on the direction and inclination of a ray leaving to local study areas. Unless local velocity
the event until the ray hits the station. However, models are highly smoothed, numerical solu-
a geometrical ray may not exist for some paths, tions to the equation of a propagating wave
resulting in “shadow zones” where travel time is front are the best approach for local travel-time
not defined, but diffracted waves may be calculations (e.g., Hole and Zelt 1995; Rawlin-
observed. Shadow zones cause significant diffi- son and Sambridge 2004). Eikonal solutions
culty for location algorithms because they are are computationally expensive and require
discontinuities in the domain of the travel-time precomputation from each station of the network
calculation. Shadow zones can be avoided by to a dense 3D grid of predefined points. Relying
using ray bending methods (e.g., Zhao and Lei on travel-time reciprocity, a travel time from any
2004; Simmons et al. 2012), where ray endpoints hypothesized event location to each station of
are fixed and the ray is iteratively bent to find the network can be computed using 3D interpo-
the least-time path. Bending methods have the lation of the travel-time grid. The derivative of
advantage of providing an approximation to the travel time with respect to event location is
travel time of diffracted waves, thus providing appreciably nonlinear at local distances, making
a travel time between any two points. Although nonlinear locators the best option (e.g., Lomax
regional ray paths and travel times can be et al. 2000).
674 Earthquake Location
Multiple Event Location Methods can be poorly constrained if events are tightly
clustered and event-station distances are large.
Simultaneous analysis of arrival times that are Douglas (1967) suggested determining station
associated with many events provides informa- corrections based on the residuals of one event
tion that can help to constrain unknowns whose hypocenter parameters are assumed
beyond those of the event locations. Unknowns known. All other events are then located relative
such as corrections to travel-time predictions, to the fixed event. This technique becomes
arrival-time measurement precision, and phase known as the master event location method and
names may be inferred if the data set contains is frequently used to constrain absolute locations
sufficient information. Because the unknown in an event cluster.
parameters include travel-time prediction A number of methods improve the robustness
corrections, uncertainties for each datum, and and efficiency of solving the multiple-event sys-
hypocenters, it is clear that the number of tem of equations. Dewey (1972) expanded on
unknowns in the most general formulation of JED to include other phases and data weighting
the multiple-event inversion will always out- based on estimated pick uncertainty. Dewey’s
number the data. Therefore, the most general formulation is the widely used Joint Hypocenter
formulation will always be underdetermined. Determination (JHD) method. Jordan and Sver-
The choice of parameters to be included in the drup (1981) and Pavlis and Booker (1983)
inversion and the approaches to constrain the employed matrix projection operators to opti-
inversion distinguish the various multiple-event mally compute unbiased estimates of station
applications. corrections. Jordan and Sverdrup (1981) solved
Douglas (1967) introduced multiple-event the system of equations including event loca-
location with the joint epicenter determination tions and station corrections in one iteration
(JED) formulation. If the data set contains one (hypocentroidal decomposition, HDC). Pavlis
phase type, the multiple-event formulation is and Booker (1983) located each event indepen-
dently, determined unbiased station corrections
@tij @tij @tij based on residuals, and repeated the process
d ij ¼ DT oi þ Dzi þ Dy þ Dxi þ Dsj
@zi dyi i @xi until convergence (progressive multiple-event
(15) location, PMEL). HDC is better suited for use
with regional and teleseismic networks for
where i specifies the event, j specifies the station, which partial derivatives with respect to event
d is the travel-time residual, z is event depth, location are relatively linear. PMEL has the
y and x are local event coordinates in the direction advantage being more robust to highly nonlinear
of latitude and longitude, T o is event origin time, derivatives, which is generally true at local
t is phase travel time, and S is a station-specific distances.
travel-time correction also called station correc- Recognizing the extreme precision of relative
tion. The linear system of equations is iterated to arrival times afforded by waveform correlation,
convergence. There is a need to impose con- Got et al. (1994) directly utilized differential
straints on the solution to Eq. 15 to remove ambi- arrival times to obtain highly precise relative
guity in the determination of the station terms event locations. The method minimizes the dif-
S. Douglas (1967) chose to force the sum of ference between observed and predicted differen-
station corrections equal to zero, which remedies tial arrivals. Thus, the approach is popularly
a trade-off between a shift in origin time of all referred to as double difference. Waldhauser and
events and the average station correction. Zero- Ellsworth (2000) extended the double difference
mean station corrections make the implicit method to larger regions by introducing an inter-
assumption that base-model travel-time predic- event length over which the implicit station/phase
tion errors are zero mean. Still, absolute locations corrections decorrelate (HypoDD program).
Earthquake Location 675
Computational limitations prevent application of thousands of associated phase arrivals. Even just
HypoDD to an arbitrarily large data set. a decade ago, it would not have been possible to
Myers et al. (2007, 2009) introduced perform near-real-time waveform correlation
a Bayesian nonlinear inversion framework to and double difference location in a dense local
multiple-event analysis (Bayesloc). The network (Waldhauser 2009) or simultaneously
nonlinear Markov chain Monte Carlo method relocate thousands of globally distributed events
enables Bayesloc to simultaneously assess the with a multiple location algorithm (Simmons
joint posterior distribution spanning event loca- et al. 2012). Hence, it can be expected that earth-
tions, travel-time corrections, phase names, and quake location methodologies will continue
arrival-time measurement precision. Myers to evolve with ever-increasing computational
E
et al. (2011) demonstrated that Bayesloc can be power.
applied to data sets of tens of thousands of events
and millions of arrivals because the solution does
not involve direct inversion of a matrix and com- Cross-References
putation demands grow linearly with the number
of arrivals. ▶ Earthquake Magnitude Estimation
Multiple location methods when used prop- ▶ Seismic Event Detection
erly will ordinarily produce precise relative ▶ Seismic Network and Data Quality
earthquake locations. However, these locations ▶ Seismometer Arrays
will almost always have an unresolved bias,
depending on the scale of lateral heterogeneity
in the underlying Earth model. Lateral heteroge- References
neity, especially in subduction zones, will also
set limitations on the maximum inter-event dis- Bolt BA (1960) The revision of earthquake epicentres,
focal depths and origin times using a high-speed com-
tance when using multiple-event location
puter. Geophys J Roy Astron Soc 3:433–440
methods. If the inter-event distances are too Bondár I, McLaughlin K (2009) Seismic location bias and
large, then travel-time predictions from events uncertainty in the presence of correlated and
to stations will not be precisely correlated, non-Gaussian travel-time errors. Bull Seism Soc Am
99:172–193
potentially degrading the precision of relative
Bondár I, Storchak D (2011) Improved location proce-
locations. dures at the International Seismological Centre.
Geophys J Int 186:1220–1244. doi:10.1111/j.1365-
246X.2011.05107.x
Bratt SR, Bache TC (1988) Locating events with a sparse
Summary network of regional arrays. Bull Seism Soc Am
78:780–798
The history of earthquake location methods is Buland R (1986) Uniform reduction error analysis. Bull
intertwined with the evolution of computers. Seism Soc Am 76:217–230
Buland R, Chapman C (1983) The computation of seismic
Even though Geiger formulated the problem
travel times. Bull Seism Soc Am 73:1271–1302
just 4 years after the great San Francisco earth- Dewey JW (1972) Seismicity and tectonics of western
quake, it was not implemented in routine opera- Venezuela. Bull Seism Soc Am 62:1711–1751
tions until the first mainframe computers became Douglas A (1967) Joint epicentre determination. Nature
215:47–48
available. Similarly, direct search methods
Engdahl ER, Gunst RH (1966) Use of a high speed com-
became computationally viable only in the late puter for the preliminary determination of earthquake
twentieth century. Before the 1980s location hypocenters. Bull Seism Soc Am 56:325–336
algorithms typically were able to deal with Engdahl ER, Villaseñor A (2002) Global seismicity:
1900–1999. In: Lee WHK, Kanamori H, Jennings
a few dozen first-arriving P phases; nowadays, PC, Kisslinger C (eds) International handbook of
owing to the much higher station density, it is not earthquake and engineering seismology, part A, chap-
uncommon that an event is reported with several ter 41. Academic Press, Amsterdam, pp 665–690
676 Earthquake Location
Engdahl ER, van der Hilst R, Buland R (1998) Global Myers SC, Begnaud ML, Ballard S, Pasyanos ME, Phillips
teleseismic earthquake relocation with improved WS, Ramirez AL, Antolik MS, Hutcheson KD, Dwyer
travel times and procedures for depth determination. JJ, Rowe CA, Wagner GS (2010) A crust and upper-
Bull Seism Soc Am 88:722–743 mantle model of Eurasia and North Africa for Pn
Flanagan MP, Myers SC, Koper KD (2007) Regional travel-time calculation. Bull Seism Soc Am
travel-time uncertainty and seismic location improve- 100:640–656
ment using a three-dimensional a priori velocity Myers SC, Johannesson G, Simmons NA (2011) Global-
model. Bull Seism Soc Am 97:804–825 scale P-wave tomography optimized for prediction of
Geiger L (1910) Herdbestimmung bei Erdbeben aus den teleseismic and regional travel times for Middle East
Ankunftszeiten. Nachrichten von der Gesellschaft der events: 1. Data set development. J Geophys Res
Wissenschaften zu Göttingen, Mathematisch- 116, B04304. doi:10.1029/2010JB007967
Physikalische Klasse 4:331–349 Pavlis GL, Booker JR (1983) Progressive multiple event
Got J-L, Fréchet J, Klein FW (1994) Deep fault plane location (PMEL). Bull Seism Soc Am 73:1753–1777
geometry inferred from multiplet relative relocation Rawlinson N, Sambridge M (2004) Wave front evolution
beneath the south flank of Kilauea. J Geophys Res in strongly heterogeneous layered media. Geophys
99:15375–15386 J Int 156:631–647
Gutenberg B, Richter CF (1954) Seismicity of the earth Ritzwoller MH, Shapiro NM, Levshin EA, Bergman EA,
and associated phenomena. Princeton University Engdahl ER (2003) Ability of a global three-
Press, Princeton dimensional model to locate regional events.
Hole JA, Zelt BC (1995) 3-D finite-difference reflection J Geophys Res 108(B7):2353. doi:10.1029/
travel times. Geophys J Int 121:427–434 2002JB002167
Jeffreys H (1932) An alternative to the rejection of obser- Rodi WL (2006) Grid-search event location with
vation. Mon Not R Astr Soc Geophys Suppl 2:78–87 non-Gaussian error models. Phys Earth Planet Int
Jeffreys H, Bullen KE (1940) Seismological tables. British 158:55–66
Association of the Advancement of Science, Gray- Sambridge MS, Kennett BLN (1986) A novel method of
Milne Trust, London hypocenter location. Geophys J Roy Astr Soc
Jordan TH, Sverdrup KA (1981) Teleseismic location 87:679–697
techniques and their application to earthquake clusters Sambridge MS, Kennett BLN (2001) Seismic event loca-
in the South-central Pacific. Bull Seism Soc Am tion: nonlinear inversion using a neighbourhood algo-
71:1105–1130 rithm. Pure Appl Geophys 158:241–257
Kennett BLN, Engdahl ER (1991) Travel times for global Schweitzer J (2001) HYPOSAT – an enhanced routine to
earthquake location and phase identification. Geophys locate seismic events. Pure Appl Geophys
J Int 105:429–465 158:277–289
Kennett BLN, Engdahl ER, Buland R (1995) Constraints Simmons NA, Myers SC, Johannesson G, Matzel
on seismic velocities in the Earth from traveltimes. E (2012). LLNL-G3Dv3: global P wave tomography
Geophys J Int 122:108–124 model for improved regional and teleseismic travel
Lee WHK, Lahr JC (1972) HYPO71: a computer program time prediction. J Geophys Res 117(B10).
for determining hypocenter, magnitude and first doi:10.1029/2012JB009525
motion pattern of local earthquakes. U.S. Geological Storchak DA, Di Giacomo D, Bondár I, Engdahl ER,
Survey Open File Report, Menlo Park, pp 72–224 Harris J, Lee WHK, Villaseñor A, Bormann P (2013)
Lienert BR, Havskov J (1995) A computer program for Public release of the ISC-GEM global instrumental
locating earthquakes both locally and globally. Seism earthquake catalogue (1900–2009). Seism Res Let
Res Let 66:26–36 84:810–815
Lomax A, Virieux J, Volant P, Berge C (2000) Probabi- Waldhauser F (2009) Near-real-time double-difference
listic earthquake location in 3D and layered models: event location using long-term seismic archives, with
introduction of a Metropolis–Gibbs method and com- application to Northern California. Bull Seism Soc Am
parison with linear locations. In: Thurber CH, 99:2736–2848
Rabinowitz N (eds) Advances in seismic event loca- Waldhauser F, Ellsworth WL (2000) A double-difference
tion. Kluwer, Hingham, pp 101–134 earthquake location algorithm: method and application
Milne J (1886) Earthquakes and other earth movements. to the Northern Hayward Fault, CA. Bull Seism Soc
Appleton, New York Am 90:1353–1368
Myers SC, Johannesson G, Hanley W (2007) A Bayesian Yang X, Bondár I, Bhattachryya J, Ritzwoller M,
hierarchical method for multiple-event seismic loca- Shapiro N, Antolik M, Ekström G, Israelsson H,
tion. Geophys J Int 171:1049–1063. doi:10.1111/ McLaughlin K (2004) Validation of regional and
j.1365-246X.2007.03555.x teleseismic travel-time models by relocating GT
Myers SC, Johannesson G, Hanley W (2009) Incorporation events. Bull Seism Soc Am 94:897–919
of probabilistic seismic phase labels into a Bayesian Zhao D, Lei J (2004) Seismic ray path variations in a 3D
multiple-event seismic locator. Geophys J Int global velocity model. Phys Earth Planet Int 141:
171:193–204. doi:10.1111/j.1365-246X.2008.04070.x 153–166
Earthquake Magnitude Estimation 677
Earthquake Magnitude Estimation, Fig. 1 Summary of highlighting that they were not converted to ground
the classical procedure of measuring ML. Note that the motion displacement amplitudes A (Modified copy of
letter B was given here to record trace amplitudes, Fig. 3.28 in Bormann et al. 2013; # IASPEI)
earthquake. By plotting, for many events, the The magnitude of a shock is . . . the logarithm of the
logarithm of the recorded maximum trace ampli- calculated maximum trace amplitude, expressed in
microns, with which the standard short-period tor-
tudes Amax over the epicentral distance D (see sion seismometer . . . would register that shock at
▶ Earthquake Location), Richter derived empiri- an epicentral distance of 100 km. (Richter 1935)
cally average distance-dependent calibration
values s(D). They were needed in order to com- This implies that an amplitude of 1 mm on the
pensate for distances up to 600 km for the decay WA seismogram corresponds to a magnitude
of the reference amplitude Ao with distance from 3 earthquake and this provides the anchoring of
the seismic source. It holds s(D) = logAo. Tab- the scale.
ulated values for s(D) are given in Fig. 1. Adding Figure 1 illustrates the determination of the
them to the logarithm of the maximum trace local magnitude ML according to this procedure.
amplitude B(WA) of a seismic event recorded Since then several other magnitude scales
by a WA seismograph (see Fig. 1) makes the have been developed, pioneered with three pub-
event amplitudes measured at different distances lications by Beno Gutenberg (1945a, b, c). He
comparable and yields the event magnitude. was interested in the use of the magnitude con-
Moreover, the corrected amplitudes for different cept for classifying the size of earthquakes world-
events are then indicative for the differences in wide, based on body-wave and surface-wave
event size. Accordingly, the local magnitude ML measurements also at teleseismic distances.
has been defined as These magnitudes were later termed mB and MS
Earthquake Magnitude Estimation 679
by Gutenberg and Richter (1956) who also In view of magnitudes being logarithmic num-
published in this paper calibration values for bers this corresponds to a difference in estimated
three types of body waves: (1) in propagation short-period seismic energy release of this event
direction oscillating pressure waves P, (2) PP by a factor of almost 2,000 times. Moreover, even
waves which are P waves that have been reflected the largest of these mb values, measured around
back on the Earth surface at half way distance 2 s period for a rupture with an estimated duration
between the source and the seismic station and of 500–600 s, saturated already at 7.5. This is
(3) S waves, which are slower travelling shear almost two magnitude units less than the much
waves of larger amplitudes (see Fig. 1) that oscil- later available largest moment magnitude value
late perpendicular to the direction of wave prop- (see section Earthquake Rupture, Radiated Seis-
E
agation. Yet, already Gutenberg and Richter mic Spectrum, Seismic Moment, Energy and
realized that their original dream of a unique Related Magnitudes) of Mw = 9.3 and explains
size scaling of earthquakes was not achievable the initial underestimation of the size and related
with magnitude scales based on different types of hazard of this earthquake. These differences
seismic waves, measured at different source dis- become less for smaller earthquakes but remain
tances and frequencies. Although they mutually nevertheless often still unacceptably large
scaled ML, mB, and MS to agree at values around (Granville et al. 2005; Bormann et al. 2007).
6.5 these magnitudes differ elsewhere, up to They are a possibly source of bias or large data
about one magnitude unit (m.u.) (Fig. 15). This scatter when such data, if they enter catalogs, are
problem became even more obvious when seis- used in estimates of magnitude-dependent
mographs operating at much wider, or narrower, ▶ Earthquake Recurrence frequency, or of Deter-
ranges of frequency with much higher amplifica- ministic Seismic Hazard Analysis or ▶ Probabi-
tion of ground motion than available at listic Seismic Hazard Analysis in terms of
Gutenberg’s and Richter’s time, made it possible ▶ Seismic Hazard: Ground Shaking.
to measure magnitudes up to 9.5 and down to less The confusion that is sometimes associated
than zero. Magnitudes measured at periods much with apparently well-established common prac-
less than the duration of earthquake rupture or of tise of magnitude determination has highlighted
another seismic waves generating process were the urgent need for:
then found to systematically underestimate the
overall earthquake size and energy release and (a) Internationally agreed standards of measure-
thus also its magnitude. This effect is often ment procedures for the most widely used
referred to as “magnitude saturation.” The con- types of magnitudes and
trolling parameters are outlined in section Earth- (b) Wider published and easily accessible critical
quake Rupture, Radiated Seismic Spectrum, reviews on the physical basis of the diverse
Seismic Moment, Energy and Related approaches to magnitude measurements,
Magnitudes. their potentials, biases, and mutual
It often occurs that different agencies publish relationship.
magnitudes with the same nomenclature although
they had been measured with different filter pro- The former issue has been taken up by an
cedures within different measurement time win- international Working Group on Magnitude Mea-
dows and corrected with different calibration surement of the Commission on Seismological
functions. Thus it could happen that the first Observation and Interpretation (CoSOI) of the
available values of the most widely measured International Association of Seismology and
short-period teleseismic mb magnitude (see sec- Physics of the Earth’s Interior (IASPEI). It
tion Body-Wave Magnitudes) for the great dev- resulted in the adoption of IASPEI (2013) reso-
astating Sumatra-Andaman tsunami generating lution No. 1 (Fig. 2).
earthquake of December 26, 2004, ranged The second issue has been elaborated in detail
between 5.7 and 7.5 (Bormann et al. 2007, 2009). in Chap. 3 of the second, now electronic edition
680 Earthquake Magnitude Estimation
Earthquake Magnitude Estimation, Fig. 2 IASPEI resolution 2013 on magnitude measurement standards
(# IASPEI)
its shear modulus m, also termed rigidity. The where D = average slip over the fault and A =
frictional conditions, m and Ds also control the rupture area. All these values, with the exception
velocity or speed of the fault rupture and thus the of the rigidity m, can be estimated via model
rupture duration, which varies between the assumptions from measured spectral displace-
smallest and the largest earthquakes between ment plateau amplitudes and the corner fre-
about 1/100th of a millisecond to several minutes quency fc of the source spectrum (Baumbach
for rupture lengths between centimeters and more and Bormann 2011). Therefore, it has been suit-
than 1,000 km with related magnitudes between able to scale the displacement spectra in Figs. 3
about 2 to 9.5. An approximate relationship and 4 to seismic moment and the related velocity
between non-saturated magnitude M and average spectra to the moment rate, i.e., the time deriva-
E
rupture duration trup has been given by Bormann tive of M0.
et al. (2009): Fig. 3 shows this for an average shear rupture
model of equal stress drop for events of different
logtrup 0:6M 2:6: (1) size and Fig. 4 for earthquakes of equal size in
terms of seismic moment but different in stress
Thus, earthquake ruptures of magnitude M = 4, drop. For model details see Bormann
5, 6, 7, 8, and 9 would roughly last on average et al. (2009), but the fundamentals can be under-
0.4 s, 1.6 s, 6 s, 25 s, 100 s and 400 s, i.e., trup stood with reference to these figures without
increases by about by a factor of four per magni- knowing these details.
tude unit. However, depending on the rigidity of M0 increases linearly with the product D A,
the material in the source volume and the related also termed earthquake potency, which is a static
stress conditions the rupture velocity vrup and thus measure of the overall earthquake “size,” But the
trup of individual earthquakes may deviate from potency does not account for differences in rup-
this average by a factor of about 2–3 shorter or ture velocity or stress drop and thus related dif-
longer (see Fig. 5). ferences in the radiated frequency content and
As shown in section Magnitude Scales and the amount of seismic energy for a given M0. None
New IASPEI Measurement Standards the size the less, this lead Kanamori (1977) to propose,
estimates of different magnitude scales may dif- via relationship (2) and by assuming constant
fer for the same earthquake. These differences average values of m and Ds and thus of the ratio
(or none) usually depend on the magnitude itself. Es/M0, a non-saturating “w = work” or moment
They can only be understood when magnitude scale, termed Mw. It is most consis-
considering – besides possible differences in the tently determined by the recommended IASPEI
time windows within which magnitude is mea- (2013) standard way of writing the original
sured on the seismic record – how the measure- Hanks and Kanamori (1979) formula, and reads
ment parameters amplitude and period, used by
these scales, differ in relation to the radiated Mw ¼ ðlogM0 9:1Þ=1:5: (3)
seismic source spectrum.
Source spectra observed in the far field, i.e., M0 and thus Mw will not saturate with growing
several wavelengths away from the rupture, rupture size provided that they are measured via
depend on the rupture size, Ds, the related vrup the plateau amplitude uo of the radiated displace-
and the ratio between released seismic wave ment source spectrum. This requires, however, to
energy ES and seismic moment M0. The scalar measure uo at frequencies much smaller than the
seismic moment M0 of a shear rupture is propor- corner frequency fc, i.e., at wavelengths larger
tional to the constant displacement plateau ampli- than the length of the source rupture.
tude uo of the radiated source spectrum and M0 calculations are nowadays routine at major
defined as seismological centers, either by measuring the
displacement amplitude u0 on the long-period
M0 ¼ mD A (2) asymptote (plateau) of the source spectrum,
682 Earthquake Magnitude Estimation
Ms(BB) Ms(BB)
23
10 1023
Mw = 9.0
mB(BB) mB(BB)
1022 1022
Ms(20) Ms(20)
Mw = 8.0
= 9.0
1021 1021
Mw
mb mb
w
M
1019 1019
0
Mw = 6.0 7.
=
1018 1018 w
M
1017 1017 0
Mw = 5.0 6.
=
w
M
1016 1016
0
Mw = 4.0 5.
= fc
1015
1015 w
M
1014 1014 0
4.
fc =
w
M
1013 1013
10–3 10–2 10–1 100 101 102 10–3 10–2 10–1 100 101 102
Frequency f [Hz] Frequency f [Hz]
Earthquake Magnitude Estimation, Fig. 3 Theoretical New IASPEI Measurement Standards). The frequencies at
source spectra of radiated displacement (left) and velocity which the plateau amplitudes, respectively M0 have
amplitudes (right) scaled to seismic moment M0, respec- dropped to about 70 % of their plateau values are termed
tively seismic moment rate dM0/dt in relation to the period corner frequencies fc (Modified Fig. 1 from Bormann
ranges within which the IASPEI (2013) standard magni- et al. 2009, BSSA # Seismological Society of America)
tudes are measured (see section Magnitude Scales and the
which is proportional to M0 after correcting for values. For example, for the great 2004
the source radiation and wave propagation effects Sumatra-Andaman Mw 9.3 earthquake the
(as done in Fig. 3), or, as a spin-off of seismic reported values for Mw differed up to about one
moment tensor calculations (see ▶ Moment Ten- magnitude unit (m.u.). The most important pro-
sors: Decomposition and Visualization), by cedure, elaborated by Dziewonski et al. (1981),
fitting long-period synthetic seismograms for dif- forms, with some later modifications, the compu-
ferent source models best to long-period filtered tational basis for the authoritative Global Cen-
real seismic records, using wavelengths that are troid Moment Tensor Catalog. Values for all
larger than the rupture length. But differences in major earthquakes with Mw > 5.5 are regularly
procedures are the reason that not all yield fully published by the GCMT project (http://www.
compatible and non-saturating M0 and Mw globalcmt.org/CMTsearch.html) and available
Earthquake Magnitude Estimation 683
a b
20
10 1020
1018 1018
Moment (M0) [Nm]
5
6.
w
=
M
1017 1017
E
1016 1016
fe [Hz]
−1
10
0.1 M
1 MPa 10−2 Pa
1014 1014 10
−3
0.1 MPa
6 7 8
Mw
1013 1013
10−3 10−2 10−1 100 101 102 10−3 10−2 10−1 100 101 102
Frequency (f) [Hz] Frequency (f) [Hz]
Earthquake Magnitude Estimation, Fig. 4 Source (in MPa) (Copy of Fig. 2 from Bormann and Di Giacomo
spectra for the same model as in Fig. 10 but for earth- 2011 # Springer; with kind permission from Springer
quakes of the same size yet different in stress drop Science + Business Media)
within about 2 h after the event. Less frequently frequency much lower and drop strongly with
and incomplete the GCMT catalog now also frequencies higher than fc. As in optics, lower
reports Mw values down to about 4.5. Regional frequencies with wavelengths larger than the
centers, such as Swiss Earthquake Service, object, in our case the causing rupture, cannot
provide – with larger time delay –reliable resolve any smaller detail of the rupture. They
moment tensor and Mw solutions also down to “see” the earthquake only as an undifferentiated
significantly smaller values, e.g., when dense “point source,” integrating over the whole rup-
quality networks close to the source are available, ture length. Accordingly, there is no difference in
down to Mw 3.0 (Braunmiller et al. 2002). the respective displacement amplitudes at fre-
A new method for rapid non-saturating quencies f fc. In contrast, higher frequencies
moment tensor and Mw estimates is based on with wavelengths smaller than the rupture length
very long-period W-phase observations between are radiated by smaller rupture patches. Their
the P- and S-wave arrivals (Kanamori and Rivera generated displacement amplitudes are lower
2008 and Sect. 3.2.8.6 in Bormann et al. 2013). the smaller the size of these patches. This
First real-time implementations are operational at explains the rapid decay of amplitudes towards
the U.S. Geological Survey’s NEIC (Hayes higher frequencies. The length of rupture can be
et al. 2009) as well as at the GFZ Potsdam and estimated from fc via some model assumptions
the Indonesian Tsunami Early Warning System and corrections made about the seismic source
(InaTEWS) since 2011. and the wave propagation effects.
The question arises why the radiated displace- The most common source spectra model is the
ment amplitude spectra in Figs. 3 and 4 are flat at o2 model, in which the displacement
684 Earthquake Magnitude Estimation
3
30 μm/s
100 s
of high-frequency energy, is not the most suitable respect to their fault maturity and thus capability
magnitude for assessing the actual seismic hazard (Choy and Boatwright 2012). For more details on
associated with an earthquake, be it of its damage the relationship between Mw and Me see also
potential due to strong shaking or of its potential Choy and Boatwright (2012), Bormann and Di
to generate a tsunami. Here even the relationship Giacomo (2011) and Bormann et al. (2013).
between the classical long-period MS and mb, Similarly important it is to consider the differ-
respectively ML, may already be more telling ence between Mw and mB(BB) because the latter
than Mw alone. For example, ML, is measured at correlates very closely with Me (Fig. 6). Figure 5
frequencies of particular interest to earthquake shows that for a given Mw mB(BB) is inverse
engineers, because they are in the preferred vul- proportional to the rupture duration and thus pro-
nerability range of man-made structures. portional with rupture velocity and the related
Recent publications have shown that consid- stress drop, as does the ratio ES/M0 and thus Me.
ering the difference between Mw and Me is par- The orthogonal relationship and standard
ticularly suitable for a more realistic near real- deviation between Me and mB(BB) is:
time assessment of potential effects (e.g., Choy
and Kirby 2004: Di Giacomo et al. 2010a; Me ¼ 1:27 mB ðBBÞ 1:98 0:11: (6)
Bormann and Di Giacomo 2011). Generally
holds, suitable seismotectonic setting, site condi- From the classical magnitudes, MS relates best to
tions, and exposure of vulnerable objects pro- Mw, also due to the fact that the Mw relationship
vided, that Me Mw may be indicative for high has been derived via the log Es-M equation (20)
tsunamigenic potential and related risk, and vice in section Relationships Between Magnitude
versa, Me
Mw for high shaking damage poten- Scales and Released Seismic Energy Es. How-
tial. But the differential magnitude M = Mw – Me ever, Bormann et al. (2009) showed that the new
is also of interest for basic research, e.g., the IASPEI broadband standard MS(BB) correlates
classification of seismotectonic regions with on average better with Mw than MS(20), reducing
686 Earthquake Magnitude Estimation
Earthquake Magnitude
Estimation,
Fig. 6 Relationship
between automatically
determined values of
mB(BB) and Me for
earthquakes with different
rupture mechanism
according to procedures
developed at the GFZ
German Research Centre
for Geosciences by
Bormann and Saul (2008)
and Di Giacomo
et al. (2010a and b)
(Courtesy of D. Di
Giacomo and J. Saul
(2010))
Earthquake Magnitude
Estimation, Fig. 7 Data
and linear orthogonal
regressions between the
new IASPEI standard
MS(BB) and Mw(GCMT) as
well as the theoretical
regression lines for
MS(20) in red, blue, and
blue hatched according to
Kanamori and Anderson
(1975) (Copy of Fig. 3.38
from Bormann et al. 2013
# IASPEI)
to half the systematic underestimation of Mw by Magnitude Scales and the New IASPEI
Ms(20) for magnitudes less than about 6.7 Measurement Standards
(Fig. 7). The MS(20) bias had already been
predicted theoretically by Kanamori and Ander- Local Magnitude ML
son (1975) and empirically confirmed by The definition of ML and the traditional proce-
Ekström and Dziewonski (1988). dure of its determination have already been
Earthquake Magnitude Estimation 687
Earthquake Magnitude Estimation, Fig. 9 Calibration difference between ML according to the Richter scale
functions s(D) = logAo for ML determination in differ- and other regional scales may already amount to 0.5 m.u.
ent regions. Some of them extend well beyond the 600 km for equal measurement amplitudes Amax (Copy of Fig. 3.3
distance plotted. But even at much shorter distance, such in Bormann et al. 2013 # IASPEI)
as the 320 km chosen here for demonstration, the
Equations 9 and 10 are based on the MS cali- vertical and horizontal component records of
bration relationship proposed by Vaněk P and PP waves, respectively on horizontal com-
et al. (1962) which is IASPEI recommended stan- ponent S-wave records, were published by
dard since 1967. But in the new standards it is Gutenberg and Richter (1956) and the respective
applied to vertical motion measurements only magnitudes were termed mB. Figure 11 shows the
and in Eq. 10 the log10(Vmax/2p) term replaces Q-values for vertical component P waves as a
the log10(A/T)max term of the original. Equation 9 function of D and h. Regrettably, it is the only
is since the 1970s exclusive MS standard at the one still in wide routine use, although S-wave
US National Earthquake Information Center amplitude, period and magnitude data would be
(NEIC). However, most surface-wave amplitude of great complementary value for improved stud-
E
maxima occur at periods outside of the period ies of the physical properties of Earth matter and
range accepted by Ms(20) (Fig. 10; Vaněk seismic source mechanisms. And PP waves
et al. 1962; Bormann et al. 2009). Moreover, would allow to measure body-wave magnitude
using the standard MS calibration function exclu- still with good signal-to-noise ratio in the “core-
sively for amplitude readings around 20 s intro- shadow” range beyond 100 distance where due
duces in some ranges systematic distance and to the influence of the liquid outer Earth core
magnitude-dependent biases of up to 0.5 m.u. direct P and S waves are no longer recorded or
Therefore, the determination of MS(BB), most only recorded as weak long-period diffracted
easily measured on nowadays widely available waves.
velocity BB records, is preferable. When elaborating these scales, Gutenberg
Besides these standards there exist still other realized that body-wave magnitudes measured
surface-wave magnitudes. Bonner et al. (2006) at different sites are compatible and stable only
proposed a largely theory-based version, apply- when measuring the ratio (A/T)max in a wide
ing multi-bandpass filtering and frequency- period range instead of just measuring Amax or
dependent attenuation corrections. Yet, measuring Amax in a narrow frequency range.
according to Bormann et al. (2009), their sophis- Measuring (A/T)max corresponds to measuring
ticated MS(Vmax) procedure yields practically the Vmax/2p in the IASPEI standard formula (10) for
same results as the much simpler MS(BB) which MS(BB). Accordingly, his general body-wave
can be performed easily by any station analyst magnitude formula reads, with A measured in
(Bormann et al. 2009). Another one uses very mm:
long-period surface-wave amplitudes with
periods between 60 and 400 s, which penetrate mB ¼ log10 ðA=T Þmax þ QðD, hÞ: (11)
into the earth mantle. This “mantle magnitude”
Mm (Okal and Talandier 1989) is not prone by Gutenberg derived the Q-functions by analyzing
saturation and thus suitable also for tsunami early mostly medium- to long-period body-wave
warning (Weinstein and Okal 2005). The Com- amplitudes in the range 2 s < T < 30 s, There-
prehensive Test-Ban Treaty Organization fore he did not account for frequency-dependent
(CTBTO) applies another 20 s surface-wave cal- attenuation, which becomes significant for
ibration formula by Rezapour and Pearce (1998) P-waves with T < 4 s (see Sect. 2.5.4.2 in
which avoids the systematic distance-dependent Bormann et al. 2013). Further, Gutenberg mea-
biases of MS(20). sured always (A/T)max within the whole P-wave
train and not only within a fixed limited time
Body-Wave Magnitudes window after the P-wave onset. Yet, this was
Gutenberg (1945a, b) proposed teleseismic mag- not considered when introducing, in conjunction
nitude scales for body waves of type P, PP, and with the deployment of the WWSSN in the
S. This allowed him to assess also the seismic 1960s, another short-period P-wave magnitude
energy released by deep earthquakes. Calibration scale mb. However, as compared to mB, mb had
functions Q(D, h) for amplitude readings on both the clear advantage of enabling teleseismic
690 Earthquake Magnitude Estimation
Earthquake Magnitude Estimation, Fig. 10 Three- 1971 and therefore were not dominated by the more
component distance-period plots for (A/T)max surface narrow-band WWSSN records (Copy of Fig. 3.35 in
wave readings retrieved at the ISC from printed station/ Bormann et al. 2013 # IASPEI)
network bulletins for earthquakes that occurred before
Earthquake Magnitude Estimation 691
6.2 6.4
400 6.1
6.0 6.5
E
7.0 7.5 8.0
6.2
300
6.3
6.5
6.0
200
6.5 6.6 6.5
6.1
6.6
100
6.5 6.0 6.7 6.6
5.9 6.7 7.0 7.5
7.0 6.5 6.7 6.8 6.9 6.6
7.0 7.3
0
5° 10° 20° 30° 40° 50° 60° 70° 80° 90° 100° Δ
Earthquake Magnitude Estimation, Fig. 11 Calibra- two body-wave magnitudes mb and mB (Figure redrawn
tion values Q(D, h) for vertical (Z) component P-wave from Gutenberg and Richter (1956). For standard tabu-
amplitudes as a function of epicentral distance D and lated QPZ values see Table 2 in Bormann and Dewey
source depth h. They are used for the calculation of the (2012))
magnitudes measured down to about magnitude 4, to the first few cycles within about 6 s after the
and thus to classify the many more small earth- onset of P. It is still used by several agencies such
quakes now being recordable with the much more as the International Data Center (IDC) of the
sensitive instruments now available than those CTBTO or the China Earthquake Network Center
available at Gutenberg’s time. (CENC), despite the 1978 IASPEI recommenda-
But mb still uses the same formula and Q(D, h) tion that for small and medium size earthquakes
values for medium-period P waves as in (5) and Amax should be measured within 20 s after the first
measures Amax only on narrow-band WWSSN-SP onset and that this time window be extended up to
(see Fig. 3) or similarly filtered short-period 60 s for very large earthquakes. Even a simple
records at periods <3 s (mostly around 1 s). relationship between rupture duration and mag-
This practice became standard in all western nitude, such as Eq. 1, makes it obvious that mb
countries, which discontinued to measure mB, measurements within the very first few seconds
although mb heavily underestimates the body- necessarily tend on average to underestimate the
wave magnitude for large earthquakes (see Intro- size of events with M > 6, because it becomes
duction). This is mainly due to the earlier-cited increasingly unlikely that they are based on the
phenomenon of spectral saturation (see sec- true Amax radiated by the considered rupture. But
tion Relationships Between Magnitude Scales also the extension of the measurement time win-
and Released Seismic Energy Es). This effect of dow up to 60 s will not always be sufficient for
underestimating earthquake size has been further really great earthquakes that rupture for several
aggravated by the often too short time window for minutes. The new standards for body-wave mag-
measuring Amax after the P-wave onset. In early nitudes, measuring always the true Amax within
years of the WWSSN, this window had been fixed the whole P-wave train, avoid this measurement
692 Earthquake Magnitude Estimation
Earthquake Magnitude Estimation, Fig. 12 Difference top trace is depicted in order to make the measurement
between standard mB(BB) and mb. Plotted are the vertical- units directly compatible with the bottom trace. On the
component P-wave records of an Mw8.4 Peru earthquake bottom trace Vmax is measured, yielding mB(BB) = 8.2, in
at station MTE, Portugal (D = 83.4 ). On the top trace, mb contrast to standard mb = 7.0. If, however, Amax is mea-
is determined by measuring Amax and the related period sured on the top trace within the first 5 s then one would
T from the WWSSN short-period recording that has been get mb = 6.4 only (Copy of Fig. 1 in Bormann and Saul
simulated from the original BB velocity trace on the 2008, SRL # Seismological Society of America)
bottom. On the second trace the time derivative of the
time-window component of saturation. For the Taking the results of several related studies
great Mw 9.3 Sumatra-Andaman earthquake of into account, IASPEI (2005; latest version
December 26, 2004, the smallest reported mb was 2013) recommends two complementary body-
5.7, the new standard mb yielded 7.5. But what wave magnitude standards, namely, a short-
also standard mb cannot avoid is the spectral period version mb, or mb in the ISF format:
component of saturation which is so obvious
when comparing the lower two traces and related mb ¼ log10 ðA=T Þ þ QðD, hÞ 3:0 (12)
velocity amplitudes of 8 mm/s and 123 mm/s,
respectively, in Fig. 12. This spectral component where A = P-wave ground amplitude in nm, cal-
of saturation is strongly reduced by mB(BB) that culated from the maximum trace-amplitude in the
is measured for large earthquakes at significantly entire P-phase train (time spanned by P, the depth
larger periods (Figs. 3 and 12). phases pP, sP, and possibly also the core
Earthquake Magnitude Estimation 693
Earthquake Magnitude
Estimation,
Fig. 13 Comparison
between the original
Gutenberg and Richter
(1956) QPZ values for
vertical component
P waves from earthquakes
at depth <70 km with
preliminary data based on
broadband velocity P-wave
amplitude measurements
by Saul and Bormann E
(2007). The Saul and
Bormann curve is
associated with curves
representing +/ the
standard error, which is also
plotted separately at the top
of the figure (Copy of
Fig. 16 in Bormann and
Dewey 2012 # IASPEI)
reflection PcP and their codas, and ending pref- energy magnitude, Mw and Me (see section Earth-
erably before PP); T = period of A in seconds quake Rupture, Radiated Seismic Spectrum,
with T < 3 s, D = epicentral distance in the Seismic Moment, Energy and Related Magni-
range 20 D 100 and h = focal depth tudes; Bormann and Saul 2008; 2009a; Bormann
between 0 km < h 700 km. Both T and A are et al. 2013), it would be very beneficial to mea-
measured on output from a vertical-component sure mB(BB) already from distances 5 onwards,
instrument/filter that simulates a WWSSN-SP e.g., in tsunami early warning schemes. However,
seismograph (Fig. 8). Although the early satura- the original Gutenberg and Richter QPZ values
tion of mb has been reduced by this new defini- are not reliable enough in the regional range.
tion up to about one magnitude unit, the short- A first version of an improved regional calibra-
period body-wave magnitude still needs to be tion relationship has been developed for
complemented – at least for mb > 5.5 – by a mB(BB) (Fig. 13) and is used already operation-
broadband mB(BB), or mB_BB in ISF format: ally in the German-Indonesian Tsunami Early
Warning System (GITEWS).
mB_BB ¼ log10 ðV max =2pÞ þ QðD, hÞ 3:0: Besides these two body-wave standards there
(13) exist several other broadband P-wave magnitudes
by various authors which aim at fast estimates of a
with Vmax = ground velocity in nm/s associated non-saturating moment magnitude Mw (see Sects.
with the maximum trace-amplitude in the entire 3.2.7 and 3.2.8 in Bormann et al. 2013). The most
P-phase train (as for standard mb), recorded on a successful ones account in different ways also for
vertical-component seismogram that is propor- the rupture duration. One example is given for the
tional to velocity at least in the period range 0.2 s great Mw 9.3 Sumatra-Andaman earthquake of
< T < 30 s; Q(D, h). The ranges for D and h are 2004 in Fig. 14, analyzed by the fully automated
the same as for standard mb. As seen from Fig. 11, cumulative body-wave magnitude mBc proce-
Gutenberg and Richter (1956) had given calibra- dure (Bormann and Saul 2009b). Its principle
tion values down to D = 5 . Since mB(BB) has was proposed by Bormann and Khalturin
proven to be a very suitable earliest available (1975). The justification for the new scale was
proxy estimator for both seismic moment and that Gutenberg and Richter (1956) assumed for
694 Earthquake Magnitude Estimation
Earthquake Magnitude Estimation, Fig. 14 Plot that drop to 40 % of its maximum values, i.e., some 16 % of the
illustrates the measurement of the cumulative body-wave released high-frequency seismic energy. Blue columns
magnitude mBc for the great Sumatra-Andaman earth- give the mB values related to subsequent new energy
quake with a rupture duration of about 540 s. The strong arrivals radiated from sub-rupture events; the red curve
P-wave radiation continues well into the time after the first shows the development of mBc with time by summing up
theoretical PP onset. The rupture duration D is estimated significant sub-event amplitudes until saturation is
from the envelope of high-frequency radiation on a achieved towards the end of the rupture (Cut-out of
1–4 Hz filtered record channel (gray shaded area). D is Fig. 1 from Bormann and Saul 2009b, SRL # granted
assumed to correspond approximately to the time window by Seismological Society of America)
from the first P-wave onset until the envelop amplitude
the mB determination and the relationship derived and/or automatically triggered risk mitigation
between mB and ES that the P-wave train has on actions taken between the detection of a strong
average a simple symmetric waveform radiated earthquake at the close-up stations and the arrival
by a “point source” (see section Earthquake Rup- of strong ground motion at locations slightly fur-
ture, Radiated Seismic Spectrum, Seismic ther away. EEWS aim at minimizing the area of
Moment, Energy and Related Magnitudes). so-called “blind zones” which are left without
They explicitly stated that they were not sure advanced warning before the arrival of the
whether this still holds for very great earthquakes S waves which have usually the largest strong-
for which they had not enough data. But great motion amplitudes. This requires very dense and
earthquakes rupture in a more or less complicated robust seismic sensor networks within a few tens
sequence of sub-events. The velocity amplitudes of km from potentially strong earthquake
for mB should therefore be measured individually sources. Such dense networks are at present
for major sub-events and summed up toward the already available in several regions, e.g., in Cal-
end of the rupture process for getting a cumula- ifornia, Japan, Taiwan, Turkey (Istanbul), and
tive mB estimate, termed mBc. Italy (see ▶ Source Characterization for Earth-
quake Early Warning).
Earthquake Early Warning Magnitudes The principles of rapid magnitude estimates
In recent years great attention has been paid to the are rather different from those mentioned above
development of ▶ Earthquake Early Warning and largely based on much debated concepts such
Systems (EEWS). They allow – besides very as the hypothesis of the deterministic nature of
quick event location – near real-time magnitude earthquake rupturing. For example, Olson and
estimates from the very first few seconds of accel- Allen (2005) claim to be able to estimate, from
eration, velocity or displacement records at sta- the maximum period of the first arriving P waves
tions close to the earthquake source. Based on within the first 4 s, the size of earthquakes up to
such data, rapid public alarms may be issued magnitude 8 with an average absolute deviation
Earthquake Magnitude Estimation 695
of 0.54 m.u. using many low-pass filtered veloc- within the first three seconds of the record. These
ity records within 100 km epicentral distance. Pd magnitudes agreed with catalog magnitudes
However, Rydelek and Horiuchi (2006) could with a standard deviation of 0.18 m.u. for
not confirm that such a dominant frequency scal- events with magnitudes <6.5 but saturated for
ing with magnitude exists. Also extensive exper- larger ones and could not be used to estimate
iments run at the Taiwan EEWS, aimed at magnitudes 6.7 (Wu and Zhao 2006). For a
reducing the reporting time from 25 s down to summary of related studies and information on
about 8 s by using only the records of the first several representative EEWS, their magnitude
8 stations within less than 21 km epicentral dis- concepts and performance see Sect. 3.2.6.5 in
tance (Wu and Kanamori 2005) came to sober- Bormann et al. (2013).
E
ing results. Estimating the magnitude via the
magnitude dependence of the average period tc
of the P-wave, as well as the peak displacement Relationships Between Magnitude
amplitudes (if > 0.1 cm) within the first 3 s of Scales and Released Seismic Energy Es
the low-pass filtered displacement signal after
the P-wave onset the authors concluded: The local and teleseismic magnitudes discussed
above are not identical (see Fig. 15). They only
“. . .For EWS applications, if tc < 1 sec, the event
has already ended or is not likely to grow beyond agree at some specific magnitude values and
M > 6. If tc > 1 sec, it is likely to grow, but how differ at others. Moreover, all classical magni-
large it will eventually become, cannot be deter- tude scales show for large or very great earth-
mined. In this sense, the method provides a thresh- quakes more or less pronounced
old warning.”And: “. . .If tc > 1 sec, the event is
potentially damaging. If it is larger than 2 sec, the underestimation of earthquake size as compared
event is almost certainly damaging. Combining this to a non-saturating purely displacement
information with other data, such as the initial amplitude-based magnitude such as the moment
velocity and displacement amplitudes, would magnitude Mw. But underestimation of the
allow the damage potential of the event to be
assessed more accurately”. moment related earthquake size may also occur
if magnitudes that are based on the measurement
More can indeed not be said within the first of A/T, such as mb and MS(20), are measured
3–4 s of a record, especially not for strong and outside of the period range of maximum seismic
great earthquakes. This fully agrees with rela- energy release. This is obvious from the
tionship (8) between average rupture duration nonlinear relationships of these classical magni-
trup and M. For earthquakes with M 6 and tudes with Mw (Fig. 15). The reasons have been
trupp 6 s it can indeed be expected that, explained in section Earthquake Rupture, Radi-
within a measurement time window of 3–4 s, ated Seismic Spectrum, Seismic Moment,
the maximum amplitude and related period of Energy and Related Magnitudes. This notwith-
the P-wave train can be measured and are repre- standing, most classical relationships between
sentative for the earthquake magnitude but not different types of magnitudes have been
for magnitudes much larger than 6 and rupture published only as simple linear standard regres-
durations of several 10 to several 100 s. Then, sion relationships that maybe rather biased by
inevitably, as for the current near real-time mag- noisy data because they assume that only one of
nitude estimates in tsunami early warning sys- the compared magnitudes is afflicted with mea-
tems (TEWS), much larger measurement time surement errors. Only recently has it become
windows up to about one minute (e.g., Lomax common to calculate general orthogonal linear
and Michelini 2012) or even longer for mega- and even nonlinear regression relationships. For
earthquakes may be required. This general con- a summary on this issue see Sect. 3.2.9 in
clusion is fully confirmed by results from South- Bormann et al. (2013). Only the most important
ern California where the magnitude has been classical and a few recently derived relation-
estimated by measuring the P-wave amplitudes ships of interest are presented here.
696 Earthquake Magnitude Estimation
Earthquake Magnitude Estimation, Fig. 15 Average Meteorological Agency (Modified Fig. 1 of Utsu (2002).
relationships of different common types of magnitude Reproduction of Fig. 4 in Bormann and Saul 2009a #
scales with respect to the non-saturating moment magni- Springer; with kind permission from Springer Science +
tude Mw. MJMA is determined by the Japanese Business Media)
According to Gutenberg and Richter (1956) has become common to convert classical magni-
tudes via empirical conversion relationships into
mB ¼ 0:63 MS þ 2:5: (14) Mw which is often considered as the physically
best defined and with smallest measurement
This relationship has been very well reproduced errors of about 0.1 m.u. afflicted magnitude.
between the new standard magnitudes Such a unification of magnitude estimates is
mB(BB) and MS(BB) (Bormann et al. 2009) in desirable especially for seismic hazard assess-
an even wider magnitude range than that investi- ment (see ▶ Probabilistic Seismic Hazard
gated by Gutenberg and Richter (1956). But Models). The latter strongly depends on the
Eq. 14 differs strongly from the relationship parameters a and b derived from a Gutenberg
between short-period mb and MS(20) (Gordon and Richter (1954) relationship logN = a + bM
1971): between magnitude M and the number N of earth-
quakes that occur on average in a given period of
mb ¼ 0:47MS ð20Þ þ 2:79: (15) time, usually normalized as of per year (see
▶ Earthquake Recurrence). However, reliable
The relationship between the two new IASPEI instrumentally measured Mw data as in the Global
standard magnitudes for body waves and surface- Centroid Moment Tensor catalog (http://www.
waves, respectively, are depicted in Fig. 16 and globalcmt.org) are available and complete only
their formulas are given in the figure caption. since 1975 and for Mw 5.5. This time span is
Regression relationships between magnitudes too short for reliable long-term seismic hazard
are frequently used to convert data of one type assessment. To include earlier events into the
of magnitude into another one. For example, logN-Mw analysis one has to try the conversion
Gutenberg used – in an effort to unify different of classical MS, or of the most frequently deter-
magnitude estimates per event into a single mined mb, into Mw. The rather good correlation
number – relationship (9) for converting MS between MS and Mw for larger earthquakes is no
data into mB which he considered superior wonder because the Mw formula (3) has been
to MS, better physically defined, applicable also scaled by Kanamori (1977) and Hanks and
to deep earthquakes and less affected by uncer- Kanamori (1979) via the log ES – MS relationship
tainties in estimates of source depth. Nowadays it (20) to MS. The most recent, both linear and
Earthquake Magnitude Estimation 697
Earthquake Magnitude Estimation, Fig. 16 Relation- regression relations are mb = 0.91 mB(BB) + 0.20 and
ships between the two new IASPEI standard body-wave MS(20) = 1.10 MS(BB) – 0.75 (Cut-out of Fig. 11 in
magnitudes (left panel) and surface-wave magnitudes Bormann et al. 2009, BSSA # Seismological Society of
Ms(20) and Ms(BB) (right panel). The orthogonal America)
Earthquake Magnitude Estimation, Fig. 17 Training black polygon line represents the median values for sepa-
data sets for deriving the ISC-GEM MS – Mw (left) and rate magnitude bins (Figure modified and compiled
mb – Mw (right) general orthogonal linear (green) and according to Figs. 14 and 17 of Di Giacomo et al. 2014
exponential (blue) conversion relationships. The dashed # PEPI)
exponential regression lines. The related equa- logEs ¼ 1:5 MS þ 4:8 (20)
tions for Mw MS are given in (16–18),
One realizes, when compared with MS Mw, with Es in units of Joule and Ms meaning an
the much larger data scatter and thus much larger MS(20). However, 20 s displacement amplitudes
uncertainty of estimating Mw via mb. Therefore, are not very representative for the radiated
the latter conversion should be avoided whenever seismic energy, especially not for smaller earth-
MS data are available. But even such proxy Mw quakes that radiate dominatingly higher-
estimates via MS are afflicted with at least twice frequency energy.
the error of direct high quality instrumental Mw Nevertheless, Eq. 20 has become since the
estimates. For most practical applications, how- 1960s the most widely used rough estimator for
ever, it may be still acceptable to use for MS seismic energy release via magnitude and it was
conversion into Mw instead of the more accurate instrumental for deriving the non-saturating
exponential relationship moment magnitude scale Mw (see section Earth-
quake Rupture, Radiated Seismic Spectrum,
Mw ¼ eð0:222þ0:233Ms Þ þ 2:863 (16) Seismic Moment, Energy and Related Magni-
tudes). Moreover, it has been modified by Choy
and Boatwright (1995) when correlating direct Es
the two orthogonal linear relationships
measurements with NEIC MS(20) data and
assuming the same slope as in (20) [see relation-
Mw ¼ 0:67Ms þ 2:13 for MS 6:47 (17)
ship (22)].
Another relationship has been derived by
and Kanamori et al. (1993) between ML and Es for
Southern California, valid in the range
Mw ¼ 1:10Ms 0:67 for MS > 6:47 (18) 1.5 < ML < 6.0:
Earthquake Magnitude Estimation, Fig. 18 Orthogo- respectively Ms(20) data of the NEIC (right-hand panel).
nal (OR) and standard regressions (SR and its inverse ISR) GR Gutenberg and Richter (1956) relationship, C&B
of logEs data for different earthquake source mechanisms, Choy and Boatwright (1995) relationship (Compiled
determined by the procedure of Di Giacomo et al. (2008 from Figs. 3.85 and 3.86 in Bormann et al. 2013 #
and 2010b) plotted over mB(BB) data (left-hand panel), IASPEI)
Choy and Boatwright 2012; Di Giacomo and of the potential strength of earthquake shaking to
Bormann 2011). Regressing such modern energy be expected. Moreover, in the long run, they also
measurements with different magnitudes it could permit to quantify the seismic activity and its
be shown that indeed standard mB(BB) correlates, variability in space and time, which is a prereq-
as expected by Gutenberg, much better with log uisite for seismic hazard assessment. This neces-
Es than MS(20). This allows rather reliable sitates, however, the compatibility and long-term
energy estimates via mB(BB). But the classical stability of magnitude estimates of the various
relationship (19) could not be reproduced well types on a global scale. The essay shows that
with modern Es data. On the other hand, although just this is often not the case. Procedural differ-
with double data scatter than with mB(BB), the ences in magnitude determination, even of sup-
Choy and Boatwright (1995) relationship posedly the same type, may occasionally reach
even more than one magnitude unit, especially
logEs ¼ 1:5Msð20Þ þ 4:4 (22) for the most seldom but therefore most important
and devastating great earthquakes. In view of the
could be well reproduced by orthogonal regres- logarithmic scaling of magnitude values this may
sion (Fig. 18). correspond to wrong estimates of released seis-
mic energy by a factor of 1,000 times or even
more. This highlights the urgency of international
Summary standards in magnitude measurements. For the
most important classical magnitudes such stan-
Magnitudes are one of the most important earth- dards have recently been agreed upon by IASPEI
quake parameter data. When properly determined and recommended for global operational applica-
and applied they may allow quick estimates of the tion. Their principles and underlying physical
tectonic effect of actual earthquakes in terms of considerations, advantages and limitations are
rupture area and slip and/or of the released seis- outlined in the essay, also in relation to other
mic energy in different frequency ranges and thus types of magnitudes such as modern Mw and Me
700 Earthquake Magnitude Estimation
as well as of their near real-time proxy estimates distances, part II: application and Ms-mb performance.
in the earthquake and tsunami early warning con- Bull Seismol Soc Am 96(2):678–696
Bormann P, Khalturin VI (1975) Relations between dif-
text. Also given are some of the most important ferent kinds of magnitude determinations and their
correlation relationships between different types regional variations. In: Proceedings XIVth general
of magnitudes but with cautioning remarks on assembly of the European seismological commission,
their often uncritical application in efforts to Trieste, pp 16–22 September 1974. Published by the
National komitee Geod. Geophys, AdW der DDR,
homogenize earthquake catalogs in terms of mag- Berlin, pp 27–39
nitude. In fact, there is no single type of magni- Bormann P, Saul J (2008) The new IASPEI standard
tude that can sufficiently well describe both the broadband magnitude mB. Seismol Res Lett 79:
geometric size and kinematic-dynamic effects of 698–705
Bormann P, Saul J (2009a) Earthquake magnitude. In:
earthquakes. What is needed is a deeper under- Meyers A (ed) Encyclopedia of complexity and sys-
standing and conscious use of the beneficial com- tems science, vol 3. Springer, Heidelberg/New York,
plementarity of different magnitude scales. For pp 2473–2496
comprehensive information on this complex Bormann P, Saul J (2009b) A fast, non-saturating magni-
tude estimator for great earthquakes. Seismol Res Lett
issue reference is made to the online IASPEI 80:808–816
New Manual of Seismological Observatory Prac- Bormann P, Di Giacomo D (2011) The moment magni-
tice (NMSOP-2; http://nmsop.gfz-potsdam.de). tude Mw and the energy magnitude Me: common roots
and differences. J Seismol 15:411–427
Bormann P, Dewey JW (2012) IS 3.3: The new
IASPEI standards for determining magnitudes from
Cross-References digital data and their relation to classical magnitudes.
In: Bormann P (ed) (2012) http://nmsop.gfz-potsdam.de
▶ Earthquake Location Bormann P, Liu R, Ren X, Gutdeutsch R, Kaiser D,
Castellaro S (2007) Chinese national network magni-
▶ Earthquake Mechanism and Seafloor tudes, their relation to NEIC magnitudes, and recom-
Deformation for Tsunami Generation mendations for new IASPEI magnitude standards. Bull
▶ Earthquake Mechanisms and Stress Field Seismol Soc Am 97:114–127
▶ Earthquake Mechanisms and Tectonics Bormann P, Liu R, Xu Z, Ren K, Zhang L, Wendt S (2009)
First application of the new IASPEI teleseismic mag-
▶ Earthquake Recurrence nitude standards to data of the China National Seismo-
▶ Moment Tensors: Decomposition and graphic Network. Bull Seismol Soc Am 99(3):
Visualization 1868–1891
▶ Probabilistic Seismic Hazard Models Bormann P, Engdahl ER, Kind R (2012) Chapter 2: Seis-
mic wave propagation and Earth models. In: Bormann
▶ Seismic Actions Due to Near-Fault Ground P (ed) (2012) http://nmsop.gfz-potsdam.de
Motion Bormann P, Di Giacomo D, Wendt S (2013) Chapter 3:
▶ Seismic Instrument Response, Correction for Seismic sources and source parameters In: Bormann
▶ Seismic Network and Data quality P (ed) (2012) http://nmsop.gfz-potsdam.de
Braunmiller J, Kradolfer U, Baer M, Giardini D (2002)
▶ Sensors, Calibration of Regional moment tensor determinations in the
▶ Site Response: Comparison Between Theory European-Mediterranian area – initial results.
and Observation Tectonophysics 356(1–3):5–22
Brune JN (1970) Tectonic stress and the spectra of shear
waves from earthquakes. J Geophys Res
75:4997–5009
References Choy GL, Boatwright J (1995) Global patterns of radiated
seismic energy and apparent stress. J Geophys Res
Baumbach M, Bormann P (2011) EX 3.4: Determination 100:18205–18228
of source parameters from seismic spectra. In: Choy GL, Kirby S (2004) Apparent stress, fault maturity
Bormann P (ed) (2012) http://nmsop.gfz-potsdam.de and seismic hazard for normal-fault earthquakes at
Bonner JL, Russel DR, Harkrider DG, Reiter DT, Herr- subduction zones. Geophys J Int 159:991–1012
mann RB (2006) Development of a time-domain, Choy GL, Boatwright J (2012) IS 3.6: Radiated seismic
variable-period surface-wave magnitude measurement energy and energy magnitude. In: Bormann P (ed)
procedure for application at regional and teleseismic (2012) http://nmsop.gfz-potsdam.de.
Earthquake Magnitude Estimation 701
Weinstein SA, Okal EA (2005) The mantle wave magni- of a given magnitude. This entry reviews past
tude Mm and the slowness parameter Y: Five years of methodologies and current developments of
real-time use in the context of tsunami warning. Bull
Seismol Soc Am 95:779–799 seismogenic tsunami generation models.
Wu YM, Kanamori H (2005) Experiment on an onsite Subduction zones are the tectonic plate bound-
early warning method for the Taiwan early warning aries where most seismogenic tsunamis are
system. Bull Seismol Soc Am 95(1):347–353 sourced. Three factors combine to explain this
Wu YM, Zhao L (2006) Magnitude estimation using the
first three seconds P-wave amplitude in earthquake observation: (1) most subduction zones occur
early warning. Geophys Res Lett 33. doi:10.1029/ beneath oceans, (2) most of the world’s largest
2006GL026871 earthquakes occur along subduction zones, (3) the
dominant mechanism of subduction zone earth-
quakes is thrusting, resulting in significant verti-
cal displacement of the seafloor. Normal-faulting
Earthquake Mechanism and Seafloor earthquakes in the outer rise and outer trench
Deformation for Tsunami slope of subduction zones and thrust faulting in
Generation the back arc can also generate tsunamis. Other
tectonic environments where seismogenic tsu-
Eric L. Geist1 and David D. Oglesby2 namis can occur are oceanic convergence bound-
1
U.S. Geological Survey, Menlo Park, CA, USA aries as distinguished from subduction zones by
2
Department of Earth Sciences, University of Bird (2003) and local, submarine dip-slip fault
California, Riverside, Riverside, CA, USA zones in intraplate settings.
To simulate the tsunami generation process,
elastic deformation models are needed to deter-
Synonyms mine the magnitude and pattern of seafloor dis-
placement from earthquake source parameters.
Seismogenic tsunami; Tsunami forcing; Tsunami To calculate the initial tsunami amplitude distri-
initial conditions; Tsunamigenic bution on the ocean surface, a 1/cosh(kh)
low-pass filter is applied to the seafloor displace-
ment field, where k is the wave number and h is
Introduction the water depth (Kajiura 1963). This filter pri-
marily has an effect for small earthquakes and
Tsunamis are generated in the ocean by rapidly surface-rupturing earthquakes that generate a
displacing the entire water column over a signif- scarp or short-wavelength displacement. Other-
icant area. The potential energy resulting from wise, the initial wave field of a tsunami is almost
this disturbance is balanced with the kinetic identical to the vertical seafloor displacement
energy of the waves during propagation. Only a field. Horizontal displacements also contribute
handful of submarine geologic phenomena can to tsunami generation in regions of steep bathym-
generate tsunamis: large-magnitude earthquakes, etry (Tanioka and Satake 1996). Although the
large landslides, and volcanic processes. Asteroid rupture process of large subduction zone earth-
and subaerial landslide impacts can generate tsu- quakes is extremely complex, simple mechanical
nami waves from above the water. Earthquakes models have traditionally been used to simulate
are by far the most common generator of tsu- seafloor displacement and tsunami generation up
namis. Generally, earthquakes greater than mag- until recently. These models assume that the
nitude (M) 6.5–7 can generate tsunamis if they earthquake occurs instantaneously relative to the
occur beneath an ocean and if they result in pre- phase speed of the tsunami, often referred to as
dominantly vertical displacement. One of the static or coseismic models.
greatest uncertainties in both deterministic and A significant advancement in simulating tsu-
probabilistic hazard assessments of tsunamis is nami generation is the use of dynamic rupture
computing seafloor deformation for earthquakes models. These models simulate the rupture process
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 703
Earthquake Mechanism
and Seafloor
Deformation for
Tsunami Generation,
Fig. 1 Perspective view
(NW) of vertical
displacement field
associated with the 2004
Mw 9.2 Sumatra-Andaman
earthquake, showing uplift
(red) near the trench and
subsidence (blue) toward
the Sumatran shore E
along the fault itself, as well as dynamic deforma- Sumatra-Andaman earthquake (Geist et al.
tion of the surrounding rocks, including dynamic 2007). In the near field along subduction margins,
stress transfer from seismic waves as well as the the field is characterized by negative amplitudes
final static seafloor displacement. The time- toward shore (i.e., toward the island of Sumatra in
varying displacement of the seafloor is extracted Fig. 1) and positive amplitudes toward the open
from these models and is used to force the tsunami ocean. The zero displacement line is often termed
in the source region. The time-varying aspect of the hinge line. In most cases, the hinge line is
the seafloor motion is not likely to be a large factor offshore, such that coastal subsidence occurs with
in tsunami generation, but as will be shown below, large-magnitude subduction zone earthquakes,
the final earthquake size, slip distribution, and and the first motion of the tsunami broadside
rupture path do depend critically on the spatiotem- from the source is negative (trough phase). Con-
poral details of the rupture process. versely, at locations in the open ocean, the first
First, the most common class of static mechan- motion is positive (peak phase). In some cases,
ical models based on dislocation theory is the hinge line is onshore, such that there is a
discussed. Static crack models are then described coastal uplift and no initial withdrawal of the
that have several advantages over traditional dis- ocean (e.g., along the Andaman Islands in
location models. General methods for incorporat- Fig. 1). In other cases where the rupture area is
ing heterogeneous slip, particularly from near the trench, there is no subsidence or uplift of
stochastic slip models, are next discussed. the coastal region (Geist 1999).
Finally, recent advances in using dynamic rup- Dislocation models were first used to model
ture models for tsunami generation are reviewed. tsunami generation in the 1970s (e.g., Abe 1973;
Because details among the different tsunami gen- Fukao and Furumoto 1975; Kanamori 1972). It is
eration models are smoothed out in the far field, a curious historical circumstance that crack
the focus in this entry is on near-field effects on models were already well established to address
the tsunami wave field. various geophysical problems by this time (e.g.,
Eshelby 1957), but were never widely used to
model tsunami generation. The main advantage
Dislocation Models of Tsunami of using uniform-slip dislocation models, termed
Generation Volterra dislocations (Bilby and Eshelby 1969),
is in reference to the mechanical definition of
Figure 1 shows an example of the coseismic seismic moment (m0)
vertical displacement field for a thrust fault:
the Sunda thrust during the 2004 Mw 9.2 m0 ¼ mds S;
704 Earthquake Mechanism and Seafloor Deformation for Tsunami Generation
a 0
b
100
x (km)
200 uz/δ
0.1
0.0 100
uz/δ –0.1
0.4
0.2 0 0 y (km)
100
0.0
x (km) 200 –100
0 300
100
y (km) 200
Earthquake Mechanism and Seafloor Deformation up-dip edge of dislocation is 19 km. Maximum normal-
for Tsunami Generation, Fig. 2 Comparison of verti- ized vertical displacements (uz/d) are 0.43 and 0.12 for (a)
cal displacement field from (a) thrust fault (dip=18 ) and and (b), respectively
(b) strike-slip fault (dip=90 ). In both cases, depth to
where m is the average shear modulus of rocks developed for a layered media (e.g., Ma and
surrounding the rupture and ds is the average static Kusznir 1994) and for spherical geometry (e.g.,
slip over the area S of the fault that ruptured during Pollitz 1996).
the earthquake. However, Madariaga (1979) The effect that individual earthquake source
shows that m0 can also be related to the static stress parameters have on seafloor displacement varies
drop for a crack. The area S of a rupture can be considerably. The amount of uniform slip
estimated from the aftershock distribution or assigned to a Volterra dislocation has the largest
scaled from earthquake magnitude (e.g., Blaser effect on the magnitude of seafloor displacement,
et al. 2010; Geller 1976). Although mean slip is with an approximately linear relationship
indicated in the definition of m0, this is often taken between slip and maximum vertical displacement
as the value for uniform slip in Volterra dislocation (Geist 1999). The dip of the fault primarily
models (i.e., duniform ¼ m0 =ðm SÞ). affects the polarity of the tsunami wave. With
Analytical expressions have been developed increasing dip of a thrust fault, the amplitude of
to calculate the coseismic displacement field the negative polarity phase decreases over the
u from dislocations with slip vector s
hanging wall and increases over the footwall.
u+i ui , where the plus and minus sign indicate Geist (1999) gives a complete description of the
opposite sides of the dislocation. Okada (1985) effect that these and other earthquake source
reviews the expressions derived through 1983 parameters have on seafloor displacement and
and derives compact expressions applicable to the associated near-field tsunami.
all fault types and non-Poisson solids. Most The effectiveness of strike-slip faults in gen-
importantly for tsunami generation, both finite- erating tsunamis relative to dip-slip faults has
fault and point-source representations are given recently been debated. Although the slip vector
for dip-slip faults, the latter useful for computing is parallel to the surface, there is vertical dis-
displacement from heterogeneous slip as placement associated with strike-slip faults,
described below. The finite-fault expressions occurring in a quadrupole pattern (see also
can be compared to those developed earlier by Fig. 2b). Using a dislocation model, the vertical
Savage and Hastie (1966) for dip-slip faults. Most displacement pattern for dip-slip and strike-slip
of the analytic expressions used for tsunami gen- faults is compared in Fig. 2 for the same amount
eration are applied to a homogeneous elastic half of slip and depth to the up-dip edge of the dislo-
space, although expressions have also been cation. For this case, the maximum vertical
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 705
cm
55
50
Homogeneous Heterogeneous 45
40
–122° 30′ –122° 20′ –122° 10′ –122° 30′ –122° 20′ –122° 10′ 35
30
25 D
20 I
15 S
10 P
5 L
0 A
47° 40′ 47° 40′ 47° 40′ –5
C
–10
–15 E
–20
–25
M
E
E
–30 N
–35 T
–40
47° 30′ 47° 30′ 47° 30′ –45
–50
–55
–122° 30′ –122° 20′ –122° 10′ –122° 30′ –122° 20′ –122° 10′
cm
10
Difference
8
–122° 30′ –122° 20′ –122° 10′
6 D
I
4 S
P
2 L
A
0
47° 40′ 47° 40′ C
–2 E
M
–4 E
N
–6 T
–8
47° 30′ 47° 30′
–10
Earthquake Mechanism and Seafloor Deformation rupture of Seattle fault through high shear modulus rocks.
for Tsunami Generation, Fig. 3 Effect of structural Dashed rectangle: up-dip rupture through low shear mod-
heterogeneity of elastic constants on vertical displacement ulus sedimentary rocks of the Seattle sedimentary basin
field in the Puget Sound region. Solid rectangle: down-dip
displacement associated with the strike-slip Masterlark 2003; Zhao et al. 2004). Static elastic
dislocation is approximately a quarter of that deformation models developed for the Nankai sub-
for the thrust dislocation. A similar comparison duction zone were adapted to calculate tsunami
is made using normal-mode theory by Ward generation along the Cascadia subduction zone
(1980) and Okal (1988), which is useful to (Geist and Yoshioka 1996). The development of
gauge the far-field response to the different thrust faulting in the presence of subducted sea-
mechanisms. mounts, also at the Nankai subduction zone, is
Rather than using an elastic half-space or examined using a finite element model by Baba
layered-earth model, realistic geologic structures et al. (2001). The effect of sedimentary basins on
and heterogeneous physical properties can be used static elastic displacements is shown for the Seattle
to calculate seafloor displacement, most often by basin in Fig. 3 (Geist and Yoshioka 2004). The
finite element methods (Geist and Yoshioka 1996; effect of rupturing into lower-than-average shear
706 Earthquake Mechanism and Seafloor Deformation for Tsunami Generation
Earthquake Mechanism
0.5
and Seafloor h=–1
Deformation for h=–3
0.4 h=–7
Tsunami Generation,
Fig. 4 Effect of decreasing h=–15
0.3
depth to up-dip edge of
rupture (h in kilometers) on
0.2
vertical displacement (uz)
profile (normalized with
uz/
respect to slip, d) for a 0.1
shallow thrust Volterra
dislocation 0
–50 –30 –10 10 30 50 70 90
–0.1 x (km)
h
–0.2
–0.3
Earthquake Mechanism
and Seafloor
Deformation for
Tsunami Generation,
Fig. 5 Coordinate system
and parameters for crack
model
modulus is to increase vertical displacement at the these and other internal singularities further.
surface (blue region, bottom of Fig. 3), consistent Application of the 1/cosh(kh) low-pass filter
with the results of Ma and Kusznir (1994) and will attenuate this artifact but will lead to a
Savage (1998). Finite element models are also general underestimation of tsunami amplitude.
critical for calculating dynamic displacements as This artifact can be avoided by using crack
will be described in the Dynamic Rupture Models models for tsunami generation as described in
section. the next section.
Finally, it is important to recognize a promi-
nent artifact associated with dislocation models
of shallow imbedded thrust faults, as shown in Crack Models of Tsunami Generation
Fig. 4. Although the expressions for surface dis-
placement developed by Okada (1985) do not Unlike dislocation models for tsunami generation
include the dislocation edge singularity for in which slip is prescribed, the slip distribution
faults that rupture to the surface, the short- associated with crack models is calculated and
wavelength effect of the singularity is evident physically consistent with the applied tectonic
for cases where the up-dip edge is near the sur- stresses and frictional conditions on the fault.
face, relative to the down-dip width of the rup- For dip-slip faults of geometry shown in Fig. 5,
ture. In a later study, Okada (1992) discusses the slip gradient in the dip direction d0 (x) is given
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 707
by the following singular integral equation subduction zone earthquakes and tsunami gener-
(Dmowska and Kostrov 1973): ation (e.g., Geist and Dmowska 1999; Wang and
He 2008). The slip gradient function prescribed
ðb ðb by Freund and Barnett (1976) is given by
1 d0 ðsÞ 1 2ð1 nÞ
ds þ K ðx, sÞd0 ðsÞds ¼ f ðxÞ;
p sx p m 8
> 12
a a >
< 3 qg g , g < q
2
q
d0 ¼ D ;
where l and m are Lame’s constants, n is > 12 g2 gð1 þ qÞ þ q, g > q
>
: 3
Poisson’s ratio, and f(x) represents the initial ð 1 qÞ
shear stress distribution. In terms of principal
E
stresses, f(x) is given by where D is a scaling coefficient, g
(x a)/(b
a), and q (0 < q < 1) controls the skewness of the
1 0 slip distribution. This function satisfies the
f ðxÞ
s0x ¼ s s0zz sin 2a þ s0xz cos 2a
2 xx smooth closure condition d0 (a) = d0 (b) = 0.
For the specific one-dimensional version of
with the zero superscript indicating initial stress. the fault geometry shown in Fig. 5, the displace-
The effect of decreased normal stress on a shal- ment profile associated with cracks can be calcu-
low imbedded crack makes it more likely to rup- lated analytically. As indicated by Freund and
ture to the surface above a critical depth as shown Barnett (1976), the displacement is calculated
by Rudnicki and Wu (1995). The Coulomb friction by integration based on the slip gradient
condition is introduced such that slip occurs when
|sx| = ks, where k is the apparent coefficient of ðb
friction, including effects of pore-pressure differ-
uz ¼ U z ðx, sÞd0x ðsÞds;
ences in the fault zone. The right-hand side of
a
integral equation above is modified such that
where
f k ðxÞ ¼ f ðxÞ ks0
Earthquake Mechanism
and Seafloor
Deformation for
Tsunami Generation,
Fig. 6 Comparison of
displacement profiles
between imbedded crack
(dashed line) and
dislocation (solid line)
models with same average
slip (uniform slip for the
dislocation model). See
Fig. 5 for fault geometry
results in more seafloor displacements over the by the location vector r(x, y) in response to slip on
center of the rupture when compared to the dis- a fault at location r0(x, y) (Fig. 8). The three
placement field from the dislocation. components of seafloor displacement in two
One can also compare a surface-rupturing dimensions can then be given by (Rybicki 1986)
crack with an equivalent very shallow imbedded
ðð
crack to show the effect of surface rupture
um ðrÞ ¼ Di ðr0 Þnj ðr0 ÞU ijm ðr, r0 ÞdS m ¼ 1, 2, 3;
(Fig. 7). The dip-directed slip profile radically
S
changes from a skewed half ellipse to a quarter
ellipse with twice the maximum slip when the
surface is ruptured (Fig. 7a). The stress drop for where n(r0) is the surface normal of a fault plane
the surface-rupturing case is lower under the tec- with arbitrary geometry and S is the area of the
tonic stress conditions given above, and the ver- fault plane. For an isotropic medium, the six
tical displacement is similar for the two cases. Green’s functions are given by
The reason for the latter is that the up-dip slip
gradient is higher for the imbedded case than for U ijm ¼ U jim ¼ ldij Gnm, n þ m Gim, j þ Gjm, i ;
the surface-rupturing case. The surface-rupturing
case results in a fault scarp at the seafloor that is where l is Lamé’s first constant and dij is the
filtered through the water column as mentioned in Kronecker delta. A review of the static elastic
the Introduction. As the depth to the top of the Green’s functions Gim under different assump-
imbedded crack increases, the peak vertical dis- tions can be found in Segall (2010).
placement decreases. The displacement field can also be approxi-
mated by small dislocation fault elements.
Because the elastic deformation is linear at large
Heterogeneous Static Slip scales, the displacement field from individual
small elements can be superimposed to construct
The surface displacement field associated with the full displacement field from a rupture. For
cracks and other heterogeneous slip models can example, Okada (1985) provides point-source
generally be computed using the static elastic analytic expressions for an element of surface
Green’s functions. Consider three components area DS. As shown in Fig. 9, this technique pro-
of displacement at a point on the seafloor given vides a good approximation to the analytic
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 709
Earthquake Mechanism
and Seafloor
Deformation for
Tsunami Generation,
Fig. 7 Comparison of slip
profiles (a) and
displacement profiles (b)
between a surface-
rupturing crack (solid line)
and a very shallow
imbedded crack (dashed
line). Static stress drop for
the two models is 4.8 and E
6.3 MPa, respectively. See
Fig. 5 for fault geometry
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation, Fig. 8 Coordinate system and
parameters for heterogeneous slip model
710 Earthquake Mechanism and Seafloor Deformation for Tsunami Generation
Earthquake Mechanism
and Seafloor
Deformation for
Tsunami Generation,
Fig. 9 (a) Smooth closure
slip profile approximated
by 1 (Volterra dislocation),
4, and 10 elements. (b)
Associated vertical
displacement profile. See
Fig. 5 for coordinate system
(a=20 )
displacement profile calculated for a crack if local tsunami approaching shore (left) and the
small enough elements are used. Linear superpo- outgoing tsunami (right). For all marigrams, the
sition of subfault elements also provides the basis maximum amplitude is higher for the crack
for finite-fault inversion using tsunami wave- model, even though the scalar seismic moment
forms (Satake and Kanamori 1991). is identical. The dominant period of the tsunami
The difference in the local tsunami wave field associated with the crack model is also shorter,
between crack and dislocation models is shown leading to a higher wave steepness and larger
by example in Figs. 10 and 11. In both cases, a run-up compared to the tsunami from the dislo-
270 km long and 70 km wide, pure thrust rupture cation model.
is used, with a dip of 20 and average slip of 1 m. Earthquake rupture is often sufficiently com-
For the crack model, the slip distribution is that plex that it cannot be adequately described by
derived from Freund and Barnett (1976) under either uniform slip (in the case of the dislocation
smooth closure conditions (q=0.3). Details of the model) or uniform static stress drop (in the case
modeling procedure are given in Geist and of the crack model). For this reason, the slip
Dmowska (1999). Marigrams (time series plot distribution along a fault is difficult to predict
of wave amplitude) are displayed for both the without modeling the rupture process itself
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 711
Dislocation Model
Local Marigrams Outgoing Marigrams
0.4 0.4
0 0
20 40 60 t(min) 20 40 60 t(min)
–0.4 –0.4
Initial Wavefield
reflected phase E
0.8 η(m) 0.8 η(m)
0.4 0.4
0 0
20 40 60 t(min) 20 40 60 t(min)
–0.4 –0.4
Slip
Distribution
0 0
20 40 60 t(min) 20 40 60 t(min)
–0.4 –0.4
Dip
Earthquake Mechanism and Seafloor Deformation wave field, assuming instantaneous rupture, shown in the
for Tsunami Generation, Fig. 10 Tsunami wave field center. Marigrams on the left side are time series plots of
using a dislocation model. Uniform slip within a rectan- local tsunami approaching shore. Marigrams on the right
gular rupture zone shown at the bottom middle. Initial side represent propagation toward the ocean basin
(next Section). Several stochastic slip models such models is the constitutive law, which deter-
have been developed to simulate static slip in a mines how the shear stress on the fault depends
probabilistic sense (Andrews 1980; Lavallée on quantities such as the normal stress across
et al. 2006). the fault, the instantaneous slip or slip rate on
the fault, as well as the integrated history of the
rupture. Such models use the laws of motion and
Application of Dynamic Rupture Models elastodynamics to calculate the time-dependent
to Tsunami Generation propagation of rupture and the evolution of slip at
each point on the fault, as well as the time-
Dynamic models of the earthquake source push dependent deformation of the earth’s surface. In
the physical assumptions of dislocation models this way, these models are more similar to the
back a step; rather than assuming an instanta- crack models mentioned above, but with the
neous distribution of slip on the fault, dynamic added feature that the rupture path, slip
spontaneous earthquake rupture models start with distribution, and final size of the earthquake
a distribution of stress and frictional parameters are calculated as a result of the interaction
on the fault, along with material properties of dynamic stress waves with the frictional
throughout the medium. A key assumption in and geometrical characteristics of the fault.
712 Earthquake Mechanism and Seafloor Deformation for Tsunami Generation
Crack Model
Local Marigrams Outgoing Marigrams
0.4 0.4
0 0
20 40 60 t(min) 20 40 60 t(min)
–0.4 –0.4
Initial Wavefield
reflected phase
0.4 0.4
0 0
20 40 60 t(min) 20 40 60 t(min)
–0.4 –0.4
Slip
Distribution
0 0
20 40 60 t(min) 20 40 60 t(min)
–0.4 –0.4
Dip
Earthquake Mechanism and Seafloor Deformation instantaneous rupture, shown in the center. Marigrams
for Tsunami Generation, Fig. 11 Tsunami wave field on the left side are time series plots of local tsunami
using a crack model. Heterogeneous slip within a rectan- approaching shore. Marigrams on the right side represent
gular rupture zone shown at bottom middle (lighter colors propagation toward the ocean basin
represent higher slip). Initial wave field, assuming
While dislocation models may assume slip distri- portions of a fault) tend to cause rupture and
butions that are not consistent with any physically slip to slow down and die out (Okubo 1989).
plausible stress distribution, such consistency is Spontaneous dynamic rupture models of
at the heart of dynamic models. Dynamic models megathrust earthquakes can shed considerable
have been used extensively in earthquake model- light on both the physics of the earthquake pro-
ing for both basic physics and ground motion cess and the generation of tsunamis. For example,
purposes (e.g., Andrews 1976; Das 1981; Day in such earthquakes, reflected waves from the
1982; Harris et al. 1991; Oglesby et al. 1998). free surface can reduce the normal stress on the
A common feature of such models is that if the fault during slip, facilitating rupture (Brune 1996;
frictional stress on the fault drops rapidly enough Oglesby et al. 1998). Kozdon and Dunham
as a function of slip, slip rate, or time, unstable (2013) show that in the case of the 2011 Tohoku
motion is a result. All other factors being equal, Earthquake, this effect can explain the rupture
high stress drops lead to fast rupture propagation and high slip all the way to the trench, even in
and rapid fault slip, along with high final slip. the near-surface zone in which the frictional coef-
Zones in which the frictional stress remains con- ficient is expected to increase with slip rate.
stant or increases over the course of the earth- Somewhat similarly, Noda and Lapusta (2013)
quake (such as in the very deep and shallow show that ruptures can tunnel through stably
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 713
Earthquake Mechanism and Seafloor Deformation branches off the plate boundary thrust (Fault A) close to
for Tsunami Generation, Fig. 12 Fault geometry of shore and at a significantly steeper dip than the plate
Wendt et al. (2009). The splay fault (Faults B and C) boundary thrust
E
sliding regions of a fault (i.e., zones with low transfers to the steeper splay fault. A comparison
seismic coupling) if pore-fluid pressurization between the vertical surface displacement fields in
takes place. The possibility of earthquake rupture these two cases is shown in Fig. 13.
and slip propagating into areas that appear to be Because of the more vertically directed
stably sliding can have important implications for hanging-wall movement, the splay faulting case
both the distribution of slip in an earthquake and produces a significantly larger vertical seafloor
its overall size, with obvious consequences for displacement than the homogeneous case. For
tsunami generation. Recent dynamic models by this reason, the local tsunami in the barrier case
Ma and Hirakawa (2013) have argued that failure produces a significantly larger wave in the near-
of material in the hanging-wall wedge might lead source region (in this case, along the shore of
to slow rupture propagation and slow slip near the Japan) than the homogeneous case, as seen in
trench, leading to low-radiation, high-tsunami Fig. 14, and it also reaches the near shore much
“tsunami earthquakes.” sooner.
Fault dynamics can also play an important role It should be noted that in this simple model,
in determining rupture path and partitioning of the moment of the barrier model (6.8 1020 Nm)
slip between different fault segments in a geo- is less than half that of the homogenous case (1.8
metrically complex fault system. An issue of 1021 Nm), so the difference in tsunami height is
great importance for tsunami generation is that purely due to the rupture path and slip distribu-
of splay faults – whether a splay fault is activated tion, not to the overall size of the earthquake.
coseismicly. In a standard dislocation or crack Another advantage of the dynamic modeling
model, the decision of whether a splay fault will method in this case is that the partitioning of
slip is made a priori and is unconnected with the slip between the plate boundary thrust and the
slip process elsewhere on the fault system. How- splay fault is a calculated result of the problem
ever, Wendt et al. (2009) show that the ability of rather than an assumption; the method correctly
rupture to propagate to a splay fault is a result of accounts for the fact that slip on the splay sup-
the dynamic rupture process and can have a first- presses slip on the plate boundary thrust, and vice
order effect on tsunami generation and near-field versa.
run-up. Their fault geometry, motivated by the The examples given above raise a tantalizing
Nankai Trough, is shown in Fig. 12. possibility: if enough is known about the geology
The researchers used the dynamic finite element and fault geometry of a subduction zone,
method (FEM) (Oglesby 1999; Whirley and dynamic modeling has the promise to help pre-
Engelmann 1993) to perform spontaneous dynamic dict (or at least constrain) the size and severity of
rupture models of potential earthquakes on this potential earthquakes and tsunamis in the region.
fault system. They found that if stress on the faults Someday it may be possible to constrain the like-
is largely homogeneous, rupture proceeds only on lihood of rupture to the trench or whether rupture
the plate boundary thrust. However, if a barrier is is likely to propagate to a splay fault rather than
present on the plate boundary thrust, rupture (or in addition to) the main plate boundary thrust
714 Earthquake Mechanism and Seafloor Deformation for Tsunami Generation
Earthquake Mechanism and Seafloor Deformation produces much higher vertical displacement than the
for Tsunami Generation, Fig. 13 Comparison of ver- homogeneous case and that this displacement is closer to
tical seafloor displacement for homogeneous-stress case shore
(left) and barrier case (right). Note that the barrier case
Earthquake Mechanism and Seafloor Deformation the barrier case (right). The rupture of the splay produces a
for Tsunami Generation, Fig. 14 Comparison of higher initial tsunami as well as higher wave amplitude as
wave amplitude (m) for the homogeneous case (left) and the tsunami approaches the shore (insets)
fault. In this manner, dynamic models may help account important interactions and processes on
to compensate for lack of historical data, espe- and near faults. As such, they can serve an impor-
cially in the case of extremely large (M 9+) tant complement to existing standard methods of
events that happen only once or twice a millen- tsunami source estimation.
nium. Dynamic models are not a panacea, how-
ever Like dislocation and crack models, they rest
on assumptions about physical properties (such Summary
as stress and friction) that are quite poorly
constrained in many cases. However, they do Tsunamis are generated from earthquakes by
have the advantage of producing physically self- vertical displacement of the seafloor during earth-
consistent slip and rupture patterns that take into quake rupture. Four representations of the
Earthquake Mechanism and Seafloor Deformation for Tsunami Generation 715
earthquake rupture process have been described, Bilby BA, Eshelby JD (1969) Dislocations and the theory
in increasing levels of consistency with earth- of fracture. In: Liebowitz H (ed) Fracture. Academic,
New York, pp 99–182
quake physics, for use as tsunami generation Bird P (2003) An updated digital model of plate bound-
models. First, simple uniform-slip dislocation aries. Geochem Geophys Geosyst 4. doi:10.1029/
modeling, scaled to the seismic moment of an 2001/GC000252
earthquake, has historically been the most com- Blaser L, Kr€ uger F, Ohrnberger M, Scherbaum F (2010)
Scaling relations of earthquake source parameter esti-
mon representation used for tsunami modeling. mates with special focus on subduction environment.
Second, crack models avoid artifacts that dislo- Bull Seismol Soc Am 100:2914–2926
cation models have associated with shallow, Brune JN (1996) Particle motions in a physical model of
imbedded thrust faults. Crack models also shallow angle thrust faulting. Proc Indian Acad Sci
105:L197–L206 E
directly relate slip distribution to tectonic stress Chambon G, Rudnicki JW (2001) Effects of normal stress
conditions and static stress drop of an earthquake. variations on frictional stability of a fluid-infiltrated
Third, heterogeneous static slip models account fault. J Geophys Res 106:11353–311372
for the observed complexity of distribution slip Das S (1981) Three-dimensional spontaneous rupture
propagation and implications for the earthquake
(and stress drop) throughout a rupture zone, pri- source mechanism. Geophys J Roy Astron Soc
marily through the use of small dislocation ele- 67:375–393
ments. Finally, dynamic rupture models provide Day SM (1982) Three-dimensional simulation of sponta-
the most complete and physically consistent rep- neous rupture: the effect of nonuniform prestress. Bull
Seismol Soc Am 72:1881–1902
resentation of the tsunami generation process. Dmowska R, Kostrov BV (1973) A shearing crack in a
Each of these models has been used in both deter- semi-space under plane strain conditions. Arch Mech
ministic and probabilistic hazard assessments, 25:421–440
although advanced tsunami generation models Eshelby JD (1957) The determination of the elastic field of
an ellipsoidal inclusion, and related problems. Proc
provide more accuracy and consistency with the R Soc Lond A Math Phys Sci 241:376–396
known processes of earthquake physics. Freund LB, Barnett DM (1976) A two-dimensional anal-
ysis of surface deformation due to dip-slip faulting.
Bull Seismol Soc Am 66:667–675
Cross-References Fukao Y, Furumoto M (1975) Mechanism of large earth-
quakes along the eastern margin of the Japan Sea.
Tectonophys 25:247–266
▶ Earthquake Mechanisms and Stress Field Geist EL (1999) Local tsunamis and earthquake source
▶ Earthquake Mechanisms and Tectonics parameters. Adv Geophys 39:117–209
▶ Earthquake Mechanism Description and Geist EL, Bilek SL (2001) Effect of depth-dependent
shear modulus on tsunami generation along subduc-
Inversion
tion zones. Geophys Res Lett 28(7):1315–1318
▶ Probabilistic Seismic Hazard Models Geist EL, Dmowska R (1999) Local tsunamis and distrib-
▶ Site Response for Seismic Hazard Assessment uted slip at the source. Pure Appl Geophys 154:485–512
▶ Tsunamis as Paleoseismic Indicators Geist EL, Yoshioka S (1996) Source parameters control-
ling the generation and propagation of potential local
tsunamis along the Cascadia margin. Nat Hazards
13:151–177
References Geist EL, Yoshioka S (2004) Effect of structural hetero-
geneity and slip distribution on coseismic vertical dis-
Abe K (1973) Tsunami and mechanism of great earth- placement from rupture on the seattle fault. Open-File
quakes. Phys Earth Planet In 7:143–153 Report 2004–1010 Rep 2004–1010, U.S. Geological
Andrews DJ (1976) Rupture velocity of plane strain shear Survey
cracks. J Geophys Res 81:5679–5687 Geist EL, Titov VV, Arcas D, Pollitz FF, Bilek SL
Andrews DJ (1980) A stochastic fault model 1. Static case. (2007) Implications of the December 26, 2004
J Geophys Res 85:3867–3877 Sumatra-Andaman earthquake on tsunami forecast
Baba T, Hori T, Hirano S, Cummins PR, Park J-O, and assessment models for great subduction zone
Kameyama M, Kaneda Y (2001) Deformation of a earthquakes. Bull Seismol Soc Am 97:S249–S270
seamount subducting beneath an accretionary prism: Geller RJ (1976) Scaling relations for earthquake source
constraints from numerical simulation. Geophys Res parameters and magnitudes. Bull Seismol Soc Am
Lett 28:1827–1830 66:1501–1523
716 Earthquake Mechanism Description and Inversion
Harris RA, Archuleta RJ, Day SM (1991) Fault steps and Roering JJ, Cooke ML, Pollard DD (1997) Why blind
the dynamic rupture process – 2-D numerical simula- thrust faults do not propagate to the Earth’s surface:
tions of a spontaneously propagating shear fracture. numerical modeling of coseismic deformation associ-
Geophys Res Lett 18(5):893–896 ated with thrust-related anticlines. J Geophys Res
Kajiura K (1963) The leading wave of a tsunami. Bull 102:11901–911912
Earthq Res Inst 41:535–571 Rudnicki JW, Wu M (1995) Mechanics of dip-slip faulting
Kanamori H (1972) Mechanism of tsunami earthquakes. in an elastic half-space. J Geophys Res 100:
Phys Earth Planet In 6:346–359 22173–22186
Kozdon JE, Dunham EM (2013) Rupture to the trench: Rybicki R (1986) Dislocations and their geophysical
dynamic rupture simulations of the 11 march 2011 application. In: Teisseyre R (ed) Continuum theories
Tohoku Earthquake. Bull Seismol Soc Am 103(2B): in solid earth physics. Elsevier, Amsterdam, pp 18–186
1275–1289 Satake K, Kanamori H (1991) Use of tsunami waveforms
Lavallée D, Liu P, Archuleta RJ (2006) Stochastic model for earthquake source study. Nat Hazards 4:193–208
of heterogeneity in earthquake slip spatial distribu- Savage JC (1998) Displacement field for an edge disloca-
tions. Geophys J Int 165:622–640 tion in a layered half-space. J Geophys Res
Ma S, Hirakawa ET (2013) Dynamic wedge failure 103:2439–2446
reveals anomalous energy radiation of shallow sub- Savage JC, Hastie LM (1966) Surface deformation associ-
duction earthquakes. Earth Planet Sci Lett ated with dip-slip faulting. J Geophys Res 20:4897–4904
375:113–122, 110.1016/j.epsl.2013.1005.1016 Segall P (2010) Earthquake and volcano deformation.
Ma XQ, Kusznir NJ (1994) Effects of rigidity layering, Princeton University Press, Princeton
gravity and stress relaxation on 2-D subsurface fault Tanioka Y, Satake K (1996) Tsunami generation by hor-
displacement fields. Geophys J Int 118:201–220 izontal displacement of ocean bottom. Geophys Res
Madariaga R (1979) On the relation between seismic Lett 23:861–865
moment and stress drop in the presence of stress and Wang K, He J (2008) Effects of frictional behavior and
strength heterogeneity. J Geophys Res 84:2243–2250 geometry of subduction fault on coseismic seafloor
Masterlark T (2003) Finite element model predictions of deformation. Bull Seismol Soc Am 98:571–579
static deformation from dislocation sources in a sub- Ward SN (1980) Relationships of tsunami generation and
duction zone: sensitivities to homogeneous, isotropic, an earthquake source. J Phys Earth 28:441–474
Poisson-solid, and half-space assumptions. J Geophys Wendt J, Oglesby DD, Geist EL (2009) Tsunamis and
Res 103. doi:10.1029/2002JB002296 splay fault dynamics. Geophys Res Lett 36, L15303,
Noda H, Lapusta N (2013) Stable creeping fault segments 15310.11029/12009GL038295
can become destructive as a result of dynamic weak- Whirley RG, Engelmann BE (1993) DYNA3D: a
ening. Nature. doi:10.1038/nature11703 nonlinear, explicit, three-dimensional finite element
Oglesby DD (1999) Earthquake dynamics on dip-slip code for solid and structural mechanics – user manual.
faults. PhD thesis, University of California, Santa University of California, Lawrence Livermore
Barbara, 223 pp National Laboratory, California
Oglesby DD, Archuleta RJ (2000) Dynamics of dip-slip Zhao S, M€ uller RD, Takahashi Y, Kaneda Y (2004) 3-D
faulting: explorations in two dimensions. J Geophys finite-element modeling of deformation and stress
Res 105(13):643–613653 associated with faulting: effect of inhomogeneous
Oglesby DD, Archuleta RJ, Nielsen SB (1998) Earth- crustal structures. Geophys J Int 157:629–644
quakes on dipping faults: the effects of broken sym-
metry. Science 280:1055–1059
Oglesby DD, Archuleta RJ, Nielsen SB (2000) The three-
dimensional dynamics of dipping faults. Bull Seismol Earthquake Mechanism Description
Soc Am 90:616–628
Okada Y (1985) Surface deformation due to shear and
and Inversion
tensile faults in a half-space. Bull Seismol Soc Am
75:1135–1154 Doug Dreger
Okada Y (1992) Internal deformation due to shear and Department of Earth and Planetary Science,
tensile faults in half-space. Bull Seismol Soc Am
College of Letters and Science, University of
82:1018–1040
Okal EA (1988) Seismic parameters controlling far-field California, Berkeley, CA, USA
tsunami amplitudes: a review. Nat Hazards 1:67–96
Okubo PG (1989) Dynamic rupture modeling with
laboratory-derived constitutive relations. J Geophys Synonyms
Res 94:12321–12335
Pollitz FF (1996) Coseismic deformation from earthquake
faulting on a layered spherical earth. Geophys J Int Radiation pattern; Source mechanism; Moment
125:1–14 tensor; Source inversion
Earthquake Mechanism Description and Inversion 717
⇀ ⇀ ⇀ €⇀
ðl þ 2mÞ∇ ∇ u m∇ ∇ u þ f ¼ ru (1)
ðdn1 gn g1 Þg3 M_ t r=b Earthquake Mechanism Description and Inversion,
uDC-Swave ðr, tÞ ¼ x^n Fig. 2 P- and S-wave radiation patterns for a double
n
4prb3 r couple
ðdn1 gn g3 Þg1 M_ t r=b
þ x^n
4prb3 r hemispherical representation, the horizontal
(6) angle represents the azimuth from north, and the
vertical angle represents the direction that a
Figure 2 compares the P- and S-wave radiation P-wave ray leaves the source, where an angle of
patterns for a double couple in the x1x3 plane zero indicates a ray leaving straight down and an
using Eq. 6. Both the P- and S-waves display angle of 90 indicates a wave leaving horizon-
two-lobed radiation patterns consistent with tally. This representation is useful for conveying
observation. As shown in Aki and Richards earthquake mechanism information in map view
(1980), the spatial derivatives may be carried for analysis of the tectonics of a region.
out on all terms leading to a solution to Eq. 1 The solutions given by Eq. 6 are for a homo-
comprised of near-field,
intermediate, and geneous and isotropic elastic whole space. In the
far-field terms ( 1 r4 , 1 r2 , 1 =r ), all of which next section it is shown that more general source
can be represented in terms of direction cosines, solutions can be characterized by the seismic
strike, rake, and dip parameters of a double- moment tensor for more complex elastic structure
couple mechanism, or more generally as a seis- using appropriate numerical methods to compute
mic moment tensor (Aki and Richards 1980). Green’s functions.
It is common to represent the mechanism of an
earthquake in terms of the P-wave radiation pat-
tern using a focal sphere representation. An Generalized Seismic Moment Tensor
example of such a representation is given in
Fig. 3 in which the lower hemisphere is The impulse response to Eq. 1 or more generally
presented. There are four quadrants as found in for 1-D, 2-D, and 3-D earth structures is referred
Eq. 6 and Fig. 2. The shaded quadrants indicate a to as the Green’s function solution to the elasto-
P-wave first arrival with a vertical component dynamic equation of motion. Green’s functions
polarity that is up, whereas the unshaded quad- may be found for volume distributed forces, for
rants indicate a first motion that is down. In this distributed shear dislocation on an interior
720 Earthquake Mechanism Description and Inversion
Depth
T
radiation pattern and
Fault
relationships with block- plane
faulting models. Shaded P
e y
an ar
Fa roj
pl xill
quadrants represent vertical
ul ec
p
“Beach ball”
t-p tio
u
A
component P-waves with
la n
ne
positive polarity. The Focal sphere
P (compression) and
T (tension) axes are shown. b
This figure was prepared by Strike slip
the USGS
T
P
Normal
T
P
Reverse
T
P
Oblique reverse
T
P
surface, or for traction acting on an exterior sur- shear dislocation over an interior surface (S)
face. The general representation theorem for a within an elastic body is written as
ð ðð h i
⇀ ⇀ ⇀ ⇀
un x , t ¼ dt ui t, z n^j Cijkl Gnk, l t t, x ; 0, z dS
ð ðð ⇀ ⇀
(7)
⇀ ⇀
un x , t ¼ dt Mij t, z Gni, j t t, x ; 0, z dS
where [ui(t,z)] is the displacement discontinuity Green’s tensor of single-couple solutions where
across a surface oriented with the unit vector n is the ground motion component, k is the
normal, n^j . Cijkl is the elastic tensor, and Gnk,l is direction of a unit force, and l is the direction
Earthquake Mechanism Description and Inversion 721
Earthquake Mechanism Description and Inversion, Note that the values of Mij, representing the strength of
Fig. 4 Schematic showing the system of single couples these source terms, are both real and symmetric
and vector dipoles comprising the seismic moment tensor.
of the spatial derivative of the single force. See even 3-D earth models, the spatiotemporal dis-
Aki and Richards (1980) for a thorough discus- tribution of fault slip, [u], may be obtained
sion of the general representation theorem of through inversion (e.g., Hartzel and Heaton
seismic sources. Equivalently, the representa- 1983). There are numerous linear, linearized,
tion theorem can be written in terms of the and nonlinear approaches to solve this system
moment tensor (Mij) distributed spatially of equations each considering different approx-
and temporally over S (Eq. 7). Figure 4 illus- imations on how to kinematically represent
trates the system of single couples and linear the spatiotemporal integration of the seismic
vector dipoles represented by both Gni,j and Mij. source.
Given a number of observations, the represen- If a source is considered to be a point in
tation theorem provides a system of equations space, in which the wavelengths of observed
that if given Green’s functions for 1-D, 2-D, or seismic waves are considerably larger than the
722 Earthquake Mechanism Description and Inversion
source dimension, then the spatial integral is linear system of equations employed in seismic
unnecessary, and moment tensor analysis (Eq. 8c).
The strength of the individual single-
ð
⇀ ⇀ couple and vector-dipole Green’s function solu-
un x , t ¼ Mij ðtÞGni, j t t, x dt ðaÞ
ð tions (Gni,j) is given by Mij which is a
⇀ ⇀ second-order tensor that is both real and symmet-
un x , t ¼ Mij SðtÞGni, j t t, x dt ðbÞ,
⇀ ⇀ ric (M12 ¼ M21 ) enabling the decomposition into
un x , t ¼ Mij Gni, j t, x ðcÞ three orthogonal eigenvectors:
(8)
0 1
M11 M12 M13
which accounts for the general case in which each
Mij ¼ @ M21 M22 M23 A (9)
term of the moment tensor can have a different
M31 M32 M33
source time history (Eq. 8a), as well as the sim-
plified case in which each element is assumed to
have the same, or synchronous, time history Equation 8c is a linear system of equations that is
(Eq. 8b). overdetermined with three-component seismic
The most commonly applied solution is a syn- waveform data, even with low-pass filters
chronous source in which the time history is applied. Writing this in terms of a system of
considered to be a delta function yielding the linear equations, the following is obtained:
GM ¼ u 0 1 0 1
0 1 M11 u1
G11, 1 G11, 2 G11, 3 G12, 2 G12, 3 G13, 3 B B M 12
C B u2 C
C B C
B ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ C B C B C
B CB M13 C ¼ B ⋮ C (10)
@ ⋮ ⋮ ⋮ ⋮ ⋮ ⋮ A B C B C
B M22 C B ⋮ C
GN1, 1 GN1, 2 GN1, 3 GN2, 2 GN2, 3 GN3, 3 @ M23 A @ ⋮ A
M33 uN
Earthquake Mechanism
Description and
Inversion,
Fig. 5 Relationships
between eigenvalues for the
three fundamental moment
tensor types
vector dipoles (CLVD; Knopoff and Randall There are several methods for determining the
1970), and combinations of DCs and CLVDs scalar seismic moment either directly from the
that sum to yield the same moment tensor. moment tensor or from its eigenvalues (Aki and
Figure 5 illustrates combinations of vector dipoles Richards 1980). Traditionally the method of
leading to spherically symmetric isotropic radia- Dziewonski and Woodhouse (1983) has been
tion, a double couple, and a CLVD. Julian used where
et al. (1998) give a detailed description of the
different decomposition possibilities and the radi- jm3 j þ jm2 j
ation characteristics of each type of source, and MDW
0 ¼ : (15)
2
Jost and Herrmann (1989) give a thorough discus-
sion of the moment tensor theory, seismic wave- Alternatively Bowers and Hudson (1999) define
form inverse procedure, and decomposition the total scalar moment (MT0 ) as the sum of the
schemes for several common cases. isotropic moment and the largest deviatoric
Here, the isotropic, double-couple, and CLVD eigenvalue, which provides an estimate equiva-
decomposition are given. For the general lent to MDW0 for double couples, but provides
6 parameter moment tensor (M), the isotropic better estimates for sources that deviate from a
moment tensor (M0I I) representing volume simple double couple for tectonic earthquakes
change is subtracted leaving the deviatoric such as the CLVD and tensile crack.
moment tensor (MD), where
MT0 ¼ MI0 þ jm3 j (16)
MD ¼ M MI0 I (12)
and I is the diagonal matrix. The isotropic scalar As Julian et al. (1998) point out, the decomposi-
moment is defined as tion of the seismic moment tensor is necessary for
interpretation but is also non-unique. Today the
MI0 ¼ ðM11 þ M22 þ M33 Þ=3: (13) most common decomposition approach is one in
which the deviatoric moment tensor is
MD is then solved for its three orthogonal eigen- decomposed into a double couple and a CLVD,
vectors (ai) and eigenvalues (mi), which are then each with the same compression (P) and tension
ranked. (T) axes, following Knopoff and Randall (1970).
To decompose the moment tensor in this way,
jm3 j jm2 j jm1 j (14) after first removing the isotropic component
724 Earthquake Mechanism Description and Inversion
Explosion (k = 1)
T=
1
−1
=
T
TC (−1, 5/9) k=0
.5
(1, 1/5)
T = 0.0
0.5
(−1, −1/5)
T=
T
T=
−0
−1
.5
Implosion (k = −1)
Earthquake Mechanism Description and Inversion, –CLVD in compression, TC and –TC are tensile and
Fig. 6 Equal area source-type plot from Bowers and closing cracks, LVD are linear vector dipoles, and the
Hudson (1999). Specific (T, k) values are plotted in the top and bottom of the plot show the position of explosion
source-type space with the corresponding fundamental and implosions. All other points can be interpreted as
source types, where the DC point represents all double mixtures of these fundamental source types
couples, CLVD has a major vector dipole in tension and
(Eq. 12), ranking the deviatoric eigenvalues each of the fundamental components. Based on
(Eq. 14), the ratio of the maximum to minimum the total scalar moment (Eq. 16), the following
deviatoric eigenvalue is found where defines the percentages of the isotropic, double
couple, and CVLD for the decomposition given
m 1 by Eq. 18:
F ¼ (17)
m3 I
M
Piso ¼ T0 100%
F is used to scale the scalar moment of the M0 (19)
double-couple and CLVD components, and the Pdc ¼ ð100 PisoÞ ð1 2FÞ%
complete moment tensor (M) is then represented Pclvd ¼ ð100 PisoÞ 2F%
as
Representation Methods
M ¼ MI0 I þ ð1 2FÞ m3 Because the choice of the particular decomposi-
ð a3 a3 a2 a2 Þ þ F m 3 tion method is non-unique (e.g., Jost and Herr-
ð2a3 a3 a2 a2 a1 a1 Þ; (18) mann 1989; Julian et al. 1998), there is a growing
interest in using graphical representations of the
a3a3, etc. are dyadics of the eigenvectors, matri- moment tensor eigenvalues (Hudson et al. 1989;
ces describing the orientations of the principal Riedesel and Jordan 1989; Tape and Tape 2012).
vector dipoles. For example, the source-type diagram of Hudson
The resulting decomposition can also be char- et al. (1989) examines ratios of the three eigen-
acterized in terms of the percentage of moment in values of the moment tensor decomposition to
Earthquake Mechanism Description and Inversion 725
Earthquake Mechanism Description and Inversion, predominantly double couple. The NSS shows a wide area
Fig. 7 Comparison of long-period (20–50 s) three- of acceptable solutions due to the relatively small number
component, complete waveforms for a tectonic earth- of stations used. The distribution of best solutions is cen-
quake on the Hayward Fault, California. The solution is tered on a double couple
find a scaling for a volumetric component (k) and epistemic and aleatoric uncertainties in the
a deviation from a pure double-couple deviatoric solutions (e.g., Ford et al. 2010):
solution (T) (Eq. 20). The values are plotted in a
Cartesian coordinate system that is deformed to
2m1 MI0
provide an equal area representation of k and T. T¼ , k¼ (20)
jm3 j MI0 þ jm 3 j
The advantage of this approach and the others is
that a solution may be analyzed in terms of its
proximity to idealized, fundamental solution Figure 6 illustrates the source-type projection
types on the diagram, as well as providing a with the position of fundamental source types
representation enabling the analysis of both identified (DC, CLVD, TC, LVD, and explosion).
726 Earthquake Mechanism Description and Inversion
Earthquake Mechanism Description and Inversion, a large volume increase component. The NSS shows the
Fig. 8 Comparison of long-period (20–50 s) three- best solutions are substantially shifted from the deviatoric
component, complete waveforms for an earthquake in line (+CLVD-DC-CLVD). The improvement in fit
Long Valley Caldera, California. The deviatoric solution afforded by the volume increase component is statistically
has a large CLVD, and the full moment tensor (shown) has significant based on the F-test
The DC point represents all possible double- (TC) solutions plot in distinct locations on the
couple solutions. Positive and negative CLVD graph. All other points can be considered to be
sources are for the cases with a major vector mixtures of these fundamental types. Shear ten-
dipole in tension (example given in Fig. 5) and sile cracks (e.g., Šı́lený, J., and J. Kozák (1988)
compression, respectively. Similarly, the linear- can be thought to lie along the line connecting the
vector-dipole (LVD) and tension crack TC and –TC source types that passes through the
Earthquake Mechanism Description and Inversion 727
origin (DC). Julian et al. (1998) demonstrate that deformation, and CO2 gas flux is presented
other locations can be interpreted as shear tensile (Dreger et al. 2000; Minson and Dreger 2008).
cracks in which the two elements of the crack are The data processing is the same as in the previous
not coplanar and have different geometries. example. The solution indicates a large volume
increase component that has been found to be
statistically significant using an F-test. The NSS
Example of a Seismic Moment Tensor shows the best fit region is shifted substantially
Inversion for Double-Couple and from the deviatoric line.
Non-Double-Couple Cases
E
This section illustrates the inversion of three-
component waveform data for the seismic References
moment tensor. A full moment tensor is solved
Aki K, Richards PG (1980) Quantitative seismology,
for an earthquake located on the Hayward Fault
vol 1. W. H Freeman and Company, San Francisco
in California. The broadband velocity data were Bowers D, Hudson JA (1999) Defining the scalar moment
instrument corrected, integrated to displacement, of a seismic source with a general moment tensors.
and filtered with an acausal, bandpass, Bull Seismol Soc Am 89:1390–1394
Dreger DS, Tkalcic H, Johnston M (2000) Dilational pro-
Butterworth filter with corners set at 0.02 and
cesses accompanying earthquakes in the Long Valley
0.05 Hz. Green’s functions were computed for Caldera, Science 288:122–125
an appropriate layered velocity structure using a Dziewonski AM, Woodhouse JH (1983) An experiment in
frequency-wavenumber integration method and systematic study of global seismicity: centroid-
moment solutions for 201 moderate and large earth-
were bandpass filtered with the same filter
quakes. J Geophys Res 88:3247–3271
applied to the data. Figure 7 shows the fit to the Ford SR, Dreger DS, Walter WR (2010) Network sensi-
data, the P-wave radiation pattern, and pertinent tivity solutions for regional moment-tensor inversions.
source information. This earthquake is a tectonic Bull Seismol Soc Am 100:1962–1970
Gilbert F (1970) Excitation of the normal models of the
earthquake and the solution is predominantly
Earth by earthquake sources. Geophys J Roy Astron
double couple. The small CLVD and volume Soc 22:223–226
increase components are due to noise and episte- Hartzell SH, Heaton TH (1983) Inversion of strong ground
mic uncertainty. motion and teleseismic waveform data for the fault
rupture history of the 1979 Imperial Valley, California,
The network sensitivity solution (NSS; Ford
earthquake. Bull Seismol Soc Am 73(4):1161–1171
et al. 2010) is also presented. An NSS represents a Hudson JA, Pearce RG, Rogers RM (1989) Source type
maximum fit surface in the source-type space. It plot for inversion of the moment tensors. J Geophys
is computed by finding the best fit solution for Res 94:765–774
Jost ML, Herrmann RB (1989) A student’s guide to and
each T and k point. The fit is measured by a
review of moment tensors. Seismol Res Lett 60:37–57
variance reduction. The NSS for this event Julian BR, Miller AD, Foulger GR (1998) Non-double-
shows a relatively wide distribution that is cen- couple earthquakes 1. Theory. Rev Geophys
tered at a double couple but extends substantially 36:525–550
Kanamori H, Given JW, Lay T (1984) Analysis of seismic
into regions with either non-double-couple
body waves excited by the Mount St. Helens eruption
deviatoric solutions or solutions that have volume of May 18, 1980. J Geophys Res 89(B3):1856–1866
change components. This is due to the fact that Kawakatsu H (1989) Centroid single force inversion of
only three stations were used and demonstrates seismic waves generated by landslides. J Geophys Res
94(B9):12363–12374
uncertainty in the solution in terms of the more
Kikuchi M, Kanamori H (1991) Inversion of complex
difficult to recover non-double-couple compo- body waves – III. Bull Seismol Soc Am 81:2335–2350
nents. In this case non-double-couple and volu- Knopoff L, Randall MJ (1970) The compensated linear-
metric sources are not statistically significant. vector dipole: a possible mechanism for deep earth-
quakes. J Geophys Res 75:4957
In Fig. 8 a moment tensor solution for an
Miller AD, Foulger GR, Julina BR (1998) Non-double-
earthquake from Long Valley Caldera, Califor- couple earthquakes 2. Observation. Rev Geophys
nia, during a period of heightened seismic, 36:551–568
728 Earthquake Mechanisms and Stress Field
Minson S, Dreger D (2008) Stable inversions for complete tectonic stress is mostly horizontal and originates
moment tensors. Geophys J Int 174:585–592 in forces driving the plate motions (Heidbach
Riedesel MA, Jordan TH (1989) Display and assessment
of seismic moment tensors. Bull Seismol Soc Am et al. 2008). These forces typically cause “ridge
79:85–100 push” processes where lithospheric plates are
Šı́lený J, Kozák J (1988) Shallow seismic events with com- pushed to move away from spreading ridge or
bined source mechanism. Tectonophysics 152:283–296 “slab pull” processes where two plates are in
Stump BW, Johnson LR (1977) The determination of
source parameters by the linear inversion of collision and one of them is subducting into the
seismograms. Bull Seismol Soc Am 67(6):1489–1502 asthenosphere (Fowler 1990). The global plate
Tape W, Tape C (2012) A geometric comparison of tectonics successfully explained concentration
source-type plots for moment tensors. Geophys J Int of large earthquakes on margins of the plates
190:499–510
characterized by the presence of low strength
zones exposed to large stresses. On regional and
local scales, stress in the Earth’s crust is also
affected by topography and its compensation at
Earthquake Mechanisms and Stress depth, sediment loading or presence of any het-
Field erogeneity caused by variations of density, rigid-
ity and rock rheology, by fluid flow in rocks or by
Václav Vavryčuk presence of faults, cracks and micro-cracks in
Institute of Geophysics, Czech Academy of rocks.
Sciences, Prague, Czech Republic Properties of the stress field and of the associ-
ated fracture processes in the Earth’s crust are
closely related (Scholz 2002). The most common
Synonyms type of fracturing is shear faulting but under
special stress conditions also tensile faulting can
Dynamics and mechanics of faulting; Earthquake be observed (Julian et al. 1998; Vavryčuk 2011b;
source observations; Focal mechanism; Rock see entry “▶ Non-Double-Couple Earthquakes”).
failure; Shear faulting; Tectonic stress; Tensile Type of faulting depends on the stress field but
faulting also on the orientation of activated fractures or
faults with respect to the stress (Vavryčuk
2011a). In addition, the slip vector for shear
Introduction faulting is close to or coincides with the direction
of the maximum shear stress acting on the fault
Earthquakes are processes associated with the (Wallace 1951; Bott 1959). Therefore, type of
sudden rupture of rocks along cracks, fractures faulting, orientation of activated faults and direc-
or faults exposed to stress field in the Earth’s tion of slip along activated faults serve as an
crust and lithosphere. If stress reaches a critical important source of information about the stress
value exceeding strength of faults or fractures in field and its spatial and lateral variations within
rock, accumulated energy of elastic deformation the Earth’s crust.
is partially spent for anelastic deformations in the In this entry, the basic concept of stress is
focal zone and partially released and radiated in introduced and its relation to earthquake mecha-
the form of seismic waves. Stress in the Earth’s nisms is explained (for description of focal mech-
crust causing earthquakes can be of tectonic or anisms, see entry “▶ Earthquake Mechanism
non-tectonic origin (Ruff 2002; Zoback 2007; Description and Inversion”). Mohr’s circle dia-
Zang and Stephansson 2010). The main source gram and simple failure criteria are described and
of non-tectonic stress within the Earth is gravita- used for defining the fault instability, principal
tional loading. This stress is vertical with largest faults and principal focal mechanisms. Methods
lateral variations near the Earth’s surface being of determining stress from observed earthquake
more homogeneous at depth. On the other hand, mechanisms are reported and their robustness is
Earthquake Mechanisms and Stress Field 729
x3
a b
x3 σ3
τ33
τ31
τ32 σ1
τ13
τ23
x1
τ11
τ21 τ12 σ2
τ22 x1
E
x2
x2
Earthquake Mechanisms and Stress Field, Fig. 1 Definition of the stress tensor in original Cartesian coordinate
system (a) and in the rotated system of principal stress directions (b)
exemplified on numerical tests. Finally, several of forces in a body which causes its deformation but
applications of stress inversions from earthquake not rotation, the stress tensor must be symmetric
mechanisms are listed.
tij ¼ tji ; (5)
Earthquake Mechanisms and Stress Field, Tectonic Stress and Faulting Regimes
Fig. 2 Mohr’s circle diagram. Quantities sn and t are
the normal and shear stresses along a fault, s1, s2 and s3 Principal stress directions in the Earth’s crust are
are the principal stresses. All permissible values of sn and
frequently close to vertical and horizontal direc-
t acting on a fault must lie in the shaded area of the
diagram tions. This lead Anderson (1951) to developing a
simple scheme connecting the basic stress
regimes in the Earth’s crust with type of faulting
The normal and shear components sn and t of
on a pre-existing fault in the crust (see Fig. 4).
traction T (also called the normal and shear
Anderson (1951) distinguishes three possible
stresses) read in the system of principal stress
combinations of magnitudes of principal
directions
stresses: the vertical stress is maximum, inter-
mediate or minimum with respect to the hori-
sn ¼ s1 n21 þ s2 n22 þ s3 n23 ; (8) zontal stresses. If the vertical stress is maximum,
the hanging wall is moving downwards with
t2 ¼ s21 n21 þ s22 n22 þ s23 n23 s2n : (9) respect to the foot wall and the normal faulting
is observed along a deeply steeping fault. If the
If principal stresses s1, s2 and s3 are fixed, then vertical stress is minimum, the crust is in hori-
normal and shear stresses sn and t are just func- zontal compression and the hanging wall is mov-
tions of normal n of a fault and can be plotted in the ing upwards with respect to the foot wall and
Mohr’s circle diagram (see Fig. 2). All permissible reverse faulting is observed along a shallow dip-
values of sn and t must lie in the shaded area of the ping fault. Finally, if the vertical stress is inter-
diagram (Jaeger et al. 2007; Mavko et al. 2009). If mediate, the foot and hanging walls are moving
the sign of shear stress t is important in the stress horizontally and the strike slip faulting is
analysis, the Mohr’s diagram is not plotted using observed along a nearly vertical fault. Obvi-
semi-circles but full circles, see below. ously, the Anderson’s classification is simple
When stress conditions in real rocks in the and does not cover all observations but still it
Earth’s crust are studied, the Mohr’s diagram is proved to be valid for many seismically active
modified. Rocks are typically porous and contain regions and helpful for rough assessment of
pressurized fluids which influence overall stress stress regime (Simpson 1997; Hardebeck and
in the rock (Scholz 2002). For example, fluids Michael 2006; see entry “▶ Earthquake Mecha-
present inside a fault reduce the normal stress nisms and Tectonics”).
along the fault. If the fluid pressure is sufficiently
high, fluids can even cause opening of the fault
(see Fig. 3a). For this reason, effective normal Rock Failure and Earthquakes
stress s is introduced as a difference between
normal stress sn and pore-fluid pressure p If a rock is critically stressed in the Earth’s crust,
the rock is fractured and this fracturing is associ-
s ¼ sn p; (10) ated with an earthquake. In principle, an
Earthquake Mechanisms and Stress Field 731
a bτ
n
increasing pore-fluid pressure p
u
Σ α
pore-fluid
pressure p
s3 s2 s1 σ
E
Earthquake Mechanisms and Stress Field, the slip vector and a is the deviation between the slip
Fig. 3 Role of pore-fluid pressure in the rock. (a) Open- vector and the fault (the so-called slope angle, see
ing of the fault caused by fluids. (b) Shift of Mohr’s circles Vavryčuk 2011b)
due to increase of pore-fluid pressure. S is the fault, u is
sV
Normal
s V > s H max > s H min
P
T
s H min
s H max
sV Strike-slip T
s H max > s V > s H min
s H max
P
s H min
sV Reverse
s H max > s H min > s V
T
s H max P
s H min
Earthquake Mechanisms and Stress Field, (right). The focal mechanisms with the P and T axes are
Fig. 4 Anderson’s classification scheme of stress in the shown in the lower-hemisphere equal-area projection
Earth’s crust (left) and corresponding faulting regimes
732 Earthquake Mechanisms and Stress Field
a b
tc
t
tc
q t 3
1 2θ
2θ
T0 0 s
T0 0 s3 s1
2
q 3
Earthquake Mechanisms and Stress Field, faulting regimes on the Mohr failure envelope tc(s).
Fig. 5 Griffith failure criterion. (a) Scheme of a cylindri- Point 1 – pure tensile faulting, points 2 – transition
cal specimen with a fracture created by loading with between tensile and pure shear faulting, points 3 – pure
stresses s1 > s2 ¼ s3 (left) and the corresponding shear faulting. Angle y defines the deviation of the fracture
Mohr’s circle diagram (right). (b) Position of different from the s1 direction, T0 is the uniaxial tensile strength
or equivalently
A variety in possible orientations of unstable
fault planes is demonstrated in Fig. 7. The left-
tc ¼ C þ mðsn pÞ; (15) hand plot of Fig. 7a shows the Mohr’s diagram,
the failure criterion and the positions of randomly
where sn is the normal stress and p is the pore distributed unstable fault planes satisfying the
pressure. Friction m for fractures was measured failure criterion. The middle and right-hand
on rock samples in laboratory and ranges mostly plots of Fig. 7a show the nodal lines and the
between 0.6 and 0.8 (Byerlee 1978). The values P (pressure) and T (tension) axes for the
of friction of faults in the Earth’s crust are sim- corresponding focal mechanisms. The nodal
ilar (Vavryčuk 2011a) but lower values like lines and P/T axes inform about the predominant
0.2–0.4 have also been reported for some large- type of faulting and about the scatter in orienta-
scale faults like the San Andreas fault (Scholz tions of the unstable fault planes. Predominant
2002). faulting and its scatter are also projected into a
If the Mohr-Coulomb failure criterion is satis- scatter of the P/T axes, which form clusters of a
fied (red area in Fig. 6), the fault becomes unsta- specific shape and size (Fig. 7a, right-hand plot).
ble and an earthquake occurs along this fault. The The form of the clusters is best demonstrated
higher the shear stress difference, Dt ¼ t tc , using the projection of the Mohr-Coulomb failure
the higher the instability of the fault and the criterion (Fig. 7b, blue line in the left-hand plot)
higher the susceptibility of the fault to be acti- onto the focal sphere (Fig. 7b, right-hand plot).
vated. A fault most susceptible to failure is called The failure criterion curve splits into two closed
the “principal” fault (Vavryčuk 2011a) being curves for the P axes as well as for the T axes on
defined by the point in which the Mohr-Coulomb the focal sphere corresponding to the failure con-
failure criterion touches the Mohr’s circle dia- ditions in the upper and lower half-planes in the
gram (blue point in Fig. 6) Mohr’s circle diagram. The two P/T failure
734 Earthquake Mechanisms and Stress Field
a N N
σ1
σ3
σ3 σ2 σ1 σ2
b N N
σ1
σ3
σ3 σ2 σ1
σ2
Earthquake Mechanisms and Stress Field, Mohr’s diagram (left), corresponding nodal lines (in the
Fig. 7 Focal mechanisms associated with unstable fault middle) and the P/T axes (right). The P/T axes form the
planes. (a) Randomly distributed fault planes inside the P (red)/T (blue) failure curves displaying the two-wing
unstable area of the Mohr’s diagram (left), corresponding pattern. The s1, s2 and s3 stress axes have directions
nodal lines (in the middle) and the P/T axes (right). The (azimuth/plunge): 308 /44 , 209 /9 and 110 /44 . The
P (pressure) axes are marked by the red circles, the T - shape ratio R is 0.5. The azimuth angle is measured clock-
(tension) axes by the blue crosses. (b) Fault planes wise from north (denoted as N), and the plunge angle from
corresponding to the Mohr failure envelope in the the horizontal plane
curves may intersect each other or be separated. the range from 0 to 1 (see Fig. 8) using the
The area inside the failure curves define positions following formula (Vavryčuk et al. 2013):
of the P/T axes of focal mechanisms conceivable
under the given stress regime. The pattern when t mð s s1 Þ
the P/T axes form two distinct sub-clusters is I¼ ; (16)
t c mð s c s1 Þ
called as the “two-wing” or the “butterfly-wing”
pattern (Vavryčuk 2011a). The butterfly wings
are well separated provided that friction is high where tc and sc are the shear and effective normal
(0.5 or higher). If friction is low, the wings come stresses along the principal fault (blue dot in
closer or they overlap. Fig. 6, red dot in Fig. 8), and t and s are the
shear and effective normal stresses along the
Fault Instability analysed fault (black dot in Fig. 8).
Since differently oriented faults have a different Since Eq. 16 is independent of absolute stress
susceptibility to be activated and thus being dif- values, the fault instability I can be evaluated just
ferently unstable in the given stress field, a quan- from friction m, shape ratio R,
tity which measures this instability can be
introduced. For example, the fault instability s1 s2
R¼ ; (17)
I of all fault orientations can be defined in s1 s3
Earthquake Mechanisms and Stress Field 735
E
s3 s2 s1 s
Earthquake Mechanisms and Stress Field, oriented fault with instability I. Quantities t and s are the
Fig. 8 Definition of the fault instability in the Mohr’s shear and the effective normal stresses, respectively; s1,
diagram. The red dot marks the principal fault character- s2 and s3 are the effective principal stresses
ized by instability I ¼ 1. The black dot marks an arbitrarily
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
t¼ n21 þ ð1 2RÞ2 n22 þ n23 n21 þ ð1 2RÞn22 n23 : orientations of unstable faults under stress field
(21) with inclined principal stress directions.
The fault instability can be calculated using Eqs. 20 Principal Focal Mechanisms
and 21 for all orientations of fault planes and As mentioned above, the higher the instability,
projected into the focal sphere (see Fig. 9). Such the higher the probability of the fault being acti-
figure is, in particular, helpful when analysing vated. Obviously, a prominent fault plane is the
736 Earthquake Mechanisms and Stress Field
Earthquake Mechanisms
a b N
and Stress Field,
Fig. 10 Principal focal
mechanisms. (a) Full
Mohr’s circle diagram, (b)
principal nodal lines and s1
o
P/T axes, and (c, d) o
o
principal focal
mechanisms. The blue dot +
in (a) marks the principal + s3
fault plane, the arrows in +
+
(c) and (d) denote the nodal
lines corresponding to the
principal fault plane. Point
B in (c) and (d) denotes the
neutral axis, N denotes s2
north (After Vavryčuk
(2011a))
c N d N
P
P o
o
T
T
B B
principal fault (Vavryčuk 2011a). This plane is The orientation of the principal fault planes
optimally oriented in the stress field and has the and principal focal mechanisms depends on stress
highest instability I = 1 (see Figs. 6 and 8). Each and fault friction. The B (neutral) axes of both
stress allows for the existence of two distinct principal focal mechanisms coincide with the s2
principal faults and two distinct principal focal axis (Fig. 10b). The P/T axes lie in the s1–s3
mechanisms (Fig. 10). The reason for the exis- plane: the s1 axis is in the centre between the
tence of two principal mechanisms is the same two P axes, and the s3 axis is in the centre
as for observing the two-wing pattern of the between the two T axes. The principal fault
failure curves and two clusters of focal planes are those nodal planes whose deviation
mechanisms – one principal fault lying in the from the s1 axis is less than 45 . Deviation y
upper half-plane and the other principal fault in between the two principal fault planes is
the lower half-plane of the Mohr’s circle diagram. expressed by a formula similar to that for the
The P/T axes of the principal focal mechanisms 2-D stress:
lie within the two P/T butterfly wings. If the area
of unstable faults (see Fig. 6) is decreased, for m1 ¼ tan 2y: (22)
example, by decreasing pore pressure, the wings
become smaller being shrunk to the P/T axes of Since the relation between the principal focal
the principal focal mechanisms in the limit. mechanisms and the orientation of stress is
Earthquake Mechanisms and Stress Field 737
simple, the inversion for stress orientation from background stress field. In this case, inverting
principal focal mechanisms is straightforward. for stress should be performed from earthquakes
On the other hand, the shape ratio cannot be clustered not only in space but also in time and
retrieved from the principal focal mechanisms. periods between and after a large earthquake
The shape ratio constrains the mutual relation should be distinguished.
between the size and shape of the P and If the above-mentioned assumptions are rea-
T butterfly wings. Therefore, a rather large set sonably satisfied, the stress inversion methods are
of focal mechanisms is needed to map the wings capable to determine four parameters of the stress
accurately. Methods for inversion of stress tensor: three angles defining the directions of the
including the shape ratio from a set of focal principal stress directions, s1, s2 and s3, and
E
mechanisms are described below. shape ratio R. The methods are unable to recover
the remaining two parameters of the stress tensor.
Therefore, the stress tensor is usually searched
Inversion for Stress From Earthquake with the normalized maximum compressive
Mechanisms stress
Eq. 23 is not applied. Subsequently, Eq. 25 is t ¼ ½ t11 t12 t13 t22 t23 T , (27)
expressed in the matrix form
At ¼ s; (26)
A is the 3 5 matrix calculated from fault nor-
where t is the vector of stress components mal n,
2 3
n1 n22 þ 2n23 n2 1 2n21 n3 1 2n21 n1 n22 þ n23 2n
1 n2 n3
4 n2 n2 þ n2 n1 1 2n22 2n 2 5
1 n2 n32 n2 n1 2þ 2n3 2 n3 1 2n22 ;
2 2
1 3 (28)
n3 2n21 n22 2n1 n2 n3 n1 1 2n3 n3 n1 2n2 n2 1 2n3
and s is the unit direction of the slip vector. fault orientations by applying the fault instabil-
Extending Eq. 28 for focal mechanisms of ity constraint. According to this constraint, the
K earthquakes with known fault normals n and fault is identified with that nodal plane which
slip directions s, a system of 3 K linear equa- has a higher value of instability calculated
tions for five unknown components of stress ten- using Eq. 20. The stress is calculated in itera-
sor is obtained. The system is solved using the tions and overall friction m on faults is also
generalized linear inversion in the L2-norm (Lay determined. Numerical tests show that the iter-
and Wallace 1995, their Section 6.4) ative stress inversion is fast and accurate and
performs better than the standard Michael’s
t ¼ Ag s: (29) inversion.
a b
τ τ
Tresca Mohr-Coulomb
E
s3 s2 s1 s s3 s2 s1 s
Earthquake Mechanisms and Stress Field, Fig. 11 Tresca and Mohr-Coulomb failure criteria. The Tresca failure
criterion is unphysical and corresponds to a fault with zero friction. The red dot marks the principal fault plane
tij XK
ðkÞ ðkÞ
0.6 reduces the maximum shear stress by 13 %
ttotal ¼ n s ; (31) only. Secondly, numerical tests show that if focal
s
tmax k¼1 i j
mechanisms inverted display a sufficient variety,
maximizing the SSSC value leads to a small bias in
where
the retrieved stress directions. Thirdly, maximiz-
s1 s3 ing the SSSC value is quite robust with respect to
tmax ¼ : (32) the errors present in focal mechanisms.
2
and the optimum stress tensor is calculated ana- Gephart and Forsyth Method
lytically (see Angelier 2002). An alternative Another way how to solve the problem with ambig-
approach defines the total SSSC value in the uous fault plane orientations in stress inversions is
form: proposed by Gephart and Forsyth (1984). The
method is based on minimizing the deviation
X
K
ðkÞ ðkÞ between direction N of shear stress and slip direc-
ttotal
s ¼ tij ni s j ; (33) tion s along the fault. This deviation is calculated for
k¼1
both options of the fault orientation and the lower
value enters into the total sum over all K focal
with the normalization condition
mechanisms minimized during the inversion:
s21 þ s22 þ s23 ¼ 1: (34)
X
K
S¼ acos NðkÞ sðkÞ : (35)
and with the zero trace condition Eq. 24. In this k¼1
case, the inversion for the optimum stress tensor
can be performed using a grid search through the The method was further modified and improved
principal stress directions and shape ratio R. The by Lund and Slunga (1999) who applied a more
fit function is maximized in the L1-norm. physical criterion for identifying the true fault
Although the condition of maximizing the planes. The criterion is based on the evaluation
SSSC value looks as a rather rough approximation, of instability of the fault using Eq. 20 and the
its use in practice is advocated by Angelier (2002) fault plane is identified with that nodal plane
for the following reasons. Firstly, the activated which is more unstable and thus more susceptible
fault planes are located mostly close to the top of to failure under the given stress field. The misfit
the Mohr’s circle diagram. For example, friction of function is minimized in the L1-norm by using
740 Earthquake Mechanisms and Stress Field
Earthquake Mechanisms a b
and Stress Field,
Fig. 12 Example of data
used in numerical tests of
stress inversions. The plots
show 200 noise-free focal
mechanisms selected to
satisfy the Mohr-Coulomb
failure criterion. Left/right
plots – dataset with a full/
reduced variety of focal
mechanisms. (a, b) Mohr’s
circle diagrams, (c, d) P/T
axes and (e, f) c d
σ3
corresponding nodal lines.
The P axes are marked by σ3
the red circles and T axes
by the blue crosses in (c) σ1 σ1
and (d). The s1, s2 and s3
stress axes are (azimuth/
plunge): 115 /65 , 228 / σ2 σ2
10 and 322 /23 ,
respectively. Shape ratio
R is 0.70, cohesion C is
0.85, pore pressure p is zero
and friction m is 0.60. The e f
minimum instability of
faults is 0.82
the robust grid search inversion scheme. It is also any stress inversion are the number of focal
possible to exclude a predefined number of out- mechanisms inverted, their accuracy and vari-
liers from the misfit function the result of the ability. The inversions yield orientations of the
inversion to be less sensitive to large errors in principal stress axes and shape ratio R.
input data. Since the misfit function is usually Numerical tests performed with sets of 25–250
smooth and simple, the time-consuming grid focal mechanisms are presented. The stress ten-
search can be substituted by any non-linear inver- sor is fixed for all datasets. The focal mechanisms
sion method. For example, the gradient or sim- are selected to satisfy the Mohr-Coulomb failure
plex methods can be applied with a success. This criterion (see Fig. 12a, b) and subsequently they
applies also to the Angelier method defined by are used for the calculation of moment tensors.
Eqs. 33 and 34. The moment tensors were contaminated by uni-
form noise ranging from 0 to 50 % of the norm
of the moment tensor (calculated as the maxi-
Accuracy of Stress Inversion Algorithms: mum of absolute values of the moment tensor
Numerical Tests eigenvalues). The noisy moment tensors were
decomposed back into strikes, dips and rakes
The robustness and accuracy of the stress inver- of noisy focal mechanisms inverted for stress.
sion methods can be efficiently tested using The deviation between the true and noisy fault
numerical modelling. The key parameters of normals and slips attained values from 0 to 25
Earthquake Mechanisms and Stress Field 741
M ic h a e l [°] M ic h a e l [%]
15
50 50 30
100 10 100
20
N
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
100 10 100
20
N
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
100 10 100
20
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
100 10 100
20
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
Earthquake Mechanisms and Stress Field, the method of Michael (1984), Angelier (2002), Gephart
Fig. 14 Mean error of principal stress directions (left) and Forsyth (1984) and Lund and Slunga (1999). The
and of the shape ratio (right) as a function of the number of method of Angelier (2002) employs Eqs. 33 and 34. The
inverted focal mechanisms and noise in moment tensors. errors are calculated from 50 random realizations of noise.
The data set with the full variety of focal mechanisms The errors in the stress directions are in degrees, the errors
(two-wings data) is inverted. The errors are shown for in the shape ratio are in percent
Earthquake Mechanisms and Stress Field 743
100 10 100
20
N
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
E
Angelier [°] Angelier [%]
15
50 50 30
100 10 100
20
N
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
100 10 100
20
N
150 150
5
10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
100 10 100
20
N
N
150 150
5 10
200 200
250 0 250 0
0 20 40 0 20 40
Noise [%] Noise [%]
Earthquake Mechanisms and Stress Field, Fig. 15 The same as for Fig. 14, but for the data set with the reduced
variety of focal mechanisms (one-wing data)
The mentioned stress inversions can be exem- stress regimes, was presented, for example, by
plified on several applications. The simplest Vavryčuk (2006) for deep earthquakes in the
approach, when the area under study is Tonga subducting slab (see Fig. 16). The slab is
subdivided just into two subareas of different considerably deformed at depths between
744 Earthquake Mechanisms and Stress Field
25° S 60
50
σ° 1
30° S σ2 40
°
30
σ3
°
20
35° S
Earthquake Mechanisms and Stress Field, (plus signs) of focal mechanisms distributed over the focal
Fig. 16 Inversion for stress in the Tonga subduction sphere for earthquakes from the northern (upper plot) and
zone. Left plot – the map view with epicentres of earth- southern (lower plot) slab segments. Right plots – the misfit
quakes. Green dots mark earthquake foci at depths between function for the principal stress axis s1, defined as the
100 and 500 km, blue/red dots mark deep-focus earth- average deviation (in degrees) between the predicted
quakes at depths greater than 500 km in the northern/south- shear stress directions and the observed slips at the faults.
ern cluster. The grey arrows show the horizontal The Gephart and Forsyth inversion method (1984) was
projections of the maximum compression in the northern/ applied. The optimum directions of the principal stresses
southern cluster. Middle plots – P axes (circles) and T axes are marked by circles (Modified after Vavryčuk (2006))
500 and 670 km and can be subdivided into two revealed systematic rotations of stress related to
segments of different orientations characterized the occurrence of the Izmit earthquake. The rota-
by different principal stress directions. The stress tion of the principal stress directions have been
inversion shows that the maximum compressive observed also after the 1999 Mw 7.6 Chi-Chi
stress is directed along the down-dip motion of earthquake in Taiwan (Wu et al. 2010) and for
the slab in both segments. A more detailed map- the 2011 Mw 9 Tohoku-oki earthquake in Japan
ping of lateral variations of stress obtained from (Yang et al. 2013). Spatial and temporal varia-
focal mechanisms was published by Hardebeck tions of stress from focal mechanisms have been
(2006) for Southern California who employed the studied also for New Zealand (Townend
stress inversion code SATSI developed by et al. 2012) with focus on the 2010 Mw 6.2
Hardebeck and Michael (2006). This code inverts Christchurch earthquake, and for other
simultaneously for stress in all subareas of the seismoactive regions including geothermal fields
region under study and minimizes the difference in which seismicity is induced by fluid injections
in stress between adjacent subareas to suppress (Martı́nez-Garzón et al. 2013).
over fitting of noisy data (see Fig. 17). A modifi-
cation of this code, called MSATSI and working
in the Matlab environment, was applied by Summary
Ickrath et al. (2014) to analyse spatial and tem-
poral variations of stress prior and after the 1999 The stress field and fracture processes in the
Mw 7.4 Izmit earthquake in Turkey. This analysis Earth’s crust are closely related phenomena.
Earthquake Mechanisms and Stress Field 745
Earthquake Mechanisms
and Stress Field,
Fig. 17 Stress inversion
for southern California.
A data set of 6957 focal
mechanisms is inverted for
the stress tensor at points on
a 0.1 grid. The bar
orientation shows the
direction of the maximum
horizontal stress axis, and
the shading of the bar
indicates the stress regime, E
following Simpson (1997).
Light grey lines show
mapped fault traces, and the
black line shows the San
Andreas Fault. The damped
inversion method of
Hardebeck and Michael
(2006) is used with
damping parameter e = 1
(After Hardebeck and
Michael (2006), their figure
5c)
Fracturing is controlled by shear stress on the fault, can be successfully recovered only by the Gephart
pore-fluid pressure, cohesion and fault friction. In and Forsyth (1984), Lund and Slunga (1999) and
addition, the orientation of the fault with respect to Vavryčuk (2014) methods provided that a high
the principal stress directions governs susceptibil- number of accurate focal mechanisms is available.
ity of the fault to be ruptured. An essential role in
the stress analysis play principal faults – the faults
which are optimally oriented for shear faulting and Cross-References
thus being the most unstable under the given stress
conditions. The type of faulting, the orientation of ▶ Earthquake Mechanism Description and
the activated faults and the direction of the slip Inversion
along the faults can be inverted for stress and its ▶ Earthquake Mechanisms and Tectonics
spatial and temporal variations within the Earth’s ▶ Non-Double-Couple Earthquakes
crust. The stress inversions work best if the fault
orientations are known. If the fault plane cannot be
uniquely identified from the focal mechanisms, the
References
inversions are less accurate. The inversions are Anderson EM (1951) The dynamics of faulting. Oliver &
capable to retrieve four stress parameters: the prin- Boyd, Edinburgh
cipal stress directions and the shape ratio. The Angelier J (2002) Inversion of earthquake focal mecha-
accuracy of the results depends on the applied nisms to obtain the seismotectonic stress IV – a new
method free of choice among nodal lines. Geophys
inversion method, on the number of inverted J Int 150:588–609
focal mechanisms, their errors and variety. The Bott MHP (1959) The mechanics of oblique slip faulting.
most critical parameter is the shape ratio, which Geol Mag 96:109–117
746 Earthquake Mechanisms and Tectonics
Byerlee J (1978) Fiction of rocks. Pure Appl Geophys Simpson RW (1997) Quantifying Anderson’s fault types.
116:615–626 J Geophys Res 102:17909–17919
Fischer T, Guest A (2011) Shear and tensile earthquakes Townend J, Sherburn S, Arnold R, Boese C, Woods L (2012)
caused by fluid injection. Geophys Res Lett Three-dimensional variations in present-day tectonic
38, L05307. doi:10.1029/2010GL045447 stress along the Australia–Pacific plate boundary in
Fowler CMR (1990) The solid earth. An introduction to New Zealand. Earth Planet Sci Lett 353–354:47–59
global geophysics. Cambridge University Press, Vavryčuk V (2006) Spatially dependent seismic anisotropy
Cambridge in the Tonga subduction zone: a possible contributor to
Gephart JW, Forsyth DW (1984) An improved method for the complexity of deep earthquakes. Phys Earth Planet
determining the regional stress tensor using earth- Inter 155:63–72. doi:10.1016/j.pepi.2005.10.005
quake focal mechanism data: application to the San Vavryčuk V (2011a) Principal earthquakes: theory and
Fernando earthquake sequence. J Geophys Res observations from the 2008 West Bohemia swarm.
89:9305–9320 Earth Planet Sci Lett 305:290–296. doi:10.1016/j.
Hardebeck JL (2006) Homogeneity of small-scale earth- epsl.2011.03.002
quakes faulting, stress, and fault strength. Bull Seismol Vavryčuk V (2011b) Tensile earthquakes: theory, model-
Soc Am 96:1675–1688 ing, and inversion. J Geophys Res 116, B12320.
Hardebeck JL, Michael AJ (2006) Damped regional-scale doi:10.1029/2011JB008770
stress inversions: methodology and examples for Vavryčuk V (2014) Iterative joint inversion for stress and
southern California and the Coalinga aftershock fault orientations from focal mechanisms. Geophys
sequence. J Geophys Res 111, B11310 J Int 199(1):69–77. doi:10.1093/gji/ggu224
Heidbach O, Tingay M, Barth A, Reinecker J, Kurfeß D, Vavryčuk V, Bouchaala F, Fischer T (2013) High-
M€uller B (2008) The world stress map database release resolution fault image from accurate locations and
2008. doi:10.1594/GFZ.WSM.Rel2008 focal mechanisms of the 2008 swarm earthquakes in
Ickrath M, Bohnhoff M, Bulut F, Dresen G (2014) Stress West Bohemia, Czech Republic. Tectonophysics
rotation and recovery in conjunction with the 1999 590:189–195. doi:10.1016/j.tecto.2013.01.025
Izmit Mw 7.4 earthquake. Geophys J Int 196:951–956 Wallace RE (1951) Geometry of shearing stress and relation
Jaeger JC, Cook NGW, Zimmerman RW (2007) Fundamen- to faulting. J Geol 59:118–130. doi:10.1086/625831
tals of rock mechanics, 4th edn. Blackwell, Oxford Wu Y-M, Hsu Y-J, Chang C-H, Teng LS, Nakamura
Julian BR, Miller AD, Foulger GR (1998) Non-double- M (2010) Temporal and spatial variation of stress
couple earthquakes. 1. Theory. Rev Geophys field in Taiwan from 1991 to 2007: insights from
36:525–549 comprehensive first motion focal mechanism catalog.
Lay T, Wallace TC (1995) Modern global seismology. Earth Planet Sci Lett 298:306–3016
Academic, New York Yang,Y-R, Johnson, KM, Chuang, RY (2013) Inversion
Lund B, Slunga R (1999) Stress tensor inversion using for deviatoric absolute crustal stress using focal mech-
detailed microearthquake information and stability anisms and coseismic stress changes: The 2001 M9
constraints: application to Ölfus in southwest Iceland. Tohoku-oki, Japan, earthquake, J Geophys Res
J Geophys Res 104:14947–14964 118:5516–5529
Martı́nez-Garzón P, Bohnhoff M, Kwiatek G, Dresen Zang A, Stephansson O (2010) Stress field of the earth’s
G (2013) Stress tensor changes related to fluid injec- crust. Springer, New York
tion at The Geysers geothermal field, California. Zoback MD (2007) Reservoir geomechanics. Cambridge
Geophys Res Lett 40:2596–2601 University Press, Cambridge
Maury J, Cornet FH, Dorbath L (2013) A review of
methods for determining stress fields from earthquake
focal mechanisms: application to the Sierentz 1980
seismic crisis (Upper Rhine graben). Bull Soc Geol Earthquake Mechanisms and
Fr 184(4–5):319–334
Mavko G, Mukerji T, Dvorkin J (2009) The rock physics
Tectonics
handbook. Tools for seismic analysis in porous media.
2nd edition. Cambridge University Press, Cambridge Jochen Braunmiller
Michael AJ (1984) Determination of stress from slip data: University of South Florida, Tampa, FL, USA
faults and folds. J Geophys Res 89:11517–11526
Michael AJ (1987) Use of focal mechanisms to determine
stress: a control study. J Geophys Res 92(B1):357–368
Ruff LJ (2002) State of stress within the earth. In: Lee Synonyms
WHK, Kanamori H, Jennings PC, Kisslinger C (eds)
International handbook of earthquake and engineering
seismology. Part A. Academic, New York
Earthquake source mechanisms; Fault plane solu-
Scholz CH (2002) The mechanics of earthquakes and tions; Focal mechanisms; Plate tectonics; Seis-
faulting. Cambridge University Press, Cambridge mic moment tensors; Seismicity and tectonics
Earthquake Mechanisms and Tectonics 747
Most earthquakes occur at the boundary of earthquake data are in the 30 –60 range.
between two tectonic plates. These are interplate Thrust (reverse) faults strike perpendicular to
earthquakes. Fewer earthquakes occur inside the convergence directions and often have a
plates, and these are intraplate earthquakes. shallow-dip angle (<30 ); this is typical for
The depth of earthquakes within the litho- megathrust subduction zone earthquakes. Trans-
sphere is usually restricted to the shallow part of form faults are oriented parallel to the relative
the lithosphere that deforms in a brittle fashion. In motion direction and have right- or left-lateral
the continental crust, the depth cutoff is often strike-slip mechanisms with near-vertical fault
associated with the 350 C isotherm which is planes. Earthquake slip vectors describe the rela-
reached at mid-crustal depth; the actual cutoff tive motion of two fault surfaces relative to each
depth (between about 10 and 25 km) depends on other. They have been used extensively in the
the temperature gradient. At temperatures above past to estimate relative plate motion directions;
350 C, the bulk material of continental crust however, we know now that spreading directions
deforms in a ductile, plastic fashion. In the oce- and slip on subduction zone thrust earthquakes
anic lithosphere, due to different mineral compo- are not necessarily parallel to plate motion
sition, ductile deformation sets in above about directions.
600 C, which, in oceanic lithosphere, is usually Earthquake fault plane solutions provide an
reached in the uppermost mantle (the depth overall, average description of the seismic source
depends on lithosphere cooling and can be esti- effectively treating it as a point source. Effects of
mated using a simply half-space cooling model). slip on a finite fault or complex ruptures involv-
Some examples exist where earthquakes occur in ing several faults or fault segments are observed
the lower crust or uppermost mantle of continen- seismically and in the field. However, these
tal lithosphere, such as in parts of the East African topics are not addressed in this entry.
Rift System and of the Tibetan Plateau. Some Plate boundaries in oceanic lithosphere are
researchers suggest that these events are located often narrower and simpler than in continental
in the uppermost mantle, which, at about the lithosphere. They will be discussed first with
same depth and temperature, is stronger than Fig. 2 showing a simple conceptual model.
crustal material, while others argue that under
certain circumstances earthquakes can occur
throughout the continental crust; this is a topic Mid-Ocean Ridges and Oceanic
of active research. Deep earthquakes within the Transform Faults
oceanic lithosphere of subducting plates are
another exception; they are discussed below. New oceanic lithosphere at mid-ocean ridge
The source mechanisms (also called fault spreading centers is due to extrusive and intrusive
plane solutions and focal mechanisms or, using volcanism. The extensional nature of spreading is
a more general description, seismic moment ten- reflected in the observed normal-faulting earth-
sors) of tectonic earthquakes are generally due to quake mechanisms. Figure 3 shows the global
shear faulting on a fault plane. In the moment distribution of spreading ridges and the normal-
tensor description, this is a pure double-couple faulting mechanisms of larger (moment magni-
source, i.e., no volume change or other non-shear tude Mw about 5 and larger) earthquakes along
components (such as a compensated linear vector them. The mechanisms have nodal planes parallel
dipole) are present at the source. The earthquake to the ridge axes.
mechanisms that correspond to extension, com- By design, Fig. 3 shows the distribution of
pression, and lateral motion are normal, thrust M 5 earthquakes along the global ridge system.
(reverse), and strike-slip fault plane solutions; Most of these events occur along ridges with slow
basic examples are shown in Fig. 1. Normal faults (55 mm/year) spreading rates. Few normal-
strike perpendicular to the extension direction, faulting earthquakes on ridges exceed
and fault dips commonly found from the analysis Mw = 6 and the largest event, which occurred
Earthquake Mechanisms and Tectonics 749
Earthquake Mechanisms
and Tectonics,
Fig. 2 Earthquake
mechanisms in oceanic
environments. Earthquakes
occur mainly on active
faults (double line:
spreading center; single
line: transform fault;
barbed line: subduction
zone, barbs point in
direction of subducting
plate). Dashed line is an E
inactive fracture zone.
Arrows show relative
motion sense
40 20 0 20 40
60
60
50
50
50
50
40
40
30 20 10 0 10 20 30
30 20 10 0 10 20 30
50
50
50
50
60
60
5 6 7 normal
Earthquake Mechanisms and Tectonics, project. Note lack of earthquakes along the fast-spreading
Fig. 3 Global distribution of spreading ridges (heavy ridges (e.g., East Pacific Rise) compared to slow-
lines) and oceanic normal-faulting focal mechanisms spreading ridges (e.g., Mid-Atlantic) (From Bird
(beach balls) for 1977–1998 from the Global CMT et al. 2002) from AGU publications
750 Earthquake Mechanisms and Tectonics
Earthquake Mechanisms and Tectonics, Table 1 Some examples of significant earthquakes for different tectonic
environments with source mechanism type and size (moment magnitude). Type: NF normal faulting, TF thrust faulting,
and SS strike-slip faulting. Events are largest (or among largest) of a given type since about 1900. Name gives source
region. Ref refers to literature (for further details on a given earthquake). Unless noted otherwise, all earthquake sizes
(Mw) in this entry are from the GCMT or USGS sources
Environment Date Type Size (Mw) Name References
Mid-ocean ridge 21 Mar 1988 NF 6.4gcmt Gakkel Ridge
Ridge-to-ridge oceanic transform 14 Mar 1994 SS 7.1gcmt Romanche 1
Transform
Megathrust subduction 22 May 1960 TF 9.5usgs Chile 2
Outer rise (subduction) – intraplate 29 Sep 2009 NF 8.1usgs Samoa Islands 3
Overriding plate 25 Apr 1992 TF 7.2gcmt Petrolia, California 4
(subduction) – intraplate
Intermediate deep 9 Dec 2007 TF 7.9gcmt Fiji-Tonga
(subduction) – intraplate
Deep (subduction) – intraplate 24 May 2013 NF 8.4gcmt Sea of Okhotsk 5
Continental – collision 15 Jan 1934 TF? 8.1usgs Bihar-Nepal 6
Continental – extension 28 Mar 1970 NF 7.17 Gediz, Turkey 7
Continental – strike-slip 14 Nov 2001 SS 7.8gcmt Kunlunshan, Tibet 8
Continental – intraplate 26 Jan 2001 TF 7.6gcmt Gujarat, India 9
Oceanic – intraplate 11 Apr 2012 SS 8.6usgs Indian Ocean 10
gcmt size from Global Centroid Moment Tensor (GCMT 2014) project, usgs size from the U.S. Geological Survey
(USGS 2014), 1 Abercrombie and Ekström (2003), 2 Kanamori and Cipar (1974), 3 Lay et al. (2010), 4 Oppenheimer
et al. (1993), 5 Ye et al. (2013), 6 Bilham et al. (2001), 7 Braunmiller and Nabelek (1996), 8 Antolik et al. (2004),
9 Antolik and Dreger (2003), 10 Meng et al. (2012)
along the ultraslow (20 mm/year) Gakkel Ridge ridges accounts for only a small percentage of
separating the North American and Eurasian the relative plate motions indicating the impor-
plates in the Arctic region, reached only tance of aseismic slip.
Mw = 6.4 (Table 1). Maximum earthquake size The mid-ocean ridge segments are connected
and the frequency of large earthquake occurrence to each other by oceanic transform faults. Ridge-
are related to spreading rate. Note the continuous transform intersections are often nearly perpen-
band of earthquake mechanisms along the slow- dicular. Motion along transforms is lateral and
spreading Mid-Atlantic and Central and South- parallel to relative plate motion directions. Trans-
west Indian ridges and the lack of events along forms offset ridge segments, and the offset direc-
the fast-spreading (here set as 80 mm/year and tion (i.e., whether the next segment is to the right
faster) ridges south of Australia and the East or left) determines motion sense and thus whether
Pacific Rise. Along the East Pacific Rise, a fault is right lateral or left lateral (Fig. 2). The
normal-faulting earthquakes rarely exceed offset-parallel nodal plane is usually the fault
M = 5. Differences between slow- and fast- plane, and both nodal planes are usually near
spreading ridges are due to the thermal structure vertical. Parts of a transform fault not between
of the lithosphere. two ridge segments become seismically inactive
Microearthquake activity along the entire and are called fracture zone. The Mendocino
ridge system – particularly along fast-spreading fracture zone, extending halfway across the
ridges – is probably vigorous based on observa- northeast Pacific Ocean, is a famous example;
tions from still relatively few, temporary seismic its active transform fault part is only about
observations from the ocean floor. Even after 200 km long and connects a spreading ridge
considering smaller earthquakes, the seismic with the Mendocino triple junction in northern
moment released by all earthquakes along the California.
Earthquake Mechanisms and Tectonics 751
50
60
60
50
50
50
40 30 20 10 0 10 20 30 40
40 30 20 10 0 10 20 30 40
50
50
140 160 180 160 140
50
50
60
60
5 6 7 8 thrust normal
Earthquake Mechanisms and Tectonics, project. Note earthquakes occur along almost all trans-
Fig. 4 Global distribution of oceanic transform faults forms independent of plate motion rate (From Bird
(heavy lines) and oceanic strike-slip focal mechanisms et al. 2002) from AGU publications
(beach balls) for 1977–1998 from the Global CMT
Earthquake Mechanisms and Tectonics, for interplate megathrust earthquake and state of stress for
Fig. 5 Schematic overview of some earthquake types intraplate earthquakes (diverging arrows, extension; con-
observed in subduction zones. Not all event types occur verging arrows, compression). Intermediate-deep event is
at each subduction zone. Focal mechanisms are in cross- an example for in-plane extension; some plates show
sectional view and are roughly scaled according to max- in-plane compression throughout the slab (then interme-
imum observed size. Arrows show relative motion sense diate and deep mechanisms are the same)
subducting plate beneath the overriding plate to cause destruction far away from the earth-
along a shallow-dipping fault. They occur land- quake rupture area itself.
ward from the trench (or, in general, beneath the As an oceanic plate starts to descend beneath
leading edge of the overriding plate). The mech- another plate, it has to bend to accommodate the
anisms have nodal planes parallel to the strike of change in plate motion from horizontal to an
the trench (and coast), and slip vectors, which inclined direction. The bending stresses are larg-
indicate the relative motion sense on the fault, est seaward of the trench at the outer rise and the
are often subparallel to the relative plate motion outer trench slope. The stresses due to bending
directions. and slab-pull forces cause earthquakes. Plate
The largest earthquake measured with seis- bending in the upper part of a subducting litho-
mometers is the 1960 Chile (Mw = 9.5) subduc- sphere plate is extensional resulting in normal
tion zone earthquake. The rupture dimensions faulting; such earthquakes are observed at many
corresponding to an earthquake of this size are subduction zones. Bending stresses in the lower
about 1,000 km rupture length, 200 km down-dip part are compressional resulting in thrust
rupture width, and 10–30 m average slip on the faulting; such earthquakes occur less frequently.
entire area that ruptured. Additional infamous The depth of transition from extension to com-
examples include the 1964 Alaska (Mw = 9.2), pression within a plate is an indicator for the
the 2004 Sumatra-Andaman (Mw = 9.1), and the stress state within oceanic lithosphere. T-axes of
2011 Tohoku, Japan (Mw = 9.0), earthquakes. normal faulting and P-axes of thrust earthquakes
Such gigantic earthquakes can be very devastat- are usually pointing in the direction of plate
ing, and they are often accompanied by large, motion with nodal planes subparallel to the strike
deadly tsunamis that travel across ocean basins of the trench system.
Earthquake Mechanisms and Tectonics 753
10 10
1991 628
World
World
5 5
0 0
100 150 200 250 300 350 400 450 500 550 600 650 700 100 150 200 250 300 350 400 450 500 550 600 650 700
27 29
26 28
World
World
27
25
26
24 25
23 24
100 150 200 250 300 350 400 450 500 550 600 650 700 100 150 200 250 300 350 400 450 500 550 600 650 700
Earthquake Mechanisms and Tectonics, percent of total. Note rapid decay of activity from 150 km
Fig. 6 Depth versus frequency of earthquake occurrence to 300 km depth, a small local maxima of events near
(top) and cumulative seismic moment release (bottom) for 400 km, and strong increase below about 530 km depth.
earthquakes below 100 km depth binned in 10 km steps. Scale at the bottom is in units of log(M0) dyne cm (subtract
See reference for data source. Left earthquakes with 7 for units of [Nm]) (From Estabrook 2004) from AGU
Mw < 5.75 and right larger events. Numbers in the top publications
are total number of earthquakes shown; vertical axis is
Away from plate boundaries, loading (stressing) new oceanic crust, cooling of oceanic litho-
rates are low and intraplate earthquakes conse- sphere, and geometrical complexities in plate
quently should occur less frequently than plate motions; observed focal solutions include nor-
boundaries events – their recurrence intervals mal, thrust, and strike-slip mechanisms. These
should be longer. earthquakes are usually not very large.
Seismicity in the New Madrid, Missouri, area A different type of oceanic intraplate earth-
seems to utilize zones of weakness that were quakes is possibly associated with initial forma-
originally generated by the Precambrian failed tion of new oceanic plate boundaries. The largest
Midcontinent Rift System. Three Mw 7.5 earth- strike-slip earthquake ever recorded (Mw = 8.6,
quakes occurred in the area in 1811–1812. The Table 1) occurred in April 2012 in the Indian
source mechanisms are not directly known but Ocean a few hundred kilometers off the west
probably involved strike-slip and thrust motions coast of northern Sumatra, Indonesia. The epi-
on several faults. center is located in a broad zone of deformation
Source mechanisms of intraplate earthquakes that marks the diffuse India-Australia plate
can be normal, thrust, or strike-slip depending on boundary. Plate boundary faults are not (yet)
the stresses inside the lithosphere plate. As such, developed and the stresses due to the slow rela-
intraplate source mechanisms are indicators of tive plate motion causes ruptures along zones of
the stress field, and their analysis is one of the preexisting weakness such as fracture zones. The
main tools available to estimate stresses in the 2012 sequence (including a Mw = 8.2 aftershock)
crust. In many regions, such as North America, shows a complex rupture pattern involving rup-
internal stresses are compressional, and most tures on several faults. The 1998 Mw = 8.1 earth-
earthquakes have thrust and strike-slip source quake near the Balleny Islands, which occurred a
mechanisms. The 2001 Mw = 7.6 Gujarat, few hundred kilometers from the Australia-
India, earthquake (Table 1) is an example of Antarctic plate boundary, also showed complex
intraplate thrust faulting that is probably a conse- faulting behavior and may represent a similar
quence of collisional plate boundary forces that case of reactivation or rupturing of new faults as
are exerted several hundred kilometers away the 2012 Indian Ocean earthquakes.
from the epicenter. A counterexample is the
European Cenozoic Rift System in the foreland
of the Alps that formed in response to the Alpine Summary
orogeny; most earthquakes north of the Alps have
either normal or strike-slip mechanisms. Intra- Focal mechanisms of tectonic earthquakes gen-
plate regions pose a challenge to earthquake pre- erally follow the deformation style that is
paredness because large, potentially destructive expected from the plate tectonic settings they
earthquakes occur very infrequently (on the order occur in. Thrust mechanisms are observed in
of thousand years or more) but could possibly subduction zones, collision zones, and other
occur almost anywhere within a plate. regions of compressional deformation. Normal-
Intraplate earthquakes also occur in oceanic faulting earthquakes occur along mid-ocean
plates. In many cases these earthquakes are ridges and in zones of continental extension.
located along the inactive fracture zones of trans- Strike-slip earthquakes are observed along trans-
form fault systems or relatively close to the oce- form faults in the ocean and on land where plates
anic ridge-transform system. The former are or crustal blocks slide laterally past each other.
possibly caused by stresses due to differential Most earthquakes occur at plate boundaries, but
cooling on either side of a fracture zone, and some do occur inside plates (intraplate earth-
mechanisms are either strike-slip or vertical quakes) away from boundaries. Plate boundaries
dip-slip (one nodal plane is vertical with up or in continental environments are usually broader
down motion along it). The latter likely represent than their oceanic counterparts. Deep earth-
local stress field variations due to formation of quakes are restricted to regions where ocean
Earthquake Mitigation Strategies Through Insurance 757
lithosphere is recycled back into the mantle at earthquake: rupture transfer through a large exten-
subduction zones. The largest earthquakes are sional step-over. Bull Seismol Soc Am 94:1173–1194
Bilham R, Gaur VK, Molnar P (2001) Himalayan seismic
shallow-dipping subduction zone thrust hazard. Science 293:1442–1444. doi:10.1126/
earthquakes. science.1062584
This entry only provides a general overview of Bird P, Kagan YY, Jackson DD (2002) Plate tectonics and
tectonic earthquakes and their source mecha- earthquake potential of spreading ridges and oceanic
transform faults. In: Stein S, Freymueller JT (eds)
nisms. Analysis of seismological data of individ- Plate boundary zones. American Geophysical Union,
ual earthquakes and earthquake sequences has in Washington, DC. doi:10.1029/GD30p0203
many cases shown complex rupture behavior Braunmiller J, Nabelek J (1996) Geometry of continental
involving several active faults, which may have normal faults: seismological constraints. J Geophys
Res 101:3045–3052 E
different orientations or even a different sense of Estabrook CH (2004) Seismic constraints on mechanisms
motion (such as strike-slip and normal faulting or of deep earthquake rupture. J Geophys Res
strike-slip and thrust faulting on nearby faults). In 109, B02306. doi:10.1029/2003JB002449
addition, slip on faults is not necessarily equally GCMT (2014) Global CMT web page, http://www.
globalcmt.org. Accessed 20 March 2014
distributed on a fault patch that ruptures during an Kanamori H, Cipar JJ (1974) Focal process of the great
earthquake; instead slip may show large spatial Chilean earthquake May 22, 1960. Phys Earth Planet
variations. The reader is referred to the seismo- Inter 9:128–136
logical literature for more information on these Lay T, Ammon CJ, Kanamori H, Rivera L, Koper KD,
Hutko AR (2010) The 2009 Samoa-Tonga great earth-
topics. quake triggered doublet. Nature 466:964–968.
doi:10.1038/nature09214
Meng L, Ampuero J-P, Stock J, Duputel Z, Luo Y, Tsai
Cross-References VC (2012) Earthquake in a maze: compressional rup-
ture branching during the 2012 Mw 8.6 Sumatra earth-
quake. Science 337:724–726. doi:10.1126/
▶ Earthquake Mechanism and Seafloor science.1224030
Deformation for Tsunami Generation Oppenheimer D et al (1993) The Cape Mendocino, Cali-
▶ Earthquake Mechanism Description and fornia earthquake of April 1992: subduction at the
triple junction. Science 261:433–438
Inversion USGS (2014) Magnitude 8 and greater earthquakes
▶ Earthquake Mechanisms and Stress Field since 1900, http://earthquakes.usgs.gov/earthquakes/
▶ Long-Period Moment-Tensor Inversion: The eqarchives/year/mag8/magnitude8_1900_date.php.
Global CMT Project Accessed 20 March 2014
Ye L, Lay T, Kanamori H, Koper KD (2013) Energy
▶ Mechanisms of Earthquakes in Aegean release of the 2013 Mw 8.3 sea of Okhotsk earthquake
▶ Mechanisms of Earthquakes in Iceland and deep slab stress heterogeneity. Science
▶ Mechanisms of Earthquakes in Vrancea 341:1380–1384. doi:10.1126/science.1242032
▶ Moment Tensors: Decomposition and
Visualization
▶ Non-Double-Couple Earthquakes
Earthquake Mitigation Strategies
Through Insurance
References
Guillermo Franco
Abercrombie RE, Ekström G (2003) A reassessment of the Head of Catastrophe Risk Research – EMEA,
rupture characteristics of oceanic transform earth-
Guy Carpenter & Company Ltd., London, UK
quakes. J Geophys Res 108:2225. doi:10.1029/
2001JB000814
Antolik M, Dreger DS (2003) Rupture process of the
26 January 2001 Mw 7.6 Bhuj, India earthquake from Synonyms
teleseismic broadband data. Bull Seismol Soc Am
93:1235–1248. doi:10.1785/0120020141
Antolik M, Abercrombie RE, Ekstrom G (2004) The Natural catastrophe risk management; Risk
14 November 2001 Kokoxili (Kunlunshan) Tibet financing; Seismic risk mitigation
758 Earthquake Mitigation Strategies Through Insurance
Retail Reinsurance
Broker Broker
Buy
Individual s
Sells Buys Sells Retrocedes
Policy Holder
Primary Reinsurance
Policy Ce Contract
de
s
Inv E
s
est
de
s
Ce
Corporate
Policy Holder
Buys Sells
ILS
Capital Markets (Cat Bond) SPV (Special
Puropse Vehicle)
Earthquake Mitigation Strategies Through Insurance, Fig. 1 A simplified schematic view of the (re)insurance
market
reinsurance industry can be traced back to the Since catastrophe events are, however, rela-
seventeenth and eighteenth centuries (Swiss Re tively rare, there is a pervasive lack of reliable
2013; Wirtz 2013). During that period, large cat- experience data on their likelihood and conse-
astrophic events such as the Great Fire of London quences. Therefore, assessing and pricing catas-
of 1666 or the Lisbon earthquake of 1755 trophe risk constitute challenges that would
cemented the importance of insurance principles persist until the appearance of numerical simula-
considered fundamental today, such as the need tion models during the second half of the twenti-
for global diversification of risk and the necessity eth century. Thanks to this innovation, the
of understanding the underlying physical mecha- quantification of probabilities and effects of
nisms. Probability theory, pioneered by Blaise physically complex catastrophes slowly starts
Pascal, Jacob Bernoulli, and Pierre de Fermat, becoming a standard practice.
marked the beginning of risk quantification Today, the (re)insurance industry is a sophis-
using mathematical approaches and establishing ticated amalgam of specialties and entities
the foundation of modern actuarial practices. connected through a variety of global links. To
As the concentration of valuable assets has understand how this risk management engine
increased in locations susceptible to natural works, it is useful to consider a simplified view
disasters, the risk exposure has also grown. The of the transactional relationships that exist in the
1906 San Francisco earthquake and the ensuing (re)insurance market (Fig. 1).
widespread fire constituted the largest catastro- Insurance is an arrangement by which a party
phe loss the (re)insurance sector had experienced (the insurer) agrees to compensate another party
at the time. Since then, catastrophes have grown (the insured) for a set of specific future losses in
more and more costly reaching the estimated exchange for a periodic payment (premium).
US$210 billion loss caused by the 2011 Tohoku Earthquake insurance policies may cover dam-
earthquake in Japan (Munich 2013). Moreover, a ages to the built environment, injury, casualty,
highly connected international economy has and business interruption. Both private individ-
uncovered new vulnerabilities derived from sup- uals and companies purchase insurance products
ply chains that can be disrupted by disasters once to protect their assets and in so doing cede their
considered distant (Nanto et al. 2011). risk to an insurance company, which effectively
760 Earthquake Mitigation Strategies Through Insurance
acts as a risk aggregator and diversifies the risk Earthquake Mitigation Strategies Through Insur-
across the population. ance, Table 1 Largest reinsurers by 2012 net reinsur-
ance written premium (US$ m)
An insurance company may in turn cede part
of its risk to one or several reinsurance compa- Reinsurer Premium
nies, who trade globally and therefore diversify Munich Reinsurance Co. (Germany) 35,797.2
risks further. To share some of its acquired risks, Swiss Reinsurance Co. (Switzerland) 25,344.0
Hannover Rueckversicherung AG 16,346.0
a reinsurer may also “retrocede” parts of their
(Germany)
portfolios to other reinsurance companies. Berkshire Hathaway Reinsurance Group/ 16,145.0
Insurance or reinsurance brokers are interme- General Re Corp. (US)
diaries whose objective is to optimize access to Lloyd’s of London (UK) 11,372.7
affordable coverage. A retail broker assists indi- SCOR (France) 11,285.6
vidual policyholders in obtaining the most appro- Reinsurance Group of America, Inc. (US) 7,906.6
priate policy to fulfill their needs from the most Partner Re (Bermuda) 4,573.0
adequate and affordable insurance provider. Sim- Everest Re (Bermuda) 4,081.1
ilarly, a reinsurance broker assists insurance MS&AD Holdings (Japan) 3,446.3
companies to access reinsurance worldwide and Korean Re (South Korea) 3,390.2
find the most appropriate coverage at the best Transatlantic Holdings, Inc. (US) 2,840.7
prices. They also work with reinsurers who wish Nipponkoa Sompo Japan Holdings (Japan) 2,717.4
Tokio Marine Group (Japan) 2,626.5
to obtain retrocessional coverage. In addition,
General Insurance Corp. of India (India) 2,533.6
brokers may provide a series of services such as
Source: Standard & Poor’s (2014)
claims processing, analytics, and strategy con-
sulting among others.
Today’s reinsurance market has a very strong Chile in 2010, as the country faced an economic
link with the broader financial and capital mar- loss estimated at US$30 billion arising from the
kets, through what is known as alternative risk February 27, MW 8.8 Maule earthquake off the
transfer or ART. The most common ART tech- coast of Concepción (Siembieda et al. 2012).
nique consists of securitizing the risk in such a Chile enjoys a strong presence of insurance com-
way that it can be sold in the form of a bond to panies that were able to assume close to one third
financial investors, who in return for providing of the total economic loss, about US$8.4 billion
capital, obtain an interest on their principal. Sov- in insured damages (Marsh 2014). These insur-
ereign governments can also interact with the ance companies were in turn well connected to
capital markets to obtain reinsurance funds the global reinsurance markets that provided
through catastrophe bonds. This niche “conver- them coverage to an astounding 95 % (Aon
gence market” brings capital from financial insti- Benfield 2011). This means that most of the
tutions into reinsurance operations. Note that, as insured loss was swiftly transferred to the inter-
shown in Fig. 1, individuals and companies not national reinsurance markets, where it had little
only participate actively in the subscription of impact in the finances of large capital providers.
insurance but, through hedge funds that purchase With this robust system in place, Chile’s private
catastrophe bonds, also contribute to the provi- market effectively transferred about 30 % of the
sion of capital for catastrophe reinsurance. total economic loss caused by the Maule event
Today, the reinsurance sector includes large, outside its borders without any direct government
diversified companies that actively trade in a intervention.
variety of risks worldwide. Table 1 shows the This success story worked hand in hand with
15 largest reinsurers according to the net reinsur- Chile’s strong enforcement of good construction
ance premium written in 2012 (Standard and practices over the decades coupled with compul-
Poor’s 2014). sory subscription of earthquake insurance for all
An example of how insurance systems provide properties purchased with a mortgage. Without a
earthquake risk mitigation was recently seen in strong legacy of proactive construction
Earthquake Mitigation Strategies Through Insurance 761
regulation, insurers would not be easily drawn to The paper by Clark (1986) brought the con-
this market. And without the compulsory system, cept of Monte Carlo simulation applied to the
the growth of insurance penetration would be assessment of hurricane losses to the actuarial
severely hampered. Indeed, many owners who community and paved the way for the foundation
had purchased their property outright with cash, of the first catastrophe modeling company,
a well-established investment strategy in Chile, Applied Insurance Research, in 1987. Clark
had no insurance protection at the time of the relied on the visionary numerical approaches
event. developed by Friedman (1975) and others. Early
Regulation and social policies stimulated the earthquake simulation models were simulta-
growth of the insurance market, and the combi- neously being developed in California supported
E
nation of these “hard” and “soft” resilience strat- by new probabilistic methods (Cornell 1968;
egies was identified as highly effective during the Esteva 1970; McGuire 2008).
early recovery period (Franco and Siembieda Earthquake models have become extraordi-
2010). Much of the success of the insurance sys- narily complex, and teams that develop them
tem in the Maule event was ultimately attributed have grown to include a wide array of specialists
to the fast claims settlement process. Ninety per- such as seismologists, geologists, engineers, stat-
cent of the claims were reported by affected isticians, and financial and software experts.
households within the first 30 days after the Most commercial earthquake models fundamen-
event, and more than 75 % had been settled tally consist of five important modules as shown
within less than a year (Marsh 2014). in Fig. 2. The first generates an event catalog, a
set of simulated events compatible with the avail-
able knowledge on the seismicity of a specific
Quantifying Earthquake Risk and the territory. This knowledge may include fault loca-
Rise of CAT Models tion data, recurrence intervals, fault geometry,
kinematic analyses, etc.
The transfer of earthquake risk from Chile to the The second module uses ground motion pre-
rest of the world is possible at a global level in diction equations, soil data, and fault-related
great part due to the existence of tools that help to effects, among other parameters, to generate
quantify it. These tools are known today as catas- fields of hazard intensities associated with each
trophe or CAT risk models. of the earthquakes contained in the event catalog.
The insurance and reinsurance industry has These ground motions are in turn fed into a pro-
traditionally used historical loss records to price cess that calculates possible damage distributions
fire, life, theft, and other risks on which there is across the affected area for different types of
plenty of experience data, and catastrophes were structures. The crux of this vulnerability module
originally considered in conjunction with more consists of a set of mathematical relationships in
common hazards. Ignoring their specific poten- the form of damage functions or fragility curves
tial loss contribution resulted in unforeseen insol- derived from structural engineering techniques
vencies for the insurance industry in events like such as the capacity spectrum method or
hurricane Andrew in 1992. In order to better nonlinear dynamic analysis. Exposure data from
assess catastrophe risks, companies needed expe- the user serves to define the types of properties
rience data that wasn’t available. for which the analysis needs to be executed.
During the late 1980s and 1990s, new risk Lastly, the financial module applies the terms
assessment platforms were being developed and contained in (re)insurance policies in order to
deployed in the market. Grossi and Kunreuther obtain a monetary estimate of insured loss based
(2005) present a comprehensive view of the his- on the computed physical damage.
tory and evolution of numerical CAT models that Only three model development companies,
simulate the physical processes involved in catas- AIR Worldwide, EQECAT, and Risk Manage-
trophe events. ment Solutions (RMS), had a serious foothold in
762 Earthquake Mitigation Strategies Through Insurance
Earthquake Mitigation Strategies Through Insurance, Fig. 2 Basic anatomy and elements of an earthquake risk
model
the industry during the early years of CAT model- limitations and uncertainties. As models become
ing. Later, other groups started operations in the more ubiquitous, the quantification of uncertainty
market occasionally specializing in regional or has gained in importance. Several studies have
peril-specific models such as Empresa de Riesgos addressed this topic in the context of hurricane
Naturales (ERN) based in Mexico and with a very (Miller 1999; Major 1999a, b) and earthquake
strong local relevance in Latin America or Risk models (Spence et al. 2003). These tools are
Frontiers based out of Macquarie University in also still of limited use to address the occurrence
Australia and specializing in Asia-Pacific perils. of very large or very rare events. When the sci-
Initiatives like the Oasis Loss Modeling Frame- entific knowledge is insufficient, all models tend
work aim at lowering the barrier to entry for to fail, as was evident in Japan when the MW 9.0
academics, consultants, and other market partic- Great Tohoku earthquake struck in an area where
ipants. The philosophy of this effort is to create a only up to MW 8.4 events were thought possible.
standardized platform that liberates potential The numerical estimates from CAT models
entrants from the common pains of software make it possible to calculate the frequency of
development and from the relatively specialized loss, which is expressed in the industry as an
financial computations involved in reinsurance exceedance probability curve or “EP curve,” a
analysis. In this way, scientists, engineers, and relationship between the value at risk and its
risk experts may be able to deploy their knowl- probability. This information is critical to price
edge more seamlessly onto channels easily acces- an insurance policy or a reinsurance contract. The
sible by the industry. pure premium of an insurance policy is defined as
During this technological revolution that its associated average annual loss, and solvency
started in the late 1980s, catastrophe risk (re) requirements are often expressed as probable
insurance has changed enormously. Rare would maximum losses at specific return periods, typi-
be the case today that an insurance or reinsurance cally at 100 years or 250 years.
company managed their catastrophe risks without By using numerical models to estimate the
any usage of CAT models. In the absence of necessary metrics, the industry has managed to
actual loss experience data, these numerical overcome the immediate problem of lack of his-
tools have been successful at leveraging scientific torical experience. However, the battle is not
information to produce a large body of simula- over. Data on physical and economic losses
tions from which the market can now objectively that would help produce better models continues
assess and price risks – albeit with significant to be scarce, often at disparate resolutions, and
Earthquake Mitigation Strategies Through Insurance 763
of varied quality. Therefore, the industry con- the actual loss with the aim of obtaining a finan-
tinues to look at diverse scientific or professional cial advantage. When the covered party experi-
initiatives such as the Global Earthquake Model ences a damaging event that causes loss, it issues
or the Learning from Earthquakes initiative a claim to the (re)insurer. This, in turn, assesses
spearheaded by the Earthquake Engineering the damage through a claim-adjusting process
Research Institute (EERI) as extremely valuable and establishes the corresponding payment, and
contributions. the (re)insured is compensated accordingly and
depending on the policy conditions of the agree-
ment. In addition to being very expensive and
Alternative Risk Transfer Solutions cumbersome, claim-adjusting processes provide
E
opportunities for the claimant to exaggerate the
CAT models have improved the way in which the loss or for the insurer to downplay it. Capital
market manages and prices its catastrophic risks, providers who do not have the mechanisms to
and moreover, they have had a great impact on assess the quality of the data provided during
the industry by opening new avenues to transfer this process have a vested interest in avoiding it.
risks through the capital markets with more With the aim of reducing moral hazard, new
transparency. techniques for risk transfer became popular dur-
The financial markets have increasingly ing the 1990s (Bisping 2013). Indexed, modeled
shown interest in catastrophe reinsurance opera- loss, or parametric solutions remove the need to
tions as a source of diversification for investment assess particular losses to a portfolio by tying the
portfolios. Through a series of new instruments, payment of recoveries to an independent proxy.
investors have traded in catastrophe risks thus PROPERTY CLAIM SERVICES (PCS), a
acting as an alternative source of reinsurance company operating in the USA, or PERILS, a
capital. These alternative risk transfer (ART) similar entity operating in Europe, collect loss
solutions are today a common way to interact data from the major insurance companies after a
with the “convergence market,” the space where catastrophe and aggregate them into an overall
finance meets insurance (World Economic loss estimate. Risk transfer solutions based on
Forum 2008). these market estimates tie particular recoveries
CAT bonds have probably been the most suc- to the overall loss reported by the insurance mar-
cessful instrument used to “package” catastrophe ket. By removing the direct connection between
risks into a securitization. This solution consists recoveries and particular portfolio losses, moral
of creating a special purpose vehicle, which acts hazard is greatly diminished. If the event created
as a reinsurer to a single entity and which holds a large loss for the insurance industry as a whole,
the bond principal invested in safe, low-risk, it seems intuitive that recoveries for a particular
short-term securities. Cummins (2008) presents insurer should be somewhat proportional to the
a comprehensive view of the history and types of overall loss. Naturally, this type of index works
securitizations and derivatives that have been best for large companies whose portfolios mimic
used in the market. the distribution of the entire industry. For com-
Capital providers have had an interest in iso- panies that operate within an industry niche, these
lating catastrophe risk to avoid other components types of solutions need to be tailored somehow
of risk that are already present in usual financial since they may involve too much “basis risk,” the
portfolios. Fully collateralized CAT bonds, for potential difference between actual losses and
instance, have attracted interest since they elim- recoveries.
inate the risk of default of the parties involved. Modeled loss solutions base their payments on
Similarly, moral hazard present in traditional the loss produced not by the event itself but on the
indemnity-based (re)insurance processes is also loss estimated by a numerical CAT model. The
undesirable for the capital markets. This problem model is run after an event assuming certain
arises when a party can potentially misrepresent parameters that match measurements from the
764 Earthquake Mitigation Strategies Through Insurance
event. For instance, an earthquake model that polynomials, evaluated at a set of locations
accounts for epicentral location, magnitude, using a ground motion metric as the main vari-
depth, and rupture length of the event may be able. Typical metrics are peak ground accelera-
run setting these parameters equal to estimates tions, spectral acceleration at a particular natural
provided by the United States Geological Survey period, or some other descriptors of site-specific
(USGS). Further constraining of the model may shaking like Shindo intensity in Japan.
be accomplished by generating a series of sto- Second-generation parametric indices are gen-
chastic footprints and identifying those that erally considered “better” than first-generation
match more closely the observations obtained solutions because the correlation between losses
from a network of seismometers, if these data and local shaking intensity is assumed to be stron-
are available. ger than between losses and the main event
Although modeled loss solutions bypass the parameters of magnitude and epicentral location
claims process, ambiguities may still creep up (Goda 2013). However, in a real scenario, the
into the post-event loss calculation (PELC), a series uncertainties affecting ground motions used in
of activities undertaken by a consultant or calcula- second-generation parametric indices are typi-
tion agent in order to run the model and determine cally much greater than those affecting the main
trigger conditions and index values. These steps parameters of the event. Quality comparisons
often involve models that contain significant intel- between triggers often remain difficult to pursue
lectual property from their developers and whose since both solutions are dependent on a plethora
technical details may not be fully disclosed. There- of design parameters such as resolution used,
fore, the intimate workings of modeled loss indices number of stations, etc.
are usually not well understood by parties external The continued appearance of both, first- and
to the transaction, and outside investors may feel second-generation solutions, suggests that there
that they don’t have enough information about the is space in the market for a variety of products that
likely range of outcomes. may coexist and complement one another. Table 2
Parametric (re)insurance bases payments nei- collects all the parametric earthquake transac-
ther on the direct observed damages nor on tions of first and second generation (trigger
modeled estimates of loss. Rather, these solutions type) since 1997 (compiled from www.artemis.
use measurable parameters of the physical event bm). Cases in which there is high uncertainty in
to determine the payouts. Within this class, two the exposure distribution could favor first-
types of solutions have been used in the market generation approaches, whereas cases in which
successfully: first-generation CAT-in-a-box trig- reliable networks of seismometers are present and
gers and second-generation parametric indices. the exposures are well known could tend to sup-
CAT-in-a-box triggers associate discrete pay- port second-generation approaches. Edmonds
ments to a set of physical event parameters such (2013) provides a market perspective on some
as magnitude, epicenter location, and focal depth. of the transactions used in the recent past, and in
They are discontinuous in nature and typically particular the Muteki Ltd. CAT Bond that was
involve significant basis risk. Attempts to reduce issued in 2008 to protect the large mutual
the basis risk associated with these transactions Zenkyoren in Japan, whose trigger mechanism
have consisted of optimizing the trigger thresh- was activated during the MW 9.0 March 11th
olds (Franco 2010) or discretizing the parameter 2011 Tohoku event and which provided US$300
domain (Franco 2013). million in recoveries. In some occasions, the lack
Transactions that rely on geographically dis- of a reliable network of seismometers has been
tributed parameters of the event, such as ground circumvented by simulating the response of a set
motions, have also appeared in the market in the of “virtual” stations (identified as “V” in the trig-
form of second-generation parametric indices. ger type of Table 2) through ground motion
These indices are computed as a weighted sum models such as the USGS’ ShakeMap or
of some mathematical expression, often PAGER systems (Wald et al. 2008).
Earthquake Mitigation Strategies Through Insurance 765
Earthquake Mitigation Strategies Through Insurance, Table 2 List of parametric earthquake transactions (www.
artemis.bm). Note: Details on the transactions are approximately inferred from information provided through public
sources, and they may not always be accurate or complete
Issuer – cedant/sponsor EQ region Size in US$m Trigger type Year
Parametric Re – Tokio Marine & Nichido JPN 100 1st 1997
Mitsui Marine & Fire Swap – Mitsui M&F JPN 30 1st 1998
Concentric – Oriental Land Co. JPN 100 1st 1999
Prime Cap. CalQuake/EuWind – Munich Re US 135 1st 2001
Atlas Reinsurance II – SCOR JPN/US 150 1st 2001
Fujiyama – Nissay Dowa Gen. Ins. Co. JPN 70 1st 2002
PIONEER 2002 – Swiss Re JPN/US 255 1st/2nd 2002 E
Studio Re – Vivendi Universal SA US 175 2nd 2002
Phoenix Quake – Zenkyoren JPN 193 2nd 2003
Phoenix Quake Wind – Zenkyoren JPN 85 2nd 2003
Phoenix Quake Wind II – Zenkyoren JPN 193 2nd 2003
Arbor I – Swiss Re JPN/US 95 2nd 2003
Arbor II – Swiss Re JPN/US 27 2nd 2003
Sequoia Capital – Swiss Re US 23 2nd 2003
Sakura Capital – Swiss Re JPN 15 2nd 2003
Arbor I (2) – Swiss Re JPN/US 60 2nd 2003
Arbor I (3) – Swiss Re JPN/US 9 2nd 2003
Sequoia Capital (2) – Swiss Re US 12 2nd 2004
Arbor I (4) – Swiss Re JPN/US 21 2nd 2004
Arbor I (5) – Swiss Re JPN/US 18 2nd 2004
GI Capital – Swiss Re JPN 125 2nd 2004
Sequoia Capital (3) – Swiss Re US 11 2nd 2004
Arbor I (6) – Swiss Re JPN/US 32 2nd 2004
Arbor I (7) – Swiss Re JPN/US 15 2nd 2004
Arbor I (8) – Swiss Re JPN/US 20 2nd 2005
Cascadia – FM Global US 300 1st 2005
Arbor I (9) – Swiss Re JPN/US 25 2nd 2005
Arbor I (10) – Swiss Re JPN/US 18 2nd 2005
Australis – Swiss Re AUS 100 2nd V 2006
CAT-Mex – FONDEN MEX 160 1st 2006
Successor – Swiss Re US 950 2nd 2006
Cascadia II – FM Global US 300 1st 2006
Successor I – Swiss Re US 29 2nd 2006
Redwood Capital IX – Swiss Re US 300 2nd 2006
Australis (2) – Swiss Re AUS 50 2nd V 2007
Successor II – Swiss Re US 100 2nd 2007
MedQuake – Swiss Re EU-Med 100 2nd V 2007
Fremantle – Brit Insurance JPN 200 2nd 2007
Fusion 2007 – Swiss Re MEX 140 1st 2007
Midori – East Japan Railway Co. JPN 260 1st 2007
Successor II (3) – Swiss Re US 100 2nd 2007
Redwood Capital X – Swiss Re US 499 2nd 2007
Muteki – Zenkyoren JPN 300 2nd 2008
Vega Capital (2008–1) – Swiss Re US 150 2nd 2008
Topiary Capital – Platinum Underwriting JPN 200 2nd 2008
(continued)
766 Earthquake Mitigation Strategies Through Insurance
Earthquake Insurance Pools known as the Swiss Natural Perils Pool. Although
this pool does not yet cover damages from earth-
It is convenient for the effective usage of risk quakes across all Swiss cantons, it covers a vari-
transfer tools that risks are properly aggregated, ety of other perils such as windstorm, hail,
or “pooled,” in the first place. Individual property avalanche, and snow pressure (Schweizerisches
risks are grouped into larger portfolios through Versicherungsverband or SVV 2014). The Swiss
the standard operations of insurance companies, pool is managed by the association of insurers
but penetration of voluntary catastrophe insur- acting in coordination with the national and
ance is often low. To boost the take-up rate and regional governments.
aggregation of individual risks, other instruments The Consorcio de Compensación de Seguros,
can be used, most prominently the creation of risk another early example of a pool, was created in
pools. 1941 after the Spanish Civil War (1936–1939) to
The main purpose of a pool is to create a compensate losses derived from rioting during
compensation fund that works on the principle the war years. In 1945, the Consorcio expanded
of solidarity. If an entire community is exposed to its mission and turned into a more general com-
a set of perils, all members in the community pensation pool that includes damages from cata-
contribute a premium to the pool, which later strophic perils. It paid, for instance, a total of
can be used to pay for damages incurred. Pools €489 million to compensate losses from the
differ by their premium collection mechanism, by Lorca MW 5.1 2011 earthquake in southern
the split of financial responsibilities between par- Spain (CCS 2014). The Spanish pool is a public
ticipating actors, by their relationship with entity that operates independently from the insur-
broader reinsurance networks, and by the level ance industry but that receives the appropriate
of involvement of public versus private entities contribution from policies underwritten in the
(Consorcio de Compensación de Seguros 2008). market. While it does not seek reinsurance in
On this last item, it should be noted that incen- the international market (thus providing no inter-
tives and subsidies provided through public enti- national risk transfer), other pools actively look
ties for the management of risks may have strong for this financial support.
beneficial effects on society, but they may also New Zealand’s Earthquake Commission
lead to difficulties in assessing the actual costs (EQC), for instance, expects to obtain about
and value associated with such measures. NZ$4.5 billion in reinsurance recoveries for the
Switzerland’s insurance sector formed the first Christchurch earthquake sequences that started in
formal catastrophe insurance pool in 1939, February of 2011 (EQC 2013). In June 2013, the
Earthquake Mitigation Strategies Through Insurance 767
Earthquake Mitigation Strategies Through Insurance, Table 3 Existing national catastrophe pools (World Forum
of Catastrophe Programs 2014)
Country or Year of
region inception Pool name Summary of characteristics
California 1996 California Earthquake Provides direct insurance for earthquake residential
(USA) Authority (CEA) exposures, participation of insurers and reinsurers is
voluntary
France 1982 Catastrophes Naturelles State-owned system, compulsory participation,
(CatNat) premiums collected through commission on standard fire
policies
Iceland 1975 Vidlagatrigging Islands State-owned system, compulsory participation,
premiums collected through commission on standard fire E
policies
Japan 1966 Japan Earthquake Privately owned by association of reinsurers,
Reinsurance Co. Ltd. compulsory excess of loss cover, residential exposures,
(JER) state backed
Mexico 1996–1999 Fondo de Desastres State-owned system, publicly financed, for ex ante and
Naturales (FONDEN) ex post activities
New 1993 Earthquake Commission State-owned system, compulsory participation,
Zealand (EQC) premiums collected through commission on standard fire
policies
Norway 1979 Norsk Naturskadepool/ State-owned system with participation of all nonlife
Statens Naturskadefond insurers covering natural disaster damages, premium
established by pool, collected by insurers
Romania 2009 Natural Disaster Insurance Jointly owned company by insurers, compulsory
Pool (PAID) dwelling cover for earthquake, flood, and landslide
Spain 1941–1945 Consorcio de State-owned system, compulsory participation,
Compensación de Seguros premiums collected through commission on standard fire
(CCS) policies
Switzerland 1939 Swiss Natural Perils Pool System promoted by the association of Swiss insurers,
premiums collected in conjunction with standard
policies
Taiwan 2001 Taiwan Residential State-owned system, compulsory participation,
Earthquake Insurance premiums collected through commission on standard fire
Fund (TREIF) policies
Turkey 2000 Turkish Catastrophe State-owned system, compulsory participation, TCIP
Insurance Pool (TCIP) policies distributed by accredited insurers
EQC had disbursed a total of NZ$5.7 billion in similar mechanism to EQC in New Zealand,
claims payments and had collected a total of through a compulsory fee on standard fire poli-
NZ$1.8 billion in reinsurance. The pool still cies. Many of the national pools currently in
contained NZ$2.9 billion then although the operation (summarized in Table 3) show distinct
expectation was that it would be completely degrees of involvement of different governments
depleted by the time all claims have been settled. (Nguyen 2013). The design criteria of each pool
The fund would have been further impaired to depend on history, on the existing insurance
respond to the Christchurch events without the mechanisms, on government capacity, and on
support of reinsurance. the objectives of each particular effort.
In Asia, two prominent programs in Japan and While the main purpose of the pool mecha-
in Taiwan address the pooling of risks through nism is to provide monetary compensation, it can
two distinct strategies, the first through the par- have a strong socioeconomic effect as well. The
ticipation of private companies in a state-backed Turkish Catastrophe Insurance Pool (TCIP),
reinsurance provider and the second using a established in 2000 after the Kocaeli and D€uzce
768 Earthquake Mitigation Strategies Through Insurance
earthquakes of 1999, has been a conduit to been strongly guided by the Hyogo Framework
accomplish many objectives such as international for Action 2005–2015, which identified the need
risk transfer, reduction of liability of the govern- to promote the development of financial risk-
ment, prevention of further taxation, promotion sharing mechanisms, particularly insurance and
of good construction practices, and an increase of reinsurance against disasters (Cummins and
social cohesion. Similarly to the EQC, the pool Mahul 2009). The impact that the Global Facility
has also financed education and research projects for Disaster Reduction and Recovery (GFDRR)
in order to foster the understanding and commu- and the World Bank have had on advances in this
nication of risks. field is patent in the solutions implemented dur-
TCIP has typically accessed the traditional ing the last decade.
international markets for reinsurance. In 2013, it In 2006, the Government of Mexico (through
decided to complement this coverage through an FONDEN) sponsored CAT-Mex Ltd., a paramet-
ART tool called Bosphorus 1 Re Ltd. This para- ric catastrophe bond designed to finance rescue
metric earthquake transaction is based on a and rebuilding costs after an earthquake. The
ground motion index that relies on a network of bond successfully transferred US$160 million of
seismometers around Istanbul managed by the Mexico’s earthquake risk to the global financial
Kandilli Observatory and Earthquake Research markets on a parametric basis over a period of
Institute. The CAT bond was oversubscribed 3 years. Several geographic zones were defined
when it hit the market, showing the great interest and associated with threshold conditions on the
of global financial entities to participate in the magnitude and location of the earthquake that
reinsurance of TCIP as a diversifying risk need to be met in order for the coverage to be
(Artemis 2013). It also provided TCIP with cov- triggered.
erage at advantageous prices. The transaction was deemed successful and
The example of Turkey’s securitization appeal the experience was repeated after the bond
is often compared to the success of a series of matured in 2009. Mexico then issued the
CAT bonds issued by FONDEN in Mexico. The MultiCat Mexico 2009 bond expanding its cov-
FONDEN system is fully backed by the Mexican erage to hurricane risk for a total of US$290
government, without any direct participation of million until 2012, at which point the transaction
the insurance industry. Originally established as was reissued for US$315 million under the name
an agency to provide funds after catastrophic MultiCat Mexico Ltd. (Series 2012-1) maturing
events, it has grown to incorporate risk financing in 2015.
among its responsibilities. Mexico has been tre- The 2009 bond was oversubscribed, with
mendously innovative in governmental risk demand for the Mexican bonds outpacing supply.
transfer and is an often-cited success story of The bond originally required $250 million of
sovereign risk transfer, an alternative solution to coverage but was upsized to $290 million to
traditional reinsurance for governments. satisfy market demand, which in turn translated
into more attractive pricing for Mexico.
Over the last decade, Mexico has established
Sovereign Earthquake Risk Transfer itself as the first sovereign country that has effec-
tively transferred some of its catastrophe risk
The cases of Chile, New Zealand, or Turkey outside of its borders via these novel instruments.
illustrate how countries can effectively transfer The Mexican experience has been pioneering and
catastrophe risk to the global markets via their has attracted much attention and interest. It has
domestic insurance systems. A country may seek also been the subject of several studies that aimed
financial protection from the global markets in a to describe the features of the first transaction
more direct fashion in what is referred to as (Cardenas et al. 2007), optimize the efficiency
“sovereign” risk transfer. The development of of these tools in the context of other reinsurance
insurance solutions in the sovereign space has strategies (Härdle and López Cabrera 2010), or
Earthquake Mitigation Strategies Through Insurance 769
give a more comprehensive perspective on the needs to be in place. After the MW 7.0 earthquake
construction, implementation, and lessons of January 2010, the country was devastated
learned from these solutions (Michel-Kerjan (DesRoches et al. 2011). The government fell in
et al. 2011). In addition, trade journals such as complete disarray, and chaos ensued. When
the Bermuda-based specialist website www.Arte- CCRIF communicated that it was ready to make
mis.bm have followed the activity related to the a payment of about US$7.7 million (about
Mexican bond issuances and have made available 20 times the yearly premium paid by Haiti) as
plenty of interesting information as to how mar- per the pool arrangement, it took some time, in
kets have reacted to these transactions over the the absence of an operational government and
years. banking community, to find mechanisms to
E
The financial markets, by welcoming Mexican deploy the funds.
catastrophe risk, have proved that the issuance of Insurance-pooling schemes have provided
these bonds is a plausible strategy for a country to developing nations with a foothold in the global
transfer its risks. However, the case of Mexico reinsurance networks. As CCRIF has established
involves one single country with a strong legal itself as a successful operation, similar initiatives
and financial system in place and with a long have been set in motion in Africa (African Risk
history of insurance experience. Other emerging Capacity 2012) and in the South Pacific (Pacific
countries with fewer systems in place require Disaster Risk Financing and Insurance Program
other strategies to successfully leverage risk 2011). Recently, cyclone Ian triggered the pay-
transfer solutions. ment of US$1.27 million to the government of
This was the motivation for the creation of the Tonga, which participates in this newly created
Caribbean Catastrophe Risk Insurance Facility system.
(CCRIF) in 2007. This institution was formed
by 16 Caribbean countries, under guidance from
the World Bank, to pool and transfer regional Remarks on Micro-Insurance
hurricane and earthquake risks to the global
financial markets (World Bank 2008). The trigger Low-income populations in developing nations
mechanisms were also parametric, conceptually frequently use community-based financing or
similar to those used in Mexico. However, the insurance schemes, often with great creativity
compensations were, in general, not designed to (Rutherford 1999). However, due to the covariant
finance long-term operations but to provide a nature of catastrophes, community-based financ-
cash injection to address what the CCRIF defined ing techniques are not well suited to respond to
as the “liquidity gap,” the period of time directly events that may affect the entire participating
after the emergency response and before other community at once. This has spawned an explora-
assistance funds can be accessed for long-term tion into whether formal insurance tools can over-
recovery. Countries like Haiti with nonexistent or come this challenge for developing nations
feeble financial systems in place would have (Linnerooth-Bayer and Mechler 2009).
found it hard to access insurance on a stand- In general, the few experiences that exist and
alone basis like Mexico did. However, CCRIF the analyses carried out suggest that insurance
has been able to access the reinsurance commu- can indeed be a good complement to other risk
nity and has effectively responded to several mitigation initiatives. However, when analyzing
events during its existence, benefitting the gov- the benefits of insurance in a developmental set-
ernments of Anguilla, St. Vincent and the Gren- ting, it should not always be considered evident
adines, Barbados, Dominica, St. Lucia, Turks and that cash by itself is helpful for long-term recov-
Caicos Islands, and Haiti. ery or for building future resilience. Monetary
The case of Haiti also brought to light that in support often needs to be accompanied by other
order for insurance mechanisms to be effective, initiatives in order to have a truly positive impact
some basic governance and finance infrastructure on society.
770 Earthquake Mitigation Strategies Through Insurance
In a recent reconstruction survey in Sri would be especially costly in locations with dif-
Lanka, eight and a half years after the island ficult access and lack of skilled adjustors.
nation was hit by the devastating Indian Ocean In this context, India has made important
tsunami of December 2004, Franco et al. (2013) advances by implementing indexed schemes
visited several communities along the coast. Dif- for crop insurance, successful in great part thanks
ferences in recovery outcomes were observed to the governmental policy that requires the
between sites that had been managed by third- insurance sector to service the low-income
party donors versus settlements where owners populations (Mechler et al. 2006; Giné et al.
had been empowered to rebuild through a grad- 2010). Other countries in Africa, Asia, and
ual injection of cash. Through this process, Latin America are making use of similar schemes
owners were provided a set of cash packages (Ibarra 2009).
that allowed them to rebuild on their original Earthquake-specific insurance programs have
plot in subsequent phases, assuming that the been studied for implementation in China, where
land remained fit for settlement. At each phase, these types of solutions have the potential to
an inspector would survey the construction qual- fit well. China has a sophisticated insurance
ity and would give approval for the next stage, market, well connected to the reinsurance indus-
at which point the owner would obtain a transfer try (Wang et al. 2009) but with many needs in
of cash through prearranged “tsunami bank low-income communities in earthquake-prone
accounts.” The outcomes of settlements built areas. In particular, parametric solutions have
by donors on land provided by the government been studied for deployment using modern
varied in quality, but the feedback on the owner- CAT modeling techniques (Del Re et al. 2008).
driven construction scheme was almost univer-
sally positive, which suggests that in the
presence of other supporting mechanisms, Summary
access to cash can truly play a powerful role in
a successful recovery. Insurance successes, large or small, have been
Micro-insurance initiatives have started witnessed in events like the Maule earthquake
addressing part of this problem by tying insur- in Chile or the Christchurch earthquakes in
ance products to other broader services that inte- New Zealand. Many experiences remain highly
grate the population in formal financial imperfect, but in the meantime, countries like
networks. Mechler et al. (2006) discuss several Turkey and Mexico have made great strides in
property insurance “bundles” connected to other advancing their earthquake risk management
social and financial services in Nepal, Bangla- strategy. Embracing ex ante risk management
desh, India, Pakistan, and Malawi. Insurance can tools, many countries are slowly assuming
have greater societal impact if its purchase links more and more responsibility over their future.
communities with educational initiatives, con- Earthquake (re)insurance systems offer excit-
struction oversight, or risk mitigation, as echoed ing opportunities for development along all the
in the objectives of Turkey’s TCIP program. dimensions presented in this paper. At the core of
Since micro-insurance solutions made their national economies, insurance industries need to
appearance in the early 2000s, the experience so be established and strengthened in order to for-
far is very limited. In the meantime, significant mally manage existing risks. In countries that
challenges have been identified in the marketing lack a strong socioeconomic foundation, multi-
and deployment of insurance solutions within national risk pools or sovereign risk transfer
populations that are remote or where great edu- remain open options to slowly reshape or com-
cational and cultural barriers exist. This has been plement ex post response strategies with proac-
partly addressed by the usage of parametric solu- tive risk management. Technological advances
tions, which as described above, suffer from basis made in the last few decades have helped com-
risk but also help bypass claim processes that munities assess their exposures to a much larger
Earthquake Mitigation Strategies Through Insurance 771
degree than they were once able to, but refine- ▶ Earthquake Disaster Recovery: Leadership and
ments of these computational tools are continu- Governance
ously being made to improve the quantification ▶ Economic Impact of Seismic Events:
risk. In parallel, new promising paths in alterna- Modeling
tive risk transfer continue to be charted in the ▶ Economic Recovery Following Earthquakes
financial markets, and new products and distribu- Disasters
tion channels continue to be explored at the com- ▶ Insurance and Reinsurance Models for
munity level through micro-insurance initiatives. Earthquake
Experiences suggest that availability of capital ▶ Learning from Earthquake Disasters
after catastrophe events at an individual and gov- ▶ Probabilistic Seismic Hazard Models
E
ernmental level can support a better and faster ▶ Seismic Risk Assessment, Cascading Effects
recovery and can ultimately increase a society’s ▶ Site Response for Seismic Hazard Assessment
resilience. Systems that facilitate the rapid
deployment of funds can provide almost instant
liquidity at a time of need. Earthquake insurance References
can therefore be a strong component of any seis-
mic risk mitigation framework with potentially African Risk Capacity (2012) African risk
capacity – sovereign disasetr risk solutions, a special-
synergistic effects with other coexisting mea-
ized agency of the African Union. Accessible on-line
sures such as risk education, construction regula- at http://www.africanriskcapacity.org/documents/
tion, and socioeconomic policies. 350251/371107/ARC_Overview_Brief_EN.pdf
The (re)insurance sector and society as a Aon Benfield (2011) Chile: one year on. Chicago. Acces-
sible on-line at www.aon.com/attachments/reinsur-
whole are still learning important lessons from
ance/201102_chile_one_year_on_report.pdf
earthquake events today, and much work remains Artemis (2013) Bosphorus 1 Re CAT bond parametric
to be done. Earthquake and tsunami perils are trigger details emerge. Accessible on-line at http://
often perceived as “distant” in time and space, www.artemis.bm/blog/2013/04/04/bosphorus-1-re-cat-
bond-parametric-trigger-details-emerge/
and this often hinders the implementation of risk
Bisping M (2013) Risk transformation through capital
management policies and measures. However, markets. In: Lane M (ed) Alternative (re)insurance
the combination of documented growing eco- strategies, 2nd edn. Risk Books, Chicago
nomic losses, small overall global earthquake Clark K (1986) A Formal Approach to Catastrophe Risk
Assessment and Management, in the Proceedings of
insurance penetration rates, and a large supply
the Casualty Actuarial Society, Vol. 73, Part 2,
of capital means that an increasing portion of No. 140
the financial burdens imposed by earthquakes Consorcio de Compensación de Seguros or CCS
worldwide could be assumed by global financial (2008) Natural catastrophes insurance cover – a diver-
sity of systems. Accessible on-line: http://www.
networks rather than by often fragile local econ-
wfcatprogrammes.com/c/document_library/get_file?
omies. This represents a challenging task but also folderId=13442&name=DLFE-553.pdf
an exciting prospect. Consorcio de Compensación de Seguros or CCS (2014)
Estadistica Riesgos Extraordinarios 1971–2013.
Accessible on-line at http://www.consorseguros.es/
web/c/document_library/get_file?uuid=548d4f59-b6c5-
Cross-References 40dd-b06b-98dbcefd790f&groupId=10124
Cardenas V, Hochrainer S, Mechler R, Pflug G, and
▶ “Build Back Better” Principles for Linnerooth-Bayer J (2007) Sovereign financial disas-
ter risk management: The case of Mexico. Environ-
Reconstruction
mental Hazards 7, pp. 40-53. Elsevier
▶ Building Disaster Resiliency Through Disaster Cornell CA (1968) Engineering seismic risk analysis. Bull
Risk Management Master Planning Seismol Soc Am 58(6):1583–1606
▶ Community Recovery Following Earthquake Cummins JD (2008) CAT bonds and other risk-linked
securities: State of the market and recent develop-
Disasters
ments. Risk Manag Insur Rev 11(1):23–47
▶ Damage to Buildings: Modeling Cummins JD, Mahul O (2009) Catastrophe Risk Financ-
▶ Damage to Infrastructure: Modeling ing in Developing Countries –Principles for Public
772 Earthquake Mitigation Strategies Through Insurance
on-site utility services were easy to achieve with- became additional demand on the medical and
out a significant cost premium, most facilities rescue resources of other facilities. The Veterans
would already be designed for this performance Administration complex of buildings in Sylmar
level, whereas in fact, comprehensive protection was the scene of two complete structural col-
of functionality of a facility is relatively rare. lapses. See Fig. 1. At the Olive View Medical
Center nearby, the 750-bed main hospital build-
ing suffered such severe soft story damage that it
Historic Examples was evacuated and torn down, only 2 months
after it had been built (Fig. 2). In addition, the
There have been many examples of earthquake two-story Olive View Psychiatric Building at that
disruption of essential facility operations, dating complex experienced a complete ground-story
back decades and continuing to the present. One collapse (Fig. 3); several unreinforced masonry
of the earliest particularly instructive earthquakes buildings on the grounds collapsed; damage to
was the 1971 San Fernando earthquake in the Los the fire station located on the grounds required
Angeles area. Seven major medical facilities lengthy extrication of the equipment; reinforced
were badly damaged: Foothill Medical Center, concrete canopies collapsed onto ambulances
Pacoima Lutheran Medical Center, Pacoima (Fig. 4); damage to the backup power system
Memorial Lutheran Hospital, Holy Cross Hospi- (Fig. 5) and to the entire array of nonstructural
tal, Indian Hills Medical Center, the Sylmar Vet- components – elevators, windows, partitions,
erans Administration (VA) Medical Center, and ceilings, piping, and medical gasses – occurred
Olive View Medical Center (Jennings 1971; Lew (Ayres and Sun 1973); and the equipment and
et al. 1971; Johnston and Strand 1973). As piping in the on-site power plant building was
a result, 1,553 hospital beds were out of service so damaged that its functions were not available
(Oakeshott 1975), and instead of being able to (Fig. 6).
provide the post-earthquake supply of medical Following the San Fernando earthquake, the
care for injuries in the area, damaged hospitals Veterans Administration adopted seismic
Earthquake Protection of Essential Facilities 775
Earthquake Protection
of Essential Facilities,
Fig. 4 Destruction of
ambulances by collapsed
concrete canopy, Olive
View Medical Center, 1971
San Fernando earthquake
(Source: Karl
V. Steinbrugge Collection,
NISEE-PEER library,
University of California,
Berkeley)
Many more examples could be cited – see 1972 earthquakes also approximately date the
FEMA (2007) – but in this brief review, these beginning of the era when focused attention
two earthquakes are adequate to illustrate the has been given to protecting essential facility
range of disruptive effects, and the 1971 and functions. Also during this post-1971 or
Earthquake Protection of Essential Facilities 777
Earthquake Protection
of Essential Facilities,
Fig. 5 Damage to
unrestrained batteries,
necessary to start the
emergency power
generator, Olive View
Medical Center, 1971 San
Fernando earthquake
(Source: Karl
V. Steinbrugge Collection,
NISEE-PEER library,
University of California, E
Berkeley)
Earthquake Protection
of Essential Facilities,
Fig. 7 Collapsed fire
station in Managua, 1972
Nicaragua earthquake
(Source: Karl
V. Steinbrugge Collection,
NISEE-PEER library,
University of California,
Berkeley)
design forces. The Uniform Building Code, drift from reaching levels that result in permanent
which was the common code of the Western distortion of the overall structure, a damage state
United States until the inauguration of the Inter- that triggers evacuation.
national Building Code in 2000, an I or Impor- Although it carries a cost premium, seismic
tance Factor of 1.5 was introduced in the 1976 isolation has been increasingly employed since it
edition having the effect of increasing the design was introduced in modern form in the 1980s, the
forces by 50 %. Because inelastic levels of first isolated building in the world being the Wil-
response are still anticipated even for these stron- liam Clayton Building in Wellington, New
ger buildings if severe ground shaking occurs, Zealand, completed in 1982. Typically, reducing
only increasing the elastic-based design forces the response of the building to a third of what it
is currently thought to be helpful but not suffi- would otherwise be, isolation can provide high
cient in attaining the performance level of levels of functional protection, assuming the
remaining operational. Nuclear power plants are structure and nonstructural systems are still
typically designed with the objective of providing designed to appropriate levels for that degree of
so much strength that their structural, as well as motion. As noted below, because nonstructural
key nonstructural, components remain elastic, components are typically more fragile than the
but that specialized topic is covered in another structure, a level of earthquake motion in
article and that level of protection is generally a building “only” one-third as great as that of
infeasible – too expensive and too restricting on the ground’s unreduced level of motion can still
the configuration of ordinary buildings. Some be damaging unless appropriate design measures
inelastic behavior in the structure does not nec- are implemented.
essarily mean that the operation of the facility has
been greatly curtailed, if the design has confined
the inelasticity, i.e., damage, to portions of the Nonstructural Protection
structure that can absorb that earthquake punish-
ment while still allowing the structure to remain While it is necessary for the structure of the
occupiable and operational. A key criterion, and essential facility to remain safe and serviceable
a difficult one to meet, is preventing interstory if the facility’s functionality is to be maintained,
Earthquake Protection of Essential Facilities 779
it is not sufficient. There are many nonstructural a rack containing equipment; anchoring refers to
items that must be protected from damage as anchor bolts connecting the rack to the floor and
well, such as emergency power generators, ceil- or wall.) Adding to the problem is the fact that
ings, fire sprinkler pipes, and elevators, to name many kinds of equipment are introduced by the
but a few. The bare structure of a building, or the occupant after the building is built and outside the
supporting framework or tower of a utility com- purview of the building code or a design profes-
ponent, by itself provides no essential service. sional, such as the furnishings, file cabinets, and
The degree of difficulty in protecting essential computer equipment in an office setting or the
functions thus is greatly increased beyond the chemicals and related apparatus in a laboratory.
already formidable performance level of confin- Making the problem even more difficult is the
E
ing any structural damage to a negligible level, fact that even if the design professionals collec-
especially because most nonstructural compo- tively cover all of the necessary nonstructural
nents are more damageable, that is, they have protective measures, there are then multiple spe-
higher fragilities than the structure itself. cialty construction trades involved who build
Unsecured reagents can fall off the shelves in those features into the facility. This entry cannot
a laboratory, which can cause a hazardous mate- delve deeply into the topic of nonstructural pro-
rial release or fire, or both, at very low levels of tection, other than to cite in passing a basic eval-
shaking, to cite one example. uation and design manual on the subject,
The first comprehensive study of the effects of Reducing the Risks of Nonstructural Earthquake
an earthquake on nonstructural components was Damage (FEMA 2011), and a study of the diffi-
for the 1964 Alaska earthquake (Ayres culties encountered in developing a coordinated
et al. 1973), and that study can still be consulted approach to implementing nonstructural protec-
for a valid overview of the range of damage. tion (Masek and Ridge 2009). The importance as
Because of the difficulty of providing compre- well as the difficulty of ensuring adequate
hensive nonstructural protection, unfortunately, nonstructural performance in an essential facility
many subsequent earthquakes have provided the can be briefly illustrated here using the example
same basic sorts of examples of damage. of the backup power system.
Protecting nonstructural components, a topic Picturing a modern telecommunications or
discussed in a separate article, requires a major medical facility, we immediately note that if the
effort to go beyond the objective of safety for the electrical power from the local utility is disrupted
occupants and prevention of serious damage to by the earthquake – a very plausible earthquake
the structure. In a sense, structural engineers deal planning assumption – the facility cannot func-
with one coherent entity in the design of the tion, unless its own backup power system
structure, an integrated structure shown entirely operates reliably. There are at least six compo-
on the structural design drawings and documents, nents to an emergency electrical power system;
whereas the nonstructural items number in the the disabling of any one of which will cause the
hundreds or thousands and they are not all under entire backup power system to fail. Batteries are
the purview of one design professional. The needed to start the motor-generator set, and these
architect, electrical and mechanical engineers, heavy batteries, which thus experience large iner-
and specialty designers, such as for fire protection tial forces in the earthquake, are only strongly
systems or elevators, each separately design braced if earthquakes are specifically taken into
nonstructural components. The design profes- account. The massive motor-generator set itself,
sional with the most expertise with regard to often weighing as much as a truck in the case of
bracing and anchorage of nonstructural compo- systems that must supply large amounts of power
nents, the structural engineer, is only employed such as for a hospital, is typically mounted on
when regulations or special attentiveness on the vibration isolation springs. The springs are
part of the client set that requirement. (Bracing, designed to filter out the very high frequency
for example, refers to diagonal struts bracing vibration of the motor, which would otherwise
780 Earthquake Protection of Essential Facilities
propagate through the building and bother occu- design in that industry have been gradually
pants, but when the floor is subjected to the lower adapted for use with less critical facilities. In
frequency motions with large displacements of some cases today, key kinds of equipment have
the earthquake, spring-mounted equipment can been subjected to shake table qualification test-
violently lunge off its supports. Seismically ing, an approach only relatively recently intro-
designed products that solve both the motor duced into some codes and standards, such as the
vibration and seismic response problem are the 2005 edition of ASCE 7, the standard referred to
typical solution today, and existing mounts have by building codes in the United States for deter-
been retrofitted with restrainers (“snubbers”). If mining design loads.
the batteries remain stable and the motor genera-
tor is adequately restrained, the system still needs
to have a functional exhaust flue – breakage of Backups for Essential Systems
which has resulted in filling the generator room
with fumes that choke off the engine. In some Essential buildings are normally dependent on
cases, a generator has been reliably run in tests or one or more essential systems to provide ade-
for brief power outages, but when run continu- quate functionality after an earthquake. Almost
ously for a day or more, it overheats. The heat all such buildings require electrical power to be
from the motor must be dissipated, in the case of delivered to its communications, computer, and
large units with water-fed radiator systems. The other equipment. Hospitals are a special case and
electrical panel that distributes the power through in addition require water, medical gasses, and in
the facility is another critical component, and some cases, waste disposal to continue care for
on-site fuel, such as oil for a diesel generator, their patients and incoming injured.
sufficient for several days or longer is necessary Community or regional power systems are
as well. In places and eras where providing pro- susceptible to damage and interruption from
tection of essential functions is a new approach earthquake shaking. On-site backup power gen-
that is being phased in, singling out the emer- eration systems, including the generator, starting
gency power system has been a strategic place batteries, fuel and day tanks, cooling systems,
to start (Applied Technology Council 1990). and exhaust systems, are typically provided at
Another way to look at the seismic protection of facilities such as hospitals and fire stations for
backup power systems is as a means to evaluate this reason. However, beyond the anchorage and
the effectiveness of seismic protection programs. bracing of this equipment for seismic purposes,
If discussions and planning meetings have gone discussed above, the amount of on-site fuel must
on for more than a year about how to deal with all be considered. The California Hospital Program
the various potential kinds of damage and outages requires enough fuel to run essential systems for
in an essential facility, but if nothing has been 36 hours. The Pan American Health Organization
done to evaluate and if needed retrofit its backup (PAHO), considering hospitals in less-developed
power system, the program’s effectiveness could countries, recommends fuel for five full days.
be questioned. Almost all hospitals have emergency generators
Mentioned again here, although it is the topic to cover their essential systems for normal system
of a separate article, is the nuclear power plant. interruptions, which are frequent in some regions,
The typical seismic design basis is aimed at keep- but these outages are typically only hours long,
ing both structural and nonstructural components and sufficient fuel for multiple days is often not
elastic, a design approach prohibitively expen- provided or locally available. In addition, when
sive for most essential facilities. Expensive generators are installed to run for hours, they will
shake table testing has qualified or certified key often overheat if run continuously for a full day or
types of equipment for that industry, and more. The adequacy of the cooling system there-
although testing standards vary, some of the fore must also be considered for generators for
sophisticated types of analysis, testing, and essential buildings.
Earthquake Protection of Essential Facilities 781
Uninterrupted Power Supply (UPS) systems disabled by on-site damage, can be temporarily
are often required for computer and/or commu- supplied by redundant facilities located
nications systems and for certain medical equip- elsewhere.
ment in hospitals. These systems are heavy and
must be carefully braced and anchored to main-
tain function during and after earthquake Emergency Response
shaking.
Water is another utility that is often necessary The severity of ground motion from an earth-
for the continued function of an essential facility quake is typically unevenly distributed over
and also one that is susceptible to interruption of a region, based on distance from the rupturing
E
the off-site supply. Besides drinking water, water fault, direction and sense of fault rupture, local
may also be required at the site of essential facil- soils and deeper geology, and topography. Vul-
ities for cooling or operation of equipment and nerabilities of facilities are also highly variable, if
for fire sprinklers. For hospitals, PAHO recom- only because the facilities in a region were built at
mends that a hospital site have an emergency different times to varying seismic criteria. This
supply of 300 l per bed per day, for 3 days. leads to the likelihood that some essential facili-
Pumps and on-site filtration, if required, may ties will be significantly damaged and
also require power and supplies that must be nonfunctional, while others will remain intact.
provided in the post-earthquake setting. Mutual aid planning and emergency response
Most essential facilities are dependent on agreements are an effective technique for coping
effective communication after an earthquake. with the outages that do occur and capitalizing on
Hard line and cell telephone may not be func- those facilities that remain online.
tional, and backup systems should be provided. In parallel with the engineering approach of
Equipment racks used for communication equip- preventing damage is the emergency response
ment must be able to withstand the shaking. approach of quickly controlling damage and pro-
Radio frequencies should be checked for compat- viding alternate means of providing service. The
ibility with outside entities that must be key difference between an exercise of a hospital
contacted. Shortwave radio and satellite tele- disaster plan for an earthquake scenario as com-
phones are possible backup systems. Large or pared to one based on a presumed plane crash or
multi-site facilities may also need to maintain other incident is that the hospital that is expected
internal communication, and the viability of the to suddenly provide an enhanced level of emer-
existing systems should be considered. gency service to the community will simulta-
Access to off-site equipment and supplies may neously find itself in a disruptive situation. The
also be cut off after an earthquake. Vehicles that hospital that cannot protect itself from disruptive
may be needed in the post-earthquake environ- damage in the earthquake is of little use and
ment must be protected and available. Adequate perhaps a source of demand for emergency ser-
quantities of supplies needed for post-earthquake vices rather than a supplier. Thus, to be realistic,
function should be stored on-site or reliably avail- seismic emergency response procedures cannot
able. For example, supplies of medical gasses merely be copied from procedures for other kinds
(oxygen, nitrogen, compressed air) are needed of disasters that do not simultaneously affect the
for many functions in a hospital, and medical region and the individual facility.
stores for 3–5 days should be available. Multi- Three examples of communication outages
facility systems, for example, a collection of hos- and their effect on emergency response are cited
pitals run by the same organization or agency or to illustrate the issues involved. In the 1971 San
several electronics plants or data processing cen- Fernando earthquake, it took 1 h and 22 min
ters of a company, can include in their planning before the key response agency, the fire depart-
how the services of one or more facilities isolated ment, knew that two buildings had collapsed at
by a transportation or other off-site outage, or the Veterans Administration Hospital, a hospital
782 Earthquake Protection of Essential Facilities
that was part of the urbanized Los Angeles area post-earthquake service. Structural retrofit is often
and was not in an isolated area (Reitherman expensive, and replacement may take years to plan
1986). Telephone service was out, radio commu- and build. But honest evaluation of expected struc-
nication was not functional, and although the tural and nonstructural performance will lead to
in-person messenger sent to the nearest fire sta- necessary and often affordable improvements, and
tion to report the need for rescue operations at the more importantly, development of plans to offset
hospital delivered that message, it was then con- functional difficulties.
fused with the problems at the nearby Olive View
Hospital. In retrospect, the single biggest emer-
gency response need in that earthquake was at the Summary
VA hospital, where 49 of the 58 fatalities in the
earthquake occurred and a massive rescue opera- Earthquakes have caused disruption of the ser-
tion was needed. vices provided by essential facilities such as hos-
Eighteen years later, in the 1989 Loma Prieta pitals, emergency communications centers, and
earthquake in northern California, the most badly fire stations. For the functions of these facilities to
damaged hospital was Watsonville Community be available in the post-earthquake setting, the
Hospital, which had to evacuate its buildings and structure must perform safely and key
provide services for in-patients and arriving nonstructural components must be protected. In
emergency patients in the parking lot. Outside addition to these engineering approaches, emer-
aide was not requested for several hours, because gency response planning is necessary to identify
initial news reports that the hospital heard made it off-site lifeline systems that could be disrupted by
seem that the San Francisco metropolitan area the earthquake, because a facility may need
100 km away was totally devastated and hence backup systems and supplies to contend with out-
aid resources would be focused there. In fact, ages to the region’s water, power, and transpor-
there were many outside resources that eventu- tation systems.
ally made their way to the hospital when its needs
were accurately known.
In the 1994 Northridge earthquake in the Los References
Angeles area, water lines at the Sepulveda Vet-
ATC (1990) General acute care hospital earthquake sur-
erans Administration Medical Center ruptured,
vivability inventory for California, funded by the
which with efficient emergency response should Office of Statewide Health Planning and Develop-
result in prompt water shutoffs of affected zones to ment. Applied Technology Council, Redwood City
minimize water damage and disruption. Instead, Ayres JM, Sun TY (1973) Nonstructural damage. The San
Fernando, California Earthquake of February 9, 1971.
patients and personnel, including facilities engi-
National Oceanic and Atmospheric Administration,
neers, were evacuated and did not reenter the Washington, DC
building until structural engineers arrived several Ayres JM, Sun TY, Brown FR (1973) Nonstructural dam-
hours later and declared the building safe for emer- age to buildings. The great Alaska earthquake of 1964:
Engineering. National Academy of Sciences, Wash-
gency operations such as turning off the water.
ington, DC
Effective emergency response planning should Bolt B, Johnston RG, Lefter J, Sozen MA (1975) The
include checklists for rapid utilities actions, such study of earthquake questions related to Veterans
as water and gas shutoffs if certain observable Administration hospital facilities. Bull Seismol Soc
Am 65(4):937–949
criteria are met, and also for rapid safety inspec- FEMA (2007) Design guide for improving hospital safety
tions by engineers under prearranged plans. in earthquakes, floods, and high winds, FEMA 577.
New facilities intended for critical post- Federal Emergency Management Agency, Washing-
earthquake functions can be designed in accor- ton, DC
FEMA (2011) Reducing the risks of nonstructural earth-
dance with recommendations discussed here and
quake damage – a practical guide, FEMA E-74. Fed-
international standards, but older existing facilities eral Emergency Management Agency, Washington,
are much more difficult to prepare for adequate DC
Earthquake Recurrence 783
Holmes WT (2002) Background and history of the seismic shaking at a specified site will exceed some
hospital program in California. In: Proceedings of the intensity-measure level of engineering interest
seventh national earthquake engineering conference,
Earthquake Engineering Research Institute, Oakland in a given time period t. The results of a PSHA
Jennings PC (ed) (1971) Engineering features of the San have a broad spectrum of end users, ranging from
Fernando Earthquake of February 9, 1971, California decision-makers who have to provide new codes
Institute of Technology, Earthquake Engineering for designing safe buildings to engineers who
Research Laboratory, Pasadena, EERL 71-02
Johnston RG, Strand DR (1973) Olive view hospital. In: must predict potential damage to critical struc-
Murphy L (ed) San Fernando, California, earthquake tures. Among all the variables involved in the
of February 9, 1971, vol I-A. National Oceanic and PSHA computation, this section will focus on
Atmospheric Administration, Washington, DC the description of the probability models used to
Lew HS, Leyendecker EV, Dikkers RD (1971) Engineer- E
ing aspects of the 1971 San Fernando Earthquake. build earthquake rupture forecast (ERF) from
National Bureau of Standards, Building Science series earthquake rate model in the area under study.
40. U.S. Government Printing Office, Washington, DC ERF allows a modeling of the probability of
Masek J, Ridge R (2009) Identification of methods to occurrence of all possible damaging earthquakes
achieve successful implementation of nonstructural
and equipment seismic restraints. Earthquake Engi- in a given region during a specified time span and
neering Research Institute, Oakland represents the basis for PSHA. In practice, an
Oakeshott G (ed) (1975) San Fernando Earthquake, Feb- incorrect or incomplete representation of the seis-
ruary 9, 1973. California Division of Mines and Geol- micity in space, time, and size may bias the anal-
ogy, Sacramento
Reitherman R (1986) How to prepare a hospital for an ysis leading to underestimation or
earthquake. Disaster Manag 4:119–131 overconservatism in the final hazard assessment
Veterans Administration (1972) Post earthquake emer- and consequential direct and nontrivial implica-
gency utility services and access facilities, CD-54. tions for the compilation of regulations for seis-
Veteran’s Administration, Washington, DC
Wright R, Kramer S (1973) Building performance in the mic design.
1972 Managua Earthquake. National Bureau of Stan-
dards, Washington, DC
The Physics of the Problem: From
Deterministic Toward Probabilistic
Perspective
Earthquake Recurrence
The elastic rebound theory proposed by Reid
Vincenzo Convertito1 and Licia Faenza2 (1910) after the 1906 San Francisco earthquake
1
Istituto Nazionale di Geofisica e Vulcanologia, suggests that large tectonic earthquakes occur
Osservatorio Vesuviano, Napoli, Italy when the stress exceeds a critical value of the
2
Istituto Nazionale di Geofisica e Vulcanologia, rock strength, producing fractures and reliving
Centro Nazionale Terremoti, Bologna, Italy crustal strain. In a given area, stress builds up
due to tectonic processes ultimately connected
to the plate movements. Thus, assuming a con-
Synonyms stant rate in the stress buildup, the critical value
and the consequent earthquake nucleation will be
Hazard function; Recurrence time; Seismic haz- reached at more or less regular intervals. This
ard; Time-dependent recurrence models; Time- periodic buildup and release of stress on a given
independent recurrence models fault represents what is known as seismic cycle
whose duration is given by the ratio of event-
strain release to tectonic strain rate. More pre-
Introduction cisely, a seismic cycle controls the long-term
timing of large earthquakes and comprises three
Probabilistic seismic hazard analysis (PSHA) different stages, namely, (i) the interseismic stage
aims at estimating the probability that ground during which there is an accumulation of strain
784 Earthquake Recurrence
energy; (ii) coseismic stage, which corresponds earthquakes, in particular of their magnitude
to the stage of principal rupture; and (iii) post- should be defined. Two recurrence relationships
seismic stage during which a new equilibrium is are generally used: the Gutenberg-Richter
reached. However, comparison of seismic slip (GR) relationship (Gutenberg and Richter 1944)
distribution with potential slip inferred from geo- and the characteristic earthquake model
detic motion measurements demonstrates that not (Schwartz and Coppersmith 1984). The GR
all the stored energy is released through seismic model assumes that in a given seismic zone,
slip but there is a deficit which is attributed to events of all magnitudes (from a minimum to a
fault creep and aseismic deformation (Bakun maximum magnitude value) may occur and are
et al. 1986). expressed through the relationship:
In this respect, earthquake recurrence refers to
the time between subsequent events rupturing on log N ðmÞ ¼ a bm (1)
a given segment fault, hence to the period of the
loading cycle. Thus, in theory one would think where N(m) is the cumulative number of events
that it is possible to predict the occurrence of having magnitude larger than m occurring in a
future events if one knows the previous history. given time. The parameter a depends on the num-
Indeed, the scheme described so far provides a ber of events in the selected time and region. The
deterministic representation of the occurrence of slope parameter b, also known as “b-value,” pro-
a single earthquake on a given structure and may vides an estimate of the ratio between the number
be too simple to depict the real nature of the of small and large events and is generally about
earthquake recurrence. In this respect, Nishenko one. It can be shown that, when the GR is
and Buland (1987) described three possible fac- assumed as recurrence relationship, the probabil-
tors for the observed variation in the recurrence ity density function of magnitude is the exponen-
period. A first factor is a possible change in tial function f(m) = e(abm) with a = a ln10 and
source size from event to event which would b = bln10. As this function covers an infinite
require a different time for restoring the initial range of magnitude, from 1 to +1, for practi-
stress value, a second factor is a nonlinear strain cal applications, a lower and upper bounded ver-
accumulation in which the strain field initially sion is used, which requires the identification of
recovers rapidly soon after the event and then both a minimum and a maximum magnitude
progressively lowers down until the next event, value. The definition of the minimum magnitude
and a third factor is given by the faults’ interac- depends on the catalog completeness, while the
tion in which strain energy released on one fault definition of the maximum magnitude is still an
segment affects the strain on nearby faults. In open question, and its quantification may play an
addition, the uncertainties of estimations of the important role in PSHA for the mitigation against
earthquake parameters, and the relatively short- very large and rare events.
ness of the seismic records available, make the The characteristic earthquake model is instead
definition of the seismic cycle even more com- based on the assumption that individual faults and
plicated, leading to a chaotic behavior of the fault segments tend to generate similar size or
sequences. In this perspective, models of earth- characteristic earthquakes having a relatively
quake occurrence are developed in the frame- narrow range of magnitudes near the maximum
work of the probability theory, on which also with frequency of occurrence larger than that
the PSHA is established. predicted by simply extrapolating the GR
From a practical point of view, when (Schwartz and Coppersmith 1984). The resulting
performing a PSHA, once the seismic zones probability density function of magnitude to be
which represent the potential threat for the site used in the PSHA has a more complex formula-
of interest have been identified and modeled tion with respect to the exponential probability
(as points, lines, areas, volumes, or single faults) density function derived from the GR (Young and
based on the available data, the recurrence of the Coppersmith 1985).
Earthquake Recurrence 785
Earthquake Recurrence,
Fig. 1 Definition of the
variables generally used to
analyze time distribution of
earthquake occurrence
(upper panel). Lower
panels depict a schematic
representation of a possible
distribution of observed
inter-event times together
with the estimated
probability density function
(pdf) and the corresponding E
cumulative function (cdf)
(see text for the definition
of pdf and cdf)
One of the assumptions of the PSHAs, derived consecutive shocks. Statistical distribution of Dti
from the Reid (1910) theory, is that the seismicity is analyzed to determine if the observed seismic-
is stationary; as a consequence, the recurrence ity can be modeled by using a specific function
relationships derived for the current available (Fig. 1).
seismic catalog are valid also in the future. How- For simplicity, the inter-event time, Dt,
ever, as mentioned above, this assumption may will be indicated as t throughout the text. Inspec-
not be valid on the geological time scale, thus tion of the distribution of the observed t may
requiring models that account for nonstationary help to establish some features of the seismicity,
behavior. As a fully deterministic approach is not such as the presence of any regularity in time
yet feasible, the solution is to refer to the earth- and possibly the influence of different tectonic/
quakes as the outcome of stochastic processes. In physics factors that regulate the spatial occur-
particular, the earthquake occurrence is regarded rence of earthquakes. For example, a relatively
as renewal process (Cornell and Winterstain large percentage of small values of inter-event
1988) in five dimensions: time, magnitude, and times may suggest that, in the area under study,
location. It is obvious that, given this assumption, earthquakes tend to cluster in time and thus
the physical processes governing rupture initia- cannot be modeled as statistically independent
tion, propagation, and stopping are neglected. On events.
the other hand, real nature of the earthquake Formally, the inter-event time distribution is
process is too complex to be modeled by using defined unambiguously by its probability density
deterministic approaches. function f(t) (pdf). By definition, the pdf has to be
In a renewal process, the times between sub- positive in the range of existence of the random
sequent events are considered independent and variable and its integral under the whole space
identically distributed random variables. This must be one. The integral between t = 0 and t = T
implies that the expected time till the next event defines the cumulative density function F(t)
does not depend on any of the details of the last (cdf), while integral between t = T and t = +1
event but its time of occurrence. Once the earth- defines the survivor function S(t), which repre-
quake occurs, the renewal process resets from the sents the probability that the system will survive
beginning. beyond time T, i.e., at least time T will elapse
Although a complete statistical analysis of between consecutive events. The cumulative
earthquakes should consider the distribution of and the survivor functions are defined,
all the five dimensions (i.e., time, magnitude, respectively, as
and location), the physical variable largely used ðT
to make statistical inferences is the inter-event FðT Þ ¼ Pft T g ¼ f ðuÞdu (2)
time, Dti = titi1, i.e., the time between two 0
786 Earthquake Recurrence
property, it is not adequate to model the occur- A counting process is said to be a Poisson process
rence of earthquakes on an individual fault with rate l, l > 0, if it is characterized by:
where non-Poisson models perform better. In
the following section, the most common statisti- (i) Stationarity. The probability of an event in a
cal models for earthquake occurrence are short interval of time t to t + h is approxi-
described. Because of the intrinsic complexity mately lh, for any t.
of the problem of earthquake forecasting, in
PSHA usually several models are used at the PfN ðhÞ¼1g ¼ lh þ oðhÞ (6)
same time, the contributions of which are then
weighted using a logic tree scheme. On one (ii) Nonmultiplicity. The probability of two or
E
hand, this approach represents the scarce knowl- more events in a short interval of time is
edge about earthquake occurrence process, and negligible compared to lh.
it is highlighted by the fact that, in some cases,
antithetical models are simultaneously used. On Pf N ð h Þ 2 g ¼ o ð h Þ (7)
the other hand, it is a way to contemplate the
complexity of the earthquake occurrence, since (iii) Independence. The number of events in any
different probability models incorporate differ- interval of time is independent of the num-
ent physical attributes of the earthquake process ber in any other (nonoverlapping) interval
that a single statistical distribution could not of time. The process is thus said to be
incorporate. Generally speaking, in any model “memoryless.”
there may be two types of uncertainty: aleatory (iv) The number of events in any interval of
uncertainty and epistemic uncertainty. The ale- length t is Poisson distributed with mean
atory uncertainty is associated to the intrinsic lt. That is, for all s, t 0,
stochastic nature of the process resulting in an
unavoidable impossibility of predicting deter- ðlsÞk els
ministically its evolution. The epistemic uncer- PfN ðtþsÞ N ðtÞ ¼ kg ¼
k! (8)
tainty represents the limited knowledge of the for k ¼ 0, 1, 2, ; l > 0
system. To the extent that a process is knowable,
in principle its epistemic uncertainty is reduc- where l is the rate of occurrence of events, i.e.,
ible, while aleatory uncertainty is not. The inclu- the number of events in unit of time.
sion of different models for earthquake Starting from (iv), it can be shown that, for a
forecasting, and the adoption of a logic tree Poisson process, the inter-event times t (or Dt)
scheme, is a way to quantify the epistemic uncer- are exponentially distributed with cdf F(t) and
tainty of the process. pdf f(t) given by
of time remains constant regardless of the elapsed the observed data, the hazard function can be
time since its occurrence or the size of the last obtained noting that
event (in case of spatial Poisson model).
The shape of the pdf and the hazard function f ð tÞ d
hð t Þ ¼ ¼ fln½1 FðtÞg t 0
for three different values of l are shown in Fig. 2, 1 FðtÞ dt
panel a. As previously noted, the hazard function (14)
is a constant that depends only on the rate
of occurrence of the events in the area under As an example, this model has been applied by
study or on the fault, i.e., the number of target Reasenberg (1985) to model aftershock
earthquakes in the unit of time. Areas or sequences and formulate a declustering proce-
faults with higher rate of occurrence have thus dure for seismic catalogs. Concerning the l(t)
higher probability to experience the next earth- function, Hong and Guo (1995) by analyzing
quake, regardless the time occurred since the last the seismicity near Taiwan have demonstrated
one. that a U-shaped function is more suited to
A direct generalization of the Poisson process model the occurrence of clustering earthquakes
is the nonhomogeneous Poisson process. As the with respect to a constant rate or a monotonic
regular Poisson process, it is based on the increasing or decreasing function. The U-shaped
assumption of independent increments and at function prescribes a constant hazard during the
most one occurrence at any instant of time. Dif- waiting period between cluster of events and
ferent from the homogeneous Poisson process, higher values at times immediately after the last
rate parameter l depends on time, i.e., l(t). For event or before the next sequence of events.
a nonhomogeneous Poisson process, the proba-
bility of having n events in a given time interval
(t1, t2) is given by Time-Dependent Models
ð t2 n Unlike the time-independent models, in time-
lðuÞdu dependent models, the probability of occurrence
t1
P½N ðt2 Þ N ðt1 Þ ¼ n ¼ of the next event (either in time or space) is
ð t2 n!
modified by the features of the preceding events.
lðuÞdu In particular, a large class of statistical models
e t1 n ¼ 0, 1, 2, . ..; assumes that the probability of future events may
(12) increase with elapsed time since the last earth-
quake and is in general applied to define the
where l(t) is the instantaneous rate of occurrence distribution of earthquakes on single faults.
ð t2 Moreover, for a fault that is in the early stage of
and lðuÞdu is the expectation value over the the seismic cycle, these models prescribe a lower
t1
considered time interval. probability of occurrence compared to the
The inter-event time random variable for a set Poisson model, whereas for faults in the late
of events modeled as a nonhomogeneous Poisson stage of the seismic cycle, they predict a higher
process has the following cumulative probability of next event with respect to the
distribution: Poisson model. Other models imply that earth-
quake occurrence probability is higher immedi-
ðt ately after a previous event, and it decreases with
lðu þ ts Þdu the time elapsed; this behavior represents the
F ð tÞ ¼ 1 e 0 t0 (13) clustered renewal models, and it is more suitable
to represent the seismicity of extended areas,
where ts is the time since the last event. Once the which enclose more than a single fault. In both
appropriate l(t) has been identified by analyzing cases, after the occurrence of each event at a
Earthquake Recurrence 789
Earthquake Recurrence, Fig. 2 Probability density (panel d). In each panel, line colors refer to different
functions and hazard rate functions for the Poisson combinations of the parameters that characterize the spe-
model (panel a), the Weibull model (panel b), the cific distribution
gamma model (panel c), and the lognormal model
given time t, the system is brought back to the independent, then the sequence of earthquakes is
conditions it presented at the time of the last a perfect periodic process, and the model is a
renewal (i.e., the system is perfectly renewed). characteristic recurrence model (Fig. 3a). On
Early examples of time-dependent models the other hand, when t1 or t2 or both are not
applied to single fault as an alternative to the constant, the process will be no longer
standard Poisson model are dating back to the periodic. In the assumption that fault strength is
1970s (Bufe et al. 1977). Analyzing data relative constant and that the fault will always rupture
to large thrust-fault earthquakes in Japan, when the stress reaches the maximum stress
Shimazaki and Nakata (1980) observed that the capacity of the fault t1 (i.e., t1 is constant), if t2
larger the coseismic slip displacement measured is not constant, slip on the fault can vary with
in an earthquake, the longer is the following quiet each event. Of course, greater slip releases more
period. Based on simple assumptions about the energy and the recurrence time (time between
initial stress t1 and the final stress t2 of faulting, slip events) is longer because those stresses
the authors suggested two recurrence models need to build up once again. This model is said
known as time-predictable model and slip- to be time predictable, because knowing the
predictable model. In the following, the main amount of slip during the past earthquake, it
features of both models are described. Consider- allows prediction of the time of the next earth-
ing for simplicity that cyclic strain, which is quake, but not its magnitude (Fig. 3b). Differ-
accumulated in time with a constant rate due to ently, suppose that t1 is not constant, that is, the
tectonic processes is relieved on the same fault, fault does not rupture at the same stress level each
then the stress drop Dt is assumed to be propor- time. If t2 is constant, each event reduces the
tional to the coseismic slip Du : stress on the fault to the same stress level. After
an earthquake, stress on the fault will increase at a
Du constant rate from t2. The potential fault slip at
Dt ffi m (15) any time is proportional to the shear stress on the
L
fault. Thus, if the time of the last rupture is
where m is the shear modulus and L is a charac- known, the shear stress on the fault and the poten-
teristic fault dimension. Now, if t1 and t2 are time tial displacement can be determined at any
Earthquake Recurrence 791
Earthquake Recurrence, Fig. 3 Earthquake occurrence level during the earthquake (t2). (b) Time-predictable
models. For each model, the upper panels show the stress model illustrating stress buildup to a certain level and
pattern during different seismic cycles, while the lower nonuniform stress drops. (c) Slip-predictable model show-
panels show the corresponding structure of fault slip. (a) ing nonuniform stress buildups and stress drops to a cer-
Perfectly periodic model of constant stress increase to a tain level
certain level (t1) and constant stress drop back to a certain
particular time of interest. This model is said to times of great earthquakes at the Pacific subduc-
be slip predictable, in that, it cannot be used to tion margins. The formula for the probability
predict when rupture will occur, but it can be used density function of the general Weibull distribu-
to predict the magnitude of the earthquake that tion is
would occur at any given time (Fig. 3c). The
application of these models to earthquake fore- b t mb1 ðtm
e a Þ for t m; b, a > 0
b
f ðtÞ ¼
cast has encountered criticism in the seismologi- a a
cal community. As an example, Mulargia and (16)
Gasperini (1995) have tested the two models on
the Italian seismic catalog and found no signifi- where b is the shape parameter, m is the location
cant correlation between the coseismic slip of an parameter, and a is the scale parameter. When
event and that previous or subsequent inter-event m = 0 and a = 1, the distribution is called the
time. standard Weibull distribution. The case where
In the following, the most common applied m = 0 is called the 2-parameter Weibull distribu-
statistical models to model earthquake occur- tion which is generally used for seismological
rence are described. applications. For the 2-parameter distribution,
the corresponding cdf and the hazard function
Weibull Model are defined as follows, respectively:
One of the first applications of the Weibull dis-
FðtÞ ¼ 1 eðaÞ
t b
tribution in geosciences is the one proposed by (17)
Hagiwara (1974). Based on the assumption that
the Earth crust is strained with a constant speed, b t b1
he modeled crustal-rupture occurrence time of hðtÞ ¼ (18)
a a
the South Kanto District in Japan by using the
Weibull model. The same distribution was also Equations 16,17, and 18 show that the shape
used by Rikitake (1976) to model the recurrence parameter b has a marked effect on the shape of
792 Earthquake Recurrence
the functions. However, some interesting insights distribution allows to combine a power law,
can be drawn from the analysis of the hazard which is usually used to model short inter-event
function (Eq. 18) as shown in Fig. 2b, in which times, that is, clustered events, and long inter-
the pdf and the hazard functions are displayed for event times through the exponential function
three different values of b (i.e., 0.5, 1.0, and 2.0) which is typical of Poisson seismicity. The
and an a value of 3.0. Indeed, if the inter-event corresponding cdf is
time distribution has a Weibull best-fit model,
then the value of b can be used to depict some gða, t=bÞ
FðtÞ ¼ (20)
features of the analyzed seismicity. If, for exam- G ð aÞ
ple, b is equal to 1, the pdf reduces to an expo-
nential function with constant hazard function where
and rate equal to 1/a; the process is thus a random
process. If b < 1, then the system exhibits early- ð
1 ðt
type failures, with a decreasing hazard function, G ð aÞ ¼ ua1 eu du gða, tÞ ¼ ua1 eu du
with time elapsing since the most recent event, 0 0
while for b > 1 the system exhibits wear-out type (21)
failures, with an increasing hazard function. The
analog for the earthquake generation process for while the hazard function is given by
b < 1 is a clustering property of the earthquake
occurrence, while for b > 1 the pdf can be ta1 eb
t
specific faults or plate boundary segments, hazard functions have zero value t = 0, i.e., the
can be described by a lognormal distribution. probability of having an earthquake is zero at
Combining both historic and geologic recur- the beginning of the cycle, then they reach their
rence data, the authors found that the lognormal maximum, and finally decrease asymptotically
distribution fits the observed data significantly to zero.
better than either the Gaussian or the Weibull
distributions. Brownian Passage-Time Model
Since the study by Nishenko and Buland Brownian passage-time model (BPT model
(1987), it has become common practice to use hereinafter) has been developed by Matthews
the lognormal distribution in several applications and coworkers in the early 2000s in the frame-
E
in seismology. One of the most relevant was in work of earthquake forecasting. The aim was to
computing earthquake probabilities in California make a connection between the rebound theory,
(Working Group on California Earthquake Prob- with the main characteristics of the stress and
abilities 2007 and previous works) and for time- strain accumulation process, and the observable
dependent seismic hazard analysis (e.g., Cramer distribution of the real inter-event data and build
et al. 2000). up a physically motivated renewal model for
The pdf of the distribution is given by the earthquake recurrence (Ellsworth et al. 1999;
equation: Matthews et al. 2002). In BPT model, the con-
stant tectonic loading is superimposed by a
ðlntmÞ2 Brownian perturbation, which makes the stress
1 2
f ðtÞ ¼ pffiffiffiffiffiffi
2s
e lnT (23) loading a Brownian relaxation oscillator (BRO).
2p sln T t
The Brownian perturbation could be interpreted
as a physical representation of the interaction
where sln T is the standard error deviation of the
of external perturbation of the cycling stress
logarithm of the recurrence time. The
load of a fault. In the BRO model, a new earth-
corresponding cdf and the hazard function h(t)
quake will occur when a state variable (or a
have the following expression:
set of them) reaches a threshold (Xf); it relaxes
! the load state to the characteristic ground
1 1 ln t m level (X0) and begins a new cycle (panel a in
FðtÞ ¼ þ er f pffiffiffiffiffiffiffiffiffiffiffi (24)
2 2 2s2ln T Fig. 4).
In the BPT model, the loading of a fault is the
" !!#1 composition of two elements: a constant-rate
pffiffiffiffiffiffi lnt m
ðlntmÞ2
2
hðtÞ ¼ 2 2p slnT t 1 er f pffiffiffiffiffiffiffiffiffiffi e 2s
lnT loading component, lt, and a random component,
2s2lnT ϵ(t) = sW(t), as a Brownian motion (where W(t)
(25) is a standard Brownian motion and s is the non-
negative scale parameter). Standard Brownian
One of the main features of the lognormal dis- motion is simply the integration of stationary
tribution is that it is a heavy-tailed distribution, increments where the distribution of the incre-
that is, there is a larger probability of getting ments is Gaussian (which might be motivated
next earthquakes with respect to the exponential by central-limit arguments if the perturbations
function for the same time window. This feature are considered as the sum of many small, inde-
makes the distribution suitable for fitting inter- pendent contributions), with zero mean and con-
arrival times of those events whose return stant variance:
period is larger than the average return period.
This is visually explained in Fig. 2d, where the XðtÞ ¼ lt þ s W ðtÞ (26)
pdfs and the hazard functions for m = 0 and
three different values of slnT (e.g., 0.5, 1.0, and The BRO model belongs to the family of stochas-
1.5) are shown. The pdfs are peaked, and the tic renewal processes identified by four
794 Earthquake Recurrence
Earthquake Recurrence, Fig. 4 Panel (a) shows three ground state X0 and the failure state Xf, respectively.
examples of loading state variable trend for the three Panel (b) shows the pdf and the hazard rate as a function
different aperiodicity values indicated in the inset of of three different aperiodicity values
panel (b). The levels indicated possible values of the
parameters: the mean loading (drift) rate (l), the m 1=2 ð2ma
tmÞ2
1 1
alternative mechanisms (seismic or aseismic) and seismicity contained in the United States Coast
not through the occurrence of the event, and so a and Geodetic Survey (USCGS) catalog from
decreased failure rate may be appropriate. On the 1961 to 1968, by using the generalized Poisson
contrary, models with hazard function reaching a model. Assuming that cluster occurrence is
constant value as time goes to infinity, as for the governed by a Poisson process and that in each
BPT and the gamma distribution, imply that after cluster the number of subsequent earthquakes
the time corresponding to the mean recurrent follows a negative power law distribution, Shlien
interval is passed, the distributions become and Toksöz (1970) provided a recursive relation
“time independent,” and the time since the most to compute the distribution of the number of days
recent event does not influence any more the with n earthquakes based on two main parameters
occurrence of the next event. The only factors named k0 and E. The probability distribution has
able to control the earthquake occurrence are in the following formulation:
the model parameters, as, for example, the mean
and the aperiodicity of the BPT distribution. kt X n
pq ðn þ 1, tÞ ¼ ð n j þ 1Þ
n þ 1 j¼0 (31)
Trigger Models qðn j þ 1Þ pq ðj, tÞ
One of the most important characteristics of the where q(n) is the selected negative power law
earthquake time series, which must be considered distribution also known as the Pareto distribution,
in the models of recurrence, is the tendency of the given by
events to clustering. A possible definition of clus-
ter is a set of earthquakes originating from a 1 E
qð nÞ ¼ n (32)
relatively small seismogenic volume and sepa- z ð EÞ
rated in time by very small intervals. Thus, the
model should contemplate a finite probability of where the normalization factor z(E) is the
more than one event to occur in a given time Riemann Z function.
increment. This feature is not taken into account The k0 parameter controls the mean rate of
by the Poisson process. A possible solution, cluster occurrence. The parameter E is a measure
which allows for events grouping, is the use of a of the departure from the simple Poisson process.
compound Poisson process. This process permits For large value of E (let us say larger than 5), the
that groups of independent and identically dis- generalized Poisson model tends to a simple
tributed events occur together at the instants of a Poisson distribution. Figure 5 shows the shape
simple Poisson process. However, the compound of the pdf for k0 = 1 and three different values
Poisson process may be not satisfactory because of E (i.e., 2.0, 3.0, and 4.0). In the same figure, the
groups of events tend to spread or to diffuse over right panel shows the hazard function obtained by
space and time. As an alternative class of models, numerical integration of the pdf for the same
those named trigger models have been proposed values of the two parameters. Among the several
by Vere-Jones and Davies (1966). Basically, the applications, the generalized Poisson model has
models assume that events initiating groups of been also used to model earthquake sequences
shocks – triggering events – occur by following and compute seismic hazard in volcanic contexts
a Poisson process with a constant rate, while the (De Natale and Zollo 1986; Convertito and Zollo
number of events in each group decays by fol- 2011).
lowing a prescribed function.
The ETAS Model
Generalized Poisson Model The spatiotemporal clustering of small-to-
In the framework of the trigger models, Shlien moderate size seismicity is a concept worldwide
and Toksöz (1970) proposed to model the accepted in seismology. Among the few
Earthquake Recurrence 797
Earthquake Recurrence, Fig. 5 Probability density function and hazard rate function for the generalized Poisson
model. The selected values of the parameters k0 and E are reported in the inset of the right panel
spatiotemporal models proposed in literature, the of the Omori-Utsu law, which describes the
epidemic type aftershock sequence (ETAS) temporal behavior the aftershock sequence.
model is by far the most studied and applied, The 5 parameters of the ETAS model are
and, in some cases, it has been defined as the generally estimated by using a maximum like-
best way to describe short-term seismicity lihood procedure. In literature, different
(Console et al. 2007; Zhuang et al. 2011, 2011) parameterizations of the ETAS model are
since its formulation in 1988 by Ogata (1988). available, especially when also spatial distri-
ETAS model is grounded on empirical statistics bution of the events is taken into account. For
and represents a stochastic implementation of the the purpose of this encyclopedia, the main
physically based model of Dieterich (1994), features common to the ETAS models devel-
based on the interaction among faults. It assumes oped by different researchers are summarized
that there is no difference in triggering seismicity in five points (Zhuang et al. 2011):
among foreshocks, mainshocks, and aftershocks,
and event triggers its own offspring indepen- • The rate of background events is assumed to
dently according to some probability rules. The be a function of space and magnitude; the
hazard rate of the ETAS model is defined as background events are generated following a
specific pdf function (e.g., Poisson, BPT, etc.),
X e½aðmm0 Þ and so they are considered as “immigrants” in
lð t Þ ¼ m þ K 0 (33)
ti <t
ðt ti þ cÞp the branching process.
• Each event produces direct offspring events
where m is the background rate (unit of time) independently of the others, and the produc-
of the seismicity and m0 is the magnitude of tivity depends on its magnitude.
completeness of the seismic catalog. The • The pdf of the time until the appearance of a
parameter a is a measure of the efficiency of child is a function only of the elapsed time
a shock in generating an aftershock activity since its parents.
and it is relative to its magnitude; K0 repre- • The pdf of the location and magnitude of chil-
sents the productivity of an event of threshold dren are a function of the location and magni-
magnitude m0, whereas c and p are parameters tudes of the parents.
798 Earthquake Recurrence
• Magnitude, of both offspring and background identified by a vector u, the best model parame-
events, is a random variable, with a certain ters y^ are those maximizing the L function:
distribution, the most common of which is
the Gutenberg-Richter relation. y^ ¼ maxLðN; yÞ (34)
ETAS model is widely adopted in short-term Generally speaking, the MLE method maximizes
hazard studies, for its ability in the forecasting of the agreement of the selected model with the
aftershock sequence; also, it is adopted as observed data.
declustering method for separating the back- When several competing models are used to fit
ground seismicity to the triggered one (Zhuang the same data set, the best model can be discrim-
et al. 2002, 2004). Moreover, it has been found inated by using the Akaike information criterion
that the inclusion of a more realistic behavior of (AIC), which uses the following statistic to eval-
the seismicity in the temporal parameterization of uate the fitting performance of each model:
earthquake occurrence in PSHA, with the well-
known structure of the sequence, provides an AIC ¼ 2 max logLðyÞ þ 2N p (35)
y
increment of almost 10 % in the hazard, in low
seismicity area, where the identification of active
where Np is the number of parameters involved in
faults or background seismicity is trivial (Faenza
each investigated model. The model with the
et al. 2007).
lowest AIC value is identified as the optimal
model. AIC balances between the goodness of
fit of the model and its complexity, and it pro-
Estimation of Parameters
vides penalties to models with a lot of numbers of
It is worth noting that the models described in
parameters, to discourage overfitting. AIC does
this entry may fit the available data set collected
not provide a test of the model in absolute sense,
in the region under study equally well. This is
but provides only a tool for the comparison of
especially true in case of large earthquakes,
different models; if all models have poor fitting,
where the amount of data available may be not
AIC will not give any warning of it.
sufficient to test any hypotheses due to the lim-
The residual analysis allows the evaluation of
ited time extent of the available seismic catalog.
the degree at which the adopted model fits the
Before performing the parameter estimation,
data. The analysis uses a transformation to map
the completeness of the catalog has to be deter-
general event times, in times having the same
mined. This could be trivial in case of historical
properties as homogeneous Poisson time with
earthquakes; an erroneous evaluation of the
unit rate parameter (Zhuang et al. 2012). Thus,
completeness can affect the value of the model
comparing the observed seismic rate with that
parameters and their uncertainties. It must be
obtained in the transformed time domain allows
reminded that large events are rare, especially
to evaluate the quality of the fit. The comparison
in area of moderate seismicity, and therefore a
allows also to recognize periods of decreased or
larger part of the seismic catalog are derived
increased seismic activity (Ogata 1988; Zhuang
from pre-instrumental time.
et al. 2012).
In practical applications, the maximum likeli-
hood estimate (MLE) is by far the most used
technique to evaluate the parameter’s models.
The MLE is based on the use of the likelihood Summary
function L for the estimation of parameters. The
L function represents the joint probability density Seismic hazard in earthquake-prone regions can
of the analyzed variable, for example, the waiting be expressed in several forms. The exceedance
times or the inter-arrival times. Given N data of a given ground-motion threshold value or the
points and a model whose parameters are probability of having a given number of future
Earthquake Recurrence 799
earthquakes that can affect a specific structure Finally we have discussed the generalized
is of primary interest in seismic engineering. Poisson model and the ETAS model which, in
As a consequence, being able to model present addition to model single large-to-moderate
and past seismicity that occurred in the area events, allow to account for sequences of events
under study may be essential to make reliable clustering in time and space.
forecast. As previously mentioned, the selection of “the
In this framework, the present work discusses best model” among several competitors is not
the basic assumptions and the origin of the main straightforward. In general, dealing with large
statistical models used in the literature to fit the earthquakes implies working with small numbers
observed seismicity. Among the five dimensions of data available for the fitting; this is especially
E
which concur to basically characterize the earth- true in case of recurrence on single fault or in area
quakes (i.e., time, magnitude, and location in 3D with moderate seismicity. In this perspective, in
space), the physical observable of interest for recent years, the seismological community has
statistical modeling is the earthquake recurrence, developed rigorous and truly perspective forecast
which is the time between subsequent events. In experiments, as, for example, the CSEP
this respect, earthquakes are considered as spe- experiment – Collaboratory for the Study of
cific stochastic processes named renewal process Earthquake Predictability, for different regions
(Cornell and Winterstain 1988) in which after the worldwide.
occurrence of an event, the system process
restarts from the beginning (i.e., is completely
renewed).
The models used to describe the recurrence of Cross-References
the earthquakes can be divided in two main clas-
ses: the “memoryless” model, i.e., Poisson ▶ Earthquake Mechanisms and Tectonics
model, and the time-dependent models. While ▶ Probabilistic Seismic Hazard Models
Poisson model is extensively used in PSHAs, its ▶ Site Response for Seismic Hazard Assessment
adequacy is limited since the mechanism of
earthquake interactions at regional scale has
highlighted the tendency to a spatiotemporal
References
clustering of earthquake occurrence, even for
moderate earthquakes. Moreover, due to the Bakun WH, King GCP, Cockerham RS (1986) Seismic
“memoryless” property, it is not adequate to slip, aseismic slip, and the mechanics of repeating
model the occurrence of earthquakes on single earthquakes on the Calaveras fault, California. In:
Das S, Boatwright J, Scholz CH (eds) Earthquake
faults in which case non-Poisson models perform
source mechanics, vol 37, Geophysical monograph.
better. American Geophysical Union, Washington, DC
As a consequence, since the work of Bufe Bufe CG, Harsh PW, Burford RO (1977) Steady-state
et al. (1977), a suite of models able to account seismic slip-A precise recurrence model. Geophys
Res Lett 4:91–94. doi:10.1029/GL004i002p00091
for earthquake interaction and clustering has been
Console R, Murru M, Catalli F, Falcone G (2007) Real
introduced. We have presented those most used time forecasts through an earthquake clustering model
in practical applications such as Weibull, gamma, constrained by the rate-and-state constitutive law:
and lognormal and described their main features comparison with a purely stochastic ETAS model.
Seismol Res Lett 78:49–56. 3, 15
and differences in terms of their probability den- Convertito V, Zollo A (2011) Assessment of pre-crisis and
sity functions and hazard functions. In addition syn-crisis seismic hazard at Campi Flegrei and
we described a more physical model, that is, the Mt. Vesuvius volcanoes, Campania, southern Italy.
Brownian passage time, which is currently used Bull Volcanol 73:767–783. doi:10.1007/s00445-011-
045-2
to perform time-dependent seismic hazard anal-
Cornell CA, Winterstain SR (1988) Temporal and magni-
ysis in several areas, tectonic and volcanic areas tude dependence in earthquake recurrence models.
such as California and Italy. Bull Seismol Soc Am 78:1522–1537
800 Earthquake Recurrence Law and the Weibull Distribution
Cramer CH, Petersen MD, Cao T, Toppozada TR, Reichle Shimazaki K, Nakata T (1980) Time-predictable recur-
MS (2000) A time-dependent probabilistic seismic- rence model for large earthquakes. Geophys Res Lett
hazard model for California. Bull Seismol Soc Am 7:279–282
90:1–21 Shlien S, Toksӧz MN (1970) A clustering model for
De Natale G, Zollo A (1986) Statistical analysis and earthquake occurrences. Bull Seism Soc Am 60:
clustering features of the Phlegraean fields earthquake 1765–1787
sequence (May 1983–May 1984). Bull Seismol Soc Udias A, Rice J (1975) Statistical analysis of microearth-
Am 76(3):801–814 quake activity near San Andreas Geophysical Obser-
Dieterich JH (1994) A constitutive law for rate of earth- vatory, Hollister. Calif Bull Seism Soc Am 65:
quake production and its application to earthquake 809–827
clustering. J Geophys Res 99:2601–2618 Vere-Jones D, Davies RB (1966) A statistical survey of
Denker M, Woyczynski WA (1998) Introductory statistics earthquakes in the main seismic region of New
and random phenomena. Birkha€ user, Boston/Basel/ Zealand, 2, time series analysis. N Z J Geol Geophys
Berlin 9:251–284
Ellsworth WL, Matthews MV, Nadeau RM, Nishenko SP, Working Group on California Earthquake Probabilities,
Reasenberg PA, Simpson RW (1999) A physically 2007, Uniform California earthquake rupture forecast,
based earthquake recurrence model for estimation of USGS open-file report 2007-1437
long-term earthquake probabilities. U.S. Geological Youngs RR, Coppersmith K (1985) Implications of fault
Survey open-file rept. 99–522 slip rates and earthquake recurrence models to proba-
Faenza L, Hainzl S, Scherbaum F, Beauval C (2007) Sta- bilistic seismic hazard estimates. Bull Seism Soc Am
tistical analysis of time-dependent earthquake occur- 58:939–964
rence and its impact on hazard in the low seismicity Zhuang J, Ogata Y, Vere-Jones D (2002) Stochastic
region Lower Rhine Embayment. Geophys J Int declustering of space-time earthquake occurrences.
71(2):797–806 J Am Stat Assoc 97:369–380
Gutenberg B, Richter CF (1944) Frequency of earth- Zhuang J, Ogata Y, Vere-Jones D (2004) Analyzing
quakes in California. Bull Seismol Soc Am earthquake clustering features by using stochastic
34:185–188 reconstruction. J Geophys Res 109(B5):B05301.
Hagiwara Y (1974) Probability of earthquake occurrence doi:10.1029/2003JB002879
as obtained from a Weibull distribution analysis of Zhuang J, Werner MJ, Hainzl S, Harte D, Zhou S (2011)
crustal strain. Tectonophysics 23:313–318 Basic models of seismicity: spatiotemporal models,
Hong L-L, Guo S-W (1995) Nonstationary Poisson model community online resource for statistical seismicity
for earthquake occurrences. Bull Seismol Soc Am analysis, doi:10.5078/corssa-07487583. Available at
85:814–824 http://www.corssa.org
Matthews MV, Ellsworth WL, Reasenberg PA Zhuang J, Harte D, Werner MJ, Hainzl S Zhou S (2012)
(2002) A Brownian model for recurrent earthquakes. Basic models of seismicity: temporal models commu-
Bull Seismol Soc Am 92:2233–2250 nity online resource for statistical seismicity analysis.
Mulargia F, Gasperini P (1995) Evaluation of the applica- doi:10.5078/corssa-79905851. Available at http://
bility of the time- and slip-predictable earthquake www.corssa.org
recurrence models to Italian seismicity. Geophys
J Int 120(2):453–573
Nishenko SP, Buland R (1987) A generic recurrence inter-
val distribution for earthquake forecasting. Bull
Seismol Soc Am 77:1382–1399
Ogata Y (1988) Statistical models for earthquake occur- Earthquake Recurrence Law and the
rences and residual analysis for point processes. J Am Weibull Distribution
Stat Assoc 83:9–27 3
Reasenberg P (1985) Second-order moment of central
California seismicity, 1969–82. J Geophys Res Jui-Pin Wang
90:5479–5495 Department of Civil and Environmental
Reid HF (1910) On mass-movements in tectonic earth- Engineering, The Hong Kong University
quakes. In: The California earthquake of April
18, 1906: report of the state earthquake investigation
of Science and Technology, Kowloon,
commission. Carnegie Institution of Washington, Hong Kong
Washington, DC
Rikitake T (1976) Recurrence of great earthquakes at
subduction zones. Tectonophysics 35:335–362
Schwartz DP, Coppersmith J (1984) Fault behaviour and
Synonyms
characteristic earthquakes: examples from Wasatch
and San Andreas faults. J Geophys Res 89:5681–5698 Earthquake; Weibull distribution
Earthquake Recurrence Law and the Weibull Distribution 801
Frequency in years
around Taiwan since 1978;
60
this statistical study offered
a tangible support to the use
of the Poisson distribution
in modeling the 40
earthquake’s randomness
in time (Wang et al. 2014) E
20
0 1 2 3
Annual occurrence
Earthquake Recurrence 4
Law and the Weibull
−1
−2
3 4 5 6 7 8
Magnitude of exceedance, m*
with Eq. 1
5
5000 0
6.0 6.5 7.0 7.5 8.0
0
3 4 5 6 7 8
Earthquake magnitude (ML)
a logarithmic scale, rather than in the arithmetic The Probability Density Function of
scale used to present the earthquake-size proba- Earthquake Size Around Taiwan
bility or frequency.
Therefore, in this technical note another As shown in Fig. 3, the probability density func-
option for the earthquake analysis is introduced. tion for earthquake size was extracted from
Different from the existing approach previously a catalog listing the earthquakes around Taiwan
mentioned, the alternative is an application of the since 1900 (Wang et al. 2011, 2012a). For exam-
Weibull distribution to earthquake statistical ple, above all ML 3.0 (local magnitude) earth-
studies. The details of the methodology and quakes, the probability density for earthquakes
a case are given in the following. larger than 7.0 is around 0.01 % among the total
Earthquake Recurrence Law and the Weibull Distribution 803
of 55,000 events. In addition, the data also show problem with the involvement of the gamma
that the earthquake-size probability distribution function. As a result, an in-house optimization
is nonlinear and asymmetrical, with the fre- procedure was developed for the task, and it was
quency or density decreasing rapidly as magni- introduced in the following section for the
tudes increase. Therefore, an asymmetrical reader’s information.
probability model such as the Weibull distribu-
tion should have more potential to model such an
earthquake statistic, than symmetrical distribu- Model Parameter Calibrations
tions like the normal distribution.
Since the solution could be difficult to be devel-
E
oped analytically, a numerical procedure to cal-
The Probability Function of the Weibull culate a and b was developed and the details
Distribution were given in the following. Using the first
given information (i.e., mX), from Eq. 3 the rela-
The probability density function (PDF) of the tionship between a and b can be rewritten as
Weibull distribution is expressed as follows follows:
(Devore 2008):
(a 1
xa1 e ðx=bÞa G 1þ
x0 a
f ðx; a, bÞ ¼ b b¼ (6)
0 x<0 mX
(2)
With this relationship, the calibration can be
where a and b are the model parameters, followed by a trial-and-error process by varying
which can be calibrated with the mean the value of a (so b changes as well) and
(i.e., mX) and SD (i.e., sX) of X. The relations substituting both into Eq. 4 to calculate the vari-
between mX, sX, a, and b are as follows: ance of X. As a result, the absolute difference,
D = f (a), between the calculated variance and
1 given variance (i.e., the second given informa-
mX ¼ bG 1 þ (3)
a tion) in each trial can be calculated.
!
With the preliminary analysis showing
2 1 2 the change in D given different a values, it was
sX ¼ b
2 2
G 1þ G 1þ (4)
a a found that this variable would converge with the
increase of a from a nominal value near zero
where G() denotes the gamma function: (a cannot be zero; see Eq. 4) to the solution of
a. Therefore, the searching process will be
ð
1 stopped when Di+1 = f (ai + Da) is greater than
GðaÞ ¼ xa1 ex dx (5) Di = f (ai). For a better resolution in results,
0
a smaller increment of a can be used, say, 0.005.
Apparently, this numerical solution is a
Therefore, given the mean and SD from obser- computer-based procedure, and the tedious iter-
vations or samples, a and b of the Weibull distri- ation should be complete with computers in
bution can be calibrated with Eqs. 3 and 4. a fraction of time. Therefore, this algorithm pro-
Afterward, the corresponding Weibull distribu- posed was programmed into two user-defined
tion can be developed for modeling earthquake- functions for Excel. Note that the two functions
size distributions. However, although this is are nearly identical except for returning a or b at
a two-equation-two-unknown problem, the ana- the end. For example, the syntax of the function
lytical solution of a and b could be fairly chal- to calibrate a was created as follows:
lenging to be developed for such a mathematical WEIBULL _ AFA(mn, sd), and obviously
804 Earthquake Recurrence Law and the Weibull Distribution
25000
Observation:
μ = 3.442, σ = 0.398
20000
Weibll distribution:
α = 1.11, β = 0.46 given
Frequency
15000 shifted mean = 0.442 (= 3.442 - 3)
10000
5000
3 4 5 6 7 8
Magnitude (ML)
Earthquake Recurrence Law and the Weibull Distri- earthquake size; this statistical analysis indicated that the
bution, Fig. 4 The comparison between the observed Weibull distribution should be suitable to the modeling of
earthquake-size distribution and the Weibull distribution earthquake-size distributions based on the empirical find-
with a = 1.11 and b = 0.46 calibrated with the observed ing with earthquake data from Taiwan
mean and standard deviation in the randomness of
“mn” and “sd” are the function input, standing Case Study with Earthquake Data
for the mean and standard deviation of the ran- Around Taiwan
dom variable, respectively. The complete VBA
(Visual Basic for Application) scripts of the As shown in Fig. 3, given magnitude
function are given in the Appendix for the threshold = 3.0, the mean earthquake size and
reader’s information. standard deviation from about 55,000 earth-
The in-house function was subject to verifica- quakes in Taiwan are equal to 3.442 and 0.398,
tion before it was used in the following statistical respectively. With Eqs. 3 and 4, a and b of the
analysis. The first attempt is to calibrate a and b Weibull distribution are 1.11 and 0.46, respec-
given m = 10 and s = 1. With the in-house tool, tively. It must be noted that such a calculation
a and b were computed at 12.16 and 10.43, was performed on the basis of shifted data,
respectively. Substituting them back into Eqs. 3 because the Weibull distribution is about the
and 4, m and s were found equal to 10 and 0.999, modeling of a random variable ranging from
which matched the given values almost perfectly. zero to infinity. As a result, the a and b values
The second attempt is to test the function’s reli- calibrated are corresponding to the shifted
ability under a large-variability condition with data with mean = 0.442 and SD = 0.398.
the standard deviation increased from 1 to 50. (The shifted data are the original data
Given m = 10 and s = 50, the corresponding a subtracting the magnitude threshold of 3.0.)
and b are equal to 0.31 and 0.25, respectively. Note that because the variability of data is not
Again substituting both into Eqs. 3 and 4, m and s affected by data shifting, the SD used for cali-
were computed at 10 and 50.45, also close brating a and b values is the same as the
enough to the given conditions. As a result, after original SD.
two verifications the in-house algorithm and user- With the model parameters calibrated from
defined function were proved reliable and observations, Fig. 4 shows both the observed
functional. earthquake-size probability density and the
Earthquake Recurrence Law and the Weibull Distribution 805
modeling with the Weibull distribution. This to see if it could provide a reasonable fit to the
analysis indicated that the earthquake-size distri- randomness of earthquake size observed in
bution in Taiwan shows a good agreement with Taiwan.
the Weibull distribution. With such a statistical study, it was found that
the Weibull distribution could fit the earthquake
data in Taiwan reasonably well. For instance,
Discussions given a magnitude threshold of 3.0, mean magni-
tude of 3.442, and standard deviation of 0.398,
Applications and Verifications with the corresponding parameters a and b of the
Seismicity in Different Regions Weibull distribution are 1.11 and 0.46, respec-
E
The statistical study of this note offered tangible tively. With the parameters calibrated with the
evidence about the earthquake size’s randomness seismicity, the observation and the model are in
potentially following the Weibull distribution, a satisfactory agreement in earthquake-size
with the earthquake data around Taiwan since distribution.
1978. Although it is believed that this statistical Although this statistical study suggests
finding should be generic and could be applied to a possible relationship between earthquake size
seismicity of different region, more earthquake and the Weibull distribution, this empirical find-
statistics studies are necessary to provide more ing offers no explanation to earthquake mecha-
support or disagreement on this empirical find- nisms and to why the earthquake’s size
ing: the earthquake-size distribution follows the distribution should follow the specific mathemat-
Weibull distribution. ical function. In other words, the Weibull distri-
bution could be useful and reliable for modeling
Curve-Fitting Parameters the earthquake-size probability distribution from
Like the G-R relationship, the new earthquake a statistical point of view, but the model itself
model on the basis of the Weibull distribution is offers no explanation to the underlying earth-
an empirical relation, without physical indica- quake mechanics.
tions to why the earthquake’s size should be
a random variable following such a mathematical
function. In other words, the model parameters
Appendix: VBA Scripts of the In-House
(i.e., a and b) hold no physical explanations or
Weibull Functions Calibrating Model
meanings to earthquake mechanisms, although
Parameters
this statistical model provide a very satisfactory
fit to the earthquake data around Taiwan. After As mentioned, the two in-house functions are
all, it was still a statistical study or curve fitting,
basically the same except for returning a or b.
which was not an analytical study derived with
Therefore, the script for calculating a is only
basic laws of physics. given in the following.
E
Period or Frequency
of that mode, to a given earthquake ground m u€ðtÞ þ u€g ðtÞ þ cu_ðtÞ þ kuðtÞ ¼ 0 (1)
motion, is obtained. Thus, the response spectrum
where a dot indicates differentiation with respect to
can be seen as a measure of the destructive poten-
time, u€g(t) is the acceleration of the ground motion,
tial of a given seismic excitation. Of course, the
and u(t) is the relative displacement between the
above argument for the fundamental mode can
mass and the ground. After dividing by the mass m
also be extended to any of the “important” modes
and rearranging terms, Eq. 1 becomes
of vibration of a structure. Furthermore, the spec-
tra of various recorded earthquake ground
u€ðtÞ þ 2xou_ðtÞ þ o2 uðtÞ ¼ €
ug ðtÞ (2)
motions have been appropriately combined into
design spectra, by which the ground excitation is where o2 = k/m is the circular frequency of the
specified in all contemporary seismic design SDOF and x = c/2mo is its damping ratio (fraction
codes. For these and other reasons, which will of critical damping). Assuming the motion starts at
become obvious later, the response spectrum is a time t = 0 while the structure is at rest, i.e.,
convenient and effective engineering tool which _
uð0Þ ¼ uð0Þ ¼ 0 , the solution of Eq. 2 can be
has been used universally in structural analysis expressed by the well-known Duhamel integral as
and design. Apparently, Biot (1933) was the first 8
ðt
1 <
to introduce the concept of response spectrum
1
and Housner (1959) was the one who applied it uðtÞ ¼ p ffiffiffiffiffiffiffiffiffiffiffiffiffi u€g ðtÞexoðttÞ
o : 1 x2
to produce earthquake design spectra. Detailed 0 (3)
discussions on these matters can be found in, )
among many others, Gupta (1990), Clough and sin ½od ðt tÞdt
Penzien (1993), and Chopra (1995).
808 Earthquake Response Spectra and Design Spectra
Earthquake Response
Spectra and Design
Spectra, Fig. 2 Single-
degree-of-freedom system
under earthquake excitation
two SDOF systems with the same ratio k/m and provide an almost complete description of the
damping ratio x would respond in an identical effects of the specific ground motion on structural
manner. Therefore, the absolute maximum response.
response of the system can be symbolically writ- The motivation for the development of the
ten as response spectrum concept is rooted on the need
to compute in a straightforward manner design
maxjuðt, o, xÞj ¼ SD o, x (6) related quantities such as maximum base shear
forces, moments, etc. For example, the maximum
where SD is a single-valued function of o and x. base shear Vmax of the SDOF system of Fig. 2 due
A plot of SD versus o or the period T = 2p/o or to a specified ground motion ug(t) is
E
the natural frequency f = 1/T = o/2p, for a fixed
value of x, is called the relative displacement V max ¼ kumax ¼ k SD (9)
(or simply the displacement) response spectrum
of the given ground motion ug(t). In view of By direct substitution from Eq. 8 and using the
Eqs. 4 and 5, similar rigorous definitions, like definition of the circular frequency o, Eq. 9 can
the one of Eq. 6, can be given for the relative be also written in another useful form as
velocity (or simply velocity) response spectrum
SV and the absolute acceleration response spec- V max ¼ m SA (10)
trum SA of the ground motion ug(t). However, the
form of Eqs. 3, 4, and 5 and the preceding com- Apparently the maximum shear force can be
ments about the small-to-negligible influence of computed directly from the SA or SD spectra.
the corresponding shift phases suggest the fol- However, it should be noted that the quantity m
lowing approximate, but quite accurate for prac- SA is the maximum elastic shear force, while
tical purposes, expressions: theoretically the maximum inertia force is
m maxu€ðtÞ þ u€g ðtÞ . As explained earlier, the
SV ¼ oSD (7) spectrum acceleration SA and the maximum
absolute acceleration maxu€ðtÞ þ u€g ðtÞ are
SA ¼ o2 SD (8) nearly equal for almost all practical cases, except
for long periods and large damping ratios. The
Due to their approximate nature, the SV and SA small difference between the elastic shear force
spectra defined in Eqs. 7 and 8 are frequently and the maximum inertia force is the damping
referred to as pseudovelocity and pseudoacce- force c maxju_ðtÞj.
leration response spectra, respectively, and, in The relative velocity response spectrum, usu-
order to underline the difference with the rigor- ally of secondary importance, is related to the
ously defined SV and SD, the symbols PSV and maximum strain energy of the SDOF system,
PSA are also used. It should be noted that the i.e., SV is equal to the velocity that gives the
prefix “pseudo” (from the Greek ceύdo, false) SDOF system kinetic energy equal to the maxi-
should not be misleading, as in this case it is mum strain energy due to the earthquake ground
used to indicate an approximation and not an motion. This follows from the definition of strain
incorrect concept. In this entry the symbols SV energy and substitution from Eq. 7, i.e.,
and SA will be used as defined by Eqs. 7 and 8.
Figure 3 shows the Lefkada 2003 earthquake 1 1
record (strong motion, T component) and the Es ¼ k SD2 ¼ m SV 2 (11)
2 2
corresponding SD, SV, and SA response spectra
for two values of the damping ratio, x = 5 % and For a wide range of periods, the values of SV are
x = 10 %. Such plots illustrate the maximum quite close to those of the maximum velocity.
response of all SDOF oscillators of practical However, a considerable difference exists
interest to a given ground motion and, thus, between the response spectrum SV and the
810 Earthquake Response Spectra and Design Spectra
Acceleration (g)
Fig. 3 Acceleration record 0.2
of the Lefkada (Greece)
2003 earthquake (14/08/ 0.0
2003 5:18:48 (GMT), OTE
−0.2
building, Lefkada,
T horizontal component, −0.4
PGA = 0.42 g), and the
corresponding response −0.6
spectra 0 2 4 6 8 10 12 14 16 18 20
Time (s)
20
18
16
14
12
SD (cm)
10
8
6
4
2
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Period (s)
180
160
140
120
SV (cm/s)
100
80
60
40
20
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Period (s)
2.0
1.5
SÁ (g)
1.0 ξ=5%
ξ = 10 %
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Period (s)
Earthquake Response Spectra and Design Spectra 811
Sa
(g
Spectra, Fig. 4 Tripartite
’ s)
response spectra plot of the
0
0.
1940 El Centro (comp. NS)
10
record (After )
(cm
Pseudovelocity Sv (cm/sec)
Anagnostopoulos (1997)) Sd
100 10
.0
10
0.
0
10
.0
1 .0
E
10 1.
0
0.1
0. ξ=0.00
1
1 ξ=0.05
0.0
ξ=0.10
1
0.01 0.1 1 10
Period (sec)
maximum relative velocity maxju_ðtÞj for long- motions. A detailed discussion on these matters
period structures. This can be seen on the basis of can be found in several sources, e.g., Chopra
physical grounds, since for very flexible struc- (1995). As explained in the following section,
tures (large T), the elastic forces are small, the these common trends permit the approximation
mass will exhibit very small displacements, and, of average response spectra and the subsequent
thus, the relative displacements between the mass development of design spectra with high proba-
and the ground will be almost identical to the bility of accounting for the general characteristics
ground
displacements,
i.e., for T ! 1,
SD ! of future seismic events.
maxug ðtÞ, and maxju_ðtÞj ! maxu_g ðtÞ , but,
because of the definition of SV in Eq. 7,
SV ¼ oSD ¼ 2p T SD ! 0 , since SD reaches a Elastic Design Spectra
finite value.
From the definitions of the three response The development of earthquake design spectra is
spectra provided by Eqs. 6, 7, and 8 follows the the most important application of the response
exact mathematical relationship spectra concept. As defined by Anagnostopoulos
(1997), a design spectrum is a specification of
SD ¼ oSV ¼ o2 SA (12) seismic actions for design purposes, aimed at
accounting for the general characteristics of
which implies that each of the three response expected ground motions at the site and at mini-
spectra provides the same information as the mizing the dependence of earthquake resistant
others about a specific ground motion. Equa- design on the peculiarities of specific ground
tion 12 also permits plotting, e.g., motions (that will never be repeated in the
Anagnostopoulos (1997), all three response spec- future). At a first glance, an obvious characteristic
tra by a single curve on a three-way logarithmic of the spectral plots produced from various
graph, as shown in Fig. 4 for the El Centro record ground motions is their jagged shape with suc-
and three damping values x = 0, 0.05, and 0.10. cessive sharp peaks and valleys within small
Such plots become useful because they reveal intervals of periods. In addition, each ground
some general trends, common to all ground motion yields a unique spectral plot with
812 Earthquake Response Spectra and Design Spectra
Earthquake Response
Spectra and Design
Spectra, Fig. 5 Newmark
and Hall (1982) elastic
design spectrum
(ag = 1.0 g)
substantial differences to the respective spectral increased, a number of similar design spectra
plots of other ground motions, e.g., valleys at one became available representing a wider range of
of them may appear as peaks at another, while conditions. Modern design spectral shapes are
their magnitudes and periods of occurrence are constructed as smoothed averages of response
usually completely unrelated. Therefore, a struc- spectra of many strong ground motions, grouped
tural design based on any single ground motion according to few broad categories of soil condi-
or, equivalently, its response spectrum would tions and normalized with respect to their proba-
assume a substantial risk. bilistically estimated PGA, e.g., Newmark
However, spectral plots corresponding to et al. (1973), Seed et al. (1976), Mohraz (1976),
ground motions recorded under similar condi- and Newmark and Hall (1982).
tions, e.g., similar soil profiles, comparable earth- The set of rules for the construction of design
quake magnitudes, nearly equal epicentral spectra proposed by Newmark and Hall (1982),
distances, etc., exhibit, in spite of their obvious the NH spectra, have played a central role in the
differences, some characteristic common trends. development of modern seismic design codes. It
In addition, it is well known that spectral plots, in is based on the specification of a ground motion
general, normalized by their respective peak envelope, at any given site, defined in terms of
ground acceleration (PGA) tend to similar values three parameters, i.e., peak ground acceleration
for very short or very long periods. These obser- ag, peak ground velocity vg, and peak ground
vations allow the development of design spectra displacement ug. On a three-way logarithmic
representative of broad classes of soil, epicentral plot, the ground motion envelope becomes a trap-
distances, etc. Thus, design ground motions are ezoid, as shown by the dashed line in Fig. 5.
specified in the form of smooth design spectra The values of all three ground parameters can
disregarding the usual peaks and valleys pro- be computed by a seismic hazard analysis, but
duced by individual seismic records. The first usually only ag is computed this way, while vg
design spectrum was produced by Housner and ug are estimated on the basis of ag. For exam-
(1959) on the basis of the four strongest ground ple, Table 1, proposed by Newmark et al. (1973),
motions recorded in the USA up to that time. is frequently used to estimate the peak values of
Later, as the number of recorded ground motions ground velocity and displacement.
Earthquake Response Spectra and Design Spectra 813
Earthquake Response Spectra and Design Spectra, Table 2 Newmark and Hall (1982) spectrum amplification
factors
One sigma (84.1 %) Median (50 %)
Critical damping ratio (%) A V D A V D
0.5 5.10 3.84 3.04 3.68 2.59 2.01
1 4.38 3.38 2.73 3.21 2.31 1.82
2 3.66 2.92 2.42 2.74 2.03 1.63
3 3.24 2.64 2.24 2.46 1.86 1.52
5 2.71 2.30 2.01 2.12 1.65 1.39
7 2.36 2.08 1.85 1.89 1.51 1.29
10 1.99 1.84 1.69 1.64 1.37 1.20
20 1.26 1.37 1.38 1.17 1.08 1.01
814 Earthquake Response Spectra and Design Spectra
Earthquake Response
Spectra and Design
Spectra,
Fig. 6 Normalized
spectral shapes for a
number of soil conditions
(After UBS (1994))
Earthquake Response Spectra and Design Spectra, Table 3 EC-8: parameters involved in elastic response spectra
Type 1 Type 2
Soil type S TB(s) TC(s) TD(s) S TB(s) TC(s) TD(s)
A 1.0 0.15 0.4 2.0 1.0 0.15 0.4 2.0
B 1.2 0.15 0.5 2.0 1.2 0.15 0.5 2.0
C 1.15 0.20 0.6 2.0 1.15 0.20 0.6 2.0
D 1.35 0.20 0.8 2.0 1.35 0.20 0.8 2.0
E 1.4 0.15 0.5 2.0 1.4 0.15 0.5 2.0
corner points. The recommended values of the scaling factor (the design ground acceleration,
parameters S, TB, TC, and TD, for five soil types obtained by site seismic hazard analysis). As a
classified as A, B, C, D, and E are given in Table 3 result, design spectra defined through these pro-
for the Type 1 and Type 2 spectra. Type 2 spec- cedures fail to provide a uniform non-exceedance
trum is recommended for sites where the earth- probability for all spectral regions. As an exam-
quakes that contribute most to the seismic hazard ple, McGuire (1978) has pointed out that such
defined for the site for the purpose of probabilistic design spectra undervalue spectral velocities,
hazard assessment have a surface-wave magni- except for high frequencies, in regions affected
tude, Ms, not greater than 5.5. The five soil types by distant earthquakes. A partial cure to this
vary from rock soil profile A with a shear wave problem has been proposed in ATC-3 where a
velocity vs > 800 m/s to soft soil profile E with second variable related to the peak ground veloc-
vs < 180 m/s. A plot of the EC-8 elastic response ity is introduced, in addition to the peak ground
spectra Se, for = 1, as shown in Fig. 7. acceleration. This procedure permits the usage of
The design spectra discussed above do not a variable spectral shape but it requires the avail-
take into account the influence of key seismolog- ability of two seismic hazard maps, one for the
ical factors, such as the epicentral distance or the peak ground acceleration and another for the
magnitude of the earthquake, on the frequency peak ground velocity. However, uniform hazard
content of the motion. This is due to the fact that spectra, providing uniform probability of
their shape is fixed, while the actual ordinates of non-exceedance, are possible nowadays via
the spectra are obtained by a multiplication with a regression analyses on response spectra of large
Earthquake Response Spectra and Design Spectra 815
a 4,0 b 5,0
A: Rock Vs>800m/s
3,5 4,5
4,0 B: Stiff soil Vs=360-800m/s
Se/ag
Se/ag
3,0
3,5 C: Medium stiff soil
2,5 Vs=180-360m/s
3,0
D: Loose to medium stiff
2,0 2,5 soil Vs<180m/s
E: Soft soil Vs<180m/s
1,5 2,0
1,5
1,0
Sd/ag
Sd/ag
1,0 E
0,5
0,5
0,0 0,0
0 0,5 1 1,5 2 2,5 3 3,5 4 0 0,5 1 1,5 2 2,5 3 3,5 4
Period T(sec) Period T(sec)
Earthquake Response Spectra and Design Spectra, Fig. 7 EC-8 elastic response spectra and design spectra: (a)
Type 1 and (b) Type 2 ( = 1, q = 4)
numbers of strong ground motion records avail- maximum force Fy, as shown in Fig. 8. Provided
able on almost every region of the world, e.g., the yield quantities Fy and uy are known proper-
McGuire (1977) and Joyner and Boore (1982). ties of the SDOF under investigation, the search
for the quantity of interest umax, the maximum
displacement, is facilitated by the introduction of
Inelastic Response and Design Spectra the nondimensional ductility ratio m defined by
umax ¼ muy . Thus, the ductility ratio m, i.e., the
In a manner similar to the previously described inelastic deformation demand imposed by a
elastic response spectrum, one can define the ground motion u€g(t), becomes the primary quan-
inelastic response spectrum as a plot of the max- tity of interest even for design purposes,
imum response of all nonlinear single-degree-of- replacing the maximum force or stress used in
freedom (SDOF) systems to a given dynamic elastic design.
excitation. Of course, in this case the type of the The definitions of the inelastic displacement
nonlinearity of the oscillators of interest should (SD = umax), (pseudo)velocity (SV) and (pseudo)
be specified a priori. Inelastic response spectra, acceleration (SA) response spectra, and their
based on some nonlinear force–displacement inter-relationships, follow those of the elastic
relationship of an SDOF oscillator, are computed case. The SA is obtained from the formula pro-
similarly to the elastic response spectra. Thus, in viding the maximum elastic force, i.e., Fy = mSA,
Eq. 1 the term ku(t), the elastic resistance force, is and, as in the elastic case, is a close approxima-
substituted by the displacement-depended force tion of the maximum absolute acceleration spec-
F[u(t)] and the equation of motion (2) is now trum. In view of the relationships
written as
Fy ¼ kuy ¼ mo2 uy ; (15)
F½uðtÞ
u€ðtÞ þ 2xou_ðtÞ þ ¼ €
ug ð t Þ (14)
m the above definition of SA can be written in alter-
native forms as
The simplest and most common case refers to an
oscillator whose nonlinear resistance is assumed
to be elastoplastic with an initial stiffness k and umax SD
SA ¼ o2 uy ¼ o2 ¼ o2 (16)
yield displacement uy corresponding to a m m
816 Earthquake Response Spectra and Design Spectra
Earthquake Response
Spectra and Design
Spectra, Fig. 8 Idealized
elastoplastic system
Earthquake Response
Spectra and Design
Spectra, Fig. 9 Inelastic
response spectra (SA) of
the Lefkada (Greece) 2003
earthquake (14/08/2003
5:18:48 (GMT), OTE
building, Lefkada,
T horizontal component,
PGA = 0.42 g)
The (pseudo)velocity spectrum can be defined in counterpart Eq. 12, except for the entrance of the
exactly the same manner as for the elastic case, ductility ratio m between SD and the other two
i.e., SV ¼ SA=o , in which case Eq. 16 can be response spectra SA and SV.
expanded to include all three response spectra as For specified values of viscous damping x and
ductility ratio m, the computation of inelastic
SD
SA ¼ oSV ¼ o2 (17) response spectra, for any value of o or T, is a
m
trial-and-error process which aims at calculating
Apparently, Eq. 17 expresses the same relation the yield force Fy which corresponds to the spec-
between the three response spectra as its elastic ified ductility ratio m. In Fig. 9 the inelastic
Earthquake Response Spectra and Design Spectra 817
response spectra (SA) of the Lefkada 2003 comparison the corresponding elastic spectra
strong-motion record is shown for several values of Eq. 13) as
of the ductility factor m. Of course, the case
m = 1 coincides, as it was expected, with the 2 T 2:5 2
0 T TB Sd ¼ ag S þ
linear response spectra shown in Fig. 3. 3 TB q 3
Regarding the development of inelastic S
design spectra, Newmark and Hall (1982) pro- TB T TC Sd ¼ 2:5ag
q
posed a set of simple rules to obtain them from
S TC
the corresponding elastic design spectra. They TC T TD Sd ¼ 2:5ag
are based on observations related to the tenden- q T
E
cies of the inelastic response spectra, as com- S TCTD
TD T Sd ¼ 2:5ag
pared to the corresponding elastic ones, when q T2
they are plotted in a tripartite logarithmic graph. For all periods Sd 0:2 ag
Thus, for the various characteristic regions of
(18)
the elastic design spectra, proposed by
Newmark and Hall (1982), the inelastic design The symbols used in Eq. 18 are the same as those
spectra are obtained from the corresponding used in the corresponding elastic spectra of
elastic ones by a simple multiplication with a Eq. 13, while the values of S, TB, TC, and TD
factor which is a function of the ductility ratio m. remain the same to those listed in Table 3. The
For more details the reader is directed to the behavior factor q, depending on the type of the
original publication by Newmark and Hall structure and the building material used, varies
(1982) and a detailed discussion by Chopra from 1.5 for unreinforced masonry structures to
(1995). Following the Newmark and Hall pro- 6.0 for the most ductile steel structures. The
posals, inelastic design spectra have been devel- viscous damping factor of Eq. 13 is removed
oped into the most popular way of estimating since the behavior factor q accounts for the influ-
the inelastic behavior of structures via standard ence of the viscous damping as well. A plot of
elastic modal analyses. Modern codes have, in the EC-8 design spectra Sd, for q = 4, is shown in
general, adopted this approach by simplifying Fig. 7.
the Newmark and Hall rules for inelastic design Design spectra based on specific values of the
spectra and, at the same time, introducing the ductility factor m are introduced in the New
so-called capacity design provisions for the pur- Zealand seismic code (1992) with certain restric-
pose of allowing control over the yield patterns tions that rule out the possibility the ductility
of a structure and avoiding excessive local duc- demand to exceed the provided capacity.
tility demands, e.g., Anagnostopoulos et al.
(1978). Thus, the response modification factor
or behavior factor is used to account for the Summary
inelastic building behavior. Almost all of the
contemporary design codes, e.g., ATC-3 The response spectrum concept is defined, and its
(1978), NEHRP (1994), UBC (1994), and key role in the characterization of seismic
EC-8 (2004), employ a constant response mod- motions, for engineering purposes, is described.
ification factor, dependent upon the building The development and the importance of seismic
material and type of structure, by which the design spectra, both elastic and inelastic, are ana-
elastic design spectrum is divided in order to lyzed and on this basis the main features of sev-
obtain the corresponding inelastic one. In par- eral contemporary design codes are outlined.
ticular, the EC-8 inelastic acceleration design A number of numerical examples, related to
spectra Sd, used in conjunction with standard response and design spectra, are shown and
modal analysis, are specified (see for discussed.
818 Earthquake Return Period and Its Incorporation into Seismic Actions
The most common approach used worldwide is the checks in order to be considered safe and reliable
design of structures with a large ductility capacity, against collapse. A typical limit state-based
thus, to be able to withstand inelastic deformation design can be viewed as a one-limit state, ulti-
during strong ground shaking with a small collapse mate strength, or a two-limit state approach, ser-
probability. However, the anticipated damage to viceability and ultimate strength. Most of the
structural and nonstructural components and con- prescriptive design codes are based on the
tents that is caused by large deformations can have assumption that structures are capable of dissi-
serious impact on post-earthquake functionality pating energy through inelastic deformations. It
and often require, if not demolition, certainly is recommended to design ductile structures pro-
costly and disruptive repairs. On the other hand, vided that ductility remains within the prescribed
E
the design of stiffer and stronger, non-ductile limits. Seismic loads can significantly be reduced
structures could minimize deformations and there- (via the so-called behavior factor) if the structure
fore damages. However, issues such as increased is designed to be dissipative.
material quantities, structural accelerations, and Furthermore, serviceability is usually checked
damage to acceleration-sensitive nonstructural to ensure that the structure will not deflect or
components and building contents have to be care- vibrate excessively during its service life. Struc-
fully assessed in such design approaches. An alter- tural assessment is typically performed consider-
native approach toward the limitation of structural ing ultimate strength limit state under the design
damages under strong ground shaking is the mod- earthquake (typically with return period of
ification of structural characteristics, in a way that 475 years and probability of exceedance 10 %
the capacity is increased and/or demand is in 50 years) and damage limitation under an
reduced. Modern technologies that can be occasional earthquake for the serviceability
implemented for these purposes are damping limit state (typically with return period of
devices, base isolation schemes, or a combination 72 years and probability of exceedance 50 % in
of both. 50 years). These two hazard levels are the most
Earthquake frequency occurrence or return generally used worldwide. In certain norms (e.g.,
period is directly incorporated into seismic US guidelines (FEMA-750 2009), it is also spec-
actions and contemporary seismic norms, since ified that structures should be capable to with-
in general, seismic-resistant structures must have stand (i.e., to avoid global collapse) the so-called
(i) the capacity to resist frequent (with return maximum considered earthquake (typically, hav-
period 50–100 years), minor earthquakes without ing return period equal to 2,475 years and prob-
damage, (ii) the capacity to resist with limited ability of exceedance 2 % in 50 years).
structural and nonstructural damage infrequent The most common approach for the definition
(with return period approximately 500 years) of the seismic input is the use of a design code
moderate earthquakes, and (iii) the capacity to response spectrum which is created for a certain
resist without collapse and life safety endanger- return period of earthquake hazard. This is
ment very strong and rare earthquakes (with a general approach, which is easy to implement
return period approximately 2,500 years). For for conventional buildings. Moreover, if higher
more critical infrastructures, even higher seismic precision is required, the use of spectra derived
demand levels are imposed which is interpreted from suitable earthquake records is more appro-
in terms of a very large return period, e.g., priate. Note that significant dispersion on the
10,000 years for maximum design earthquake structural response due to the use of different
(MDE) motions for major dams in New Zealand. records even scaled to the same earthquake inten-
The majority of seismic norms belong to the sity can be observed, since ground motions with
category of prescriptive design codes (or limit equal peak ground acceleration (PGA) but differ-
state design procedures), as their aim is to ensure ent frequency content can have very different
that the structural design satisfies a number of demands upon a structure.
820 Earthquake Return Period and Its Incorporation into Seismic Actions
Response spectrum method (RSM) has the probability of exceedance in a given time
become the basic computational tool in period (typically for a 50-year period).
earthquake-resistant design of structures, mainly
because of its simplicity as it basically uses two
parameters: spectral period and damping ratio. Seismic Hazard Analysis
The principal weakness of the RSM is that it
does not include the duration and other aspects Seismicity at a site is typically defined as
related to frequency content of a strong motion, a quantitative estimation of ground shaking,
thus, especially for higher seismic intensities, and which may be represented most commonly by
it is incapable to fully cover all aspects of peak ground acceleration (PGA) (as well as
dynamic nonlinear response. To overcome this velocity (PGV) and displacement (PGD)), or
deficiency, dynamic analysis methods (linear response spectrum ordinates (mostly accelera-
and preferably nonlinear) can be used utilizing tions). It has to be stressed that in this work, the
historic, simulated, or synthetic acceleration time emphasis on seismic hazard in terms of return
histories. Toward this goal, recent guidelines period is given on ground shaking. Nonetheless,
(e.g., FEMA-445 2006) suggest the use of more similar analysis related to kinematic distress and
accurate dynamic analysis procedures for struc- various soil failure types (fault rupture (Angell
tural design and assessment. Input motions et al. 2003), liquefaction (Kramer and Mayfield
should be representative of regional seismicity 2007), landslides (Saygili 2008), etc.) due to
and must be compatible with each earthquake earthquakes needs to be considered in certain
hazard level; thus, various methodologies have occasions, especially for lifelines that are fre-
been proposed to select and scale accelerograms quently exposed to such perils. The severity of
(e.g., Katsanos et al. 2010). these hazards (i.e., probability of exceedance)
The modern conceptual approach for struc- used in performance evaluation should be com-
tural design and assessment is the so-called patible with that used in the specification of
performance-based design (PBD) approach, ground shaking hazard levels (FEMA-350 2000).
which is based on the principle that a structure In any case, the main aim in seismic hazard
should meet certain performance objectives for analysis is the estimation of the most crucial
multiple seismic hazard levels. Regarding seis- parameters related to earthquake hazard that is
mic hazard, it ranges from frequent (i.e., having expected to occur at a given site during
a small return period) earthquakes of low inten- a predefined time interval, the so-called return
sity to more catastrophic and rare events that or recurrence period. Seismic hazard analysis
have large period of occurrence. This framework requires the knowledge of certain important data
has been introduced in order to increase struc- for the wider neighborhood of the examined site,
tural safety against natural hazards. According such as geologic evidence of earthquake sources,
to PBD, the structures should be able to resist fault activity, historical seismicity data, etc. Past
earthquakes in a quantifiable manner and to pre- earthquake data provide valuable information for
sent levels of possible damage in a prescribed predicting the magnitude and occurrence fre-
preferable manner (e.g., avoiding soft-storey quency of future earthquakes and, thus, are most
collapse). In any case, the adoption of a proper valuable in seismic hazard analysis. Recently
stochastic model (usually the Poisson distribu- a substantial effort has been devoted to the devel-
tion) for the occurrence of earthquake events has opment of a global earthquake model (GEM) and
to be used within an elaborate seismic hazard open-source software for seismic hazard and risk
analysis. In addition, the construction of assessment (for more details, visit http://www.
a seismic hazard curve is needed to associate globalquakemodel.org/openquake/about/engine/).
a particular ground motion parameter (e.g., At this point, it is necessary to clarify the
peak ground acceleration, spectral acceleration, difference between the two commonly used
etc.) with an annual frequency of exceedance or terms “hazard” and “risk.” These terms are
Earthquake Return Period and Its Incorporation into Seismic Actions 821
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 1 Basic steps in
DSHA (Modified from
FEMA-451B (2003))
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 2 Example of DSHA
(Adopted from FEMA-
451B (2003))
The first step is to identify all possible sources a source to the examined site. As the known
of ground motion, while some may be rather easy earthquakes will have occurred at a distance that
to identify (e.g., a known active fault), others may is not likely to be the same as the distance to the
be more difficult to describe. Subsequently, the site, certain correction needs to be made to take
controlling earthquake needs to be defined, and into account geometric spreading of seismic
this involves engineering judgment. For instance, waves and seismic energy absorption. This is
if one wants to design for the largest earthquake done utilizing attenuation models that have been
that could ever occur at the site (using perhaps an established for each region, as it is very important
estimate of seismic moment) or the largest to use attenuation relationships that characterize
motion that has occurred within a time period local geotectonics.
(e.g., the past 500 years). Note that as aforemen- Figure 2 presents a very simple example of
tioned, there is no inclusion of aspects related to DSHA, in which three sources were considered,
probability of occurrence related to regional while in each case the closest distance to the site
seismicity. was used. Note the different magnitudes for the
Selection of source-site distance parameter is different sources, which are consistent with the
also important for the DSHA process. In DSHA, selected “maximum earthquake”. The peak
it is common to use the closest distance from ground acceleration (PGA) at the site was
Earthquake Return Period and Its Incorporation into Seismic Actions 823
obtained from an appropriate attenuation rela- considers all magnitudes, all distances, and all
tionship. The motion with the greatest resulting effects. Furthermore, PSHA considers uncer-
PGA is chosen as the controlling earthquake. tainty in location, size, rate of earthquake occur-
Note that this is not sufficient to establish seismic rence, predictive relationship, etc., and combines
risk, since the effect of the motion on the exam- them to compute probabilities of different levels
ined structure is not addressed. It is worth men- of ground shaking. In particular, PSHA integrates
tioning that under certain circumstances, the site’s total seismicity for a given time period and
more distant earthquake, though having a lower provides strong-motion parameters within
effective PGA, could produce waves at predefined confidence levels, in the form of haz-
a frequency that could be more damaging for ard curves, uniform response spectra, etc., for
E
a specific structure. various return periods. In general, as shown in
Identification and characterization of all Fig. 3, PSHA consists of four primary steps
sources can be performed in DSHA utilizing (Kramer 1996):
slip rate approach, in which if average displace-
ment over the slip surface, D, relieves stress/ 1. Identification and characterization of all
strain builtup by movement of the lithospheric source zones (areas or faults) based on histor-
plates over a certain time period, t. Therefore, ical seismicity, tectonics, and geology.
D = S t, where S is the slip rate at the examined 2. Characterization of seismicity of each source,
fault. Subsequently, the seismic moment is i.e., to determine an upper magnitude limit and
given by develop a magnitude-recurrence relationship
for each source zone.
Mo ¼ m A S t ¼ m A D (1) 3. Determination of a predictive relationship for
seismic parameter motions (such as PGA)
where m is the rigidity or shear modulus of the from each source.
bedrock, usually taken equal to 3 1011 dyne/ 4. Perform probabilistic calculations to compute
cm2, and B is the rupture area on the fault plane the probability distribution of ground motion
where the slip occurs during the earthquake. at the site and select the design value at the
In addition, the “moment rate” can be defined as appropriate probability level.
Mo/t = m A S. The seismic moment is
transformed to earthquake magnitude or energy Ground motion parameters that can be derived
according to simple logarithmic expressions. from a PSHA include: PGA and spectral acceler-
The resulting hazard statement of DSHA is ation (Sa) hazard curves, uniform response spec-
basically a “scenario” earthquake for seismic tra (URS), de-aggregation of hazards by
design (i.e., design earthquake). The rather con- magnitude and distance, and parameters for
servative way that DSHA is commonly used pro- assisting in time-history selection for dynamic
duces a worst-case scenario. DSHA provides no analysis. A seismic hazard curve is constructed
indication of how likely design earthquake is to by plotting the rate of exceedance of the seismic
occur during lifetime of a structure. Design earth- parameter for different seismic intensity levels.
quakes may occur every 200 years in some places In order to assess that a certain engineering
and every 10,000 years in others. Furthermore, demand parameter (e.g., drift) will be exceeded
DSHA require subjective opinions of the seis- in a building located at a certain site, it is essential
mologist on input parameters, while variability to evaluate the accuracy in terms of the mean
of effects is not rationally accounted for. annual frequencies (MAF) of exceedance given
in a form of hazard curve, as in Fig. 4 in terms of
Probabilistic Seismic Hazard Analysis PGA values (or alternatively using spectral accel-
Probabilistic seismic hazard analysis (PSHA) eration values). Similar curves can be constructed
was introduced by Cornell (1968). In contrast to for a kinematic type of seismic hazard. For exam-
DSHA, it assumes many scenarios, in which it ple, utilizing the so-called probabilistic fault
824 Earthquake Return Period and Its Incorporation into Seismic Actions
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 3 Basic steps in
PSHA (Adopted from
FEMA-451B (2003))
1.0000000
Annual Frequency of Exceedance
0.1000000
0.0100000 50/50
0.0010000 10/50
2/50
0.0001000
0.0000100
0.0000010
0.0000001
0.001 0.010 0.100 1.000 10.000
Acceleration. g
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 4 Typical hazard curve of PGA
values in terms of annual frequency rate
displacement hazard analysis (PFDHA), one can a function of return periods, as shown in Fig. 5.
derive fault displacement hazard curves that rep- Tools for producing hazard curves as those
resent annual frequencies (and corresponding shown in Figs. 4 and 5 as well as response spectra,
return periods) of certain fault slippage level etc., can be found at US Geological Society
(Angell et al. 2003). (USGS) site: http://earthquake.usgs.gov/haz-
Frequently, fractile curves (e.g., 15th, 50th, ards/.
85th percentiles) of hazard curves are also plotted
together with the mean curve to illustrate data Earthquake Occurrence Frequency
scattering. Note that differences between mean According to PSHA, a given source can produce
and median curves can be observed. Equiva- different earthquakes: low-magnitude earth-
lently, seismic parameters can be interpreted as quakes occur often and cause fewer damages,
Earthquake Return Period and Its Incorporation into Seismic Actions 825
1.000
0.100
E
0.010
0.001
1 10 100 1000 10000 100000 1000000 10000000
Return Period of Ground Motion, years
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 5 Typical hazard curve of PGA in
terms of return periods
while large magnitude and more severe events Typically, the seismicity is interpreted using
occur rarely. The latter have longer return periods a specified relationship between magnitude and
and lower probability of exceedance, yet they are likelihood of occurrence and is called “recur-
more catastrophic; thus, they are referred as low rence relationship,” which indicates the average
probability – high consequence events. In rate at which an earthquake of a particular
a probabilistic analysis, the recurrence relation- magnitude, M, will be exceeded as shown in
ship (e.g., frequency of earthquakes above Fig. 6. Such relationships are determined from
a certain magnitude) needs to be included as it regression analysis of historical earthquake data.
is the major uncertainty of the process. All major The most popular is Gutenberg-Richter model
uncertainties are included in PSHA, and the result which is a rather simple mathematical formula-
is a seismic hazard curve that relates the design tion that is reliable for intermediate scale events
motion parameter to a probability of exceedance. but may be insufficient for small or for large
Hence, for the design of a bridge for a ground earthquakes; thus, more advanced models have
motion with a 2 % probability of being exceeded been proposed (Kramer 1996).
in a 50-year period (which is transformed to According to the Gutenberg-Richter recur-
annual probability), the ground motion parameter rence, law earthquake magnitudes are exponen-
(e.g., PGA) is obtained from the seismic hazard tially distributed and the mean rate of occurrence
curve (such as the one in Fig. 4). (i.e., the number of earthquake events/year equal
826 Earthquake Return Period and Its Incorporation into Seismic Actions
to or greater than magnitude M), lM, is defined in 3. The probability of more than one occurrence
log-linear or exponential form (see plots in Fig. 6) in a very short time period is negligible.
as follows:
According to the popular Poisson model, the
ðabMÞ ðabMÞ temporal distribution of earthquake recurrence
loglM ¼ a b M ) lM ¼ 10 ¼e
for a certain intensity level (magnitude, PGA,
(2)
etc.) is given by
and following the above formulation, the event
mn em ðl tÞn elt
return period is directly given by P½N ¼ n ¼ or P½N ¼ n ¼
n! n!
(4)
1 10bM
TR ¼ ¼ (3)
lM 10a where n is the number of occurrences and m = l•t
is the average number of occurrences in the time
where a and b (while a = 2.303a and interval of interest, t, and l is the annual rate,
b = 2.303b) are regional parameters determined which is directly related to the return period of
from tectonic and geological data, rate of seis- such event as shown in Eq. 3. Hence, the proba-
micity, etc.; typically, b is close to unity (b = 1.0 bility that an earthquake having a specific ground
0.5) and represents the relative likelihood of motion parameter, such as magnitude, higher
large and small events. than a certain level, will occur at least once in
t years is given by
Temporal Earthquake Uncertainty
The temporal uncertainty of earthquakes is
P ½ N 1 ¼ P ½ N ¼ 1 þ þ P ½ N ¼ 1
included in PSHA by considering the distribution
of their occurrence in time, which can be modeled ¼ 1 P ½N ¼ 0 ¼ 1 elt
by random models, such as Bernoulli, Poisson,
¼ 1 et=T R (5)
and Markov processes. In this manner, it is pos-
sible to calculate the probability of the number of
occurrences over a certain time interval. The The above expression can be easily combined
occurrence of an earthquake for a certain time with a recurrence law (as the one in Eq. 2) to
interval is usually modeled as a Poisson process predict the probability of at least one exceedance
that describes the number of occurrences of an of a specific magnitude M, lM, in a period of
event (not only earthquakes but floods and other t years as follows:
natural disasters as well) during a specified time
interval and provides also the return period of P ½N 1 ¼ 1 elM t (6)
such event. The main assumptions of Poisson
model are as follows (Kramer 1996): An example of using Eq. 5 is to calculate the
probability that an event that occurs on average
1. The number of occurrences over a given time every 1,000 year will occur at least once in a
interval are independent of the number of t = 100-year period. Thus, l = 1/1,000 = 0.001
earthquakes that occur in any other time inter- ) P = 1 – exp[(0.001)(100)] = 0.0952 =
val (i.e., the system has no memory from past 9.52 %, while the probability that one such
seismic history, while there are other time- event will occur at least once in a 1,000 year
dependent models that keep track of occurred period is P = 1 – exp[(0.001)(1,000)] =
events). 0.632 = 63.2 %. Obviously, this probability
2. The probability of occurrence in a very short value (63.2 %) is the same for all hazard level
time interval is proportional to the length of return periods when the examined time interval is
the interval. equal to the return period: t = TR.
Earthquake Return Period and Its Incorporation into Seismic Actions 827
Apart from the probability of exceedance, the not valid for higher probabilities. For instance,
level of ground shaking that has a certain proba- the actual return period of 50 % probability of
bility of exceedance in a given period can be exceedance in 50 years is 72 years, not 100 years
determined via Eq. 5 and a hazard curve as it as suggested by the approximate formula. Note
will be shown in the next section. Furthermore, also that since l depends on P and t values, there
solving Eq. 5 for annual rate (events per year), l, are many pairs of P and t that result to the same
gives return period.
Figure 7 depicts the relationship between
lnð1 PÞ return period, time interval of interest, and the
P½N 1 ¼ 1 elt ) l ¼ (7)
t related probability of exceedance. For instance,
a 2 % probability for an event of being exceeded
and the corresponding return period, TR, is given in 50 years has a return period of 2,475 years,
by while a 10 % probability of exceedance in
50 years has approximately a 475-year return
1 t t
TR ¼ ¼ (8) period. Such curves are very useful for assessing
l lnð1 PÞ P multiple hazard levels in performance-based
design approaches. For structural design,
For example, the annual rate of exceedance a 50-year base line is generally used, since this
for an event with a 10 % probability of exceed- is estimated as the service life of a typical build-
ance in 50 year, or in short 10/50 hazard level, is ing. For critical infrastructures, such as dams,
given by l = (ln(1–0.1)/50) = 0.0021/year, a longer return period is more appropriate as the
while the corresponding return period is required functioning life is much longer and fail-
TR = 1/l = 475 years. Similarly, for 2 % in ure consequences more severe.
50 year, l = 0.000404/year and TR = 2,475 years. Probability of exceedance must be associated
As indicated in the last part of Eq. 8, for low to a time interval of interest and to be connected
probabilities (or long return periods), the return to a certain return period, which is simply
period is approximately equal to t/P. For exam- described by its value. For example, an earth-
ple, for a period of 100 years and return period of quake with a 2 % probability of being exceeded
2,500 years, the 100-year probability is roughly in 100 years has a return period of TR = 100/ln
equal to P = t/TR = 100/2,500 = 4 % approxi- (1–0.02) = 4,950 years. Note that since the pre-
mately equal to 3.96 % which is the exact out- cision of knowledge of this natural phenomenon
come of Eq. 8. Nonetheless, this approximation is is not perfect; thus, the return periods can be
828 Earthquake Return Period and Its Incorporation into Seismic Actions
Earthquake Return a b
Period and Its
log λ max
log λ max
Incorporation into
log TR
Seismic Actions,
Fig. 9 Examples of using
log TR
seismic hazard curves for
calculating: (a) exceedance
probabilities, and (b) 0.0021 475 yrs
ground motion parameters
0.001
E
0.35g amax 0.24g amax
Earthquake Return Sa
Period and Its
0.6
Incorporation into 0.4
Seismic Actions, 0.2
Fig. 10 Example of 0
typical response spectrum −0.2
pattern in seismic norms −0.4
(continuous line) and −0.6
0 5 10
a response spectrum
derived from a pulse-like
accelerogram (dashed line)
and values are extracted for a certain confidence framework. Similar plots can be produced for
level (e.g., a 10 % in 50 year probability of spectral displacements or velocities.
exceedance), then a response spectrum may be Note that the information derived from PSHA
created by plotting the parameter magnitudes in may be used to assist in the development (selection
terms of spectral period, T. The plot at the lower and scaling) of suitable ground motion time histo-
left in Fig. 11a shows the first point on such ries (Katsanos et al. 2010). For this purpose, the
a spectrum, in this case: the PGA with a 10 % ground motion amplitudes is scaled in a way that
probability of being exceeded. Similarly, in the spectrum of the scaled ground motion closely
Fig. 11b, the point representing the 0.2 s (5 % matches a target UHS. Nonetheless, this can lead
damped) spectral ordinate with the same (10 %) to inaccurate results, since the UHS is a composite
probability of being exceeded in 50 years is plot- of hundreds if not thousands of earthquakes, any
ted. In this manner, the whole response spectrum two of which are very unlikely to occur simulta-
is constructed as shown in Fig. 10. neously. Typically, the short period part of the
If the process is repeated for every return period UHS is representative of moderate nearby earth-
of interest, then design spectra for different hazard quakes, while the long-period part reflects the haz-
levels can be constructed, as shown in Fig. 12. If ard from larger, more distant events. Figure 13
the aforementioned procedure is continued for illustrates more clearly this point. Although it is
many earthquake records, the result is a complete possible to generate a realistic earthquake ground
UHS that can be used in a simplified form in motion that matches small or large earthquakes,
seismic norms (see Fig. 10). Such multiple spectra a generated earthquake that matches the whole
can be used in response spectrum analysis of UHS would be unrealistic and most probably too
a structure within a performance-based design demanding for the structure.
830 Earthquake Return Period and Its Incorporation into Seismic Actions
0.2
PGA=0.33g
0.0 0.5 1.0 1.5
Period, T (sec)
b
Mean Annual Rate of Exceedance
0.8
spectrum 10−8
0.0 0.2 0.4 0.6 0.8
0.6
0.2 Sec Spectral Acceleration (g)
0.4
0.2
0.2 Sec accn = 0.55g
0.0 0.5 1.0 1.5
Period, T (sec)
Earthquake Return Period and Its Incorporation into to determine: (a) PGA and (b) Sa for 0.2 s (Adopted from
Seismic Actions, Fig. 11 Example of using seismic FEMA-451B (2003))
hazard curves for generating a design response spectrum
Proper hazard maps can be used to construct 5 %) for a variety of exceedance probabilities
uniform response spectra that can be used for (e.g., 2/50, 10/50, etc.) as illustrated in Fig. 14.
seismic design of structures. For instance, in the In many countries worldwide, there are hazard
United States (US), seismic provisions are based maps for different PGA or spectral ordinates and
on probabilistic maps developed by USGS after for various mean return periods (e.g., details for
performing elaborate probabilistic seismic haz- the USA are given at the theme issue of Earth-
ard analyses of the entire USA (Petersen quake Spectra (2000)). Such European hazard
et al. 2008). Using these maps, which provide maps have been recently developed (both for
spectral values for different spectral periods PGA and spectral accelerations for various return
(e.g., for 0.2 s in the Sa 2/50 map of Fig. 14), it periods: 101, 475, 975, 2,475 years) in the frame-
is possible to develop damped response spectra work of SHARE project (www.share-eu.org).
(for a specific structural damping ration, typically Figure 15 depicts the UHS for 10 % probability
Earthquake Return Period and Its Incorporation into Seismic Actions 831
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 12 Example of design
spectra for different hazard
levels
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 13 Example of a URS
for various types of
earthquakes (Adopted from
FEMA-451B (2003))
of exceedance in 50 years (right column plots) for in 50 years and 2 % in 50 years in Western US
Thessaloniki and Istanbul derived from their (WUS) is in general less than the difference
regional hazard curves (left column plots). between these two hazard levels in Central and
The most frequently used return periods are Eastern USA (CEUS).
475 years (corresponding to 10 % probability of In the US seismic design, provisions (IBC
exceedance in 50 years; abbreviated as 10 % in 2000; FEMA-368 2001; FEMA-450 2003; etc.)
50 years or 10/50) and 2,475 years provide, as a minimum, a margin of about 1.5
(corresponding to 2 % probability of exceedance times the design earthquake ground motion
in 50 years, or 2 % in 50 years, or simply 2/50). (denoted as the light blue curves in spectra in
The latest US hazard maps can be found at the Fig. 16b, c). More specifically, to prevent col-
USGS website http://earthquake.usgs.gov/haz- lapse for a relatively rare but catastrophic event
ards/. These maps are updated every few years. (associated with a 2 % in 50 year probability) to
The USGS hazard maps have quantitatively occur and taking into account the approximate
revealed important differences between the minimum seismic margin of 1.5 against collapse,
ground motion characteristics in different regions the IBC provisions generally define design
of the USA. In particular, they illustrate that the ground motion as 1/1.5 (=2/3) times the 2 % in
difference between the ground motions for 10 % 50-year ground motion. In other words, if
832 Earthquake Return Period and Its Incorporation into Seismic Actions
25°N 25°N
125°W 95°W
120°W 115°W 100°W
110°W 105°W
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 14 Example of uniform response
spectrum generation using probabilistic hazard maps (Adopted from FEMA-451B (2003))
a structure is subjected to a ground motion 1.5 buildings in CEUS designed for the 500-year
times the standard design level of 10/50, the event are vulnerable to collapse in a large, very
structure should have a low likelihood of col- rare earthquake (a 2,500-year period event).
lapse. Certainly, this margin depends on the The plot in Fig. 16b illustrates the UHS for
type of structure, detailing requirements, San Francisco, in which the 2/50 and 10/50 spec-
etc. Nevertheless, this 1.5 factor can be conser- tral curves are shown. Note that, in general, two
vative or not depending on the region, e.g., there third times the uniform hazard spectrum for 2 %
are important exceptions to this rule, especially in 50 years is generally comparable to the design
near active faults in coastal California. spectrum for 10 % in 50 years (corresponding
The plot in Fig. 16a presents several seismic light blue curve coincides with red curve). On
hazard curves for major cities in the USA. The the other hand, the plot in Fig. 16c depicts the
main difference to note is the slope of the curves UHS for Charleston, South Carolina in EUS,
for WUS compared to CEUS. In particular, for where a severe event (with a magnitude of 7.3)
WUS, there is a much smaller increase in seismic occurred in 1886. The differences are significant,
demand when shifting to the right side of the plot e.g., two third times the spectrum for 2 % in
(from a 10/50 to a 2/50 probability) compared to 50 years is much higher than that for 10 % in
CEUS. Obviously, frequent occasional and rare 50 years. Actually, the 10 % in 50-year spectrum
events are more probable to happen more often in provides roughly one fourth of the corresponding
seismic-prone regions compared to low seismic- values of 2 % in 50 year spectrum. Note also that
ity regions. Nonetheless, it can be easily noticed the shape of the spectrum is different, having its
that for two regions of low and high seismicity, peak at a shorter spectral period. Conclusively, by
very rare earthquakes (2/50 and less) may have comparing hazard curves and UHS in Fig. 16, it
a similar occurrence rate or even to have a higher can be easily concluded that the two third (2,500-
probability (but very small in absolute numbers) year. EQ) 500-year motions in WUS as
for the low seismicity regions. Due to the shape of suggested by US norms, while in contrast, two
the hazard curves, it can be seen that there is third (2,500-year. EQ) 1,600-year motions in
a substantial difference in CEUS between the EUS. Therefore, in EUS the application of IBC
500- and 2,500-year hazard levels. Thus, many design philosophy leads, to a certain extent, to the
Thessaloniki: Lon=22.78, Lat=40.8
1
5% Quantile 100 5% Quantile
15% Quantile 15% Quantile
Mean Mean
85% Quantile 85% Quantile
95% Quantile 95% Quantile
10−1
10% PoE in 50y
0.1
10−2
2% PoE in 50y
10−1
10% PoE in 50y
0.1
10−2
2% PoE in 50y
Spectral Acceleration / [g]
0.01
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 15 UHS for 10 % probability of exceedance in 50 years (right column) and PGA hazard curves
(left column) for Thessaloniki and Istanbul (Adopted from www.share-eu.org)
833
E
834 Earthquake Return Period and Its Incorporation into Seismic Actions
1.00
CITIES
Los Angeles
0.10 San Francisco
10% in 50 Years
5% in 50 Years
2% in 50 Years
Seattle
Salt Lake City
New York City
Charleston
Mempris
0.01
0.10000 0.01000 0.00100 0.00010 0.00001
Annual Frequency of Exceedance
b Effect of Scaling in Western United States c Effect of Scaling in Eastern United States
2.5
Spectral Response Acceleration (g)
2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Period (sec) Period (sec)
2% in 50 years 10% in 50 years 2/3 of 2% in 50 years 2% in 50 years 10% in 50 years 2/3 of 2% in 50 years
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 16 Comparison of hazard curves and
UHS in US (Adopted from FEMA-451B (2003))
possibility to overcome a more rare and cata- a 1/2,400 per annum exceedance probability,
strophic earthquake than the design action of while for Montreal the ground motions would
10/50. have a 1/1,600 annual exceedance probability.
Similar observations can be drawn in other Clearly, the same level of safety would not be
seismic-prone countries. For instance, Fig. 17 achieved. Even if different constants were used
illustrates the impact of a change in hazard level for east and west, the geographical variation
in Eastern and Western Canada. More specifi- across Canada would prevent achieving
cally, it depicts Sa(0.2 s) hazard curves for Van- a constant level of safety for medium to severe
couver and Montreal, showing how increasing hazard levels. Nonetheless, as in the USA, hazard
the 10 % in 50 years hazard by a factor of two curves indicate that the more severe (2 % in
(2) produces different changes in seismic 50 years) hazard level leads to a more uniform
demand for the two cities. In particular, for Van- seismic demand for the whole country (Ventura
couver, this would lead to ground motions with 2006).
Earthquake Return Period and Its Incorporation into Seismic Actions 835
2400 years
increasing return period c
an 2x
1600 years
(i.e., decreasing V
475 years
probability) in Canada 0.2
al
(Adopted from Ventura re
(2006)) o nt
0.1 M
E
0.05
20 50 100 200 500 1000 2000 500010000 20000
1/Probability (years)
of the weight factors is equal to unity, while the implementing EC8 should have a National
outcome of each branch is the product of the Annex containing all nationally determined
weight factors, as shown in the right column of parameters and other supplementary guidelines
Fig. 18. Hence, final outcome can be obtained as to be used for the design of buildings and civil
a weighted average of the values obtained at the engineering works. Additionally, as many Euro-
terminal branches of the logic tree. pean countries have very low seismicity, EC8
includes also the provision that in such cases
(defined in every National Annex) it is not oblig-
Incorporation of Earthquake Return atory to implement seismic design rules.
Period in Seismic Norms: The Case of In general, as prescribed in Eurocode 8 – Part
Eurocode 8 1 (EN 1998-1) – a new structure in seismic-prone
regions is checked against two limit states: the
In this section, a brief overview will be given ultimate no-collapse limit state (corresponding to
regarding the incorporation of earthquake return life safety requirement) and the damage limitation
period into seismic actions as described in the limit state (corresponding to damage limitation
provisions of European seismic design code requirement). Regarding the first limit state, in
Eurocode 8 or EC8 (EN 1998), which is com- order to prevent local failures, the design seismic
posed of six parts: action in terms of reference probability of exceed-
ance in 50 years is PNCR = 10 % (or for a reference
• EN 1998-1: General rules, seismic actions and return period of the seismic action equal to
rules for buildings TNCR = 475 years). These are the recommended
• EN 1998-2: Bridges values for PNCR (=10 %) and TNCR (=475 years);
• EN 1998-3: Assessment and retrofitting of however, each National Annex can ascribe other
buildings values. The value of the probability of exceedance,
• EN 1998-4: Silos, tanks, and pipelines PR, in TL years of a specific level of the seismic
• EN 1998-5: Foundations, retaining structures, action is related to the mean return period, TR, of
and geotechnical aspects this level of the seismic action in accordance with
• EN 1998-6: Towers, masts, and chimneys Eq. 8 as follows: TR = TL/ln(1 PR). Thus, for
a given TL, the seismic action may equivalently be
This standard provides alternative procedures, specified either via its mean return period, TR, or its
values, and recommendations with notes indicat- probability of exceedance, PR in TL years as pre-
ing where national choices may be made. There- viously discussed in section “Seismic Hazard
fore, in each country the National Standard Analysis.”
Earthquake Return Period and Its Incorporation into Seismic Actions 837
Earthquake Return Period and Its Incorporation into Seismic Actions, Table 1 Importance classes and factors
for buildings and reduction factor for damage limitation limit state according to EC8 (EN 1998-1)
Importance Importance Reduction
class Building type factor (gI) factor (v)
I Buildings of minor importance for public safety, e.g., agricultural 0.8 0.5
buildings, small warehouses, sheds, barns, etc.
II Ordinary buildings, not belonging in the other categories 1.0 0.5
III Buildings whose seismic resistance is of importance in view of the 1.2 0.4
consequences associated with a collapse, e.g., schools, assembly halls,
cultural institutions, etc.
IV Buildings whose integrity during earthquakes is of vital importance for 1.4 0.4
civil protection, e.g., hospitals, fire stations, power plants, etc. E
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 19 Influence of
importance factor on the
return period (Adopted
from Carvalho (2011))
the TLR years for which the reference seismic slightly less for importance class III compared to
action is defined via the following expression: class II. The values of n in each country are
provided in its National Annex, while different
T 1=k values of n (that can be associated to a certain
gI LR
=T L (10)
time interval similarly to gI by an exponential
Similarly, the value of the importance factor gI expression, analogous to Eq. 10) may be defined
that needs to multiply the reference seismic for the various seismic zones of each country,
action to achieve a value of the probability of depending on the seismic hazard conditions and
exceeding the seismic action, PL, in TL years on the preferred protection level.
other than the reference probability of exceed- The reduction factor, v, is applied on the
ance PLR, over the same TL years, may be esti- design seismic action in accordance with limita-
mated as follows: tion of inter-storey drifts as follows (unless oth-
erwise specified):
P 1=k
gI L
=PLR (11) (a) For buildings having nonstructural elements
of brittle materials attached to the structure,
For the damage limitation limit state, the dr n 0.005 h;
values for the probability of exceedance in (b) For buildings having ductile nonstructural
10 years and the return period are PDLR = 10 % elements, dr n 0.0075 h;
and TDLR = 95 years, respectively. In the absence (c) For buildings having nonstructural elements
of more precise information, the seismic action fixed in a way so as not to interfere with
for the verification of the damage limitation structural deformations or without
requirement is derived from AEd (=gI AEk) nonstructural elements, dr n 0.010 h;
using a reduction factor, v, equal to 0.5 or 0.4 to
account for the lower return period for damage where dr, is the design inter-storey drift and
limitation performance level. As shown in h denotes storey height.
Table 1, this factor depends on the importance Additionally to the two basic limit states, EC8
class. It is worth noticing that by applying v, the implies that a third limit state can be used to
influence of gI is reduced compared to the check the safety margin against a global struc-
no-collapse limit state, since the differences tural collapse (Carvalho 2011). In this case, spe-
among the four classes are not so substantial cific measures should be taken to reduce the
(i.e., multiplying v and gI the values of seismic uncertainty and ensure a robust behavior of the
actions are 0.4, 0.5, 0.48, 0.56 for the four classes, structure, even under seismic actions more severe
respectively). In addition, the seismic action is than the design seismic action of 10/50. This
Earthquake Return Period and Its Incorporation into Seismic Actions 839
performance level is implicitly equivalent to the each country. The reference peak ground accel-
satisfaction of a third performance requirement: eration, chosen by the National Authorities
prevention of global collapse under a very rare for each seismic zone, corresponds to the
event (having a much higher return period than reference return period TNCR of the seismic
TNCR = 475 years), equivalently to the 2/50 event action for the no-collapse requirement (or
of US norms. This performance level is denoted equivalently the reference probability of
as near collapse (NC) limit state in EC8 Part exceedance in 50 years, PNCR) chosen by the
3 (EN 1998-3), which is very close to the actual National Authorities. As aforementioned, the
collapse of the structure as it will be discussed in values of gI may be different for the various
the sequence. seismic zones of the country, depending on the
E
Regarding the other parts of EC8 and their seismic hazard conditions and on public safety
provisions for seismic return periods, EC8 Part considerations. In general, an importance factor
2 for bridges also follows general rules for refer- gI equal to 1,0 is assigned to the reference return
ence seismic action return period (there are dif- period. For other return period values, the
ferences, e.g., in importance classes as well as in design ground acceleration on type A ground
damage limitation limit state (Fardis et al. 2012), is equal to ag = gI agR.
etc.) and also includes informative Annex A with For instance, the National Annex of Greece
specific details on the reference seismic action uses the default values (10 % in 50 years event,
and the corresponding return period as well as etc.) and divides the country into three zones, in
on the design seismic action during the construc- which the seismic action, AEk, in terms of agR is set
tion of bridges. EC8 Part 3 (EN 1998-3) for the equal to: 0.16 g, 0.24 g, 0.36 g, for zones Ζ1, Ζ2,
assessment and retrofitting of buildings provides and Ζ3, respectively. An objective of the future
specific guidelines for the incorporation of return updating of EC8 (Carvalho 2011) is to create
period into seismic actions, which will be European zonation maps with spectral values for
presented in a subsequent section. EC8 Part different hazard levels, similar to provisions in the
4 (for silos, tanks, and pipelines), EC8 Part USA. As aforementioned, such European hazard
5 (for foundations, retaining structures, and geo- maps have been recently developed for various
technical aspects), and EC8 Part 6 (for towers, return periods, within SHARE project (www.
masts, and chimneys) do not differentiate with share-eu.org), as the one for 10/50 (475 years
respect to return periods and seismic actions return period) mean PGA values shown in Fig. 20.
with the basic principles and general rules of In EC8 the earthquake motion at a site is
EC8 Part 1. represented by an elastic response spectrum in
terms of spectral acceleration values (Annex
Seismic Zones and Representation of the A of EC8 Part 1 additionally represents seismic
Seismic Action action in the form of an elastic displacement
According to EC8, national territories shall be response spectrum). The shape of the elastic
subdivided by the National Authorities into seis- response spectrum for each country may be
mic zones, depending on the local seismicity, found in its National Annex, and it is the same
while the hazard within each zone is assumed to for the two levels of seismic performance, i.e., for
be constant. For most of the applications of EC8, the no-collapse requirement as well as for the
the hazard is described in terms of a single param- damage limitation requirement. It is also noted
eter, i.e., the value of the reference peak ground that in selecting the appropriate shape of the
acceleration on type A ground (i.e., rock), agR. spectrum, consideration should be given to the
Additional parameters required for specific types magnitude of earthquakes that contribute most to
of structures are given in the relevant parts of EC8. the seismic hazard as defined in PSHA, rather
The reference peak ground acceleration on than on conservative upper limits (e.g., the max-
type A ground, agR, may be derived from zona- imum credible earthquake). Regarding magni-
tion maps contained in the National Annex of tude, two types of spectra are recommended,
840 Earthquake Return Period and Its Incorporation into Seismic Actions
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 20 PGA mean values in Europe for
475 years return period (Adopted from Giardini et al. (2013))
type 1 for magnitude bigger than 5.5 and type accelerograms) representative of the earthquake
2 for smaller earthquakes. These two types have motion may be used in certain cases, taking into
the same pattern but different spectral amplifica- account spatial variability whenever needed.
tions factors that depend also on the
stromatography (of the top 30 m surface layer) Return Periods in EC8 Part 2
of the examined region. EC8 provides also rec- As aforementioned EC8 Part 2 contains Informa-
ommendations for the vertical acceleration tive Annex A that defines probabilities related to
response spectra. the reference seismic action as well as guidance
When the earthquakes affecting a site are gen- for the selection of design seismic action during
erated by widely differing sources, the possibility the construction phase of a bridge. In particular,
of using more than one spectrum, each having the reference seismic action can be defined by
different pattern, should be considered to enable selecting an acceptably low probability, p, that
the design seismic action to be adequately it can be exceeded within the design life, tL, of the
represented. In such circumstances, different bridge. Therefore, the return period of the event,
values of ag will be required for each type of TR, is given by
spectrum and earthquake. Furthermore, for
important structures (gI > 1.0), topographic
TR ¼ 1= 1 ð1 pÞ1=tL (12)
amplification effects should be taken into
account, as described in Informative Annex
A of EC8 Part 5 (EN 1998-5). Finally, it is also The reference seismic action (corresponding to
stated that time histories (at least three gI = 1,0) usually reflects a seismic event with
Earthquake Return Period and Its Incorporation into Seismic Actions 841
a reference return period, TNCR, of 475 years. Part 1 is roughly equivalent to the one that is
Such an event has a probability of exceedance defined as LS of significant damage in Part 3.
between 0.10 and 0.19 for a design life ranging The National Authorities decide whether all
between 50 and 100 years, respectively. This three or less (and which) limit states shall be
level of design action is applicable to the majority checked, as defined in National Annex. The
of the bridges of average importance. appropriate levels of protection are achieved by
Regarding the design seismic action for the selecting, for each of the aforementioned LS,
construction phase, assuming that tc is the dura- a proper return period for the seismic action.
tion of the construction phase of a bridge and p is The return periods ascribed to the various LS to
the acceptable probability of exceedance of the be checked in a country may be found in its
E
design seismic event during this phase, the return National Annex. The protection is normally con-
period TRc is given by Eq. 12, using tc instead of sidered appropriate for ordinary buildings by
tL. For the relatively small values usually associ- selecting the following values for the
ated with tc (tc 5 years), expression Eq. 12 may corresponding return periods:
be approximated by the following simpler rela-
tionship: TRc = tc/p, while it is recommended that – LS of near collapse (NC): 2,475 years,
the value of p does not exceed 0.05. corresponding to a probability of exceedance
The value of the design ground acceleration, of 2 % in 50 years
agc, corresponding to a return period, TRc, – LS of significant damage (SD): 475 years,
depends on the seismicity of the region. In corresponding to a probability of exceedance
many cases the following relationship offers an of 10 % in 50 years
acceptable approximation: – LS of damage limitation (DL): 225 years,
corresponding to a probability of exceedance
k of 20 % in 50 years
agc T Rc
¼ (13)
agR T NCR
For instance, the Greek code (KANEPE 2012)
for structural interventions on reinforced concrete
where agR is the reference peak ground accelera- buildings, after harmonization with Eurocode
tion corresponding to the reference return period 8 Part 3, defines two hazard levels. The first has
TNCR. The value of the exponent k depends on the probability of exceedance of 50 % (maximum
seismicity of the region, and values in the range tolerable) in 50 years and corresponds to an aver-
of 0.30–0.40 can be used. Finally, it is stated that age return period of 70 years, while the second has
the robustness of all partial bridge structures a probability of exceedance of 10 % in 50 years
should be ensured during the construction phases that corresponds to a 475 years return period. For
independently of the design seismic actions. the 10/50 level, the design seismic action of EC8
Part 1 is taken into account, while for the 50/50
Limit States and Return Periods in EC8 Part 3 level, 60 % of the design seismic action of EC8
As prescribed in Eurocode 8 Part 3 (EN 1998-3), Part 1 is taken into account. It is also stated that
a retrofitted existing structure is checked against a Public Authority shall define the cases where
three limit states (LS), namely, near collapse a 50 % probability of exceedance within the refer-
(NC), significant damage (SD), and damage lim- ence 50-year period will not be allowed.
itation (DL). Note that the definition of the LS of
near collapse given in Part 3 of EC 8 is closer to
the actual collapse of the building than the one Seismic Actions in Performance-Based
given in Part 1 of EC 8 and corresponds to the Design
maximum exploitation of the deformation capac-
ity of the structural elements. Actually, the LS Severe damages caused by recent earthquakes in
associated with the “no-collapse” requirement in seismic-prone areas made the engineering
842 Earthquake Return Period and Its Incorporation into Seismic Actions
Earthquake Return Period and Its Incorporation into Seismic Actions, Table 2 Various performance and hazard
levels in a typical PBD framework
community to question the effectiveness of con- hazard and structural performance are consid-
temporary seismic design codes. In general, in ered. In PBD more accurate analysis procedures
strong events the economic losses were vast, are usually employed compared to conventional
while in many cases the losses in terms of human seismic deign to represent more accurately the
lives were also excessive, especially in developing nonlinear structural response for higher seismic
countries. Given that the primary goal of contem- intensities. PBD has the following distinctive
porary seismic design is the protection of human characteristics with respect to the prescriptive
life, it is evident that additional performance tar- design codes: (i) it allows the owner and the
gets should be considered. Under this perspective, engineer to choose both the appropriate level of
performance-based design (PBD) has been quite seismic hazard and the corresponding perfor-
recently introduced in earthquake engineering and mance level of the structure, and (ii) the structure
guidelines (SEAOC 1995; FEMA-273 1997; is designed to meet a series of combinations of
FEMA-356 2000; FEMA-445 2006). PBD is hazard levels with corresponding performance
a holistic framework that includes: local site levels. As shown in PBD illustrative Table 2,
conditions examination; development of concep- the horizontal axis represents step by step to the
tual, preliminary, and final design stages; and con- right structural performance degradation, while
struction and maintenance of the building over its in the vertical axis the size of earthquake
life to ensure reliable and predictable seismic per- increases as we move downward. Although sev-
formance under varying demand conditions. Con- eral hazard and performance levels have been
sequently, the structure is designed to withstand proposed, the 3 3 PBD framework as the one
certain earthquake hazard levels with a predefined shown in Table 2 is considered as more effective
behavior for each case. (e.g., Elnashai and Di Sarno (2008)).
According to PBD procedures, the structures The first step in a performance-based seismic
should be able to resist earthquakes in design procedure is the definition of the perfor-
a quantifiable manner and to present specific mance objectives that will be used. One perfor-
levels of possible damages. They are multilevel mance objective is defined as the combination of
design approaches, where various levels of a certain performance level for a specific hazard
Earthquake Return Period and Its Incorporation into Seismic Actions 843
level. Typical structural design performance Under this perspective, the structural response is
levels are the following: altered from initial elastic to limited and highly
nonlinear for the last two performance levels,
(i) Operational performance level: the struc- respectively.
ture retains its strength and stiffness as the An initial fully operational performance level
overall damage level is very light and none can also be used as the first performance state in
permanent drift is observed. Only minor a PBD framework, if no damage is allowed to
cracking in facades, partitions, and ceilings structural and nonstructural components and full
are encountered, while all components that functionality of the structure is maintained with-
are important to normal operation are func- out any repairs. An additional performance level,
E
tional. A green placard will be placed in the the so-called damage limitation, can also be used
building by post-damage inspectors. in PBD as an intermediate state between opera-
(ii) Life safety performance level: the overall tional and life safety performance levels. In this
damage level is characterized as moderate, state, the structure may be only slightly damaged
i.e., some structural elements and compo- having negligible permanent drifts. Structural
nents may be damaged and permanent drifts elements retain their strength and stiffness prop-
may be encountered. Nevertheless, struc- erties, do not have any significant yielding, and
tural elements in all stories maintain do not need any repair measures. Nonstructural
a considerable residual strength and stiff- components, such as partitions and infills, may
ness and gravity load-bearing elements con- have several cracks; however, such damages
tinue to function. Injuries may occur, could be economically repaired.
however, the overall risk for casualties due The second step in a performance-based seis-
to structural damage is low. The damaged mic design procedure is to determine the earth-
structure shouldn’t be at imminent collapse quake hazard levels. Earthquake-related hazards
risk, but certainly temporary supports and include mainly ground shaking but also – in cer-
structural repairs will be necessary prior to tain cases as aforementioned in section “Seismic
re-occupancy (i.e., it should be possible to Hazard Analysis” – various kinematic distress
repair the structure although this may not be types (fault rupture, liquefaction, lateral spread-
economical). A yellow placard will be ing, and landslides). For instance, soil liquefac-
placed in the building by post-damage tion hazard in a PBD perspective has been
inspectors. presented by (Kramer and Mayfield 2007) to
(iii) Collapse prevention performance level: compare likelihood of liquefaction occurrence
the structure is heavily damaged and has for different seismicity level (associated with
large permanent drifts. Structural elements earthquake return period).
have low residual lateral strength and stiff- Typically, ground shaking is the earthquake
ness, although vertical elements are still hazard that provisions of the seismic design
capable of sustaining gravity loads. Most norms directly address, and it is also the focus
nonstructural components have collapsed. of this work. As described in section “Seismic
The structure is near collapse, and most Hazard Analysis,” ground shaking (as well as any
probably it will not be techno-economically other type of) hazard is usually characterized by
practical and safe to repair and reuse it. A a hazard curve, which indicates the probability
red placard will be placed in the building by that a given value of a ground motion parameter,
post-damage inspectors. for example, peak ground acceleration, will be
exceeded over a certain period of time. Hence,
The first level aims to maintain structural stiff- the earthquake design levels are expressed in
ness, the second is focused on strength, and the terms of a mean recurrence interval or
last one aims to exploit ductility of the system. a probability of exceedance for a specified
844 Earthquake Return Period and Its Incorporation into Seismic Actions
Earthquake Return Period and Its Incorporation into Seismic Actions, Table 3 PGA hazard levels (50/50 and
10/50) in Greece utilizing (Papazachos et al. 1993) approach
Occasional (recurrence interval 72 years or Rare (recurrence interval 475 years or
Seismic Parameters 50 % in 50 years) 10 % in 50 years)
zones B1 B2 PGA (g) PGA (g)
Zone I 3.28 0.61 0.08 0.14
Zone II 3.64 0.64 0.13 0.215
Zone III 4.01 0.61 0.17 0.29
Zone IV 4.64 0.55 0.29 0.46
number of years that covers the design life of the exceedance) or a very rare event hazard level
structure, e.g., typically 50 years for buildings. having an intermediate return period (i.e.,
The most commonly used ground shaking hazard 970 years, or probability of exceedance 10 % in
levels are the following: 100 years, or 0.1 % annual probability of exceed-
ance) between the two higher ones. In addition, it
(iv) Occasional hazard level: with probability of is worth noticing the distinction among maxi-
exceedance 50 % in 50 years with recur- mum earthquake (ME) level and maximum con-
rence interval equal to 72 years and annual sidered earthquake (MCE) (ATC-40 1996). The
probability of exceedance approximately ME is deterministically defined as the maximum
equal to 1.4 %. This level is the so-called level of ground shaking that may ever be
serviceability earthquake (SE), which is expected at the construction site based on
approximately 0.5 times the design earth- regional seismotectonic data. If ME is expressed
quake (DE) (ATC-40 1996). in probabilistic terms, as the aforementioned haz-
(v) Rare hazard level: with probability of ard levels, then it has a return period of about
exceedance 10 % in 50 years with recur- 1,000 years; thus, it is equivalent to very rare
rence interval equal to 475 years and annual event hazard level.
probability of exceedance approximately The above hazard levels can be associated to
equal to 0.2 %. This level is the so-called major seismic demand parameters, such as the
design earthquake (DE) or design basis peak ground acceleration (PGA). For instance,
earthquake (DBE), and (as aforementioned in the study by Papazachos et al. (1993), Greece
in section “Seismic Hazard Analysis”) for was divided into four zones, and based on this
certain norms, it is approximately the two work, Table 3 depicts the PGA levels for two
third of the more severe 2/50 level. return periods, TR, utilizing the following
(vi) Maximum considered event hazard level: log-linear formula:
with probability of exceedance 2 % in
50 years with recurrence interval equal to SI ¼ B1 þ B2 logT R (14)
2,475 years and annual probability of
exceedance approximately equal to 0.04 %. in which SI denotes a characteristic measure of
This level is the so-called maximum consid- the seismic intensity (macroseismic intensity or
ered earthquake (MCE). the logarithm of strong ground motion parame-
ters, such as PGA or PGV), and parameters B1
According to the adopted PBD formulation and B2 vary for each SI.
(e.g., SEAOC (1999)), other hazard levels can Such relationships can be derived based on
also be used. For instance, a frequent event haz- a Cornell-type attenuation formula of the form
ard level having a very low return period (i.e., (Papazachos et al. 1993)
43 years or probability of exceedance 50 % in
30 years, or 2.3 % annual probability of SI ¼ C1 þ C2 M C3 lnðD þ C4 Þ (15)
Earthquake Return Period and Its Incorporation into Seismic Actions 845
Lateral Deformation
Annual Probability of Exceedance
10−2
10−3 Δ
−4
10
Earthquake Return Period and Its Incorporation into Seismic Actions, Fig. 22 Basic steps in performance
evaluation (Adopted from FEMA-451B (2003))
recurrence periods of PBD hazard levels have cost and safety. In a holistic probabilistic
been used as probabilistic constraints in the LCCA and PBD framework, all aspects are con-
optimization process. Life cycle cost analysis sidered, and the outcome is given in terms
(LCCA) is also considered nowadays in con- of fragility curves and loss hazard curves.
junction with PBD and optimization to achieve As shown in Fig. 23, the latter are similar
best possible solutions. LCC represents the total to seismic hazard curves (e.g., Fig. 4), as the
cost, including that related to the expected dam- losses (or repair cost), expressed in terms
ages caused by earthquake events that may of monetary units, and are also associated to
occur during the design life of the structure. It earthquake return periods or probabilities of
has been proven that the objective to satisfy all exceedance.
checks imposed by contemporary norms aiming To accomplish this goal, it is necessary to
to minimize initial construction cost is not obtain curves illustrating the so-called mean
always efficient in terms of overall long-term annual frequencies (MAF) of exceedance for
Earthquake Return Period and Its Incorporation into Seismic Actions 847
Earthquake Return
Period and Its
Incorporation into
Seismic Actions,
Fig. 23 Characteristic
repair loss curve for various
hazard levels
seismic isolation, energy dissipation and active vibra- Kanellopoulos A (2007) Seismic design and retrofitting of
tion control of structures, Sendai, 24–27 Sept 2013 reinforced concrete buildings, Cubus Hellas, Athens
Cornell CA (1968) Engineering seismic risk analysis. Bull (in Greek)
Am Seismol Soc 58(5):1583–1606 KANEPE (2012) Code of structural interventions on
Earthquake Spectra (2000) Seismic design provisions reinforced concrete buildings, English version after
and guidelines. Earthq Spectra Theme Issue 16(1): harmonization with Eurocode 8 Part 3, Earthquake
1–310 Planning and Protection Organization, Athens. Avail-
Elnashai A, Di Sarno L (2008) Fundamentals of earth- able at http://ecpfe.oasp.gr/
quake engineering. Wiley, New York Katsanos EI, Sextos AG, Manolis GD (2010) Selection of
EN 1998 (2004) Eurocode 8 (EC8): design provisions of earthquake ground motion records: a state-of-the-art
structures for earthquake resistance. European Com- review from a structural engineering perspective. Soil
mittee for Standardization (CEN), Brussels Dyn Earthq Eng 30:157–169
Fardis MN, Kolias B, Panagiotakos T, Katsaras C, Kramer SL (1996) Geotechnical earthquake engineering.
Psychogios T (2012) Guide for bridge design with Prentice Hall, Upper Saddle River
emphasis on seismic aspects. Report No SEE Kramer S, Mayfield RT (2007) Return period of soil
2012–01, Report Series in Structural and Earthquake liquefaction. ASCE J Geotech Geoenviron Eng
Engineering, Department of Civil Engineering, Uni- 133(7):802–813
versity of Patras, Patras Krinitzsky EL (1995) Deterministic versus probabilistic
FEMA-273 (1997) NEHRP guidelines for the seismic seismic hazard analysis for critical structures. Eng
rehabilitation of buildings. Federal Emergency Man- Geol 40:1–7
agement Agency, Washington, DC Lagaros ND, Mitropoulou CC (2013) The effect of uncer-
FEMA-350 (2000) Recommended seismic design criteria tainties in seismic loss estimation of steel and
for new steel moment-frame buildings. SAC Joint reinforced concrete composite buildings. Struct
Venture, Federal Emergency Management Agency, Infrastruct Eng 9(6):546–566
Washington, DC Makarios TK (2013) Peak ground acceleration functions
FEMA-356 (2000) Prestandard and commentary for the of mean return period for known reliability index of
seismic rehabilitation of buildings. Federal Emergency Hellenic design earthquakes. J Earthq Eng
Management Agency, Washington, DC 17(1):98–109
FEMA-368 (2001) NEHRP recommended provisions for Papaioannou C, Voulgaris N, Karakaisis G, Koutrakis S,
seismic regulations for new buildings and other struc- Latousakis J, Makropoulos K, Papazachos B, Sokos E,
tures, part 1, provisions. Federal Emergency Manage- Stavrakakis G, Tselentis G (2008) The utilization of
ment Agency, Washington, DC new seismological data in the compilation of the new
FEMA-445 (2006) Next-generation performance-based seismic hazard map of Greece. In: Third Greek con-
seismic design guidelines, program plan for new and ference on earthquake engineering & engineering seis-
existing buildings. Federal Emergency Management mology, Athens, 5–7 Nov 2013, paper 2025 (in Greek)
Agency, Washington, DC Papazachos BC, Papaioannou CA, Theodulidis NP
FEMA-450 (2003) NEHRP Recommended provisions for (1993) Regionalization of seismic hazard in Greece
seismic regulations for new buildings and other struc- based on seismic sources. Nat Hazards 8(1):1–18
tures. Federal Emergency Management Agency, Petersen MD, Frankel AD, Harmsen SC, Mueller CS,
Washington, DC Haller KM, Wheeler RL, Wesson RL, Zeng Y, Boyd
FEMA-451B (2003) NEHRP recommended provisions: OS, Perkins DM, Luco N, Field EH, Wills CJ,
instructional and Training materials. Building Seismic Rukstales KS (2008) Documentation for the 2008
Safety Council (BSSC), Washington, DC update of the United States national seismic hazard
FEMA-750 (2009) NEHRP Recommended seismic pro- maps. US Geological Survey, Open-file report
visions for new buildings and other structures. Federal 2008–1128, p 61
Emergency Management Agency, Washington, DC Raffaele D, Fiore A (2013) A simplified algorithm for
Fragiadakis M, Papadrakakis M (2008) Performance- evaluating the seismic return period of structural
based optimum seismic design of reinforced con- capacity. In: 4th conference on computational methods
crete structures. Earthq Eng Struct Dyn 37(6): in structural dynamics and earthquake engineering,
825–844 Kos, 12–14 June 2013
Giardini D, Woessner J, Danciu L, Crowley H, Cotton F, Saygili G (2008) A probabilistic approach for evaluating
Gr€unthal G, Pinho R, Valensise G, SHARE consor- earthquake-induced landslides. PhD Thesis, The Uni-
tium (2013) SHARE European seismic hazard map for versity of Texas at Austin, TX
peak ground acceleration, 10% exceedance probabili- SEAOC (1995) Performance based seismic engineering of
ties in 50 years. SHARE European seismic harmoni- buildings, Vision 2000 Committee, Structural Engi-
zation project. Available at www.share-eu.org. Aug neers Association of California, Sacramento, CA
2013 SEAOC (1999) Recommended lateral force requirements
IBC (2000) International building code. International and commentary, Structural Engineers Association of
Code Council, Falls Church California, Sacramento, CA
Earthquake Risk Mitigation of Lifelines and Critical Facilities 851
Vamvatsikos D (2012) Accurate application and second- risk and global losses associated to damages of
order improvement of the SAC/FEMA probabilistic lifelines and critical facilities and establish effi-
formats for seismic performance assessment. ASCE
J Struct Eng. doi:10.1061/(ASCE)ST.1943-541X. cient seismic risk mitigation strategies. Mitiga-
0000774 tion actions include, among others, enhancement
Ventura CE (2006) Fundamental concepts in seismic engi- of emergency preparedness, strengthening of
neering. Short course for CSCE Calgary 2006 annual existing structures, upgrade of seismic design
conference, Canada
guidelines, and standards and improvement of
the recovery planning.
The issue of resilience of lifelines and critical
facilities is of utmost priority for the preparedness
E
Earthquake Risk Mitigation of and mitigation of seismic risk; a good evaluation
Lifelines and Critical Facilities of the resilience of the lifelines for different seis-
mic scenarios and the implementation of
Kyriazis Pitilakis and Sotiris Argyroudis pro-seismic actions to improve it are of major
Department of Civil Engineering, Aristotle importance. In other words, lifeline companies,
University, Thessaloniki, Greece public administration, and community need to be
“prepared” and less “vulnerable,” in order to
achieve a high “resilience.” In the last years, as
Synonyms the idea of the necessity of building disaster resil-
ient communities gains acceptance, new methods
Infrastructures; Readiness; Response; Restora- have been proposed to quantify resilience beyond
tion; Retrofitting; Risk reduction; Seismic risk estimating losses. Because of the vastness of the
assessment; Transportation networks; Utilities definition, resilience necessarily has to take into
account its entire complex and multiple dimen-
sions, which includes technical, organizational,
Introduction social, and economic facets (Cimellaro et al.
2010).
Lifelines and critical facilities refer to the com- Lifelines may be distinguished in two major
plex system and network assets of connected categories: (i) utility systems including potable
components usually interacting with other com- water, wastewater, natural gas, oil, electric
ponents and systems, which are performing vital power, and communication and (ii) transportation
functions that are essential to sustain the life and systems comprising roadways, railways, airport,
the growth of a community, such as producing, and port facilities. Sometimes the terms infrastruc-
transporting, and distributing goods or services. tures or critical facilities are used instead, at least
Their global value for the society and economy is for some of them. Lifeline systems present three
permanently increased in our modern, technolog- distinctive features: (a) spatial variability and
ically advanced, highly demanding, and fragile exposure to different geological and geotechnical
world. In case of strong earthquake motions, their conditions and hazards (e.g., fault crossing, land-
physical damages and the consequent disruption slides, liquefaction); (b) wide variety of compo-
of the services they provide may contribute seri- nent typologies and materials used; and (c) specific
ously to the global economic losses. At the same functionality requirements, which makes them
time, their restoration cost may be very high, highly hierarchical networks. For practical rea-
reaching in several cases to 10 %, 15 %, or even sons, they are generally distinguished to links
more in some extreme cases (e.g., the water sys- (e.g., pipelines, roads) and nodes (e.g., tanks,
tem in Christchurch, New Zealand) of the initial power substations, bridges).
construction cost of the whole system they An important issue in understanding and
belong. Therefore, in terms of resilience, it is defining the risk mitigation of lifelines and criti-
important to recognize and quantify the seismic cal facilities is the extensive interdependencies
852 Earthquake Risk Mitigation of Lifelines and Critical Facilities
Transpor-
Oil
fuels, lubricants
fueltransport, tation
access to facilities
po fuels, lubricants
access to facilities
sto wer
gas shipping
fueltransport,
f ,
ra
ge or p ng
an um n ali trol
fuel for generators,
d c pi sig con
on ng or c
and production needs
r f raffi
l s tio o w g, t
ys n p tin
lubricants
te s, lig
h power for
SCADA,
m
s compressors, Natural
storage and gas
Electric control systems
power
water fuel for generators
for cooling p
s
pum
on
co
for tions
mu DA,
ati
SC unic
r
Water
m r fo s
e a
pow lift st po
AD at
m
nic
com SCA
w
supply and
A, ion
sw
e SCADA,
itc
communications
he
r
s
water for
SCADA, cooling
communications Telecommu-
nication heating
fuel for generators
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 1 Example of infrastructure interdependencies
(Modified, Rinaldi et al. 2001)
within utility networks and infrastructures as well from 1900 to 2012 showed that around 30 % of
as infrastructure business, which are schemati- the direct losses came from infrastructure losses
cally shown in Fig. 1. Infrastructure outages due in transport, communications, pipelines, energy
to seismic damages can cascade through the supply systems, etc. Knowledge, experience, and
economy and considerably increase the overall lessons learned from past earthquakes have sig-
losses. Moreover, interactions between systems nificantly contributed to understanding the struc-
can severely impact the emergency response, tural and functional behavior of lifelines and
especially when the disruptions concern critical critical facilities and to develop more appropriate
facilities, such as hospitals, command centers, methods and tools toward effective risk
etc. Therefore, the key interactions should be mitigation.
recognized and considered in the risk mitigation The lessons learned from past earthquakes
planning. also concern the readiness of lifeline companies
Several devastating earthquakes occurred dur- and authorities to restore damages as well as to
ing the last century, causing extensive losses in understand the interactions between different
terms of fatalities, injuries, damages, and disrup- systems. In 1989 Mw6.9 Loma Prieta earthquake,
tions. The experiences from past earthquakes the damages occurred in the water network
show that the short- and long-term economic greatly affected the firefighting capacity in the
impact due to damages of lifelines can be signif- area of San Francisco, Oakland, and Berkeley.
icant. In a recent study by CATDAT (Daniell and In 1994 Mw6.7 Northridge earthquake, all of the
Vervaeck 2013), the disaggregation of direct eco- 1.5 million power customers and about 50,000
nomic losses of 61 selected major earthquakes water customers lost service, many for a week or
Earthquake Risk Mitigation of Lifelines and Critical Facilities 853
more (Schiff 1995). Earthquake-related damage selection of “acceptable” risk. Still the deci-
to electric power systems can cause considerable sion by itself is very challenging, because
direct repair and restoration costs, business seismic risk assessment includes various
interruption, inconvenience, and permanent uncertainties related to seismic hazard, vulner-
loss of data, food, and perishable goods. Simi- ability, performance, and loss estimation pro-
larly, earthquake damage to water supply cedures. Moreover the selection of acceptable
systems can have serious economic and health risk needs appropriate balance between cost,
consequences. After the 1995 Mw6.8 Hyogoken- risk, and benefits, which is different for each
Nanbu earthquake, it was observed that the network and society.
restoration period of water system in Kobe was Seismic hazard and therefore seismic risk
E
much longer compared to other major earth- are considered as variable in space and time.
quakes due to traffic congestion, street block- Risk assessment in different time scales is
ades, damaged buildings, and water that flowed illustrated in Fig. 2 with reference to the
into gas pipelines (Kameda 2000). In general, occurrence of a disastrous earthquake. Exam-
the functional restoration for the damaged net- ples of mitigation actions appropriate for each
works in Kobe needed five to ten times more time period are also given. Long-term assess-
time than those experienced in other major ment is based on the seismic hazard model
earthquakes. The final recovery period in Kobe providing, for example, PGA values for out-
was 82 days for water supply, 85 days for natural crop rock conditions with 475 mean return
gas supply, 7 days for power supply, 20 months period for Euro-Mediterranean region
for highway bridges, 6 months for major lines of (Giardini et al. 2013). Such maps can be used
railways, and about 2 years for the main port for the risk assessment of common structures;
facilities. for essential facilities (e.g., communication
towers, hospitals, major bridges), the hazard
assessment and risk require longer return
Main Actions for Risk Mitigation of periods (e.g., up to 5000 years). Short to
Lifelines and Critical Facilities medium term hazard and risk assessment
may also be computed often using defined
The general mitigation framework of lifelines scenario type earthquakes.
and critical facilities includes the following Recent technological advances in seismic
main actions: risk assessment (1), risk reduction instrumentation and telecommunication per-
(2), readiness for emergency response (3), and mit the implementation of early warning sys-
readiness for recovery (4). Each of these actions tems (Fig. 2c) and real-time hazard and risk
is shortly described in the next paragraphs. assessments (Fig. 2d), which are related to the
emergency response actions in the co-seismic
1. Risk assessment The estimation of seismic phase (Gasparini et al. 2007; Allen et al. 2009;
risk is the fundamental step in risk manage- Zschau et al. 2009; Allen 2011; Erdik and
ment and decision-making for risk mitigation. Fahjan 2006). They involve the generation of
Risk assessment refers to the probable future alarm signals directly from online instrumen-
losses (physical, performance, economic, tal data and the generation at a first stage of
social) not only due to structural damages but real-time ground estimation maps and in the
also considering the business interruption or second stage the vulnerability and risk assess-
interdependencies of lifelines. Lifeline com- ment of structures provided that fragility
panies and civil protection authorities are curves are available. The alarm signals alert
asked to determine how to best allocate the for the coming earthquake in a time window of
limited resources for risk reduction measures few seconds and enable safe shutdown of crit-
(actions 2, 3, and 4) based on the results of ical facilities (e.g., nuclear plants, gas net-
seismic risk analyses for their facilities and the work) or train and traffic stopping systems.
854 Earthquake Risk Mitigation of Lifelines and Critical Facilities
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 2 Risk assessment and examples of mitigation
actions in different time scales with reference to the occurrence of a disastrous earthquake (EQ)
The assessment of ground shaking intensity Finally, a critical issue in the risk manage-
maps and/or physical damage and casualties ment and reduction of impact after a seismic
takes part immediately after earthquake (few event is to assess the functionality of
minutes/hours) and assists the lifeline owners mainshock damaged infrastructures (e.g.,
and stakeholders for the emergency actions major bridges, hospitals) considering the
(rapid response system). In this way the aftershock hazard (Fig. 2e). This is a challeng-
impact of a disastrous earthquake can be ing issue due to limited time availability (days/
reduced due to timely and effective operations months), the lack of adequate models for the
(e.g., allocation of resources and prioritization fragility assessment of damaged structures,
of interventions). Such dense strong ground and the inherent uncertainties in aftershock
motion monitoring systems capable of com- hazard assessment.
puting damage and casualties in near real time 2. Risk reduction is related to the identification
have been implemented in regional or city of the components or parts of the networks and
scale in California, Yokohama, Tokyo, Istan- infrastructures that present particular vulnera-
bul, Taiwan, Bucharest, and Naples (Erdik and bility and can result to high loss of network
Fahjan 2006). In other cases, early warning performance due to seismic damages. The def-
systems are implemented for the monitoring inition of priorities for retrofitting should be
of individual structures and critical facilities based on vulnerability and risk assessment
(see REAKT project, http://www.reaktproject. studies. Specific risk reduction measures
eu/). should be taken before the earthquake in
Earthquake Risk Mitigation of Lifelines and Critical Facilities 855
Earthquake Risk
Mitigation of Lifelines
and Critical Facilities,
Fig. 3 Fragility curves for
a five spans bridge, with an
expansion joint in the
center span, supported by
four columns, before and
after retrofit (Kim and
Shinozuka 2004)
order to improve the seismic performance of and/or shear blocks; replacement of old
infrastructure assets. These include the bearings with ductile, multi-rotational,
following: and multidirectional bearings; retrofitting
– Strengthening of vulnerable infrastruc- of concrete columns with insufficient con-
ture assets to improve their seismic perfor- finement through steel jacketing to provide
mance. Several utility systems and networks passive confinement. Several studies have
are now outdated and often without proper been conducted showing the effect of
seismic design and protection, as, for exam- retrofitting. An example is shown in
ple, parts of water and wastewater systems Fig. 3, where the effect of retrofit by
in several cities were built more than a cen- means of steel jacketing of columns is illus-
tury ago. State-of-the-art technical guide- trated in the fragility curves of typical brid-
lines for seismic assessment, retrofit, and ges in southern California. In the USA, the
rehabilitation have been published around Federal Highway Administration (FHWA)
the world, with emphasis to buildings. has published a manual including most cur-
Retrofitting techniques are generally case rent state of practice in assessing the vul-
dependent and the description of their typol- nerability of highway structures to the
ogies is beyond the present essay. Few typ- effects of earthquakes and implementing
ical examples are given below: retrofit measures to improve performance
Geo-structures: Stabilization of (FHWA 2004, 2006).
embankments, retaining walls, and steep – Undertake countermeasures against
natural slopes; ground improvement, espe- geotechnical hazards such as liquefaction
cially in case of liquefaction susceptibility. (later spreading, settlement) and land-
Pipelines and wastewater networks: slides/rockfalls or other induced phenom-
Replacement of vulnerable segments, ena such as the tsunami waves (e.g.,
upgrade of valves, local ground improve- protecting walls).
ment, and increase of dead load of the pipes – Improve the design and construction
in particular in liquefaction-prone areas. practices for new lifeline components.
Bridges: Retrofitting for continuity to The design of critical infrastructures should
multiple simple spans through restrainers be based on advanced and detailed seismic
856 Earthquake Risk Mitigation of Lifelines and Critical Facilities
hazard analyses, considering the topo- related to the emergency needs, such as
graphical and geotechnical conditions of accessibility and supply of energy to hos-
the site at local or regional scale. Soil- pitals and critical facilities. Other specific
foundation-structure interaction effects are actions that can improve the readiness and
of high importance and should be consid- rapidity include the arrangements to have
ered in the design of critical infrastructures helicopters committed for use in post-
(e.g., major bridges). The uses of special earthquake inspections of extended net-
materials and techniques that improve the works (e.g., railroads), the regular training
seismic performance of structures (e.g., of staff, and check and maintenance of
ductile materials for pipelines, soil-rubber material and equipment. The use of
mixtures for embankments) and the use of advanced tools like GIS and remote sensing
base isolation systems are other practices techniques and continuous monitoring of
for seismic risk reduction of several lifeline the functionality of the networks are now-
components. adays indispensable.
– Relocate components if this is cost bene- – The need for coordinated actions between
ficial or more feasible instead of taking the different lifeline owners and stake-
large-scale strengthening countermeasures, holders is also emphasized due to the inter-
for example, to stabilize landslide zones. actions between different lifeline elements.
The case of active seismic fault crossing For example, the restoration of utilities
is another typical example where some- such as gas, water, wastewater pipes, and
times relocation of road or railway lines communication or electric power cables
should be examined considering all other that exist underneath urban roads requires
parameters involved. a coordinated action. A representative case
– Installing restraints to restrict the failure of interactions among different lifeline sys-
(toppling) of sensitive equipment and tems during the restoration period is from
nonstructural components, such as com- 1995 Kobe earthquake. The restoration
puter equipment, communication devices, activities of the water and gas networks in
batteries, health-care equipment, diesel the Kobe area have been delayed due to
tanks, or equipment installed on roofs and traffic congestion, street blockades, dam-
ceilings. aged buildings, and water that flowed into
– Improve the redundancy of specific com- gas pipelines. It is clear that an efficient
ponents where this is cost-effective. For rescue policy and an optimum recovery
example, redundant diesel generators and strategy require the evaluation and the effi-
diesel tanks in hospitals, main water cient governance of the interactions
pumping stations, and strategic buildings. between different lifeline systems and
3. Readiness and resourcefulness for emer- authorities.
gency response, which is related to the repairs – Measures to mitigate losses due to after-
and measures for the safety of the public and shocks, such as short-term supporting
the functionality of the critical facilities countermeasures of heavily damaged
immediately after the disaster. Particular structures.
actions are: – Installation of early warning systems, to
– Response planning, including prioritiza- alert for the coming earthquake and cut off
tion for restoration interventions, availabil- the operation of high-risk facilities such as
ity of resources (inspection and repair the gas network or to slow down high-
crews) and means (equipment, materials, speed trains. Such systems are currently
and spare parts), organization, and coordi- implemented in Bay Area Rapid Transit
nation of the emergency response. The pri- system (www.bart.gov) and Japan Rail-
oritization for repairs in this phase is ways (Nakamura and Saita 2007).
Earthquake Risk Mitigation of Lifelines and Critical Facilities 857
Earthquake Risk Mitigation of Lifelines and Critical performance level (PO) or higher (PA*). In case B, the
Facilities, Fig. 4 Effect of mitigation measures and read- system’s performance can reach lower (PB), PO, or higher
iness (A, B, C) in the recovery process after the occurrence (PB*) level, while in case C the performance will most
of a disastrous earthquake and representation of a system probably reach Po or lower (PC) level. The losses are
with high (R1) and low (R2) resilience. Ideally, in case A, represented by LA, LA + LB and LA + LB + LC for miti-
the performance of the network can reach the pre-disaster gation measures A, B, and C, respectively
4. Readiness and rapidity for recovery is same, close to, or better than the pre-disaster
related to the restoration period and until the one. Several uncertainties are associated to the
functionality of all lifelines and facilities recovery time such as the structural restoration
returns to the pre-disaster level. It can be time and the business interruption time. It
described in terms of time needed to reorga- typically depends on the earthquake intensi-
nize services, activities, and functions. An ties and on the location of the system with its
efficient planning is required for the prioriti- given resources such as capital, materials, and
zation and organization of restoration works. labor as well as public policies, following the
The post-earthquake restoration approaches major seismic event (Cimellaro et al. 2010).
are briefly described in a next section. The Figure 4 illustrates the response and recov-
implementation of pre-earthquake risk reduc- ery time of a system with lower and higher
tion measures together with the readiness for resilience in relation with the risk reduction
recovery considerably affects the evolution of and readiness measures. Three cases are
the recovery process (Fig. 4). described: with risk reduction and readiness
The resilience of the system (e.g., water measures (A), without risk reduction mea-
supply, electric power) is defined by its ability sures, but with readiness (B), and without
to withstand the disastrous event without suf- any measures (C). Obviously, in reality there
fering devastating losses and to recover are several variations of risk mitigation mea-
shortly. The recovery time is the period nec- sures, resilience levels, and recovery time.
essary to restore the functionality of the sys- Ideally, in case A, the performance of the
tem to a desired level that can operate the network can reach the pre-disaster
858 Earthquake Risk Mitigation of Lifelines and Critical Facilities
performance level (PO) or higher (PA*). In characteristics, such as local site conditions
case B, the system’s performance can reach and the potential for liquefaction or land-
lower (PB), PO, or higher (PB*) level, while in slides; and (iii) the importance of the struc-
case C the performance will most probably ture, which is evaluated based on its
reach PO or lower (PC) level. Losses are function within the network, the impact to
represented by LA, LA + LB, and LA + LB + the economy, the interdependencies with
LC for mitigation measures A, B, and C, other lifelines, the public safety, or its
respectively. Both losses and recovery time importance for the emergency response.
are increasing from A to C. Such models have been proposed for build-
ings, bridges, as, for example, in the USA by
the Federal Highway Administration
Prioritization for Mitigation Actions (FHWA 1995) and in Japan by Kawashima
and Unjoh (1990) and for utility networks
Pre-seismic (retrofitting and strengthening (Wang et al. 1995; Muhlbauer 1996).
actions), co-seismic, and post-seismic strategies (B) Advanced risk assessment studies aiming to
and policies must be based on advanced the estimation of probable direct and indirect
approaches that quantify the seismic risk of life- losses (physical, performance, economic,
lines and critical facilities considering seismic social) due to future earthquakes considering
hazard, vulnerability, and exposure of the net- also interactions among different systems.
works. The final prioritization criteria combine A general layout for the seismic risk assess-
engineering techniques, economic analysis tools ment of lifelines is outlined in Fig. 4 and is
and decision-making, or political aspects. The further described in Pitilakis and Argyroudis
available engineering techniques for seismic (2014). Inventory and taxonomy/typology
risk assessment and prioritization for mitigation definitions of all elements at risk and systems
actions are classified in two main groups: are the necessary initial steps to describe the
elements and networks exposed to seismic
(A) Screening approaches based on multi- risk. The estimation of site-specific seismic
criteria value models. A rating system is ground motion and other geotechnical haz-
used to assign a score in each attribute of ards and the selection estimation of appro-
the factors that describe the overall risk of priate seismic intensity measures (IM) is the
the network component. The scaling values basis of the seismic hazard analysis. Vulner-
are defined based on expert judgment and ability is the expected response of an element
the experience of past earthquakes, while or a network to a given seismic intensity. It is
weighting factors are commonly used in commonly assessed based on fragility
order to account for the relative contribu- models, which estimate the probability of
tion of each attribute to the overall risk of exceeding certain damage states for given
the component. This type of models has seismic intensity. Fragility functions are
been mainly developed for buildings and adapted with respect to the typological char-
bridges, and the main factors considered acteristics of each element at risk. An exten-
are (i) the level of seismicity, usually sive review of available fragility functions
described by the peak ground acceleration and state-of-the-art methods for vulnerabil-
(PGA) value as derived by national provi- ity assessment of buildings and lifeline com-
sions or based on site-specific studies ponents can be found in Pitilakis
required for important structures and infra- et al. (2014a).
structures; (ii) the vulnerability of each ele-
ment at risk, based on the typology and The expected damage and loss of the entire
other geometrical and structural character- network are estimated through appropriate sys-
istics of the element at risk and soil temic analysis, which takes into account the
Earthquake Risk Mitigation of Lifelines and Critical Facilities 859
Damage FRAGILITY
Intensity Measures (IM)
States FUNCTIONS
E
Performance VULNERABILITY& DAMAGE
Indicators assessment
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 5 General layout for the seismic risk
assessment of lifelines
physical damage and the relations and interac- Post-earthquake Restoration Modeling
tions between the different components and of Lifeline Systems
systems (Pitilakis et al. 2014b). The interdepen-
dencies among different systems may consider- Technically the restoration process is part of the
ably increase the overall impact. The results, recovery process and commonly is expressed by
which are commonly provided in terms of perfor- restoration curves, which define the recovered
mance indicators that describe the network func- functionality of a damaged network component
tional losses or the expected economic or (e.g., bridge, water pumping station, or harbor
socioeconomic losses (Khazai et al. 2014), pro- crane) as a function of elapsed time since the
vide the means for an efficient mitigation or earthquake occurrence, which normally is the
recovery planning. In particular, priorities for main event. However in case of very strong earth-
retrofitting or adding redundancy of the elements quakes, the aftershock activity is expected quite
with greater risk can be defined. The level of high for few or several months, a fact that is
correlation of damaged components with the net- seriously affecting most actions in the post-
work loss performance along with additional earthquake recovery period. For any system, res-
indicators and criteria for the characterization of toration depends on the characteristics and the
the importance of each component can be used in damage level of the individual components as
order to classify the mitigation priorities. Obvi- well as the network performance and the interac-
ously, several sources of uncertainties are inher- tions with other lifelines. In general, the estima-
ent in the various definitions, models, tools, and tion of the required time for recovery is a
methods for estimation of losses, which make the complicated task and varies significantly between
risk assessment of lifelines a complex and chal- different countries or cities, as it depends on
lenging topic (Fig. 5). numerous factors.
860 Earthquake Risk Mitigation of Lifelines and Critical Facilities
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 6 Restoration curves for reinforced concrete
bridges as a function of number of spans and damage state (Argyroudis 2010)
Restoration process is based on the available same area, based on data from the 1971 San
construction techniques, local experience and Fernando earthquake (Chang et al. 1996).
expertise, the regulations and regional politics, Figures 6 and 7 show examples of restoration
the available manpower, material and equipment curves for highway bridges and port facilities
together with the knowledge of the system, the based on experts’ opinion. In some approaches,
economic resources, the interactions of individ- resource constraints are accounted for by speci-
ual components and networks, the accessibility of fying the number of repairs that can be made in
the damaged components, and all these in relation any time period as a function of the personnel
with the intensity, typology, and spatial variabil- available, assuming that all necessary material
ity of the damages. Hence it is obvious that res- and expertise are available. In this way, the pro-
toration curves for lifelines may differ gress of the restoration across time and space is
considerably from one country or city to another. depicted, while it is possible to investigate the
Modeling of post-earthquake lifeline restoration effect of prioritization of repairs. This approach
can be achieved using experts’ opinion, empirical has been employed by Isumi and Shibuya (1985),
data, theoretical models, or a combination of Ballantyne (1990), Chang et al. (1999), and
these. In ATC-25 (ATC 1992), restoration curves HAZUS (EMA 2003) for water supply system.
are constructed using ATC-13 (ATC 1985) data The personnel and rate of repairs are assumed to
sources that are based on regression analysis of be constant or not with time. However, the actual
expert opinion data. These restoration curves restoration time of each component may be very
were adapted also in HAZUS methodology different depending on the parameters influenc-
(FEMA 2003). Examples of empirical ing the restoration process. Moreover, the recov-
approaches include the functions developed for ery of the whole system is an even more
the electric power system in Memphis based on complicated issue as it depends on the restoration
the 1994 Northridge earthquake and ATC-25 data of all individual components. Therefore, the
(Shinozuka et al. 1998) or for the water system in aforementioned models are deterministic and
Earthquake Risk Mitigation of Lifelines and Critical Facilities 861
generally of moderate reliability, as the various importance, where local authorities, government,
uncertainties related with the restoration time are and lifeline actors have a major role. It is also
not accounted. important to notice that any pre-earthquake
Dynamic simulation approaches that can be actions (upgrading, retrofitting of the most vul-
either deterministic or stochastic have been nerable elements and sectors) have a very bene-
employed recently to represent utility company’s ficial impact on the definition of the specific
decision variables (e.g., repair prioritization restoration curves.
plans, mutual aid agreements, repair resources)
and to produce spatially disaggregated restora-
tion curves, so as one can end up with restoration Example of Prioritization in the
curves for each region of the service area, rather Restoration Process
than just one for the entire lifeline system (Kozin
and Zhou 1991; Zhang 1992). Statistical variabil- An example of seismic risk assessment and pri-
ity and uncertainties associated with key param- oritization for restoration of damages is presented
eters, such as post-earthquake damage inspection for the port of Thessaloniki in Greece. In this
durations, repair durations for each damaged application, the port facilities are classified into
component, time needed to replace components main, important, and secondary through appro-
that cannot be repaired, and amount of available priate ranking of the value of the exposed ele-
resources, are taken into account by defining ments, based on various factors that describe the
these parameters as random variables with spec- role of each element in the system. In that way,
ified probability distributions (Nojima and the “global value” of each element at risk
Kameda 1992; Cagnan and Davidson 2007). depends not only on its direct specific value or
However, these approaches require several sim- content (physical and human) but also upon its
plifying assumptions and numerous input data, indirect/immaterial value that is represented by
which are not easily and always available. the usefulness and relative role in the whole
As a conclusion, the construction of restora- urban system, at a specific time. Three periods
tion curves for any component and each lifeline are identified in respect to the occurrence of an
system must be defined by local actors, using earthquake event: normal, crisis, and recovery.
basically expert elicitation, considering local “Global value” evaluation in different periods
conditions and constrains. The reduction of the could be a powerful tool for the prioritization of
time delays is actually a political decision of great pre-earthquake actions and quantification of the
862 Earthquake Risk Mitigation of Lifelines and Critical Facilities
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 8 Classification of the importance of cargo
handling equipment of Thessaloniki’s port during the crisis period
overall importance of different complex and Earthquake Risk Mitigation of Lifelines and Critical
Facilities, Table 1 Risk analysis matrix showing cargo
coupled lifeline systems. Several criteria for this handling equipment seismic retrofit priorities
are used, such as operational attributes, land use,
Urban risk/ Priorities
population influenced, human losses, economic
seismic hazard Main Important Secondary
and social weight under normal, crisis and recov-
Complete/ 1st 1st 2nd
ery circumstances, identity/radiance, environ- extensive priority priority priority
mental impact, and others. Appropriate damages
qualitative or quantitative indicators can then be Moderate 1st 2nd 3rd
defined for each period, while relevant measuring damages priority priority priority
units are used for their evaluation and the identi- Slight/minor 1st 2nd 4th
damages priority priority priority
fication of “main,” “important,” and “secondary”
elements and system’s weak points (RISK-UE
project: Mouroux and Le Brun 2006; Pitilakis
et al. 2010). could also follow these priorities aiming at effi-
The cargo handling equipment of cient seismic risk management procedures.
Thessaloniki’s port are classified based on indi- Table 1 summarizes the application of the pro-
cators that describe the capacity, location, type of posed approach for the cargo handling equip-
cargo handled, and redundancy (Pitilakis ment of Thessaloniki’s port. Figure 10
et al. 2010), and representative maps illustrating illustrates the restoration priorities for a damage
the classification of elements at risk are scenario with mean return period of 475 years
constructed (Fig. 8). (Fig. 9) (Pitilakis et al. 2010).
Combining “global value” evaluation and Restoration curves have been defined in col-
vulnerability assessment and using, if necessary, laboration with the port authority (Fig. 7), based
an “expert opinion,” it is possible to estimate on available manpower, capabilities, local expe-
priorities and to account for the economic and rience, and expertise. Figure 11 illustrates the
social losses, for a specific utility system and a functionality level of cargo handling equipment
given seismic scenario. Recovery activities at 3 and 18 months after the seismic event,
Earthquake Risk Mitigation of Lifelines and Critical Facilities 863
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 9 Distribution of damages to waterfront
structures and cargo handling equipment of Thessaloniki’s port (Tm = 475 years)
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 10 Restoration priorities of cargo handling
equipment of Thessaloniki’s port during the crisis period
864 Earthquake Risk Mitigation of Lifelines and Critical Facilities
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Fig. 11 Functionality percentage of cargo handling
equipment of Thessaloniki’s port in 3 and 18 months’ time after the seismic event
Earthquake Risk Mitigation of Lifelines and Critical Facilities 865
Earthquake Risk Mitigation of Lifelines and Critical Facilities, Table 2 Infrastructure upgrades in San Francisco
area (Zoback 2014)
Infrastructure Number clients Total cost/source
provider served Scope of upgrades of funding
Pacific Gas and 15M throughout System upgrade of underground gas, electrical $2.5B
Electric North CA components, substations, and administration Ratepayers
building
Bay Area Rapid 400,000 daily Retrofit core system-aerial structures, stations, $1.3B
Transit ridership trans-bay tube (completion system by 2018, tube by Bonds and
2023) taxpayers/ $3 M
from FEMA
East Bay 1.3M in East Bay Entire system upgrade: pipelines, fault crossings, $0.189B E
Municipal Utility dams, administration building, pumping, and Ratepayers
District treatment plants (completed 1999)
California 38.3M statewide Structurally upgraded and seismically retrofit over $13.08B
Department of 2000 bridges and overpasses, new E span Bay CA taxpayers
Transportation Bridge
San Francisco 2.6M residential, Upgrade of >100-year-old Hetch Hetchy water $4.6B
Public Utilities commercial, and system (pipelines), fault crossings treatment Bond measure
Commission industrial facilities, and reservoirs (completion by 2016)
Total investment $21.6B
assuming that the restoration process starts dramatically increased their disaster resilience
immediately after the earthquake and that there through substantial system upgrades, an invest-
are two available working groups for the restora- ment of $21 billion funded almost exclusively by
tion process. ratepayers and taxpayers – almost $1 billion/year
(see Table 2). A five-step strategy has been
employed by the main infrastructure providers to
Lifeline Seismic Retrofitting Programs increase their earthquake resilience (Zoback 2014):
the California Department of Transportation not only for life safety but also to ensure that they
(Caltrans) initiated a bridge seismic safety retrofit can return to operation shortly after a major earth-
program focused on bridge expansion joints. The quake. The upgrades will be accomplished by
mid-span hinge and support joint retrofit program using the latest seismic standards to improve the
was completed in 1989 at a cost of $55 million. structural integrity of BART facilities. The com-
With the significant loss of life and closure of pletion of all earthquake upgrades is anticipated
major routes during the 1989 Loma Prieta earth- by 2018. The total project is $1.2 billion (in 2004
quake, the Caltrans established a new seismic dollars). Preliminary upgrade concepts include:
retrofit program for the more than 12,000 bridges Vibro replacement to compact soil backfill,
in the California State Highway system. This increasing seismic joint capacity and sealing
program is focused on identifying and retrofitting around joints, and new concrete shear walls in
existing bridges statewide, bringing them up to the Oakland Ventilation Structure. Potential
the latest seismic safety retrofit standards upgrading concepts for elevated structures
established to prevent collapse during future include enlarged or reinforced concrete founda-
earthquakes. In particular, after the Loma Prieta tions, “jackets” around concrete columns, addi-
earthquake, the Department initially identified tional shear keys, and additional foundation piles
1,039 state highway bridges (mostly single- where poor soil conditions exist (http://www.
column bridges) in need of being retrofitted to bart.gov/about/projects/eqs).
meet seismic safety standards (Phase 1 Program).
After the 1994 Northridge earthquake, the Lifelines in New Zealand
Caltrans identified another 1,155 state-owned Lifeline vulnerability to natural hazards includ-
bridges (mostly multicolumn bridges) in need of ing earthquakes had been given prominence in a
being retrofitted (Phase 2 Program). The state’s comprehensive project undertaken by the Christ-
bridge earthquake strengthening program, which church Engineering Lifelines Group in the first
has been almost completed, involves more than half of the 1990s. Utility networks were
2,200 structures and costs more than $12 billion. described, broken down to component level,
A separate program was established for toll brid- with each one assessed for vulnerability to the
ges which are the largest and most complicated various natural hazards. Assessments considered
bridges in the State. An estimated $2.6 billion is “importance” to the network as well as “vulnera-
allocated for this program, which also includes bility” and “damage impact” estimated in three
the replacement of the east span of the San time frames: immediately after the event, period
Francisco-Oakland Bay Bridge (http://www.dot. following, and time for return to normality. Dif-
ca.gov/hq/paffairs/about/retrofit.htm). ferent lifeline systems were covered, namely,
energy, telecommunications, transport, and
Bay Area Rapid Transit (BART) System water/sewerage. The project also included emer-
BART is one of the Bay Area’s most vital trans- gency buildings like broadcasting, police and fire
portation links, carrying about 360,000 travelers stations, and ambulance bases. Several mitigation
every day. BART’s success in maintaining con- measures have been undertaken in the following
tinuous service directly after the 1989 Loma years. For example, in the water system restraint
Prieta earthquake reconfirmed the system’s actions taken for spare parts and equipment, seis-
importance as transportation “lifeline.” However, mic strengthening of reservoirs and pumping sta-
because of the likelihood that BART will be tions, valve shut off improvements; in the electric
subject to a major future earthquake, an Earth- power network, insecure buildings were removed
quake Safety Program has been recently initiated. and the control room was relocated, while major
The program will upgrade vulnerable portions of transformers and other essential equipment as
the original system to ensure safety for the public well as bridge approaches were strengthened to
and BART employees. Portions of the original protect cables. In transportation networks,
system with the highest traffic will be upgraded improvements were made at bridges including
Earthquake Risk Mitigation of Lifelines and Critical Facilities 867
Earthquake Risk
Mitigation of Lifelines
and Critical Facilities,
Fig. 12 Superposition of
cost curves as a function of
the system performance for
seismic retrofit cost,
expected damage cost, and
total cost
strengthening connections between superstruc- reopened within a day. Similar cases have been
tures and substructures, increasing column noted for other networks and components in the
strength and ductility, strengthening or renewing area. It is a representative example of worthwhile
retaining and approach structures, and strength- investment, as the costs of seismic risk mitigation
ening lateral/longitudinal restraint mechanisms. in Christchurch would have been repaid many
Planning was improved in relation to key roads, times over if these measures were not have been
bridges, and the important Lyttelton Tunnel. taken place (Fenwick 2012).
Improvements were made at the airport and
port. Additional equipment, such as generators Europe
and portable pumps, were also purchased. In European countries with high seismicity (i.e.,
The substantial program of seismic mitigation Italy, Greece, Turkey), risk assessment programs
undertaken by Christchurch lifeline utilities has have been undertaken in the last decade for public
served Christchurch in reducing losses and facil- buildings (i.e., hospitals, schools) and upgrades
itating emergency responses and recovery during have been implemented in some cases. For other
the 2010 (M7.1) and 2011 (M6.3) earthquakes. infrastructures (e.g., bridges), risk assessment
The damage would have been greater and the studies and screening approaches have been par-
response slower if the emergency planning and tially conducted; however, retrofitting is rather
other preparatory work had not been undertaken. limited. A short overview of the activity of bridge
For example, the electricity distribution seis- upgrading currently ongoing in Italy is given by
mic strengthening program, commenced in 1996 Pinto and Franchin (2010).
and progressed systematically each year, costs $6
million and is estimated to have saved $60–$65
million in direct asset replacement costs and Economic Efficiency of Seismic
repairs. Total repair costs from the 2011 earth- Retrofitting Measures
quake amounted to only $150,000. Seventeen
bridges in Canterbury, including many in or The economic efficiency of seismic retrofitting
near Christchurch, were identified as at risk and measures is related to benefit-cost analysis.
had been retrofitted at a cost of $2 million. These Figure 12 shows an idealization of economic
measures prevented critical damage, as only one efficiency based on the concept of performance
bridge was closed due to approach settlement and levels of the system under consideration. The cost
868 Earthquake Risk Mitigation of Lifelines and Critical Facilities
of retrofitting measures is increasing when higher employment and properties’ value, with benefits
performance levels are desired. Οn the other to the insurance and construction industry.
hand, the cost of the expected seismic damages,
when the retrofitting measures are applied, is
decreasing. The total cost (initial construction Summary
cost, without considering depreciation and
retrofitting cost) is lower in the intersection of Lifelines and critical facilities are essential public
the two curves. In order to determine the eco- infrastructures and networks that support citi-
nomic efficiency, both the repair costs after zen’s life, economy, and industry. Experience
expected earthquakes of different intensities and from past earthquakes reveals that their compo-
return periods and the costs of retrofitting mea- nents are quite vulnerable and the induced dam-
sures to achieve certain performance levels have ages can have high impact in the emergency
to be considered, usually as a function of the response as well as serious economic and health
initial construction costs. consequences. Losses can be even higher because
Benefit-cost analysis is a well-established tool of the extensive interdependencies between util-
to assess the viability of seismic risk reduction ity networks and infrastructures. Therefore, high
projects and policies. In particular, the appraisal priority should be given to secure safety of life-
of earthquake mitigation projects is crucial in line facilities against earthquakes and reduce the
making community, policy- and decision- overall impact.
makers, and insurance and construction industry The decision for mitigation of lifelines against
to understand and quantify the effects of earth- seismic hazard is a quite complex issue as differ-
quake mitigation. The analysis for the economic ent institutions from public and private sector are
feasibility of seismic retrofitting considers the involved: governmental and local authorities,
achievable loss reduction, investment return lifeline companies, industries, and insurances.
period, and retrofit cost. Overall, the organization Moreover, the decision by itself is very difficult
of mitigation policies for lifelines is a political because it should be made based on the results of
and economic decision and surely not a decision the seismic risk assessment of lifelines which
that is based on emotion or public feeling, which, includes various uncertainties owing to seismic
however, should be seriously considered. The hazard vulnerability, performance, and loss esti-
selection of the acceptable risk considering mation procedures. The risk assessment is the
appropriate seismic scenario or scenarios is fundamental step for the prioritization of mitiga-
always a major decision to make. It should take tion measures, which in general include the fol-
into account and combine different parameters lowing actions: risk reduction, readiness for
like the detail of the seismic hazard, the will of emergency response, and readiness for recovery.
central and local actors, the available funds, and Risk reduction is related to the strengthening and
the financial and technical capability of the com- upgrade of vulnerable assets, improvement of
munity and the country. Epistemic and physical design and construction practices, relocation of
uncertainties are involved in every step of any components, improvement of redundancy, and
earthquake risk reduction policy and should be security of sensitive nonstructural equipment.
taken into account. Readiness and resourcefulness for emergency
The increase in safety and reliability of life- response is related to the repairs and measures
lines and the reduction of direct and indirect for the safety of the public and the functionality
losses against future earthquakes are the main of the critical facilities immediately after the
benefits of efficient retrofitting programs. How- disaster. Typical measures include installation
ever, they are not the only ones. Retrofitting strat- of early warning systems, advanced response
egies followed by investment policies are also a planning, coordinated actions between different
serious and efficient tool for short- and long-term entities/stakeholders, and measures to reduce
growth of economic development, increasing losses due to aftershocks. Readiness and rapidity
Earthquake Risk Mitigation of Lifelines and Critical Facilities 869
for recovery is related to the restoration period ▶ Seismic Risk Assessment, Cascading Effects
and until the functionality of all lifelines and ▶ Site Response for Seismic Hazard Assessment
facilities returns to the pre-disaster level. The ▶ Strengthening Techniques: Bridges
implementation of pre-earthquake risk reduction ▶ Uncertainty Theories: Overview
measures together with the readiness for recovery
considerably affects the evolution of the recovery
References
process.
Pre-seismic (retrofitting and strengthening Allen RM (2011) Earthquakes, early and strong motion
actions), co-seismic, and post-seismic strategies warning. In: Gupta H (ed) Encyclopedia of solid earth
and policies must be based on advanced geophysics. Springer, Dordrecht, pp 226–233
Allen RM, Gasparini P, Kamigaichi O, Böse M (2009) E
approaches that quantify the seismic risk of life-
The status of earthquake early warning around the
lines and critical facilities considering seismic world: an introductory overview. Seism Res Lett
hazard, vulnerability, and exposure of the net- 80:682–693. doi:10.1785/gssrl.80.5.682
works. The final prioritization criteria should Applied Technology Council [ATC] (1985) Earthquake
combine engineering techniques, economic anal- damage evaluation data for California. Report no:
ATC-13, Redwood City
ysis, and decision-making tools. Applied Technology Council [ATC] (1992) A model
methodology for assessment of seismic vulnerability
and impact distribution of water supply systems.
Cross-References Report no: ATC-25-1, Redwood City
Argyroudis S (2010) Contribution to the seismic vulnera-
▶ Community Recovery Following Earthquake bility and risk assessment of transportation networks in
urban environment. PhD dissertation, Aristotle Uni-
Disasters
versity of Thessaloniki, Greece
▶ Damage to Infrastructure: Modeling Ballantyne DB (1990) Earthquake loss estimation model-
▶ Early Earthquake Warning (EEW) System: ing of the Seattle water system. Report to the U.-
Overview S. Geological Survey, Federal Way
Cagnan Z, Davidson R (2007) Discrete event simulation of
▶ Earthquake Disaster Recovery: Leadership and
the post-earthquake restoration process for electric power
Governance systems. Int J Risk Assess Manage 7(8):1138–1156
▶ Earthquake Early Warning System in Taiwan Chang SE, Seligson HA, Eguchi RT (1996) Estimation of
▶ Earthquake Mitigation Strategies Through the economic impact of multiple lifeline disruption:
memphis light, gas and water division case study.
Insurance
Technical report no. NCEER-96-0011, Buffalo
▶ Economic Recovery Following Earthquakes Chang SE, Shinozuka M, Svekla W (1999) Modeling
Disasters Post – disaster urban lifeline restoration. In: Elliott
▶ Emergency Response Coordination Within WM, McDonough P (eds) Optimizing post earthquake
lifeline system reliability: proceedings of the 5th
Earthquake Disasters
U.S. conference on lifeline earthquake engineering.
▶ Estimation of Potential Seismic Damage in ASCE Technical Council on Lifeline Earthquake
Urban Areas Engineering Monograph no. 16, pp 602–611
▶ Learning from Earthquake Disasters Cimellaro GP, Reinhorn AM, Bruneau M (2010) Frame-
work for analytical quantification of disaster resil-
▶ Liquefaction: Countermeasures to Mitigate
ience. Eng Struct 32:3639–3649
Risk Daniell J, Vervaeck A (2013) The CATDAT damaging
▶ Probabilistic Seismic Hazard Models earthquakes database - 2012 - the year in review,
▶ Reconstruction Following Earthquake CEDIM Research Report 2013–01, Karlsruhe,
Germany
Disasters
Erdik M, Fahjan Y (2006) Early warning and rapid dam-
▶ Resilience to Earthquake Disasters age assessment. In: Assessing and managing earth-
▶ Resourcing Issues Following Earthquake quake risk. Geotechnical, geological and earthquake
Disaster engineering, vol 2. Springer, Dordrecht, pp 323–338
Federal Emergency Management Agency [FEMA] (2003)
▶ Seismic Fragility Analysis
Multi hazard loss estimation methodology: earthquake
▶ Seismic Loss Assessment model – HAZUS-MH MR3 technical manual. FEMA,
▶ Seismic Resilience Washington, DC
870 Earthquake Risk Mitigation of Lifelines and Critical Facilities
Fenwick T (2012) The value of lifeline seismic risk miti- early warning systems. Springer, Berlin/Heidelberg,
gation in Christchurch. New Zealand Earthquake pp 249–281
Commission under its ‘Science-to-Practice’ Nojima N, Kameda H (1992) Optimal strategy by use of
Programme, ISBN 978-0-473-21631-3 tree structure for post-earthquake restoration of lifeline
FHWA (1995) Seismic retrofitting manual for highway network systems. In: Proceeding of the 10th world
bridges: report FHWA-RD-94-052, Federal Highway conference on earthquake engineering, Balkema, Rot-
Administration, Washington, DC terdam, pp 5541–5546
FHWA (2004) Seismic retrofitting manual for highway Pinto PE, Franchin P (2010) Issues in the upgrade of
structures: part 2 – retaining structures, slopes, tunnels, Italian highway structures, Journal of Earthquake
culverts and roadways, FHWA-HRT-05-067 Engineering 14(8):1221–1252
FHWA (2006) Seismic retrofitting manual for highway Pitilakis K, Argyroudis S (2014) Seismic vulnerability
structures: part 1 – bridges, FHWA-HRT-06-032 assessment: lifelines. In: Jalayer F, Galasso C (eds)
Gasparini P, Manfredi G, Zschau J (eds) (2007) Earth- Seismic risk assessment, encyclopedia of earthquake
quake early warning systems. Springer, Berlin, 362 pp engineering. Springer-Verlag Berlin Heidelberg
Giardini D et al (2013) Seismic Hazard Harmonization in Pitilakis K, Anastasiadis A, Kakderi K, Alexoudi M,
Europe (SHARE). Online Data Resource, Argyroudis S (2010) The role of soil and site condi-
doi:10.12686/SED-00000001-SHARE tions in the vulnerability and risk assessment of life-
Isumi M, Shibuya T (1985) Simulation of post earthquake lines and infrastructures. The case of Thessaloniki
restoration for lifeline systems. Int J Mass Emerg (Greece). In: 5th International conference on recent
Disaster 3(1):87–105 advances in Geotechnical Earthquake engineering
Kakderi K (2011) Seismic vulnerability assessment and and Soil Dynamics and Symposium in honor of prof.r
seismic risk management of complex transportation I.M. Idriss, 24–29 May, San Diego, Paper no SPL13
networks. Application to port facilities. PhD Disserta- Pitilakis K, Franchin P, Khazai B, Wenzel H (eds) (2014a)
tion, Aristotle University of Thessaloniki, Greece SYNER-G: systemic seismic vulnerability and risk
Kameda H (2000) Engineering management of lifeline assessment of complex urban, utility, lifeline systems
systems under earthquake risk. In: Proceeding of the and critical facilities. Methodology and applications,
12th world conference on earthquake engineering, Geotechnical, geological and earthquake engineering.
New Zealand Society for Earthquake Engineering, Springer, Heidelberg. ISBN 978-94-017-8834-2
Upper Hutt Pitilakis K, Crowley E, Kaynia A (eds) (2014b) SYNER-
Kawashima K, Unjoh S (1990) An inspection method of G: typology definition and fragility functions for phys-
seismic vulnerability of existing highway bridges, ical elements at seismic risk, vol 27, Geotechnical,
Structural Engineering and Earthquake Engineering, geological and earthquake engineering. Springer, Hei-
JSCE, 7(1):143–150 delberg. ISBN 978-94-007-7872-6
Khazai B, Daniell JE, D€ uzg€un Ş, Kunz-Plap T, Wenzel Rinaldi SM, Peerenboom JP, Kelly TK (2001) Identifying,
F (2014) Framework for systemic socio-economic vul- understanding, and analyzing critical infrastructure
nerability and loss assessment. In: Pitilakis K, interdependencies. IEEE Contr Syst Mag 21(6):11–25
Franchin P, Khazai B, Wenzel H (eds) SYNER-G: Schiff AJ (ed) (1995) Northridge earthquake: lifeline per-
systemic seismic vulnerability and risk assessment of formance and post-earthquake response. ASCE Tech-
complex urban, utility, lifeline systems and critical nical Council on Lifeline Earthquake Engineering
facilities. Methodology and applications. Springer, Monograph No. 8, New York, p 339
Heidelberg. ISBN 978-94-017-8834-2 Shinozuka M, Rose A, Eguchi RT (1998) Engineering and
Kim SH, Shinozuka M (2004) Development of fragility socioeconomic impacts of earthquakes: an analysis of
curves of bridges retrofitted by column jacketing. electricity lifeline disruptions in the New Madrid area.
Probabilistic Engineering Mechanics 19:105–112 Monograph no. 2, MCEER, Buffalo
Kozin F, Zhou H (1991) System study of urban response Wang L, Li H, Ishibashi I (1995) Prioritization for reha-
and reconstruction due to earthquake. J Eng Mech bilitation of buried lifelines. In: O’Rourke MJ (ed),
ASCE 116:1959–1972 proceedings of the 4th US conference on lifeline earth-
Mouroux P, Le Brun B (2006) Risk-UE project: an quake engineering, ASCE Technical Council on Life-
advanced approach to earthquake risk scenarios with line Earthquake Engineering Monograph No. 6, pp
application to different European towns. In: Oliveira 588–595
CS, Roca A, Goula X (eds) Assessing and managing Zhang RH (1992) Lifeline interaction and post earthquake
earthquake risk. Springer, Dordrecht, pp 479–508. urban system reconstruction. In: Proceeding of 10th
doi:10.1007/978-1-4020-3608-8_23 World conference on earthquake engineering,
Muhlbauer WK (1996) Pipeline risk management manual. Balkema, Rotterdam, pp 5475–5480
A tested and proven system to prevent loss and assess Zoback ML (2014) Written communication. Stanford
risk. Gulf Publishing, Houston University, Stanford, CA
Nakamura Y, Saita J (2007) UrEDAS, the earthquake Zschau J, Gaparini P et al (2009) The SAFER-Project and
warning system: today and tomorrow. In: seismic early warning in Europe, AGU, Fall meeting
Mandredi G, Zchau J, Gasparini P (eds) Earthquake 2009, abstract #S21C-02
Earthquake Swarms 871
the mid-Atlantic ridge, which separates Iceland and tectonic. Accordingly, several models have
over the two major lithosphere plates, the Eurasian been proposed to explain the conditions and mech-
plate to the east and the North American plate to anisms leading to the formation of earthquake
the west. An exceptionally intense earthquake swarms. Mogi (1963) proposed their origination
swarm has accompanied the Bárðarbunga volca- as consequence of stress concentration in unusu-
nic eruption. The seismic crisis started in August ally heterogeneous rock and considered magma
2014 and has been going on for 8 months. The intrusions to be the likely cause of stress accumu-
largest earthquake reached magnitude Mw = 5.6 lation. Hill (1977) proposed a model for the gen-
and more than 70 events had magnitudes eration of earthquake swarms in volcanic
Mw 5.0. Further examples of European earth- environments, in which they occur on planes
quake swarm areas somehow connected with the connecting the edges of parallel and offset inflat-
plate or microplate boundaries are western Alps, a ing dikes. The magma intrusion causes dike infla-
few areas in Apennines, and also a few areas in tion, which increases the stress difference s1–s3
Greece. A swarm-like activity preceded a severe (s1 is maximum and s3 minimum local stress) and
Mw = 6.3 earthquake in L’Aquila (Abruzzo drives slip on the shear planes. Another model
region, central Italy) in 2009. based on mechanical coupling between the rupture
Intraplate earthquake swarms occur mainly in process and fluid migration was proposed by
Quaternary volcanism areas, which are often char- Yamashita (1999). Here, numerous pores are cre-
acterized by geodynamic unrest and other phe- ated due to the rupture and this allows fluids to
nomena like diffuse degassing, geothermal migrate from the high-pressure compartments into
anomalies, and chemical or dissolution anomalies. the ruptured zone. Swarm earthquakes are gener-
A typical European intraplate swarm area is ated if the rate of the pore creation is high, and
W-Bohemia/Vogtland. Other intraplate earthquake consequently, the pore volume rapidly increases.
swarm areas affected by Quaternary volcanism are, Since the pores are not filled with fluids during
for example, the French Massif Central, Colorado, their creation, fluids flow into the fracture zone
and Long Valley in California. Vidale and Shearer causing a pore pressure decrease with time and
(2006) examined earthquake bursts in southern ceasing the rupture growth. But the consecutive
California between 1984 and 2002 and identified rise of the pore pressure reactivates the rupture
18 sequences as swarm like. An example of intra- process and the swarm activity goes on. In such a
plate earthquake swarm areas where Quaternary case, the seismic b-value (basic parameter of the
volcanism is missing is the Vosges in France. One frequency–magnitude distribution; for more
of the most intensive nonvolcanic earthquake details, see section “Statistical Characteristics of
swarm area in Japan is located in the Wakayama Swarm Earthquakes”) is unusually large, which
district in southwest Japan. The Wakayama swarm corresponds to the high b-values observed in vol-
activity was initiated more than 40 years ago and canic earthquake swarms. In case the rate of the
has continued to the present day; the occurrence pore creation is small or there are no new pores at
rate of earthquakes with magnitudes greater than all, the model provides the foreshock-mainshock
3.0 is >10 events per year. Earthquake swarms of sequences and small b-values. A prerequisite for
magmatic origin occur frequently at mid-oceanic Yamashita’s model are continuously open faults,
spreading centers, which is discussed in more detail which serve for the fluid supply.
in section “Earthquake Swarms Induced by Mag-
matic and Volcanic Activity.”
Local Networks: A Tool for
Understanding Earthquake Swarms
Primary Earthquake Swarm Models
All the models mentioned above imply that earth-
Earthquake swarms are of diverse nature; they are quake swarms are phenomena which occur in
usually considered to be of two types: volcanic diverse tectonic environments. Their underlying
Earthquake Swarms 873
causes are not easy to disclose. For the same typically exceed 1.0 and are often as high as 2.5
reasons, earthquake swarms are defined rather (e.g., Lay and Wallace 1995), which implies
empirically than deterministically and their trig- prevalence of small events against larger ones in
gering mechanisms and driving forces are also individual swarm activities. However, relevant
unclear. A promising way to get a deeper insight studies show that b-values of both tectonic and
into their driving mechanisms is a detailed anal- volcanic swarm activities vary frequently
ysis of swarm seismicity including the smallest between common values of 0.7–1.5. The b-values
possible events, over a long time period. A pre- 0.85–1.0 are characteristic for intraplate earth-
requisite for this is high-quality data from long- quake swarms in W-Bohemia/Vogtland, similar
term observations by means of a dense network of b-values were found for several earthquake
E
local seismic stations providing a sufficiently low swarms on the Reykjanes Peninsula and in the
completeness magnitude (ML <0 as an optimum). Hengill volcanic complex in SW-Iceland (Fig. 1).
Among others, similar networks are operating in Also some volcanic earthquake swarms show
the W-Bohemia earthquake swarm region b-value <1.0, e.g., the earthquake swarm prior
(WEBNET consisting of 22 stations), in Iceland to the 1995 eruption of Mt Ruapehu in New
(SIL-network of about 70 stations), and in Yel- Zealand (Hurst and McGinty 1999) having
lowstone National Park (Yellowstone seismic b-values 0.75. Note that the b-value close to
network consisting of 26 stations). 1.0 indicate that the earthquake swarm concerned
is a mixture of mainshock-aftershock sequences
that are largely controlled by a local tectonic
Statistical Characteristics of Swarm stress.
Earthquakes The b-values >1.0 are also observed quite
often, particularly for volcanic swarms. For
The Gutenberg–Richter law, log N = a bM, example, b-values of the Yellowstone earthquake
where N is the number of events having a swarms range from 0.6 to 1.5 (Farrel et al. 2009),
magnitude M and a and b (b-value) are con- b-value 1.2 are typical for swarms in Long
stants, expresses the relation between earthquake Valley Caldera and the adjacent Sierra Nevada
size and the frequency of their occurrence in a (Hill et al. 2003). The MLmax = 4.6 earthquake-
particular region and time period, hence it is the swarm activity associated with an eruption of El
most important scaling relationship in the earth- Hierro (Canary Islands) showed b-value 1.2,
quake statistics. The constant b signifies the rate however, the early pre-eruptive activity was char-
of smaller to larger events. In general b-values for acterized by a b-value of 2.25 (Ibáñez et al. 2012).
ordinary mainshock-aftershock sequences as well Those b-values >1.0 occur in nonvolcanic areas
as for seismicity in particular region and given such as the Ubaye swarm (French Alps) in
time period typically range from 0.8 to 1.0. 2003–2004, b = 1.2 (Jenatton et al. 2007) or the
The b-value of 1.0 implies that there will be a Sampeyre swarm (Western Alps, Italy) in 2010,
tenfold increase in the number of events for every b = 1.4 (Godano et al. 2013). Extremely high
unit drop in magnitude. It means that distribution b-values of 1.5–2.0 have been indicated for earth-
of earthquakes with respect to magnitudes is scale quake swarms in circum-Pacific subduction
invariant, appears to be self-similar and obeys a zones (Holtkamp and Brudzinski 2011) and an
power law. Computation of the magnitude–- even higher value of 2.3 was determined by
frequency distribution is easy to do but its inter- Einarsson (1986) for divergent segments of the
pretation may be tricky. Reliable evaluation of b- Mid-Atlantic Ridge (see also section “Earth-
value requires a complete data set over at least quake Swarms Induced by Magmatic and Volca-
three orders of magnitudes; an incompleteness of nic Activity”).
the analysed data may result in false b-value. For temporal evolution of seismicity at plate
It is generally assumed and often claimed in boundaries, it holds that the aftershock rate
literature that b-values of earthquake swarms decays with the time from the mainshock,
874 Earthquake Swarms
Earthquake Swarms, Fig. 1 Statistical characteristics fault segment (Fischer et al. 2014). (a) The magnitude–-
of interplate earthquake swarms from three different areas frequency distribution of all these swarms shows b-value
in SW-Iceland: Hengill volcanic complex in 1997, ML between 0.8 and 1.0, which suggests an occurrence of the
4.5, Ölfus area in 1998, ML 5.1, and Krı́suvı́k area on mainshock-aftershock sequences in these swarms. (b) The
the Reykjanes Peninsula in 2003, ML 4.2, and of intra- interevent-time probability distribution signifies a similar-
plate W-Bohemian earthquake swarms in 2000, ML 3.3 ity of the event-rate (rapidity) of the individual swarms
and in 2008, ML 3.8, which were located on the same
according to the power law N(t) / (t + c)p, similar depths are reported for earthquake
where t is the time after the mainshock, c is con- swarms in continental rifts of the Rio Grande
stant, and exponent p 1 (modified Omori law). and Kenya (Ibs-von Seht et al. 2008). The hypo-
Yet no adequate law was proposed for earthquake centers of the 2000 volcanic Miyakejima swarm
swarms. The reason is simply the absence of a in Japan were located in a large range of depths
single mainshock; alternatively the swarm can be between 5 and 20 km (the strongest Mw = 6.4
considered as a number of overlapping aftershock earthquake occurred at depth of 10 km), while
sequences in time and space (Fischer et al. 2014). foci of the nonvolcanic Matsushiro earthquake
To describe this behavior the epidemic type after- swarm of 1965–1967 in central Japan took place
shock sequence model (ETAS) was proposed by from near surface down to about 12 km.
Ogata (1988) which considers each earthquake Focal depths of the intraplate earthquake
being capable of generating aftershocks and its swarms in W-Bohemia/Vogtland vary from 5 to
aftershock productivity increases with its 20 km but those in the main focal zone Novy
magnitude. Kostel, where more than 35,000 ML <4.0 events
were recorded during the last 25 years, range
between 6 and 13 km. Interplate earthquake
Space and Time Distribution of the Foci swarms in SW-Iceland – on the Reykjanes Pen-
insula, (ML 4.2), in the Henghill volcanic com-
Earthquake swarms frequently originate in the plex, (ML 4.5), and Ölfus area (ML 5.1) are a
upper part of the Earth’s crust. Depths of swarm bit shallower compared to those in W-Bohemia/
earthquakes all over the world range approxi- Vogtland (Fig. 2).
mately between 2 and 20 km, while deeper earth- Extremely shallow earthquake swarms have
quake swarms exist but rather infrequently. The been observed in the caldera of Hakone volcano
majority of swarm events are located around in central Japan, the hypocenters of the 2009
10 km and shallower. For example, earthquake swarm (ML 3.2) were located at depths between
swarms in the Yellowstone volcanic field occur at 0.5 and 2.8 km (Yukutake et al. 2011). Even
depths between 2 and 12 km (Farrel et al. 2009), shallower was the Tricastin earthquake swarm,
Earthquake Swarms 875
Earthquake Swarms, Fig. 2 Space and time distribu- area (Reykjanes Peninsula) in 2003. The spatial distribu-
tion of earthquakes in the W-Bohemian earthquake tion of foci for each earthquake swarm is represented by
swarms of 2000 a 2008 and the SW-Icelandic swarms in two plots: (Top) the map view, and (Middle) depth view
the Hengill volcanic complex in 1997 and in the Krı́suvı́k along the strike. (Bottom) The time development of swarm
876 Earthquake Swarms
the Rhône Valley in France, (ML <1.7), in Spatial clustering and focal migration of
2002–2003, that took place at depths from swarms are related to each other by their origin.
200 m to 1 km (Thouvenot et al. 2009). A striking In the beginning of a particular swarm and/or
phenomenon is meteorologically triggered earth- individual swarm phases, the foci cluster tightly
quake swarms at Mt. Hochstaufen in in the vicinity of the activity nucleation point and
SE-Germany (ML 2.4), with depths from gradually migrate outwards. The extent of the
100 m down to 2.5 km (Kraft et al. 2006). focal migration depends largely on switching
Earthquake swarms in the lower crust, where the activity among several portions of the fault.
ductile deformation is expected, are relatively rare Individual swarms show quite different migration
but occasionally observed beneath active volca- patterns, even if they occur in the same place.
noes. An example is the SW-margin of Long Val- Two W-Bohemian swarms of 2000 and 2008,
ley Caldera in California (Shelly and Hill 2011), which took place on the same fault segment
where both shallow brittle-failure swarms at depth (they represent “swarm twins”), are an informa-
down to 7 km and deep brittle-failures at depths tive example of an unusually different earthquake
between 18 and 30 km occurred. Another exam- swarm development (Fischer et al. 2014). The
ple is the 2007–2008 swarm beneath former, with maximum magnitude ML = 3.3,
Mt. Upptyppingar (a part of the Iceland Northern was of a distinctly episodic character: most of
Volcanic Zone) that was of the magmatic intrusion the events occurred during eight swarm phases
origin; the focal depths range between 10 and with abrupt onsets lasting 2–4 days, which were
20 km, see Fig. 3. (It is worthy of notice that this followed by longer quiescence periods. The
swarm activity was evidenced by a significant swarm activity culminated in the last phase dur-
deformation, which was observed at permanent ing which all the ML >3.0 earthquakes occurred.
GPS stations in the region). The latter, with maximum magnitude ML = 3.8
Earthquake swarms, regardless of the tectonic showed a much faster seismic-moment release,
environment, are characterized by a dense clus- which resulted in the overlapping of the subse-
tering of the foci in time and space and in most quent swarm phases and thus the 2008 swarm
cases by pronounced migration of the foci. Indi- displayed only two distinct phases. Most of the
vidual swarms usually initiate abruptly, event seismic moment, which was more than three
rate accelerates and magnitudes of events times larger than that in 2000, was released dur-
increase. The event rate may differ from case to ing the first 8 days. Both the 2000 and 2008
case considerably, but quite often exceeds swarms initiated at the bottom of their focal clus-
100 ML >0 events per hour when swarm activity ters but they showed different patterns of focal
is culminating. Larger earthquake swarms are migration (Fig. 2).
commonly comprised of several swarm phases, The focal-migration rate in earthquake
which are quite often separated by even longer swarms depends on the driving force. In general
periods of inactivity, during which the swarm it differs for individual swarm areas and can be
activity shifts between different segments of the highly variable for a particular swarm. Extremely
fault. Large events tend to occur in later swarm high migration rates between 0.5 and 1.2 m/s
phases rather than in the beginning of the swarm. (approximately a common walking speed) are
Earthquake Swarms, Fig. 2 (continued) activity is earthquake whereas in the remaining swarms the strongest
depicted by the magnitude–time plot. Color coding from events occurred during some later swarm phase. The
dark blue to dark red corresponds to the course of seis- events were localized by applying the double-difference
micity in time. The north direction is marked by arrow. method (HypoDD program package) to data of WEBNET
Note that both W-Bohemia swarms, which happened on for the West Bohemia swarms and of SIL for Icelandic
the same fault segment, initiated on the bottom of the focal swarms (section “Local Networks: A Tool for Under-
cluster but their migration pattern is different. Further note standing Earthquake Swarms”)
that the 2003 Krı́suvı́k swarm began by the strongest
Earthquake Swarms 877
a 8 b 8
10 10
12 12
Depth [km]
Depth [km]
14 14
16 16
18 18 E
20 20
22 22
0 2 4 6 8 0 2 4 6 8 10 12 14
Southing [km] Easting [km]
c 8
10
12
Depth [km]
14
16
18
20
22
F M A M J J A S O N D J F M A
2007 2008
Earthquake Swarms, Fig. 3 A deep earthquake swarm and (b) the coordinate system is rotated 9 counter clock-
beneath Mt. Upptyppingar, ML 2.2, in 2007–2008 that wise, so the easting is along the strike of the focal belt. It is
accompanied magmatic intrusion into a dike in the lower obvious that the swarm is of planar character (the dip of
crust of Iceland’s Northern Volcanic Rift Zone (for more the fault plane is 40 ). The temporal development of the
detailed information of earthquake swarms induced by swarm activity and focal depth changes are depicted in
magmatic activity see section “Earthquake Swarms plot (c). The time-axis is scaled in months. It is apparent
Induced by Magmatic and Volcanic Activity”). The that during an 11-month period the swarm activity
swarm exhibited extremely high b-value of 2.1. The spa- migrated gradually upwards. The events are located by
tial distribution of foci is shown in two depth sections: (a) the double-difference method applied to the SIL data
view along the strike and (b) view across the strike. In (a)
observed in earthquake swarms that are closely intraplate swarms amounts to several hundreds of
connected with intrusion of dikes, such as in the meters per day. The fact that earthquake swarms
rift zone of Krafla in Iceland or in the Afar rift in mostly occur in volcanic or postvolcanic areas,
Ethiopia (Wright et al. 2012). Massin et al. (2013) which exhibit high fluid activity in crust, leads to
reports a fairly high migration rate of 1 km per a general assumption that swarming with the
day for the Yellowstone Lake swarm in seismicity migration are governed by three pos-
2008–2009. The focal-migration rate of intensive sible processes: (a) magma migration within the
878 Earthquake Swarms
Earth’s crust, (b) dike intrusion propagation (for Ordinary tectonic earthquakes are mostly
more details see section “Earthquake Swarms shear slips along a fault, i.e., the slip direction is
Induced by Magmatic and Volcanic Activity”), parallel to the fault plane, which is represented by
or by (c) pressurized hydrothermal fluids diffus- the double-couple (DC) force equivalent, while
ing within permeable fracture zone along the the non-double-couple components (ISO and
fault. Accordingly, Vidale and Shearer (2006) CLVD) are insignificant. However, earthquakes
suggest that the fluid pressure fluctuations or where crustal fluids act in the rupturing process
aseismic slip as plausible driving mechanisms need not be pure shear slips along a fault. If the
of the swarm-like sequences in southern Califor- fluid pressure is high enough, it can facilitate the
nia (section “Occurrence of Earthquake Swarms slip by reducing the normal stress on the fault.
in the World”). The pressurized fluids may also cause fracturing
by tensile cracks. Rupturing during which the
shear slip is coupled with the simultaneous open-
Source Mechanisms ing of fault results in a slip inclined to the fault
plane by the slope angle alpha (so-called tensile
Knowledge of source mechanisms is of crucial earthquake, for more details see Vavryčuk 2011).
importance for understanding of rupturing of Such earthquake is represented by a more com-
earthquake swarms and revealing the forces driv- plex force equivalent, which is manifested by
ing the swarm activity. The fact that earthquake significant non-DC components (both ISO and
swarms mostly occur in volcanic areas, ocean CLVD) of the source mechanism (for more infor-
ridges, or in areas with high activity of crustal mation refer to Julian et al., this issue).
fluids begs the question of possible differences The DC component includes information about
between forces which bring about and govern magnitude of the double-couple forces and focal
rupturing of swarm-like and ordinary earth- mechanism (DC constraint of source mechanism).
quakes. The source mechanism is an equivalent Focal mechanism is represented by three DC
of the body forces acting in the point source in an angles – strike, dip, and rake – which define geom-
elastic medium, hence it provides the first infor- etry of the fault plane and the slip angle along the
mation about rupturing process on the fault fault plane. It depends on local stress, i.e., the
(Dreger, this issue). Its unconstrained mathemat- orientation of the principal stress axes s1
ical representation is the seismic moment tensor (maximum stress), s2 (intermediate stress), and
which is in its completeness 9 dipole forces but s3 (minimum stress) relative to the fault plane.
only 6 of them are independent due to symmetry. There are three basic faulting types: strike-slip
The moment tensor is usually decomposed into (s1 and s3 are horizontal, s2 is vertical), normal
three force systems (components): double-couple (s2 and s3 are horizontal, s1 is vertical) and thrust
(DC) corresponding to shear slip on a fault, the (s1 and s2 are horizontal, s3 is vertical). Other
isotropic (ISO) representing volumetric changes focal mechanisms, the oblique normal and oblique
due to explosion or implosion, and compensated thrust, are a combination of strike slip with normal
linear-vector dipole (CLVD), which has no phys- and thrust type (for more details see Vavryčuk
ical interpretation (for more details see Vavryčuk 2, this issue). In general, all types of focal mecha-
1, this issue). For example, the CLVD model may nisms occur in earthquake swarms depending on
approximately represent an opening of cracks due the tectonic and local-stress conditions in a partic-
to the magma injection. But in the CLVD model ular area. Nevertheless, when analyzing earth-
it is implicitly assumed that material flows rap- quake swarms in the world, it appears that
idly into the opening crack from the nearby area oblique normal strike-slip faulting prevails.
so that no net volume change is involved. How- A characteristic feature of earthquake swarm is
ever, this flow may not occur very rapidly in fairly significant spatial variation of the focal
viscous magma and on the short timescale mechanisms in a particular swarm area as well as
(Minson et al. 2007). during individual swarms; very often each focal
Earthquake Swarms 879
Earthquake Swarms, Fig. 4 Distribution of the earth- mechanism, which is depicted as the second from the
quake swarms in the main focal area of W-Bohemia/ left, prevailed in the second phase of the 1997 swarm. It
Vogtland and typical source mechanisms in the individual contains significant non-DC components (both ISO and
swarms. The section along the strike is depicted on the CLVD), the pattern indicates tensile earthquakes. The
right, the cross-section is on the left. It is obvious that the events are located by the same way as in Fig. 2. The source
focal area consists of a few shorter faults. All but one mechanisms are retrieved in the full moment-tensor
source mechanisms are pure shears; the DC angles fit the description by inversion of amplitudes of the direct P-
geometry of the corresponding fault segments. The and S-waves from stations of the WEBNET network
cluster is characterized by its own focal mecha- The non-DC earthquakes have been identified
nism family (see Fig. 4). It points out a rather at a number of hydrothermal and magmatic cen-
complex system of differently oriented shorter ters, e.g., in Long Valley Caldera (Foulger
faults with stress heterogeneities at small spatial et al. 2004) or Hengill volcanic complex in
scale, which is very likely one of the prerequisites SW-Iceland (Julian et al. 1997), as well as in the
for the strain-energy release in the form of earth- intraplate area of W-Bohemia/Vogtland (Horálek
quake swarms. et al. 2002). However, there are only few papers
Existence of tensile swarm earthquakes is still dealing with a detailed analysis of non-DC earth-
a question at issue. The percentage of the non-DC quakes in individual earthquake swarm areas or in
part in the moment tensor of tensile events is a particular swarm. So an existence of tensile
fairly sensitive to the deviation of the slip vector, swarm earthquakes and their possible connection
for example, a deviation of a = 5 out of the fault to magma intrusion or to fluid transport is still
plane generates up to 20 % of the non-DC com- unclear. This can be demonstrated by the follow-
ponents (Fig. 2 of Vavryčuk 2001). However, ing examples: (1) The Long Valley Caldera,
even large non-DC components can be an artifact which is still volcanically active and hosts an
of the inconsistency of data, which are inverted extensive hydrothermal system. The source
for moment tensor. Typically, insufficient distri- mechanisms of swarm events from there were
bution of sensors around the focus gives rise to investigated by several teams of authors, how-
apparently large fictitious non-DC components, ever, the results appear ambiguous. For example,
particularly of CLVD. Further causes can be Stroujkova and Malin (2002) analyzed the
inaccurate or noisy data used for the moment- mid-1997 swarm and arrived to the conclusion
tensor inversion, inaccurate localization of the that majority of the resulting mechanisms appear
hypocenter, or a poor seismic-velocity model of to be of the DC nature, while Foulger et al. (2004)
the area under study. reported the occurrence of significantly non-DC
880 Earthquake Swarms
source mechanisms, inconsistent with simple plate motion loading and aftershocks are a result
shear faulting, for selected events of this swarm. of stress redistribution due to the main shock. Yet
Afterwards, Templeton and Dreger (2006) there is no common knowledge how it works in
showed that fluid-influenced earthquakes are the case of earthquake swarms. In the case of
quite rare in the Long Valley volcanic region earthquake swarms it appears useful to distin-
and that majority of local earthquakes are best guish the triggering mechanisms that lead to the
characterized by a simple DC source model. initiation of the activity and driving forces that
(2) W-Bohemia/Vogtland, a European intraplate keep the activity running.
area with extinct Quaternary volcanism manifests Compared to tectonic earthquakes at the plate
present geodynamic unrest by persistent swarm boundaries, the intraplate earthquake swarm
seismicity and degassing of big amount of CO2 areas miss the plate motion that could periodi-
(of about 500 m3 per hour) of the upper-mantle cally bring the fault to failure. Recent studies
origin. The moment tensors of earthquake swarms have tried to understand the possible triggering
in the 1997, 2000 and 2008 events signify fairly mechanisms by probing/imaging anomalous
different amounts of non-DC components of crustal structures within seismic swarms linked
respective source mechanisms. It appears that to fluids (e.g., Lin and Shearer 2009; Kato
swarm earthquakes, which occurred on the opti- et al. 2010). It is assumed that a fluid overpressure
mally oriented faults relative to the local tectonic builds up and acts as a triggering mechanism of
stress, are pure shears. It is the case of both 2000 earthquake swarms. The subsequent swarm activ-
and 2008 swarms in which the faulting was ity is then probably driven both by the fluid flow
governed merely by local tectonic stress along the fault zone and stress transfer from pre-
(Horálek and Šı́lený 2013). A question is the vious earthquakes. This idea is based on charac-
1997 swarm in which two types of source mech- teristic features of earthquake swarms that help us
anisms occurred (Horálek et al. 2002): pure shears to disclose their background mechanisms.
in the first swarm phase but significantly non-DC One of the most striking features of earth-
mechanisms in the second swarm phase (thrust- quake swarms is their focal migration, which is
mechanism type colored dark blue in Fig. 4), often attributed to the fluid flow along a
which may indicate the occurrence of tensile quasipermeable fault zone. The most frequent
earthquakes on a less favorably oriented fault model of this driving force is based on diffusive
segment. However, one cannot rule out that the propagation of the pore-pressure front from a
non-DC components are due to the small number fluid source within a poroelastic medium p defined
ffiffiffiffiffiffiffiffiffiffi
of stations which were used in the moment-tensor in a simplified way by the formula r ¼ 4pDt
inversion of the 1997-events. (3) Minson (Shapiro et al. 1997). Here, r and t are the dis-
et al. (2007) tried to find non-double-couple tance and time interval of the foci from the first
mechanisms of selected earthquakes in a strong event, which corresponds to the source of fluids,
Miyakejima volcanic swarm in Japan during the and D is the hydraulic diffusivity. A different
year 2000 and concluded that the source mecha- fluid-related model employs the preferential
nisms of the most analyzed earthquakes appear to upwards migration of the hypocenters, which is
be related to inflation of the dike yet their source attributed to the buoyancy of fluids that results in
mechanism solutions are ambiguous. hydraulic fracturing of the rock. This allows for
estimation of overpressure responsible for the
swarm triggering which Hainzl et al. (2012)
Possible Triggering Mechanisms and obtained an overpressure of 20 MPa for the
Driving Forces W-Bohemian swarm in 2008. Daniel
et al. (2011) analyzed the effective stress neces-
As far as the mainshock-aftershock sequences at sary to drive the 2003 swarm in Ubaye, France,
plate boundaries are concerned, there is a general and obtained a maximum excess pore pressure of
understanding that the main shock is triggered by 8 MPa.
Earthquake Swarms 881
system in 1994–1998, in Iceland (e.g., Peder- the caldera fault in response to more than 60 m of
sen et al. 2007, and references therein). These caldera subsidence.
periods may last for days, to months, to even a Ever since the discovery of mid-oceanic plate
few years. The largest swarm earthquakes may boundaries, it has been recognized that there is a
be between magnitude M = 4 and 5. fundamental difference between the seismic
2. Intrusion of dikes or sills into the crust. If activity of transform boundaries and that of diver-
increasing magma pressure fractures the sur- gent boundaries, i.e., spreading centers. The
rounding crust a dike propagates away from transform-zone seismicity is characterized by
the magma source. The dike tip changes the mainshock-aftershock earthquake sequences
local stress and thus induces fracturing and having b-values of 0.8–1.0 and strike-slip
earthquakes. The rate of propagation depends faulting mechanisms, with maximum magnitudes
on the regional stress, the width of the dike, M 6.5–7.0. The spreading-center earthquakes
and viscosity of the magma. Large variation in tend to occur in swarms, fault plane solutions
the waveforms is one of the characteristics of show normal faulting, b-values are high, as high
earthquake swarms of this origin, especially if as 2.3, and the largest earthquakes rarely exceed
the fracture front approaches the surface. magnitude M = 6.0 (e.g., Einarsson 1986).
Propagating swarms of this type have been The association here with an area of volcanic
documented in the Afar rift in Ethiopia activity is obvious, but even with all of these
(Wright et al. 2012), in the rift zone of Krafla, mentioned studies the causal relationship is still
and Bárðarbunga volcanic system in Iceland, debated. Experience from divergent boundaries
where an intense seismic crisis has accompa- on land, in Iceland and East Africa, appears to
nied the large lava eruption that began in indicate that both explanations apply. Dike intru-
August 2014 (Sigmundsson et al. 2015). In sions associated with rifting do produce earth-
all these cases the dikes are vertical and prop- quake swarms (Wright et al. 2012), but swarms
agate horizontally away from the magma that do not seem to be associated with magmatic
source. Sill propagation accompanied by activity are also common.
earthquakes preceded the eruptions of
Eyjafjallajökull in 2010.
3. Extrusion of lava domes at volcanoes, and its Conclusions
subsurface equivalent, the intrusion of
cryptodomes, is frequently accompanied by Earthquake swarms represent a specific type of
earthquakes. Domes may have steady or epi- seismicity without a mainshock, usually with a
sodic growth, with emplacement times ranging gradual increase of seismic activity, clustering in
from a few hours to many decades. The accom- time and space, and migration of their hypocen-
panying earthquakes therefore show a swarm- ters. In contrast to mainshock-aftershock
like behavior on various time scales. They typ- sequences, earthquake swarms occur in various
ically have the characteristics of B-type volca- tectonic environments including both purely vol-
nic earthquakes (Chouet 1996), which have canic areas and areas with no active volcanism.
indistinct P-waves, weak or no S-waves, and a Their recognition is still in progress and thanks to
low-frequency coda. more detailed observations it turns out that they
are more common than thought before. Their
Several cases are known of earthquake swarm diverse behavior is expressed in multiple models
activity accompanying caldera collapses. The lat- suggesting their origin. Besides the models based
est case is that of the slow collapse of the on fluid injection and migration along faults,
Bárðarbunga caldera in 2014–2015 following recent hypotheses suggest aseismic loading by
dike propagation and a large lava eruption. creeping faults. Some recent observations show
More than 70 M 5.0 earthquakes occurred on that earthquake swarms can share the same faults
Earthquake Swarms 883
with other types of seismicity. In terms of the mostly shear-slips along a fault just as ordinary
wide range of their behavior it appears possible tectonic earthquakes. In special cases of magma-
that volcanic swarms and mainshock-aftershock or fluid-influenced earthquakes, more complex
sequences represent end members of a continu- tensile rupturing might occur. However, the
ous variety of seismic activity that is governed by existence of tensile swarm earthquakes and
a combination of two basic types of external their possible connection to movements of
loading: fluid injections on one end and tectonic magma or fluids is still an open question. Also,
loading on the other end. triggering mechanisms and driving forces of
earthquake swarms remain unclear. These
mechanisms are mainly related to activity of
E
Summary pressurized hydrothermal fluids, magma injec-
tion, or dike intrusion, and also to aseismic load-
Earthquake swarms are sequences of numerous ing by creeping faults. Accordingly, numerous
events closely clustered in space and time and do models trying to explain the seismic-energy
not have a single dominant mainshock. A few of release in the form of earthquake swarms have
the largest events in a swarm reach similar mag- been proposed.
nitudes and usually occur throughout the course
of the earthquake sequence. These attributes dif- Acknowledgments We are grateful to both anonymous
ferentiate earthquake swarms from ordinary reviewers for their valuable suggestions, which helped us
mainshock-aftershock sequences. Earthquake to improve the paper substantially. Most of the work was
swarms occur worldwide, in diverse geological accomplished within the Grant Project P210–12–2336 of
the Grant Agency of the Czech Republic “Earthquake
units. The swarms typically accompany volcanic swarms and their triggering mechanisms and driving
activity at margins of the tectonic plate but also forces in diverse tectonic environments (Bohemian Mas-
occur in intracontinental areas where strain from sif, Mid-Atlantic ridge, French Alps).”
tectonic-plate movement is small, but large het-
erogeneities of stress and fault strength may
exist. It implies that earthquake swarms gener- Cross-References
ally originate from volcanic and tectonic epi-
sodes. The swarm earthquakes are usually of ▶ Earthquake Mechanisms and Stress Field
magnitudes M <5, with some exceptional events ▶ Earthquake Mechanism Description and
reaching or exceeding M 6.0. The b-value of Inversion
the frequency–magnitude distribution log N = ▶ Moment Tensors: Decomposition and
a bM of both tectonic and volcanic swarms Visualization
varies mostly between 0.7 and 1.5 but many ▶ Non-Double-Couple Earthquakes
earthquake swarms show b 1 which is a typical
b-value for mainshock-aftershock sequences.
References
Depths of swarm earthquakes range approxi-
mately between 2 and 20 km with the majority Cappa F, Rutqvist J, Yamamoto K (2009) Modeling
of swarm events located around 10 km and crustal deformation and rupture processes related to
shallower. Deeper earthquake swarms also upwelling of deep CO2-rich fluids during the
1965–1967 Matsushiro earthquake swarm in Japan.
exist but rarely occur. Earthquake swarms
J Geophys Res 114, B10304. doi:10.1029/
exhibit usually pronounced migration of the 2009JB006398
foci, particularly the swarms that are closely Chouet B (1996) Long-period volcano seismicity: its
connected with intrusion of dikes. The source sources and use in eruption forecasting. Nature
380:309–316
mechanisms, which are an equivalent of the
Daniel G, Prono E, Renard F, Thouvenot F, Hainzl S,
body forces acting in the point-source approxi- Marsan D, Helmstetter A, Traversa P, Got JL,
mation, signify that the swarm earthquakes are Jenatton L, Guiguet R (2011) Changes in effective
884 Earthquake Swarms
stress during the 2003–2004 Ubaye seismic swarm, Hurst AW, McGinty PJ (1999) Earthquake swarms to the
France. J Geophys Res 116, B01309. doi:10.1029/ west of Mt Ruapehu preceding its 1995 eruption.
2010JB007551 J Volcanol Geotherm Res 90:19–28
Einarsson P (1986) Seismicity along the eastern margin of Ibáñez JM, De Angelis S, Dı́az-Moreno A,
the North American Plate. In: Vogt PR, Tucholke BE Hernández P, Alguacil G, Posada A, Pérez N
(eds) The Western North Atlantic region, vol M, The (2012) Insights into the 2011–2012 submarine erup-
Geology of North America. Geological Society of tion off the coast of El Hierro (Canary Islands,
America, Boulder, pp 99–116 Spain) from statistical analyses of earthquake activ-
Farrel J, Husen S, Smith RB (2009) Earthquake swarm and ity. Geophys J Int 191:659–670. doi:10.1111/j.1365-
b-value characterization of the Yellowstone volcano- 6X.2012.05629.x
tectonic system. J Volcanol Geotherm Res Ibs-von Seht M, Plenefisch T, Klinge K (2008) Earthquake
188(1–3):260–276 swarms in continental rifts – a comparison of selected
Fischer T, Horálek J (2005) Slip-generated patterns of cases in America, Africa and Europe. Tectonophysics
swarm microearthquakes from West Bohemia/ 452:66–77
Vogtland (central Europe): evidence of their triggering Jenatton L, Guiguet R, Thouvenot F, Daix N (2007) The
mechanism? J Geophys Res 110:B05S21. 16,000-event 2003–2004 earthquake swarm in Ubaye
doi:10.1029/2004JB003363 (French Alps). J Geophys Res 112, B11304.
Fischer T, Horálek J, Hrubcová P, Vavryčuk V, Bräuer K, doi:10.1029/2006JB004878
Kämpf H (2014) Intra-continental earthquake swarms Julian BR, Miller AD, Foulgler GR (1997) Non-double-
in West-Bohemia and Vogtland: a review. couple earthquake mechanisms at the Hengill-
Tectonophysics 611:1–27. doi:10.1016/j. Grensdalur volcanic complex, Southwest Iceland.
tecto.2013.11.001 Geophys Res Lett 24:743–746
Foulger GR, Julian BR, Hill DP, Pitt AM, Malin P, Shalev Kato A, Sakai S, Iidaka T, Iwasaki T, Hirata N (2010)
E (2004) Non-double-couple microearthquakes at Non-volcanic seismic swarms triggered by circulating
Long Valley caldera, California, provide evidence for fluids and pressure fluctuations above a solidified dio-
hydraulic fracturing. J Volcanol Geotherm Res rite intrusion. Geophys Res Lett 37, L15302.
132:45–71 doi:10.1029/2010GL043887
Godano M, Larroque C, Bertrand E, Courboulex F, Kraft T, Wassermann J, Schmedes E, Igel H (2006) Mete-
Deschamps A, Salichon J, Blaud-Guerry C, orological triggering of earthquake swarms at
Fourteau L, Charléty J, Deshayes P (2013) The Mt. Hochstaufen, SE-Germany. Tectonophysics
October–November 2010 earthquake swarm near 424:245–258
Sampeyre (Piedmont region, Italy): a complex Lay T, Wallace TC (1995) Modern global seismology.
multicluster sequence. Tectonophysics 608:97–111. Academic Press, Inc. San Diego
doi:10.1016/j.tecto.2013.10.010 Lin G, Shearer PM (2009) Evidence for waterfilled cracks
Hainzl S (2004) Seismicity pattern of earthquake swarms in earthquake source regions. Geophys Res Lett
due to fluid intrusion and stress triggering. Geophys 36, L17315. doi:10.1029/2009GL039098
J Int 159:1090–1096. doi:10.1111/j.1365- Massin F, Farrell J, Smith RB (2013) Repeating earth-
246X.2004.02463.x quakes in the Yellowstone volcanic field: implications
Hainzl S, Fischer T, Dahm T (2012) Seismicity-based for rupture dynamics, ground deformation, and migra-
estimation of the driving fluid pressure in the case of tion in earthquake swarms. J Volcanol Geotherm Res
swarm activity in Western Bohemia. Geophys J Int 257:159–173
191:271–281. doi:10.1111/j.1365-246X.2012.05610.x Minson SE, Dreger DS, B€ urgmann R, Kanamori H,
Hill DP (1977) A model for earthquake swarms. Larson KM (2007) Seismically and geodetically deter-
J Geophys Res 82(8):1347–1352 mined nondouble-couple source mechanisms from the
Hill DP, Langbein JO, Prejean S (2003) Relations between 2000 Miyakejima volcanic earthquake swarm.
seismicity and deformation during unrest in Long Val- J Geophys Res 112, B10308. doi:10.1029/
ley Caldera, California, from 1995 through 1999. 2006JB004847
J Volcanol Geotherm Res 127(3–4):175–193 Mogi K (1963) Some discussion on aftershocks, fore-
Holtkamp SG, Brudzinski MR (2011) Earthquake swarms shocks and earthquake swarms – the fracture of semi-
in circum-Pacific subduction zones. Earth Planet Sci infinite body caused by an inner stress origin and its
Lett 305:215–225 relation to the earthquake phenomena. Bull Earth-
Horálek J, Šı́lený J (2013) Source mechanisms of the 2000 quake Res Inst 41:615–658
earthquake swarm in the West Bohemia/Vogtland Ogata Y (1988) Statistical models for earthquake occur-
region (Central Europe). Geophys J Int 194:979–999. rences and residual analysis for point processes. J Am
doi:10.1093/gji/ggt295 Stat Assoc 83:9–27
Horálek J, Šı́lený J, Fischer T (2002) Moment tensors of Pedersen R, Sigmundsson F, Einarsson P (2007) Control-
the January 1997 earthquake swarm in West Bohemia ling factors on earthquake swarms associated with
(Czech Republic): double-couple vs. non-double- magmatic intrusions; constraints from Iceland.
couple events. Tectonophysics 356:65–85 J Volcanol Geotherm Res 162:73–80
Earthquakes and Their Socio-economic Consequences 885
103
Alaska 1964
102
101
5 6 7 8 9
earthquake Magnitude, Mw
consequences of two case studies, the L’Aquila year. A direct “hit” of a MW 7 earthquake in
earthquake of 2009 in Italy and the more recent Tehran will cause more than a million deaths
Great East Japan earthquake and tsunami of (Bilham 2009). As shown in Fig. 2, the number
2011, will be examined in detail in section of megacities in this figure depicts cities with
“Two Events in the Twenty-First Century.” The over a million people are aplenty and the seismic
main objective of this section is to not only com- risk of some of these cities are high. For many
pare and contrast these events but also to draw cases traversed by active faults, this is due to what
important conclusions on the factors affecting Jackson (2006) calls it the fatal attraction. For
recovery and therefore long-term socio-eco- many centuries, an uneasy accommodation was
nomic impacts on societies affected by earth- reached between human needs and the
quakes. The intention is to draw this chapter to a earthquake-controlled landscape, sometimes bril-
close by looking to the future and examining liantly exploited by local hydrological engineer-
what as a community of seismologists, engineers, ing, as in Iran. In the past occasional moderate
social scientist, public health practitioners, econ- earthquakes would occur, killing a proportion of
omists, and politicians can be done to eliminate relatively small population. However, many
the prospect of a million-death earthquake once-small rural communities have now grown
(Musson 2012)? into towns, cities, or megacities while retaining
Since 2000, more than 1,000,000 people have their vulnerability through poor building
died from earthquakes and 2010 has seen a single standards.
event killing more than 200,000 people (Haiti As witnessed in recent years, earthquakes that
2010). Bilham suggested in the conclusions of occur in these places now kill many more than
his Mallet Milne lecture in 2009 that the global they did in the past and there are other cities like
death toll from moderate earthquakes near human capitals as well, like Kathmandu in Nepal which
settlements is likely to average 8,000–10,000 a is highly at risk. Extreme catastrophes have been
year in the next several decades and catastrophic rare only because the exposure of modern mega-
events may bring this up to 50,000 deaths per cities to earthquake hazards has been relatively
Earthquakes and Their Socio-economic Consequences 887
Earthquakes and Their Socio-economic Consequences, Fig. 2 Population map of Asia showing the megacities at
risk (Credit: GEM)
short (approx. 50 years); an increase in the num- loss, and fatalities. Apart from the two earth-
ber of such catastrophes now seems to be inevi- quakes in Japan (Kobe 1995; Tohoku 2011),
table., e.g., Tehran, as shown in Fig. 3. there is an inverse relationship between property
The most used parameter to measure the social losses and casualties – high losses and low fatal-
impact or severity of an earthquake is fatalities, ities in developed counties and low insured losses
but in terms of social consequences, it is the and high fatalities in undeveloped or developing
survivors that need the humanitarian assistance countries. The Great East Japan earthquake and
in the short term and committed recovery and tsunami as shown is the single most costly event
reconstruction plans to rebuild shattered lives in since 1980, reporting overall losses of over US
the long term. After the initial relief phase, the $210 billion. This earthquake was exceptional in
“fate” of those affected is the responsibility of many ways, not least because of the high fatality
those in governance, e.g., the establishment of the rate in a highly developed exceptionally well
Canterbury Recovery Authority (CERA) after the prepared country. It is important to note here
Canterbury Sequence in 2010–2011. Possible that with a better understanding of parameters
indicators to help measure and monitor long- like vulnerability and exposure, losses and casu-
term social and economic recovery are alties are now only loosely associated to physical
highlighted in the section “Recording Socio-eco- variables such as magnitude.
nomic Consequences.” More assessments on the economic and
Table 1 shows how earthquake severity is insured losses can be found in Chapter 29 of this
usually reported in terms of total losses, insured volume.
888 Earthquakes and Their Socio-economic Consequences
Earthquakes and Their Socio-economic Consequences, Fig. 3 The city of Tehran with the active North Tehran
Fault as a backdrop (Credit GEM)
Earthquakes are usually reported by insured transportation disruption (Boarnet 1995). In the
losses and casualties, though the effects of an future, disruption of electronic networks may
earthquake can be generally classified into the also be crucial, as there is a huge dependency of
following categories: businesses on ICT.
After the February 2011 earthquake in Christ-
1. Physical impacts church, New Zealand, the majority of businesses
2. Social impacts (the impact on people) in the center business district (CBD) had to relo-
3. Economic impacts (the impact on the wealth cate. In the week after the quake, people at the
of an area) Canterbury Earthquake Recovery Authority
4. Environmental impacts (the impact on the (CERA) described how everyone was on the
landscape) phone trying to find a place and negotiate a
lease. In some cases this involved multiple loca-
These can be immediate or long-term conse- tions. Inland Revenue staff, for example, are
quences and most are entirely interlinked. For scattered in 300 different offices. Some firms
example, damage to a transportation system due have moved to residential or semi-industrial
to ground shaking in an earthquake can take years areas and others have relocated to other cities.
to repair and therefore contribute significantly to Tourism declined and tertiary education was
the business and economic losses in the area under pressure. A City Council survey suggests
affected. In a study on transportation disruptions that over 60 % of businesses that moved out
from the Northridge earthquake in 1995, it was wanted to return to the city and the CBD was
reported that 43 % of the businesses surveyed already under pressure from a number of periph-
attributed their losses at least partially to eral shopping malls, though 51,000 people were
Earthquakes and Their Socio-economic Consequences 889
Earthquakes and Their Socio-economic Consequences, Table 1 Table of the most costly earthquakes since 1980
(MunichRe)
Overall Insured
losses losses Fatalities
US$ m, original
Date Event Affected area values
11.3.2011 Earthquake, Japan: Honshu, Aomori, Tohoku; Miyagi, 210,000 40,000 15,840
tsunami Sendai; Fukushima, Mito; Ibaraki; Tochigi,
Utsunomiya
17.1.1994 Earthquake USA: CA, Northridge, Los Angeles, San 44,000 15,300 61
Fernando Valley, Ventura, Orange
22.2.2011 Earthquake New Zealand: South Island, Canterbury, 16,000 13,000 185
E
Christchurch, Lyttelton
27.2.2010 Earthquake, Chile: Bió Bió, Concepción, Talcahuano, 30,000 8,000 520
tsunami Coronel, Dichato, Chillán; Del Maule, Talca,
Curicó
4.9.2010 Earthquake New Zealand: Canterbury, Christchurch, 6,500 5,000
Avonside, Omihi, Timaru, Kaiapoi, Lyttelton
17.1.1995 Earthquake Japan: Hyogo, Kobe, Osaka, Kyoto 100,000 3,000 6,430
29.5./3.6.2012 Earthquake Emilia-Romagna, San Felice sul Panaro, 16,000 1,600 18
Cavezzo, Rovereto di Novi, Carpi, Concordia,
Bologna
26.12.2004 Earthquake, Sri Lanka, Indonesia, Thailand, India, 11,200 1,000 220,000
tsunami Bangladesh, Myanmar, Maldives, Malaysia
17.10.1989 Earthquake USA: CA, Loma Prieta, Santa Cruz, San 10,000 960 68
Francisco, Oakland Berkeley, Silicon Valley
13.6.2011 Earthquake New Zealand: Canterbury, Christchurch, 2,000 800 1
Lyttelton
# 2013 M€unchener R€
uckversicherungs-Gesellschaft, Geo Risks Research, NatCatSERVICE. As at: March 2013
still working in the center. This real-life scenario choosing not to reinvest in the city center or
is very reminiscent of Kates and Pijawka (1977) Christchurch; the scale of the demolition and
model of recovery from earthquakes. the time taken to rebuild and owners of tall build-
The main actors helping businesses after the ings that require demolition are holding the city
earthquakes of 2010–2011 were the Chambers to ransom for months and months. The last cor-
of Commerce, Canterbury Development Corpo- dons to the CBD were lifted in June 30, 2013,
ration, Christchurch City Council and other 859 days after the earthquake. Whether Christ-
local authorities, central government, and church will recover financially from the earth-
CERA. The City Council relaxed regulations quakes remains to be seen.
allowing businesses to relocate to residential In stark contrast to pure financial costs are
areas in the west and to new retail hubs. This social and psychological consequences that are
however only aggravated the problem of hard to quantify but the effects of which will be
attracting business back to the center. In Christ- felt for a long time after the disaster. The inunda-
church it is estimated that NZ$20 billion will be tion of the tsunami in schools in the East Coast of
invested in the city and 15,000 jobs created in Japan in 2011 and the collapse of school build-
construction (Fig. 4). ings after the Wenchuan earthquake in China in
The overall vision for the city was to give 2008 killed an entire generation of school chil-
investors confidence. However, some of the dren (Fig. 5).
risks and problems the city still faces include a Effects are often ordered as primary and sec-
flight of capital in which property owners are ondary impacts. Primary effects occur as a direct
890 Earthquakes and Their Socio-economic Consequences
result of the ground shaking, e.g., buildings col- living more densely. Rapid development has
lapsing. Secondary effects occur as a result of the meant that even with the strongest of political
primary effects, e.g., tsunamis or fires due to and economic will, the pace of regeneration of
ruptured gas pipelines. The table below summa- building stock is outrunning the return periods of
rizes the range of possible consequences from earthquakes (Bilham 2009). Perhaps the only way
earthquakes (Table 2). to engage with earthquake-prone communities in
With urbanization, the world’s population is mitigation is through social and fiscal means,
increasing in hazardous areas and people are rather than purely physical interventions.
Earthquakes and Their Socio-economic Consequences 891
$284
inflation-adjusted
normalized
normalized + 1% mitigation
Total adjusted losses (billion $2005 USD)
8 $105
normalized + 2% mitigation
2
19
19
19
19
19
19
19
19
19
19
20
00
10
20
30
40
50
60
70
80
90
00
–1
–1
–1
–1
–1
–1
–1
–1
–1
–1
–2
90
91
92
93
94
95
96
97
98
99
00
9
5
Decade
Earthquakes and Their Socio-economic Consequences, Fig. 6 Graph shows normalized damages in the USA
1900–2005 from Vranes and Pielke (2009) where the 1906 was found to be the most costly event in US history
mitigation strategies for other areas. The need US$328 billion, though some would still argue
for accurate and comparable numbers cannot be about the definition of “cost.” Clear standardized
underestimated as they help a community of definitions would help to reduce this ambiguity
seismologists and engineers communicate risk and inconsistency. This question could be
to economists and politicians. What are the true answered partially by distinguishing between
costs of earthquakes? Socio-economic figures direct and indirect costs and short- and long-
are meaningless unless placed in context and term costs. Some studies report the number of
normalized in terms of population size and households affected, while others report the
demography. For example, what was the abso- number of people. The reporting of social data
lute number of people killed, as a percentage of is also dependent on the experience and training
local population? What is the cost of the earth- of interviewers and surveyors. Often samples are
quake in today’s terms? For example, a study by poorly constructed, interviewing is inconsistent,
Vranes and Pielke (2009) on twentieth-century and questionnaire forms are badly designed
US earthquakes found that if normalized to 2005 (Fig. 6).
dollars by adjusting for inflation, increased in Knowing when essential services and liveli-
wealth, and changes in population, the 1906 hood are restored would capture some key
San Francisco earthquake would most likely to aspects of recovery. But another issue with
have been the most costly US earthquake, at existing socio-economic data is that they lack
Earthquakes and Their Socio-economic Consequences 893
date stamping, for example, it is insufficient to such as liquefaction, landslide, tsunami, or fire
know the number of homes without electricity or following, or because of abandonment/relocation
water supply or the number of homeless or unem- of their neighborhood, village, or town (due to
ployed; one also needs to know the time period occurrence or high risk of occurrence of land-
of loss. slide, mudslide, rockfalls, liquefaction, fire fol-
The indicator “number of people in home- lowing, tsunami, risk of dam collapse, nuclear
less,” for example, is also time dependent. The disaster, other hazardous substance release,
term “homeless” can apply to any of the etc.), even if their homes were not destroyed.
following: An additional (and important) indicator,
related to recovery, is the speed of reconstruction
E
• People taking shelter in an organized camp or of the dwellings that were destroyed. Recovery
in an organized shelter building (e.g., school, can be expressed as a percentage of destroyed
sports hall, etc.) – these are facilities providing houses that have been reconstructed by a given
short-term shelter, having communal kitchens time after the event, e.g., 1 or 2 years after the
and bathrooms. event (high-consequence events take several
• Number of people taking or given the option years until full recovery is achieved even in the
of alternative accommodation (with or with- most developed regions of the world).
out public subsidy), e.g., rent another home, An example of the recovery of the destroyed
rent a room in a hotel or hostel, and be hosted dwellings during the first 2 years after the Sep-
by relatives or friends or volunteers (usually tember 7, 1999, Athens earthquake is given
after the first week following the event). below:
• Number of people provided temporary hous-
ing units (usually more than 1 month after the • On the first anniversary of the earthquake, the
event) these are usually prefabricated Greek Ministry of Environment and Public
homes with minimal facilities intended for Works released the following information:
occupation no longer than 1–2 years. more than 50 % of the red-tagged housing
units (i.e., housing units that are not habitable
In the simplest case, there should be two cat- or usable in its current state) are being
egories for the homelessness data: “number of reconstructed (2,503 out of 4,220).
homeless in the immediate aftermath” and “num- • On the second anniversary of the earthquake
ber of homeless for a significant duration of the Greek Ministry of Environment and Public
time.” This is a crucial differentiation. If not Works released the following information: out
directly available these two categories of home- of a total of 4,682 red-tagged buildings, 70 %
lessness can be derived from damage statistics. had received reconstruction funding, and
Long-term homelessness would be related to the 400 (8.5 %) were part of a government-funded
estimated population in destroyed houses or col- regeneration plan in the worst affected and
lapsed and red-tagged houses; homelessness in poorest municipality of Ano Liosia and in
the immediate aftermath would be related to the Nea Philadelphia municipality (expected to
estimated population in houses with serious be completed 42 months after the occurrence
structural damage that were deemed unsafe for of the earthquake).
immediate occupancy after post-earthquake dam-
age/safety assessments, e.g., the number of Records of socio-economic consequences
yellow-tagged dwellings – depending on the help track progress of recovery through time
damage reporting style in each event, country, that can be used as an accounting measure for
and region. Long-term homelessness (usually donors and provide an insight into the possible
more than 1 month) would imply people who issues contributing to delays in recovery of the
have lost their homes due to ground shaking economy (lack of housing, infrastructure, educa-
damage or due to secondary and induced hazards, tional facilities, etc.).
894 Earthquakes and Their Socio-economic Consequences
Earthquakes and Their Socio-economic Consequences, Fig. 7 Graph comparing the progress made in reduction
of deaths in earthquakes and other public health issues (up to 2000)
likely capital required to rebuild and recover is since local economies were destroyed and
not readily available after the event. Broadly young people will leave as a result (Masateru
speaking one perhaps can venture to summarize 2011).
the likely socio-economic consequences of earth- The preexisting socio-economic trends in a
quakes for different countries in the following region are likely to be amplified by a major disas-
way (Table 4). ter. Places that were shrinking before the disaster
Though perhaps the overall level of wealth of may be pushed into severe decline. Industries that
a country is not as fundamentally diagnostic of its were struggling to survive may fold and young
earthquake vulnerability as is its differential people may move elsewhere in search of work.
between rich and poor. Conversely places that were booming before the
The recent earthquake and tsunami in East disaster, or were the economic conditions were
Japan has also highlighted the need to anticipate propitious, may experience a spurt in growth as
the risk to the future population in earthquake- opportunities open up and money flows into the
prone regions. About half of victims of this earth- area, e.g., Kobe, Japan. Much of the Tohoku
quake were elderly people of 65 years old or region, for example, was in economic and demo-
more, unable to respond to warnings and evacu- graphic decline before 2011 and, despite the huge
ate in time. Based on the statistics presented by investment in the region, the earthquake has
Statistical Information Institute for Consulting exacerbated the situation.
and Analysis, depopulation has extended over The risk of vulnerability and the impacts of
the whole region, excluding large metropolitan disaster are disproportionately borne by those
areas such as Sendai. Forty percent of all munic- who are already socioeconomically and physi-
ipalities will experience a population decrease of cally disadvantaged and who have fewer
20 % or more, accentuating the pre-disaster con- resources to enable them to “bounce back” to
dition. In coastal areas of the Pacific Ocean, pop- some measure of normality. These include for-
ulation is anticipated to decrease even further eign nationals, the very young and very old, those
896 Earthquakes and Their Socio-economic Consequences
Earthquakes and Their Socio-economic Consequences, Table 3 List of 10 significant earthquakes in the last
20 years and their socio-economic consequences (high HDI, 0.758; medium HDI, 0.64; low HDI, 0.466)
Direct
economic “Cost”/
losses fatality
(million (million
Earthquake Fatalities US$)a US$) HDI Lessons learned (from reconnaissance)
Northridge, 61 44,000 721 0.937 Severe damage to hitherto deemed safe steel-
Mw 6.7 framed buildings
(1994)
Kobe, Mw 6.9 5,502 96,000 17.4 0.912 Revised building code and nationwide
(1995) (fires) retrofit Sch. of traditional timber residences
Kocaeli, Mw 17,118 11,400 0.67 0.722 Building control called into question; new
7.4 (1999) RC buildings performed worse than older
traditional ones
Athens, Mw 143 3,395 24.1 0.86 45 % of fatalities were in industrial
5.9 (1999) buildings
Bhuj, Mw 7.7 13,800 4,405 0.32 0.554 Site effects were significant on extremely
(2001) weak masonry
Bam, Mw 6.6 26,271 500 0.019 0.742 Extremely vulnerable dry mud brick
(2003) buildings. 26 % of Bam died
Wenchuan, >73,000 141,000 1.93 0.699 Bad siting of towns and schools killed many
Mw 7.6
(2008)
Haiti, Mw 7.0 >230,000 4,245 0.018 0.416 Extremely vulnerable substandard
(2010) buildings
Christchurch, 185 16,000 86.5 0.919 Collapses of two mid-rise RC buildings
Mw 6.1 accounted for 70 % of the deaths. People
(2011) were killed on the streets by out-of-plane
failure of masonry debris
Great East >18,100 211,753 11.69 0.912 Deaths from earthquake were ~150;
Japan, Mw (tsunami) tsunami heights were significantly above
9.0 (2011) design allowances. Revisit nuclear facilities
design
a
All the data has been taken from the GEM Earthquake Consequences Database (GEMECD)
impact disasters such as earthquakes, this fact Hatch and Dohrenwend (2007) also pointed to
alone does not account for gender differences in the lack of research evidence on the distribution
casualty rates. Because the gender division of of the events according to gender, age, racial and
labor keeps women in or close to homesteads, ethnic background, and socio-economic status
Byrne and Baden (1995) argued that it is gener- (SES). They suggest that such information
ally they who suffer disproportionately from the would help identify groups at greatest risk so
collapse of poorly constructed dwellings. that they could be investigated further. Although,
A clear relationship has been found between theoretically, the impacts of a disaster are not
the number of buildings that collapse in an earth- randomly distributed across socio-economic
quake and the number of fatalities caused. Mor- classes, empirical evidence of this claim is
E
tality is therefore related to the chance of being in scarce. In a population-based cohort study of the
a poorly constructed building at the time of the 1999 Taiwan earthquake, the multivariate results
disaster. Timing is clearly important. For exam- suggest that health status and socio-economic
ple, there may be a higher death toll of both men status (SES) are likely to be two key predictors
and women if a disaster happens at night when of earthquake mortality. Indeed, lower SES,
people are asleep and a higher death toll among major disease status, and moderate physical dis-
women if it happens during the day when men are ability status increased the risk of earthquake
away from the home. If death rates are higher death 1.5– 2.5-fold (Chou et al. 2004). It is nec-
among women, it could be because: essary to understand how socio-economic status
and health interact with gender in such cases. In
• Women are more likely than men to be caught one of the greatest modern disasters, the Indian
in poorly built buildings, for example, by Ocean tsunami in 2004, more women than
being more likely to be at home. men died.
• Women are more likely to be caring for young Women’s and men’s decisions in times of
children or elderly relatives who slow their crisis have been shown to differ along risk-taking
evacuation. lines, nature of coping strategies, adaptability,
• Women are likely to be older, since women and advice-taking and information-seeking
live longer. behaviors (Dankelman et al. 2008). Hazeleger
• In some societies, women tend to be less well (2013) examined gendered disaster recovery
educated and make poorer decisions in a issues relevant to Australian policy and also
crisis. found differences in disaster risk taking as well
• Women less likely to evacuate rapidly, as economic differences and inequalities that
because they are more inhibited about leaving affected survival and recovery between men and
the home. women. This calls into question the assertion that
• Women are more likely to be caring for depen- high levels of economic development iron out the
dent children or elderly relatives. risk levels between men and women.
• Again in some societies women are generally
of lower economic and social status and have
less access to resources. Two Events in the Twenty-First Century
It has been also suggested that women have a In this section the socio-economic consequences
greater propensity to sacrifice themselves, for of two real events – the L’Aquila earthquake of
example, by rescuing young children or elderly 2009 in Italy and the Great East Japan earthquake
relatives, rather than escaping on their own, but and tsunami of 2011 – are examined. These two
this has never been proven by field evidence. earthquakes represent the most socially controver-
Alternatively, there may be gender differences sial and most costly events in the past decade. The
in attitudes to risk-taking behavior that influence intention here is not to generalize about how cer-
mortality. tain decisions pre- and post-event would help a
898 Earthquakes and Their Socio-economic Consequences
society recover more effectively. The purpose of the EU, the Italian government estimated the
this section is to describe the implications of earth- cost of the earthquake at just over €10 billion
quakes on two very different communities and in while insured losses reached $US250 million.
doing so draw the readers’ attention to the complex Four months after the event, the Italian Civil
long-term social and economic consequences of Protection reported 49,000 people were still
earthquakes. The consequences of earthquakes go homeless.
far beyond the immediate aftermath of the event
and can adversely or positively impact on future Government Plans for Reconstruction
generations. Comparing the management of these Unlike the management of past earthquakes in
two disasters and assessing the key overriding Italy, the central government under Berlusconi
factors on recovery. What are the true costs of promised that the affected population would not
earthquakes and at whose expense? In reviewing be placed in transitional housing which usually
the evidence on the recovery of the, four central came in the form of mobile trailers or
themes were found namely: prefabricated housing. The homeless were
housed in camps around the affected area and in
1. Government plans for reconstruction various hotels along the coast. The process of
2. Socio-economic characteristics of the region reconstruction was entirely led by the Italian
3. Citizen involvement in decision making Civil Protection Authority (DPC), and the aim
4. Politics and the media was to reconstruct and rehouse the affected pop-
ulation within a year in 19 new towns. This was
The following sections are therefore orga- led entirely by the central government at a cost of
nized accordingly. €1.6 billion. The name of this project was the C.
A.S.E. (Complessi Antisismici Sostenibili ed
Reflections on 2009 L’Aquila in Italy Ecocompatibili (Calvi and Spaziante 2009).
The Mw 6.3 L’Aquila earthquake caused According to the Italian government, nearly €8
308 deaths and nearly 1,200 injuries of which billion has been spent on L’Aquila and the 56 sur-
202 were serious. This was the deadliest earth- rounding communities that were affected in the
quake in Italy since the November 23, 1980, quake (BBC 2013). This housing provision has
earthquake in the region of Campania and the sparked controversy though. The fact that all of
most destructive earthquake in L’Aquila since these new towns are far away from the main city
the series of three earthquakes that devastated and are not accessible by public transport has
the wider region in January and February 1703 been a source of major discontent (Alexander
that killed at least 10,000 people. In 2009 2012), though the rehousing of the affected pop-
203 people died in L’Aquila, a city of 73,000 ulation has taken place very swiftly (Fig. 8).
inhabitants, and many more thousands are stu-
dents of the L’Aquila University, not living in Socio-economic Characteristics of the Region
the city, while in the village of Onna, 38 out of L’Aquila is the largest city in the region and, with
around 350 inhabitants were killed. The hospital a GDP of €27,703.41 million, is the wealthiest
at L’Aquila, where many of the victims were region in Central Italy, hosting a university. The
brought, suffered damage in the 4.8 aftershock surrounding villages like Onna and Paganica are
which occurred one hour after the main earth- predominantly home to small businesses or sec-
quake and had to be partly evacuated. Damage ond homes. The city’s work force is mainly con-
to L’Aquila’s and wider regions’ historic build- centrated in public services such as defense,
ings and churches was also quite severe. manufacturing, service industries, and the build-
In total, almost 4,000 buildings collapsed or ing industry. The death toll might have been
were damaged beyond repair, while another much worse, if many more day visitors and stu-
70,000 buildings were damaged but repairable, dents had been in the city, inside and outside
and 65,000 were made homeless. In a report to vulnerable buildings.
Earthquakes and Their Socio-economic Consequences 899
Earthquakes and Their Socio-economic Consequences, Fig. 8 A C.A.S.E. town on the outskirts of L’Aquila
Abruzzo has traditionally relied on agricul- Unlike other Northern Italian cities, the
ture, but there has been some industrial growth charming historic center of L’Aquila does not
especially in engineering, food, transportation, attract many tourists, probably due to its geo-
and telecommunications sectors in recent years; graphical location. Despite the initial heroic
though by comparison with its northern neigh- remediation measures, the historic center still
bors, the Abruzzo region is not a major industrial lies derelict with scaffolds supporting the struc-
center and the industry is on the coast, two hours’ tures, but little else is being reconstructed or
drive from L’Aquila. The fact that there were considered. All attention was focused on building
relatively little financial losses in terms of busi- new settlements, and perhaps L’Aquila’s recon-
ness interruption may be a key factor in the slow struction was never the major focus of attention
pace of overall recovery and reconstruction. from the start since the town center does not have
Contrasting L’Aquila’s situation is the recovery the social clout to demand the attention of the
of the wealthy Emilia-Romagna region, one of central government. In a recent article by the
the most industrialized areas of the country. Two BBC (2013), it was reported that some 22,000
major earthquakes Mw5.8–5.9 caused 27 deaths of the 70,000 people of the city who were made
and more than €12 billion euros of damage in homeless are still unable to return to their homes
May 2012, seriously affecting small- and and are living in new blocks of flats on the out-
medium-sized companies. The earthquake skirts. Some 70 % of the reconstruction needed
disrupted more than 10,000 companies in the on the outskirts has been completed, but the
affected areas. There was an urgency to be back thirteenth-century town center remains crippled,
in business and private companies were the main guarded by the military who only allow those
drivers of the recovery of the region. The sym- with authorization to enter.
bolic “solidarity cheese” fund brought in around What will become of L’Aquila’s historic cen-
€1 million. People all over Italy were encouraged ter? 4 years on and the center is still littered with
to purchase the damaged wheels of Parmigiano- unusable buildings; commercial activity is much
Reggiano and Grana Padano cheese, produced in reduced and public services deteriorated; it is
and around the city of Parma, to help with the hard to imagine how its fortunes can be reversed
region’s economic recovery. (Fig. 9).
900 Earthquakes and Their Socio-economic Consequences
Citizen Involvement in Decision Making in April 2009, with the Prime Minister’s 23 visits
There has been a lot of criticism of the authorities to the region and his decision to bring the 35th G8
handling of the reconstruction process after the summit to L’Aquila in July 2009, 3 months after
L’Aquila earthquake. A recent report suggested the earthquake hit the region. The need to secure
that there had been no public participation in the votes in provincial elections by Prime Minister
reconstruction planning of the new towns around Berlusconi’s Government only weeks after the
L’Aquila. A top-down approach was used under disaster was perhaps the incentive behind a cen-
the pretext of a more rapid reconstruction strat- tralized top-down approach to the rehousing pro-
egy and recovery of the region. However “Once jects in L’Aquila, crucially ignoring local public
people have been housed in accommodation that opinion (Fig. 10).
will last for years, they can be forgotten, or at However, once the cameras and world leaders
least left to better their circumstances without left and votes for the provincial elections were
further state intervention” (Őzerdem and Jacoby secured, the town center was left alone. In a
2006). The decision to create the surrounding recent article in the BBC (2013), Mayor Massimo
new settlements seems to lack long-term vision Cialente accused the government of forgetting
and will inevitably change the residents’ daily about L’Aquila. He says the delay has been due
lives. As recounted in an interview for the BBC: to paralyzing bureaucracy, a lack of organization
between the local and national governments, and
I miss my former life, the neighbours I grew up a lack of funds. But reconstruction is more easily
with, the woman who lived upstairs who used to
treat me like a daughter . . . I want the L’Aquila we said than done in a city where many of the build-
all know to return and be even better, I want people ings are centuries old, where homes are interwo-
to come back. (L’Aquila_) is an oasis among ven with historically important buildings, and
dozens of empty, damaged buildings.” says where the narrowness of the streets means cranes
Mariella Riccobono, who now lives 10 min outside
the town. Wiping a tear from her cheek, she says: and other construction vehicles can only be
“It’s worth fighting for your home. If you give up, brought in a few at a time.
you’re abandoning the city. We’ve continued pay-
ing a mortgage all this time and it’s my livelihood, Reflections on the Great East Japan
it’s everything I’ve created.
Earthquake and Tsunami of 2011
The earthquake and tsunami devastated Tohoku
Politics and the Media and other regions. In total by the 18th month
The media and politics were never far from after the event (September 2012), around
L’Aquila immediately following the earthquake 1,230,000 buildings were recorded as having
Earthquakes and Their Socio-economic Consequences 901
been damaged or destroyed due to tsunami and According to MunichRe, the overall losses of
ground shaking (which is equivalent to 2.4 % of this event are estimated to be over US$210 bil-
all of Japan’s existing buildings) and the tally lion. The monetary cost associated with the fail-
continues. Life loss is reported as 20,721 and ure of the Fukushima Daiichi nuclear power plant
includes 2,303 earthquake-related deaths that as a result of the 14 m tsunami is still not quan-
occurred due to various causes after the day of tified to this day, though Tokyo Electric Power
the earthquake. Company (TEPCO) announced a ¥1.02 trillion
In terms of cause of death (21/4/2011): cost of natural disaster. Immediately after the
disaster, approximately 85,000 people were evac-
• 92.5 % drowned. uated from 20 km evacuation zone and 185,000
• 4.4 % crushed/died of injuries. people were screened for contamination. Less
• 1.1 % burnt. serious problems occurred in the Onagawa and
• 2.0 % unknown (Fig. 11). Fukushima Daini nuclear power plants.
902 Earthquakes and Their Socio-economic Consequences
Not all places that were affected by the tsu- Japanese Institute of Architects (AIJ) is consid-
nami are the same, however. The area around ering using this as a model of participation.
Sendai in Miyagi Prefecture is a flat plain and Unfortunately people cannot wait and they are
has a strong economy, good transport links, and a now down to 100 households and the commu-
growing population. Further north in Iwate Pre- nity may fall apart because of the delay. The
fecture, there are steep slopes and fiords, a declin- group decided they would not oppose the pro-
ing population, and a weak economy. In each posed embankment but suggested it be moved
there are differences of scale with a few larger back. Initially the city was not happy but
cities and towns and many smaller settlements changed their minds after receiving the petition.
and villages. This suggests that different places The proposed municipal plan is now for a much
face different issues of recovery. These differ- lower embankment further back, but this needs
ences in socio-economic prospects, demography, cooperation between the Ministry of Forestry,
topography, and scale suggest that approaches to Japan railways, the National Highways Agency,
both safety issues and economic development and the prefecture.
assistance might be fine-tuned to meet local In Kesennuma the citizens’ committee oppose
circumstances. the planned harbor embankment, and they are in
talks with the municipality and prefectural gov-
Citizen Involvement in Decision Making ernment. The majority of residents are against the
In the areas affected by the 2011 tsunami, proposal and it is not settled yet. Planning
consultations between governments and com- arrangements cannot proceed while there is a
munities were the rule, and community repre- dispute, but city officials are making land-use
sentatives were invited to serve alongside plans assuming the embankment will go ahead.
experts on recovery planning committees from Because there has been so much opposition, a
the earliest stages. The most common ways of new deadline has been set for October 2013. In
collecting residents’ opinions were surveys and other places plans are proceeding more rapidly.
workshops. The central government and local In Kamaishi the three community workers for
governments outside the disaster-affected area the prefecture explained that local authorities
helped affected municipalities plan their recov- have to accept what citizens want. Ideally they
ery by conducting research, seconding staff, and would simulate different heights of embankment
hiring professionals to provide technical sup- since communities in some places have opposed
port. University faculty members, architects, the plans. If people want something outside, the
engineers, lawyers, and members of NGOs par- government’s recommended solution planners
ticipated in the municipal planning process have to be careful that safety measures are in
(World Bank 2012). place and that the community has collectively
Along the Rias coast the response of the relocated.
majority is that the government has already Partly because of citizen opposition, recon-
decided so they can’t do anything. Some even struction of sea embankments, which suffered
admire the massive colossal infrastructure. But extensive damage, has been considerably
the younger generation, in their forties, is delayed. Local governments in devastated areas
opposed to large embankments and tall sea- cannot decide on the details of restoration plans,
walls, but they are not the decision makers. In as discussions continue on whether to prohibit
Japanese community associations, it is elderly people from returning to coastal areas. Recon-
men who make the decisions. In Ohyakaigan struction work has started on only 31 % of
near Kesennuma the community association destroyed embankments. According to the Fish-
meets twice a month and tries to involve chil- eries Agency, which has jurisdiction over sea
dren as well as older people. The plan is to embankments, the design of embankments will
collectively relocate the 120 households and to depend on whether people will live nearby (Daily
have the land cleared by 2015–2016. The Yomiuri 2013).
Earthquakes and Their Socio-economic Consequences 905
Local governments were tasked with recovery disaster. The central government was keen to
by the National Government who asked them to act fast and one of the first projects completed
develop local plans based on consultation. The was the 12 m high seawall along the Natori
problem is they lack the technical capacity, espe- plains by the Sendai Airport. This structure is
cially in effective methods of involving citizens twice as high as its maximum height than the
in strategic decision making. Voluntarily urban previous wall. Figure 14 shows the inundation of
planners and architects from all over Japan the tsunami into the Natori plains by the airport.
surged to provide missing capacity. Local gov- Although the need to protect the coast from
ernment has lots of problems with consultation, future risks is clear and the government would
which is time-consuming. And it is not easy to argue that it is doing what is right in saving
convince communities to relocate. Local govern- towns that have been here a long time, the public
ments want to consolidate communities to make pressure was there to spend as quickly as possi-
it more efficient and economical to deliver ser- ble and there was money in the economy to
vices, but many of these places were in decline support this. But there was a time limit for peo-
before the tsunami. The fundamental problem is ple (central and local) to decide on what to do, as
that the authorities do not really know what size the government had set stringent deadlines for
population they are reconstructing for. They have spending.
to provide facilities to each community, so the The media has mainly focused their attention
cost is enormously considerable (Bhatia et al on the Fukushima Daiichi nuclear disaster. Fol-
2013). lowing the earthquake, a 15 m tsunami disabled
the power supply and cooling of three reactors,
Politics and the Media causing a nuclear accident. All three cores melted
There was inevitably a great deal of media atten- in the first 3 days and the accident was rated seven
tion surrounding the Great East Japan earth- on the INES scale, due to high radioactive
quake, especially over the Fukushima nuclear releases over following 3 days. Eventually four
disaster. In addition to the coverage of the after- of the six reactors had failed. The days and
math of the earthquake and tsunami, there was months after the failure, the handling of the disas-
intense attention on how the leaders of the then ter by the government and TEPCO, the Tokyo
fragile government would respond to the Electric Power Company, was under heavy
906 Earthquakes and Their Socio-economic Consequences
scrutiny. To this day there is mistrust of the state- above, if initiatives scientists and engineers
ments made by TEPCO on the safety of the sur- should focus their efforts on for earthquake mit-
rounding areas, and some residents in and around igation and preparedness.
the region have instead chosen to self-evacuate or
purchased their own Geiger meters to track the Understanding Human Behaviors in
radiation levels. Emergencies
The disaster changed the national debate over Risk perception is not a precise science and even
energy policy almost overnight. Japan, a country with the best of intentions, the way one acts when
heavily reliant on nuclear power, due to its own experiencing an earthquake. In an interview with
shortage of fossil fuels had by June 2012 shut a survivor of the Great East Japan earthquake and
down all but two of the country’s 54 nuclear tsunami, he recounted how he knew it was the
reactors. This discussion continues and it largest earthquake he had personally ever expe-
remains to be seen what the true long-term pol- rienced and that he knew that a large tsunami
icies and health effects are from the nuclear would follow. However, against every rational
accident. and educated decision he could have made to
For this devastating event, the socio-economic run to higher ground, he chose deliberately to
consequences have had local, national, and global rush to the coast, to where his family was. He
effects and will continue for decades to come, survived, but many others including his family
raising the question: which is the biggest did not. This is just one of many similar stories
disaster – the earthquake (concentrated and pow- (EEFIT 2014). Understanding the needs and rea-
erful), the tsunami (vast area affected), or the soning behind actions after a disaster would help
radiation release (contamination for centuries)? us devise appropriate strategies for mitigation.
What physical and social provisions would have
changed this man’s decision?
Looking to the Future
Incorporating Socio-economic Impacts in
Earthquakes and their socio-economic conse- Earthquake Loss Models
quences will continue to dominate headlines as To better understand and address earthquake risk
highlighted in the chapter. The need to consider in its entirety, one needs to go beyond seismic
the severity and long-term impacts of socio-eco- hazard, exposure, and the vulnerability of the
nomic factors in earthquake risk reduction and built environment. The socio-economic charac-
disaster management is therefore crucial. Echo- teristics of populations exposed to earthquake
ing the recommendations drawn from investigat- threats need to be integrated within loss estima-
ing socio-economic impacts of Iranian tion models to an integrated and holistic estimate
earthquakes by Hosseini et al. (2013) includes: of risk in an area. A realistic socio-economic
impact assessment can be used to examine how
1. Promoting public participation an earthquake will change the lives of the current
2. Organizing community-level disaster man- and the future residents of a community. How-
agement groups and systems ever, it is not easy to quantify the changes in
3. Developing a natural hazards insurance social and economic conditions that will arise
system from future earthquakes. At present these are
4. Establishment of a catastrophe insurance pool done as modelled with reconstruction costs per
5. Socio-economic impact assessment for future square meter. The main indicators include
earthquake scenarios changes in demographics, characteristics of the
community, potential effects of an earthquake on
As one anticipates for a future with the possi- services and the housing market, changes in
bility of a “million-death earthquake,” below are employment and income levels, and alterations
some additional recommendations to those listed to the quality of life.
Earthquakes and Their Socio-economic Consequences 907
As part of the GEM, there has been an empha- Disasters in the past are often seen as catalysts for
sis for the first time to review how one could change in the built environment. The urgency of
assess the socio-economic impact of earth- rehousing a large number of people affected by
quakes in loss estimation. The ambition of the event mean that funds are diverted and atten-
GEM is to deliver methods, metrics, and open tion is paid to the area in need. Time is com-
source software that can be used worldwide to pressed and here is an urgency to “get back to
explore the compounded nature of earthquake normal” – to rebuild livelihoods, clear up the
events where environmental processes blend debris, repair the damage, and, among the more
with factors such as urban growth, marginaliza- farsighted, to “build back better.” A challenge in
tion, and poverty to bring about and magnify post-disaster recovery and key to reducing the
E
earthquake impacts. The integration of physical long-term social and economic consequences of
risk with (a) the social characteristics of earthquakes is balancing the need for speed and
populations and (b) the susceptibility of people deliberation.
to the adverse impacts of earthquakes leads to an Earthquake losses will continue to rise if the
encompassing perspective on global risk assess- community of practitioners working in the field
ment which: do not come together and try to mitigate these
losses together. Earthquake engineers and scien-
• Considers loss and damage as part of a tists have the knowledge to dramatically reduce
dynamic system, where interactions between the number of deaths from earthquakes. Prevent-
natural systems and societal factors redistrib- able deaths are attributed to damage and col-
ute risk before an event and redistribute loss lapses of buildings and buildings should not fall
after an event down. It is clear that earthquakes can have a
• Mainstreams socio-economic vulnerability significant impact on the social and economic
and resilience in policy discussions on earth- fabric of society, both in the short and the long
quake loss and damage term. The ability of a society to recover depends
• Evaluates loss and damage, taking social fac- very much on the precondition of the affected
tors into account at different time and space area but even more so on the immediate decisions
scales and subsequent policies made by the governing
• Uses risk assessments in benchmarking exer- authorities. Some would argue that losses mainly
cises to monitor trends in earthquake risk over rise from political factors, not from a lack of
time knowledge or inability to apply it. Mitigation is
• Recognizes that both causes and solutions for not a priority in the short term, and therefore
earthquake loss are found in human-(built) governments and communities are not necessar-
environmental interactions (GEM 2014). ily interested in devoting adequate resources to
mitigation. Several authors would further argue
that corruption is the main contributing factor to
Summary losses from earthquakes (Escaleras et al. 2007;
Ambraseys and Bilham 2011), and unless we
Recovering from a disaster involves all kinds of tackle this sensitive issue, there is little hope of
development and planning activities – land-use mitigating catastrophic losses. This is not dis-
plans, building norms, and transport plans. What puted here. However, one should and must
is unique in post-disaster situations is that all remain positive to strive for a better understand-
these activities happen in a compressed period ing of the direct and indirect causes of losses in
of time. Communities must rebuild as quickly as earthquakes and devise appropriate mitigation
possible to maintain existing social networks and strategies for the future, tailoring for local
get the economy back on its feet. But they must sociodemographics and focusing on the either
also be deliberate in trying to maximize the physical or social interventions, or both. Inaction
opportunities disasters provide for improvement. is not an option.
908 Earthquakes and Their Socio-economic Consequences
People live in earthquake-prone areas. Tectoni- The underlying rationale for applying economic
cally active landscapes may have produced analysis to understand the consequences of seis-
greater biodiversity, more food and water, mic events is concern for the well-being of those
through disturbed drainage patterns and wetland persons affected or potentially affected. For the
creation. Nevertheless over the last 200 years, most part, economic analysis does not aim to
global population growth has increasingly placed measure changes in well-being directly, but
pressure on human settlements and resulted in the rather changes in the level of activity and assets
proliferation of urban areas worldwide, many in held within the formal market-based economy.
seismically active zones. The opportunities pro- Commonly used market-based indicators of
vided by living close to a seismic area are tem- well-being include business output or turnover,
pered by the risks of loss and destruction during a value added or gross domestic product, capital
seismic event. Attitudes toward seismic risk are assets including property values, household
informed not only by understanding of the geo- income, employment, and consumption of
physical characteristics underpinning seismic goods and services.
events but also by socioeconomic conditions, Seismic events may also result in a wide range
such as historical attachment to the area, property of nonmarket social and environmental impacts.
ownership, and community connectedness. These nonmarket impacts may be both negative
Although risks of damages to property may be and positive. On the one hand, negative impacts
handled on an individual basis through insurance may result from loss of life and physical health,
markets, risks of seismic events on a societal psychological problems and stress, destruction of
level are seldom channeled through formal mar- family and social networks, inconvenience, and
kets, creating a strong mandate for policy analy- degradation of ecological systems. On the other
sis and intervention. hand, seismic mitigation measures put in place
910 Economic Impact of Seismic Events: Modeling
Economic Impact of Seismic Events: Modeling, Table 1 Typology of economic impacts associated with seismic
events
Examples of direct impacts Examples of higher-order impacts
Market- Stocks Damage to buildings, infrastructure, Reduced purchases of public assets as
based inventories finances diverted toward disaster
analyses recovery
Flows Loss of production by businesses with Loss of production by suppliers and
damaged capital customers of businesses unable to operate
Reconstruction diverts resources away
from more efficient production activities
Nonmarket Stocks Loss of life Loss of social capital via disruption of
analyses social networks and relocation of people
Loss of physical and mental health Loss of trust in authorities
Damage to natural capital
Flows Interruption of ecosystem services Stress associated with lost incomes and
Loss of leisure as a result of inconveniences jobs
experienced during the event and its recovery
may lead to positive impacts on well-being by Flow losses include reduction in the rate of output
creating a greater sense of security and safety for by businesses, reduction in the quantities of
local residents. These social and environmental goods and services consumed by households
impacts may be associated with changes in eco- over a given period, or the hours of lost leisure
nomic activity, e.g., physical or psychological as a result of various inconveniences following a
injuries sustained as a result of a seismic event disaster.
constitute a loss of human capital which may, in Direct economic impacts include damage to
turn, result in a loss of economic production. buildings, infrastructure, and inventories, loss of
However, methods based solely on analysis of production by businesses with damaged capital,
market outcomes are unlikely to capture the full loss of life, loss of health, and so on. Such impacts
extent of the effects. In such cases nonmarket are a direct physical consequence of a seismic
valuation techniques may be utilized for impact event. Higher-order impacts of earthquakes are
assessment. those not provoked by an earthquake itself but by
Table 1 provides a typology for the various its consequences. Sometimes higher-order impacts
types of economic impacts associated with seis- are referred to as “secondary” or “indirect” effects;
mic events. As with most existing typologies higher order is used here to avoid any confusion
(e.g., Rose (2004) and Hallegatte and Przyluski with “indirect effects” as used in input–output
(2010)), Table 1 distinguishes not only between (IO) analysis (Rose 2004; Okuyama 2007).
stock and flow impacts but also between direct Regardless of the particular term used, the separa-
and higher-order impacts. The distinction tion between direct and higher-order impacts is
between stocks and flows is well recognized often blurred. Particularly contentious is how busi-
within economics. A stock is a quantity of a ness interruptions are classified. On the one hand,
system variable that exists at a single point in business interruption caused directly by damage to
time, while a flow is a movement of materials or capital used in production, e.g., damage to the
information to or from a stock over time. Stock Sendai fishing fleet during the 2011 Tōhoku earth-
losses associated with seismic events include the quake and tsunami, may be considered a higher-
physical destruction of houses and other build- order impact. On the other hand, this could be
ings, infrastructure, machinery, inventories, and referred to as a direct impact, with the term
other physical assets, as well as loss of human higher-order reserved for business interruption
capital from deaths, injuries, and relocation. not directly caused by damaged capital. These
Economic Impact of Seismic Events: Modeling 911
would include, for example, a number of car fac- economic impacts associated with seismic events
tories in Korea and the United States that experi- using simulation models that do not originate
enced shutdowns following the 2011 Tōhoku solely from within the economics discipline.
event because of the disruption in supply of key These have included inter alia the use of system
components. Hallegatte and Przyluski (2010) pro- dynamics (SD), agent-based models (ABMs),
pose that higher-order impacts extend to cover and integrated assessment methodologies.
those costs or benefits that span a longer period
of time and a larger spatial scale or occur in a Engineering-Based Models for Valuing
different economic sector than the disaster itself. Capital Losses
Accordingly, business interruptions are direct Studies concerned only with the estimation of
E
losses if caused by damage to buildings, plants, flow impacts often require, as a direct modeling
machinery, or critical infrastructure or disruption input, the assessment of the value of earthquake
to employees’ availability for work, while losses damages to buildings, infrastructure, stocks, and
in production that spread to other firms via back- other physical capital. The value of lost capital is
ward and/or forward linkages are higher-order required to determine the likely changes in eco-
impacts. nomic production and the expenditure patterns
A further point of distinction in the categori- during recovery and reconstruction. A common
zation of earthquake impacts, which is not con- approach used to estimate the direct costs from
sidered in Table 1, is the separation of costs and seismic events is the use of susceptibility or dam-
benefits that arise from the occurrence, or poten- age functions. Such functions describe the rela-
tial occurrence, of an event itself and those that tionship between one or more seismic parameters
arise from mitigation measures, e.g., earthquake (e.g., ground motion) with damage for a certain
strengthening of buildings, training of urban type or use of object at risk, which can then be
search and rescue personnel, and so on. directly correlated to economic losses.
Econometrics
Market-Based Methods Econometrics focuses on providing an empirical
basis to economic relationships, through the use
Up until the 1990s, the economic impacts of of statistical analysis, for testing economic the-
earthquakes and other disasters had received rel- ory, forecasting, decision making, and ex post
atively little attention within mainstream eco- evaluation. Although perhaps best known for its
nomics (Okuyama 2007). Subsequently, a use in generating short-term forecasts of the per-
plethora of studies extended and modified con- formance of national economies, econometric
ventional market-based economic models. IO modeling is widely utilized in both microeco-
analysis, for example, has been extensively nomics and macroeconomics. Econometric
applied in the assessment of higher-order modeling is also undertaken in combination
impacts. Other conventional market-based eco- with other forms of modeling, particularly for
nomic methods used in assessment of seismic the purposes of parameter estimation, e.g., for
events include econometrics, social accounting CGE, SD, and ABMs.
matrix (SAM), and computable general equilib- In the analysis of seismic events, several econo-
rium (CGE). With largely the exception of econo- metric studies have attempted to understand the
metrics, the remaining methods have principally determinants of the direct costs (e.g., mortality,
focused on determining higher-order impacts. capital losses). According to Cavallo and Noy
Economic impact assessment undertaken using (2009), these studies typically employ a model
conventional methods must thus rely on either framework that assumes direct costs are related
empirical data, engineering models, or estimation to some measure(s) of the physical magnitude of
to calculate direct impacts. In recent years there the disaster (e.g., ground movement) and variables
has been growing interest in modeling the that capture the “vulnerability” of the region to the
912 Economic Impact of Seismic Events: Modeling
disaster (i.e., susceptibility to the impact of natural economy (Leontief 1941). Arguably, IO models
hazards). Possible determinants that may increase are the most widely applied tool for assessing
a country’s vulnerability include the level of eco- higher-order impacts of seismic events.
nomic development, population, and various polit- A significant advantage of IO analysis is its abil-
ical and institutional conditions. ity to capture economic interdependencies within
A number of authors have also investigated a regional or national economy. In its most basic
the higher-order impacts of earthquake disasters. form, an IO model consists of a system of linear
Many of these studies look at changes in conven- equations, each one describing the distribution of
tional macroeconomic indicators, such as GDP economic goods among industries and final con-
per capita, although there has also been some sumers. An IO model is typically used to evaluate
research at the sector level and to changes in the way in which changes in production by one
poverty as measured by indicators such as the economic industry may, through supply chain
Human Development Index. Other studies on linkages, ripple through an economic system.
natural disasters, although not necessarily earth- Although mainly concerned with flow impacts,
quake specific, have also focused on impacts at IO models have some potential to capture stock
the village our household level, including effects impacts when constructed as a dynamic model.
on consumption, nutrition, and capital. Macro- There are several well-known limitations
economic studies may be generally separated associated with IO analysis, including linearity,
into those that are interested in short-run (up to absence of quantity–price interactions, rigid
several years) and those that consider long-run structure with respect to input and import sub-
(at least 5 years, sometimes decades) impacts. stitutions, and neglect of resource (e.g., capital,
Examples of short-run analyses include those labor, and land) constraints. In particular, IO
authored by Hochrainer (2009) and Jaiswal and models utilize “technical coefficients,” which
Wald (2013), while long-run studies include describe, for any given industry, the quantity of
those produced by Skidmore and Toya (2002) each type of good or service required per unit of
and Crespo Cuaresma et al. (2008). The focus of production. Although technical coefficients are
much of the debate in long-run analyses is on the relatively stable under normal economic condi-
extent to which natural disasters may contribute tions, it may be unrealistic to assume that these
toward faster economic growth via the replace- will remain constant following a disaster, when
ment of damaged capital with newer and more an economy is likely to be subject to supply
productive technologies. shortages and significant adjustments. A variety
Given the highly variable nature of earthquake of ad hoc extensions and refinements have been
events, obtaining sufficient data (time series and put forward to help address IO weaknesses. For
cross sectional) upon which to build such models example, Cochrane (1997) allows for a more
poses a significant challenge. By its very nature, flexible treatment of imports to deal with short-
econometric modeling also depends upon the ages in local production. Similarly, Hallegatte
assumption that the underlying causal mecha- (2008) adapts the IO framework to incorporate
nisms do not change in form, strength, or stochas- production bottlenecks and changes in produc-
tic properties from the estimation period to the tion capacity, prices, and labor demand.
forecasting period or from one cross-sectional Recently, the inoperability IO model (e.g.,
sample to another. Econometric models are, how- Haimes et al. 2005) has received significant atten-
ever, statistically rigorous and may provide sto- tion in the context of disaster analysis. In mathe-
chastic and forecasting capabilities. matical terms, the evaluation of higher-order
impacts within the inoperability IO model is sim-
Input–Output Analysis ilar to other IO models, except that results may be
IO analysis is attributed to Wassily Leontief, the stipulated in normalized terms, with the point of
1973 Nobel Laureate for Economics, who in reference being the as-planned level of produc-
1936 published the first IO table for the US tion for each economic industry. Inoperability IO
Economic Impact of Seismic Events: Modeling 913
models are commonly used to explain the dynam- models have been identified as promising tools
ics that underlie business operation losses. For for earthquake and other disaster analysis for
example, Barker and Santos (2010) explore the some time, the number of actual examples in
role of inventory in muting disruption following the literature is still relatively few.
an event. When compared to IO and SAM-based
models, CGE models exhibit a number of advan-
Social Accounting Matrices tages. Firstly, the modeling framework provides
SAMs are closely related to IO models – both are for flexibility in the specification of functions
typically constructed from a country’s System of describing the production of each type of com-
National Accounts. While the traditional IO modity, including allowance for substitution
E
model focuses on flows among economic actors between inputs in response to relative changes
involved in production (i.e., industries), a SAM in price. For example, if a particular good
additionally describes the way in which income becomes scarce following an event due to dam-
generated by factors of production (e.g., capital, age to local production facilities, a CGE model
labor) is allocated among different economic will allow for industries requiring that good for
actors (e.g., businesses, households). Secondary production to increase purchases of the imported
distributions of income among actors through, form of that good. Secondly, the framework
for example, tax payments and government allows for consumers to alter consumption pat-
transfers, along with allocation of income to terns in response to changes in relative prices of
savings, are also included, thus providing a com- commodities and/or income levels. Despite this
plete picture of the circular flow of income generic flexibility, practitioners typically concen-
within an economy. SAM-based models thus trate on specifying production functions with
have the potential to evaluate a greater range of substitution only between factor inputs (i.e.,
higher-order impacts than an IO model. labor and capital), between imported goods and
SAM-based models are however subject to the domestic goods, and possibly between some
same limitations as IO models, including the use other key inputs (e.g., substitution between
of rigid technical coefficients. energy commodities is often included). Practi-
tioners typically rely on IO tables to specify pur-
Computable General Equilibrium chases of intermediate goods.
Starting with Leif Johansen’s pioneering Whereas IO models are mainly suited to the
(1960) multi-sectoral study of economic analysis of demand-side supply chain and some
growth, and aided by rapid advances in com- income-related effects, CGE models are also
puter technology, CGE models have now suited to the analysis of supply-side effects (i.e.,
become a standard part of the toolbox of econ- impacts on customers) and price effects. For
omists concerned with policy-oriented research. example, if half of the capital used in producing
In general terms, a CGE model seeks to simul- a particular type of commodity is damaged dur-
taneously determine commodity and factor ing an earthquake, among the impacts considered
prices within an economic market and the quan- by a CGE model will be the loss of income for the
tities of commodities produced and consumed, firms directly impacted by lost production capac-
subject to budget (e.g., household, government) ity, loss of income for suppliers of directly
and resource (e.g., capital, labor, land) con- impacted firms, gains in income for remaining
straints. Like IO models, CGE models are typi- firms supplying the commodity as a result of
cally used to assess flow impacts, for example, higher market prices, and gains in income for
changes in economic output, gross domestic firms supplying commodities that experience
product, or household consumption of commod- increasing market demand as a result of input
ities. Dynamic CGE models also have the poten- substitution. Due to the wide ambit of economic
tial to track changes in stocks (e.g., capital, impacts considered, with impacts often occurring
labor) over a recovery period. While CGE in opposite directions (i.e., counterbalancing),
914 Economic Impact of Seismic Events: Modeling
CGE models typically generate smaller estimates exacerbate dynamics away from an equilibrium)
of overall impact than IO or SAM-based models. or balancing (i.e., regulate dynamics toward a
In terms of limitations, CGE models are new equilibrium). At the heart of the SD
constrained to the evaluation of economies at approach is the modeling of feedback loops
equilibrium. A typical CGE analysis is under- through time using causal loop and stocks and
taken by “shocking” an economic system (e.g., flows diagramming.
adjusting capital stocks to reflect damage and SD has been used to model the economic
reconstruction) and, in turn, evaluating the impacts associated with disruptions to critical
change in the system necessary to produce a infrastructure. Dauelsberg and Outkin (2005),
new equilibrium. While the equilibrium nature for example, calculate economic impacts as
of CGE models is not a problem with long-run non-equilibrium events with critical infrastruc-
analysis, it is however a significant shortcoming ture interdependencies captured through event
for short-run analysis following a seismic event, propagation. Similarly, Portante et al. (2011)
when economies are likely to undergo significant describes EpFast, a simulation and impact anal-
adjustment and exhibit out-of-equilibrium ysis tool developed at Argonne National Labora-
dynamics. Rose and Guha (2004) therefore dis- tory. EpFast is a suite of simulation models
tinguish different response phases following an which focus on evaluating the vulnerability of
event and identify some of the variations in energy infrastructure to events. They describe
model parameters suitable for evaluation at each an ex ante application of the model to a high-
phase. intensity New Madrid seismic event of the US
The extensive data requirements for a CGE Eastern Interconnection. Kachali (2013) has fur-
model present a significant limitation to the ther applied SD to study the 2010 and 2011 Can-
more widespread application in earthquake eco- terbury (New Zealand) earthquakes. This work
nomic impact assessment. In addition to all of coalesce hazard, recovery, and organization dis-
the data requirements of a typical IO model, a ciplinary fields with primary data collected via
CGE model requires the input of numerous surveys, case studies, and interviews of impacted
behavioral functions and substitution parame- organizations, to depict how systemic interac-
ters. Parameter estimation is also often based tions within and between economic industries
on data from historic behavior under business- affect recovery post-disaster.
as-usual conditions, and such behavior may be
ill suited to describing behavior following a Agent-Based Modeling
seismic event. Although some advancements Closely aligned with understanding the economic
have been made toward refinement of model consequences of seismic events as complex sys-
parameters to better suit disaster analysis (Rose tems is the use of ABMs. Thomas Schelling is
and Liao 2005), the examples of such work are often attributed with the creation of ABMs in the
still relatively few. early 1970s (Schelling 1971). ABMs simulate the
actions of autonomous agents with a view to
System Dynamics assessing their consequences on the system as a
SD is a dynamic modeling approach for under- whole. Unlike the conventional economic impact
standing the behavior of complex systems. Jay assessment approaches, which typically seek to
Forrester, of the Massachusetts Institute of Tech- describe movement from, or to, an economic
nology Sloan School of Management, developed equilibrium, ABMs capture the dynamic
SD in the mid- to late 1950s. Examples of com- (equilibrium or otherwise) interaction of agents
plex systems include the environment, economy, typically across space and through time. It is the
and society. Such systems are characterized by emergence of patterns and trends in behavior,
the presence feedback loops and time delays and either as collectives of agents or for the system
typically exhibit nonlinear behavior. Feedback as a whole, which is of most interest to agent-
loops are classified as either reinforcing (i.e., based modelers.
Economic Impact of Seismic Events: Modeling 915
The generative ability of ABMs is well suited methods to better understand how recovery is
to simulating the behavior of agents at the time influenced by the underlying dynamics of
during, immediately after, and in the recovery change in the natural, built and human environ-
and reconstruction phases of a seismic event ments, notably including influences of planning
when economies are likely to exhibit out-of- (e.g., land use, built form) on disaster risk.”
equilibrium dynamics. Miles and Chang
(2007), for example, developed an ABM which
focuses not only on estimating losses but also on Nonmarket Methods
the anticipated recovery trajectories following
an event. Specifically, the draw on empirical Nonmarket valuation techniques are essentially
E
information from previous disasters, rather than concerned with measuring the extent to which
regional economic theory, focused on spatial changes brought about by an event either subtract
units or aggregate sectors, to model emergent from or add to values experienced by individuals.
trajectories of an economy. Their models are Nonmarket impacts of seismic events, such as
calibrated using the 1995 Kobe (Japan) and changes to safety and security, environmental sys-
1994 Northridge (California, USA) earthquakes. tems, human health, and cultural heritage,
As another example, the US Department are often difficult to identify and measure. Never-
of Homeland Security applies a large-scale theless, nonmarket impacts are an important com-
micro-simulation model, termed N-ABLETM, ponent of the range of effects resulting from
to capture complex supply chain, critical seismic events. Noting that the classifications of
infrastructure interdependencies, and market nonmarket valuation methods vary between
dynamics associated with seismic events. authors, three broad categories of nonmarket val-
N-ABLETM is used primarily to gain insight uation techniques are identified:
into network vulnerabilities, economic security,
and continuity. • Stated Preference Methods. These methods
utilize surveys or experiments to elicit what
would hypothetically be paid for values which
Integrated Assessment Methods are not traded on markets. The most well-
An emerging area of investigation in the eco- known stated preference methods are contin-
nomic consequences of natural hazard events, gent valuation (CV) and choice modeling
including seismic events, is the use of integrated (CM). CV uses surveys to elicit the monetary
assessment methodologies. Such methods typi- value people are willing to pay to avoid a
cally incorporate critical feedbacks between decrement of some type of nonmarket good
existing disciplinary models (e.g., scientific, or service or the value they are willing to
engineering, planning, economics, and so on), accept in exchange for its deterioration. CM
thus, broadening the boundaries of investiga- is similar to contingent valuation, except that
tion. They also tend to be spatially explicit in willingness to pay is elicited from experiments
nature, applying geographical information sys- in which respondents are asked to rank prefer-
tems to facilitate data entry and reporting. Feed- ences for different attributes of nonmarket
backs exist between land use, transport, and goods or services. Although CV is the more
economic activity which are salient to under- commonly used method, CM has gained pop-
standing how communities recover from seismic ularity over recent years. In the context of
events (Cho et al. 2001; Ham et al. 2005). seismic events, CV and CM have mainly
Implicit in integration is the view that local been used to help ascertain socially optimum
communities, cities, and regions are complex policies for risk reduction. Asgary et al.
systems requiring a holistic approach to their (2007), for example, apply CV to evaluate
study. Blanco et al. (2009, p. 206) note a key the benefits of hypothetical earthquake early
recovery phase priority “consists of developing warning systems, suggesting that their results
916 Economic Impact of Seismic Events: Modeling
could be used by policymakers and businesses option, bequest, and existence) values. However,
to determine the optimal level of investment in these studies can be very difficult to implement
such systems. and must be carefully designed to avoid biases.
• Revealed Preference Methods. These methods Additionally, the “free-rider” problem can create
estimate the value of a nonmarket good or incentives for respondents to misrepresent real
service based on observed behavior toward preferences. Travel cost and hedonic pricing
some closely connected marketed good or ser- methods are known to be sensitive to model spec-
vice. Among these the travel cost method is ification, including choice of functional forms.
frequently used to infer the value of recrea- Cost-based valuation methods are not considered
tion, including experience of cultural heritage to have a sound basis in welfare theory but nev-
sites, based on the costs of travel to these sites. ertheless may be of use when there is limited time
Although seldom employed in seismic-related and resources available and only approximate
studies to date, the method has the potential to estimates are necessary.
evaluate nonmarket losses and justify the Despite the relatively widespread recognition
replacement of public goods and services. of the opportunities for use of nonmarket tech-
Loss in beach recreation value at tourist desti- niques in the analysis of natural disasters, the
nations in Thailand following the 2004 Asian number of actual seismic-related applications is
tsunami is, for example, one possible applica- still relatively few. Consequently, the use of
tion for the method. Hedonic pricing is “value transfer,” i.e., values estimated in one
concerned with inferring the value of individ- study applied to a later study, is common among
ual attributes of a market commodity. In the practitioners of nonmarket valuation. Unfortu-
context earthquakes, it has been used to ascer- nately, there are few guidelines available upon
tain the value people implicitly pay for hous- which to evaluate the appropriateness of value
ing with reduced seismic risk, or the loss in transfer methods.
value of housing following changing percep-
tions of risk, based on information from real
estate markets. Research Challenges
• Cost-Based Valuation Methods. These
methods assume that some set of observed The economic impact typology outlined in
market costs can be used as a proxy for the Table 1, and the preceding description of assess-
value of an asset. The cost of illness approach ment methodologies, provides a general picture
has been utilized to determine expenditures on of the approaches currently being employed in
health care and lost wages from illness as a assessment of the economic impacts of seismic
proxy measure of the physical injuries and events. Nevertheless, a number of key challenges
psychological trauma of seismic events. remain to be addressed.
Replacement cost or restoration cost methods
may also be relevant to the valuation of Defining Model Scope and Reference Mode
nonmarket assets following an earthquake. Ultimately, the selection of impacts for assess-
Previous valuations of natural capital have ment, and by corollary the choice of economic
included determining the costs of building method used in evaluation, depends on the pur-
infrastructure to perform similar functions. pose of the assessment. As stated above, eco-
For example, many studies have examined nomic impact modeling is primarily undertaken
the value of wetlands in improving water qual- with a view to maintain or enhance well-being;
ity in terms of replacement costs. however, some impacts will be much more rele-
vant to a particular audience than others. Insur-
Nonmarket valuation methods are subject to ance agencies, for example, will be mainly
considerable limitations. Only stated preference concerned with losses that are potentially
methods are capable of capturing nonuse (e.g., insured. Local authorities and governments,
Economic Impact of Seismic Events: Modeling 917
however, are charged with considering wider regionally, and nationally. Until the mid-1990s
impacts on employment, poverty and inequity, most economic impact studies of seismic events
fiscal balances, and so on. Nevertheless, some focused mainly on the location directly impacted.
impacts that arise following a seismic event Since the early 2000s, with growth in the devel-
may be outside the scope of evaluation. opment of multiregional IO tables and, more
A further point to note in relation to the set- recently, spatial CGE models, spatial impacts
ting of model scope is that impact is a relative have increasingly been incorporated in economic
concept, identified by comparing an actual or a impact assessment. A number of studies (e.g.,
hypothetical trajectory of a disaster, with some Okuyama 2004, Tatano and Tsuchiya 2008)
kind of counterfactual baseline trajectory or ref- note that there are often significant economic
E
erence model, i.e., that which would have transfers between regions brought about through
occurred in the absence of the disaster. Either the relocation of people, business activity, and
trajectory may include or exclude mitigation recovery operations. While an impacted area
measures or response strategies, depending on may experience significant economic losses in
the purpose of the analysis. Although many eco- the period immediately following a seismic
nomic assessments concentrate on comparing event, locations a little farther out may experi-
situations before and after an event, this ence the opposite. Moreover, there are dynamic
approach is only likely to be valid when consid- higher-order economic feedbacks which exist
ering short-run impacts. Economic systems are between regions, for example, competition for
complex and constantly evolving over time; thus labor or capital between an earthquake-damaged
the longer the length of time for the analysis, the area in recovery and an adjoining area experienc-
more difficult it is to define an appropriate ref- ing growth. Inappropriate boundary selection can
erence mode, with many different alternatives thus easily lead to the unintentional netting out of
possible. In a survey of businesses one and a impacts.
half years after the 1994 Northridge Second, the economic consequences of seis-
(California, USA) earthquake, for example, one mic events are often felt most in the period imme-
quarter of businesses had failed to recover to the diately following an event but nevertheless
level of operation prior to the event (Tierney continue into the recovery period and longer.
1997). Some of these businesses may, however, Seismic events are also often felt as a series rather
have experienced a state of decline even without than as a single event, e.g., the 2010 and 2011
the event. Similarly, following an event it has Canterbury (New Zealand) earthquakes. These
also been proposed that new or more productive characteristics change not only economic flows
economic sectors may emerge, assisted by the but also stocks, resulting in lags that lead to
investment in reconstruction and opportunities oscillation in key economic variables through
to replace capital with newer and more produc- time. The conventional market-based economet-
tive capital (e.g., Skidmore and Toya 2002). ric, IO, SAM, and most CGE models require
significant adjustments if such time-dependent
Determining Economic Impact Assessment processes are to be included. Attempts to include
Boundaries temporal dynamics into economic assessment
It is very important in undertaken economic have been made by Okuyama et al. (2004) in the
impact assessment of seismic events that the specification of the sequential interindustry
boundaries of the study are clearly defined and model, by Barker and Santos (2010) in the for-
communicated with intended audiences. Gener- mulation of dynamic inoperability models, and
ally speaking, boundary challenges fall into three by Rose and Liao’s (2005) CGE model. It is
categories: spatial, temporal, and distributional. worth noting that dynamic analysis is also a key
First, the spatial extent of any assessment strength of SD and ABM approaches. The tem-
requires careful consideration if impacts are to poral analysis of disaster impacts is, as Rose
be appropriately accounted for locally, (2004) and Okuyama (2007) point out, still ad
918 Economic Impact of Seismic Events: Modeling
hoc in nature – lacking an underlying theory to investigating model parameters through numer-
support their application. ous experiments based on appropriate sampling
Finally, a common criticism made of many methods, i.e., Monte Carlo.
economic impact assessments is that only aggre-
gate or sector impacts are reported. There is how-
ever growing interest in understanding how Summary
economic impacts, both direct and higher order,
are distributed across businesses and communi- In the context of responding to seismic risk and
ties. Understanding economic consequences by events, the behaviors of formal economic markets,
socioeconomic and demographic variables alone, are unlikely to produce outcomes that are
including household, family, ethnic, and income overall in the best interests of communities. Since
types is potentially important to identifying the 1990s, economic modeling of seismic events,
socioeconomic vulnerability and, in turn, through and their associated responses, has risen as an
mitigation and adaptation, building resilient busi- important tool in assisting policymakers in man-
nesses and communities. It is worth noting that aging risk and adopting appropriate mitigation and
SAMs, CGE, SD, and ABM models, if appropri- response strategies. The application of economic
ately configured to include disaggregated house- methods to date can be grouped based on whether
hold types (or similar), are ideally suited to the assessment is ex post or ex ante an event and
addressing this challenge. Nevertheless, there is based on market or nonmarket analysis. The range
a current dearth of assessments that explicitly of impacts that may be considered in assessments
address distributional issues. are vast, from analysis of only direct damages to
physical assets to analysis of higher-order
Building Confidence in Economic Impact
upstream and downstream business disruptions,
Assessment
competition for scarce materials during rebuild,
The validation of an economic impact assessment
and even valuation of destruction of natural capital
can only be judged against its intended purpose.
and psychological trauma. Although conventional
Unfortunately, like all models of complex sys-
IO and econometric models dominated much of
tems, there is no mathematical or absolute proof
the early literature, during the last decade there has
that an economic impact assessment may fulfill
been a strong interest in extended IO, SAM, and
its purpose. Furthermore, we can conclude that
CGE models. Modeling techniques applied in dis-
any economic impact assessment is inherently
ciplines outside of economics, particularly SD,
imperfect. Like all models, any assessment rep-
ABM, and integrative modeling approaches, are
resents a simplification of reality. It is not so
also now being used to assess the economic con-
much whether an assessment is valid or not but
sequences of seismic events. Although the poten-
whether it is useful. Methodological pluralism is
tial for use of nonmarket valuation techniques is
required when assessing the economic impacts of
also recognized, to date their uptake has been
seismic events as there is “no one size fits all”
limited. It is important to note that “no one size
assessment method. Nevertheless, confidence
fits all” method exists and that all methods are
may be built in applying methods through the
inherently imperfect. In this regard, there are
use of sensitivity and uncertainty analysis. Sensi-
many research challenges still to be addressed.
tivity analysis is essential for learning how gen-
eral patterns of behavior exhibited by impact
models change in response to key input parame- Cross-References
ters, narrowing down key drivers of change in a
simulation. While sensitivity analysis may be ▶ Community Recovery Following Earthquake
used to identify the key exogenous drivers of Disasters
change, uncertainty analysis adds value by ▶ Damage to Infrastructure: Modeling
Economic Impact of Seismic Events: Modeling 919
the magnitude of the slowdown is only partly Public spending on reconstruction, in contrast,
a function of the nature of the disaster (e.g., can endure as the duration of the reconstruction
hydrometeorological vs. geophysical), its spatial efforts can last for more than a decade. Thus, the
extent, its magnitude, and its timing (e.g., with observed increase in the budget deficit, and the
relation to the crop cycles). The slowdown is also, consequent increase in government borrowing, is
maybe mostly, a function of the economic vul- largely due to increased spending rather than
nerabilities of the affected region (e.g., the degree decreased revenue.
and nature of diversification in the economic This picture, however, does not describe accu-
sectors that dominate the affected region). rately the ways in which very poor countries have
In high-income economies, however, even to conduct their fiscal policy in the aftermath of
E
large disasters seem to have very little macroeco- a catastrophic disaster. Low-income countries
nomic aggregate effects. Thus, even a mega- have less access to credit markets; and their
disaster, such as the triple earthquake-tsunami- high cost of borrowing and their narrow tax
nuclear event in Japan in 2011, has had only very base prohibit them from spending adequately on
temporary impact on the aggregate Japanese reconstruction. Yet, even in these cases when the
economy. In the long run, it appears that disasters government is not able to spend on an aggressive
have little measureable impact on the national recovery effort, this failure does not necessarily
economy (Cavallo et al. 2013). imply reduced public sector deficits as spending
Impacts on prices are not well documented. does not increase. In these cases, the inability to
The scant evidence that is available appears to spend on recovery implies that incomes do not
suggest that goods and services that are directly recover, and the tax revenue is decreased. This is
required for the reconstruction efforts seem to frequently compounded by the fact that in poor
experience price increases as a consequence of countries, much tax revenue comes from taxes on
increased demand and limited supply. Most typ- imports, and imports typically decline precipi-
ical is an increase in the cost of construction and tously when incomes are significantly decreased.
construction workers’ wages. These increases do Thus, in both high- and low-income countries,
not seem, however, to translate into more general earthquakes will entail increasing government
price increases unless the event itself generated budget deficits. It is only in rare cases, when, for
specific bottlenecks and supply constraints, e.g., example, much of the damage was privately
the destruction of oil refineries in the Gulf of insured or reinsured internationally that the gov-
Mexico during hurricane Katrina of 2005. ernment finances may not suffer much (e.g.,
Doyle and Noy in press).
in spending frequently is concentrated in luxury The limited evidence seems to suggest poten-
goods (less surprising) and health and education tially very long-term adverse impacts at the
(clearly more worrying as this decrease may local level that are quite distinct from the rela-
entail long-term adverse consequences). This tively benign impacts at the national level, even
pattern is not unique to low-income countries for fairly short horizons. At the very extreme,
but is also documented, for example, for house- some regions never recover at all from a large
holds in the Japanese city of Kobe after the 1995 event. Such an extreme example is the island of
earthquake (Sawada and Shimizutani 2008). Kefalonia, in which a devastating earthquake in
For trade, it is not unusual to observe declines 1953 led to mass exodus from the island, and the
in imports, and especially in imported luxury population on the island was permanently
goods, but these are typically offset by increases reduced by more than 20 %. While such an
in importing of building materials, fuels, and extreme case is unusual, earthquakes and other
other goods associated with the reconstruction geophysical events can lead to a redistribution of
effort. After an initial decline in exports, as the population and economic activity that may per-
productive capacity of the affected region is dam- sist for long periods of time.
aged, the overall and longer-term impact of The exact nature of the long-term impact of an
a typical earthquake shock on exports is largely earthquake on the regional/local economy
contingent on the types of exports the country depends on the structure of the economy of the
specializes in and the nature of the earthquake region. For example, the agricultural sector is
damage to production facilities, transportation relatively resilient to earthquakes, and even if
infrastructure, etc. In some cases, one can poten- the earthquake damages some of the processing
tially observe a counterintuitive increase in and transportation infrastructure, this can be
exports, especially if the earthquake disaster led quickly rebuilt. Other sectors may be more vul-
to a depreciation of the currency, making nerable, especially if these were already facing
exported goods more affordable for foreign difficulty before the event. In the city of Kobe, for
buyers. example, the heavy industry of the region was
While predicting the typical impact of an already facing increasing competition from else-
earthquake on exports of goods is fraught with where in Asia (in particular from China) before
uncertainty, exports of services are almost always the 1995 earthquake. The earthquake itself led to
harmed by a disaster. Especially vulnerable is the a speeding up of a process that would have poten-
tourism sector, as the tourist trade is mostly tially occurred even without the earthquake.
dependent on image branding and perception Much of the heavy industry was not rebuilt and
rather than on the actual capacity to supply ser- reopened in Kobe in the years after the quake, in
vices. Thus, tourism declined significantly in spite of massive government assistance for the
places where earthquake damages occurred in region.
tourist areas (Sri Lanka and Southern Thailand
in 2004 or Japan, Chile, and New Zealand
in 2011). Paying for Reconstruction
Economic Recovery Following Earthquakes Disasters, Table 1 Earthquake losses insured by country for some
recent earthquakes (Source: Swiss Re sigma catastrophe database)
Date Country Losses % GDP Insured losses %
April 2009 Italy $4B 0.2 14
January 2010 Haiti $8B 121 1
February 2010 Chile $30B 18 27
April 2010 Mexico $0.95 0.1 21
September 2010 New Zealand $6B 5.3 81
February 2011 New Zealand $15B 10 80
March 2011 Japan $300B 5.4 17
October 2011 Turkey $0.75B 0.1 4 E
Economic Recovery Following Earthquakes Disasters, Table 2 Principles for natural disasters’ sound workforce
strategies (Chang-Richard et al. 2013)
Principles Guidelines
1. Target and tailor measures Employment strategies (e.g., public works, employer subsidies) should be
with flexibility targeted for the most troubled sectors and/or needed groups and individuals; if
fiscal capacity is constrained, resources should be directed to sector groups that
have the most potential to generate employment
Social protection measures should be as flexible as possible to meet the needs of
beneficiaries, reducing application “red tape,” providing easy access, extending
the coverage and length of support
2. Institutional and An overarching recovery strategy with employment being one of key
organizational innovation components provides a framework for an efficient, responsive labor market
response
A truly dedicated response team in respective policy domains (regional
development organizations, rebuilding advisory services, skills specialist
officers) will drive efficiency and higher levels of engagement from the industry
and community
One-stop design is critical to the success and accessibility of employment
services
3. A diversity of multiagency A diversity of collaboration among multiple agencies is essential to ensure
collaboration coverage of assistance and combination of capacity (financial, technical, and
“know-how”) for effective service delivery
Expertise and “know-how” knowledge of nongovernment organizations,
particularly the International Labour Organization (ILO), should be drawn on,
promoting a collaborative approach to supporting employment and livelihood
assistance
4. A focus on sustainable A commitment to sustainable employment outcomes requires social protection
employment outcome measures to be coupled with training and skills development programs, job
creation activities to be combined with productivity and skills enhancement, and
disaster labor market response to be integrated in regional workforce
development strategies
Workplace-based employment retention and livelihood support play an
important role in overcoming workers’ participation barriers
Self-efficacy and employability concepts should be built into the design of self-
employment assistance and training programs, respectively, with a focus on
linking support measures with direct employment outcomes
5. Capacity-based preparedness Predesigned labor market response policies which can be quickly activated
and response ensure timeliness and efficiency of employment assistance
Existing networks and relationships, together with a strong leadership, allow
resources and capabilities to be quickly assembled and/or redirected in response
to the changing needs over the recovery time
Fiscal, technical, staffing and technological capacity is essential to ensure
efficiency within service delivery
6. Local input and place-based Economic and business recovery should increase the engagement of local
solutions organizations and people with local and industry knowledge
Employment assistance provided to displaced workers and job seekers,
particularly those disadvantaged (youth, women, people with disability, long-
term unemployed), can be based on their individual needs and opportunities
identified by themselves
7. Robust labor market An evidence-based approach, building on what has proved to be good practice in
intelligence international experience is instrumental in designing a similar measure. Business
surveys will allow for more responsive and targeted assistance
Labor market programs should be based on robust labor-force information which
will contribute to a sustainable and resilient employment services system
Economic Recovery Following Earthquakes Disasters 927
The need to replace lost homes, infrastructure, damage or safety cordons that have been put in
and commercial properties generates a surge in place. This is a major decision for an organization
demand for reconstruction that can act as a short- as it may contribute to employee decisions to stay
term stimulus to the local economy. This stimulus or leave as well as impacting on customer flows
benefits not only the construction sector but also through patterns of foot traffic and agglomeration
other sectors that directly support the construc- effects (where similar businesses are colocated
tion sector (such as lawyers and accountants, thus increasing the customer traffic for all of
office-supplies providers, and accommodation them). The decision to relocate is a complex deci-
providers) as well as those sectors that indirectly sion and is complicated further by such factors as:
benefit from the wages and salaries paid to those does the business own, rent, or lease the building;
E
working on the rebuild (such as the hospitality level of damage to the building; damage to the
sector, retailers, and arts and recreation neighboring buildings and the presence of
providers). a safety cordon (e.g., Chang et al. 2014)
While demand surge in the construction mar- preventing access to the building or to nearby
ket can provide a buffer to an economy affected buildings; and availability of infrastructure
by disaster, it can also create imbalance and dys- services.
function. The key characteristic that distin- If the business owns the building, it may have
guishes the post-disaster condition from normal more control over building access, whereas the
times is time compression which requires an landlord or the landlord’s insurer may prevent
unusual pace of capital expenditures and recon- a business in rented premises from gaining
struction (Olshansky et al. 2012). In many cir- access. This also applies to speed of repairs,
cumstances, this leads to shortages in capital how insurance payouts are used to repair or
(machinery, trucks, diggers, etc.), materials upgrade the building, and the disruption this has
(concrete, steel, timber, etc.), and skills on access to the building. These decisions are
(engineers, builders, plumbers, painters, etc.) typically in the hands of the owner (with some
(Chang-Richard et al. 2012). Such resourcing constraints from emergency services and the
bottlenecks can lead to price increases and insurer), while a renter or lessee may find that
inflows of large amounts of money, and new they have limited influence over such decisions.
players in the construction market can also lead Level of damage to the building clearly has
to increases in profiteering, corruption, and fraud. implications for the relocation decision. A badly
damaged building is a safety and reputational risk
to the business and will typically force a closure
Business Recovery or relocation. Less obvious however is the situa-
tion where the business’ building is undamaged
Business recovery following an earthquake is but where nearby buildings are badly damaged,
a crucial aspect for community recovery. Busi- closed, and cordoned. In this situation, there is
nesses provide employment, goods and services, a neighborhood effect in which customer num-
as well as social support for their employees and bers drop significantly as they see the badly dam-
for the community at large by way of gathering aged neighborhood and assume that all
places. Key aspects impacting on business recov- businesses in the area are impacted equally
ery include levels of earthquake damage to their (Chang and Falit-Baiamonte 2002). This situa-
own premises and surrounding buildings, whether tion is particularly difficult for the undamaged
they have to relocate, disruption to their supply business as business interruption insurance is
chain, changes in consumer behavior, and the typically dependent on physical damage to trig-
impact of infrastructure disruption (Tierney 1997). ger a payout. Since they are not physically dam-
In the immediate aftermath of a damaging aged, there may be no payout, but their cash flow
earthquake, a business owner may need to make is definitely impacted due to the neighborhood
the decision to relocate their business due to effect.
928 Economic Recovery Following Earthquakes Disasters
Availability of infrastructure services such as attention from the business owner, and endanger
power, water, and wastewater is a crucial feature the business further as the owner may focus too
for business recovery and the decision to relocate. much effort on the insurance difficulties and too
A building may be undamaged and the business little on keeping their business running effec-
owner desiring to reopen as quickly as possible, tively (Brown et al. 2013).
but the lack of services may make this impossi-
ble. In addition, if infrastructure services are
functioning, but at a lower than normal level Government Policy Impacts on
(e.g., electricity “brownouts” or poor quality Economic Recovery
water), this may have an impact on ability to
conduct business at a competitive quality and The impact of government policy on economic
undermine the rate of customer return and loyalty recovery can be significant. Policies such as
to the business. national or regional earthquake insurance
In the aftermath of an earthquake, it is quite schemes, direct government loans and grants,
possible that a business and its neighborhood are employment subsidies, and the establishment of
largely undamaged with full services working. government recovery bodies are all examples of
However, there may be significant impacts on government interventions that can be hugely ben-
their supply chain, both upstream (disrupted sup- eficial or undermine long-term recovery if
pliers) and downstream (customer impacts and misguided.
behavior changes). A supermarket may be fine National or regional earthquake insurance
following an earthquake with only minor impacts schemes can provide much needed capital to an
on their part of town, but their primary supplier’s earthquake-hit region following a major event.
warehouse may have collapsed and deliveries to Examples include the New Zealand EQC
the supermarket heavily disrupted, so customers (Earthquake Commission), the Japanese JER
go to other supermarkets and get in the habit of (Japanese Earthquake Reinsurance) scheme, and
doing so. Even after the supplier can once again the California CEA (California Earthquake
deliver reliably (or new suppliers are brought Authority). These schemes were all formed
on-stream by the supermarket), customers may through government legislation but vary consid-
be slow to return due to maintained perception erably on how they function. Characteristics that
and new habits. differentiate them include: whether policies
Business insurance is a two-edged sword. cover both residential and commercial buildings,
A swift settlement at expected levels of payout whether policies are mandatory for building
can be a lifesaver to a business, allowing rapid owners, the extent to which the government acts
repair or replacement of the building, inventory, as insurer of last resort (it covers further losses
and equipment, while also providing much once the scheme’s resources are exhausted), and
needed cash flow through a business interruption who is responsible for administering the scheme
policy. This is a likely scenario if the earthquake (public or private entities).
damage is not widespread and the business owner An alternative to earthquake insurance
has a thorough understanding of their insurance schemes is direct government loans and grants
policy and a good relationship with their insurer. following an earthquake. In many countries,
Unfortunately these provisos are often not met. earthquake insurance is not available as part of
Damaging earthquakes tend to be regional in a building owner’s policy and may be unavailable
nature resulting in widespread disruption. As in any form. In countries such as Italy or China,
a result, there are a large number of claims that low levels of earthquake insurance are the norm
the insurers are attempting to deal with at the (Asian Development Bank 2008). Where this is
same time. Insurers become overwhelmed with the case, a government may still intervene by
the consequent workload with the result that set- reconstruction of homes and infrastructure and,
tlements are slow in coming, require much much less often, commercial buildings. This
Economic Recovery Following Earthquakes Disasters 929
intervention may come in the form of direct gov- local government back on its feet, and supporting
ernment intervention or in the form of grants and local social needs. In a few cases, even the central
loans to the building owners. Such intervention is government may be overwhelmed by the recon-
a form of national self-insurance. This provides struction needs; that was the case in Haiti after the
the capital to rebuild regionally, while sharing the 2010 Port-au-Prince earthquake when the Interim
economic burden with the rest of the country. Haiti Recovery Commission led by former US
This risk sharing is of course a characteristic of President Bill Clinton was established jointly by
any insurance scheme, whether explicit or the United Nations and the Haitian government.
implicit.
Following an earthquake, governments may
E
intervene to support business directly. This may Metrics of Economic Recovery
include business recovery loans (such as from the
US Small Business Administration) or through How does a city, a region, or a country know they
employment subsidy schemes (such as the Earth- have recovered? What are the ways of measuring
quake Support Subsidy used following the the progress from devastation to normality? The
Christchurch earthquakes in New Zealand). The complex and multifaceted process of economic
benefit to such government intervention may be recovery has winners and losers, and it has waves
the long-term retention of businesses in a region of progress and ebbs of disappointment. Metrics
(Hatton et al. 2012). This will typically have that help us to see and measure the progress in
a number of positive flow-on effects including population change, consumer spending and eco-
higher employment, reduced risk of population nomic growth, extent of business relocations, and
dislocation, and generally an increase in the levels of reconstruction all contribute to the tap-
speed of economic recovery. The potential draw- estry that is economic recovery.
backs to such direct government intervention A key determinant of economic recovery is the
comes when the intervention is longer term state of the population. If there was significant
(thus creating a business-government depen- population dislocation and loss, to what extent
dency), where investment is inefficient (keeping has the population returned and resettled? If pop-
businesses alive that were on the verge of failing ulation did not leave, will the population be
prior to the earthquake, and the government retained through the difficult recovery period?
grant/loan simply keeps them alive until the aid As noted earlier in this entry, an important feature
runs out), or where the intervention has insuffi- of population recovery is the employment oppor-
cient accountability resulting in budget blowouts tunities available. Other key measures are avail-
and increased potential for corruption. ability and quality of utilities and amenities
In very large events, the national/central gov- (power, water, communications, shopping cen-
ernment may form a Recovery Authority such as ters, schools, and higher education) and direct
the Canterbury Earthquake Recovery Authority measures of population gain and loss
(CERA) in New Zealand following the (immigration, emigration, growth of nearby
2010–2011 earthquakes or the Badan towns, number of working visas issued, etc.).
Rehabilitasi dan Rekonstruksi (BRR) in Indone- Each of these can be measured and contribute to
sia following the 2004 earthquake-tsunami in the complex picture of economic recovery
Aceh. Such an intervention provides a central (Table 3).
authority to coordinate services, aid, and policy The condition of a city or region’s critical
during the difficult recovery period. This will infrastructure is a primary element in economic
more typically be the case where the event over- recovery. Earthquakes are damaging to road,
whelms local government capabilities or where power, water, and telecommunication networks.
central government is providing significant direct The recovery of these services can be measured
funding to support the recovery (e.g., in rebuild- in terms of total level of damage (in physical and
ing housing and critical infrastructure), getting monetary terms) and the progress of
930 Economic Recovery Following Earthquakes Disasters
Economic Recovery Following Earthquakes Disas- profitability. Are levels of temporary relocation
ters, Table 3 Metrics of economic recovery still high? Are agglomerations (groups of like
Description Measure businesses such as a fashion district) recovering
Business birth and Percentage of total number of and regathering? More general economic mea-
death rate businesses sures such as levels of exports, total retail sales,
Direct population Number of people and and other measures of economic activity round
statistics: percentage of base population
immigration, out the broader picture of business recovery.
emigration, working Measures that capture the state of the various
visas issued economic sectors will also highlight the winners
Levels of Total numbers employed and and losers in the aftermath of earthquakes and
employment and unemployed, workforce
provide a picture of sectoral recovery that can
unemployment participation rates, and
percentage of total population inform various business and government inter-
of working age ventions. For example, tourism and education
Critical Total km to be repaired, total are export sectors that are often hard hit following
infrastructure monetary value of losses, km of earthquake events. Interventions to support the
reconstruction and repairs now complete,
recovery percentage of reconstruction
recovery of such sectors (e.g., collaborative inter-
complete in km and dollars. national marketing programs) can break the cycle
Percentage of rebuild complete of bad news and accelerate the recovery of such
Residential and Number of damaged houses, hard-hit sectors.
community building number of houses repaired, The growth or shrinkage of an economy can be
reconstruction demolished, or rebuilt.
Percentage of total work measured in a multifaceted way to gain a rich
completed picture of general economic recovery. This gen-
Commercial building Number of damaged buildings, eral economic recovery is dependent on all of the
reconstruction number of buildings repaired, earlier characteristics of population, state of ame-
demolished, or rebuilt. nities, business stability, and employment and
Percentage of total work
completed. Similar but using levels of reconstruction. The state of the economy
square meters of floor space can of course be measured directly as well.
rather than number of buildings Growth of regional GDP, levels of employment,
Business growth and Total sales in retail and tax revenues, and total retail spending are all
stability wholesale, sector-specific
useful measures of the state of the overall econ-
measures of growth,
profitability, and employment omy, and the trends on these measures provide
Levels of business Percentage of businesses in a moving picture of the recovery.
relocation temporary accommodation
Overall measures of Regional GDP trends, total tax
economic recovery take, retail sales, and Summary
employment (as noted above)
countries. However, the picture for local busi- Brown C, Seville E, Vargo J (2013) The role of insurance
nesses may be far more complex. They must in organisational recovery following the 2010 and
2011 Canterbury earthquakes. Resilient Organisations
deal with the economic vulnerabilities that Research Report 2013/04. http://www.resorgs.org.nz/
existed prior to the quake compounded by the Publications/research-reports.html. Retrieved 12 May
impacts of the disaster, including the need to 2013
relocate, loss of key staff, changes in consumer Cavallo E, Noy I (2011) The economics of natural
disasters – a survey. Int Rev Environ Resour Econ
behavior, disruption to their supply chain, and 5(1):63–102
overwhelmed insurers. Cavallo E, Galiani S, Noy I, Pantano J (2013) Catastrophic
In the midst of the challenges faced locally natural disasters and economic growth. Rev Econ Stat
will be opportunities for those able to tap into 95(5):1549–1561
Chang SE, Falit-Baiamonte A (2002) Disaster vulnerabil- E
the changes caused by the disrupted economy ity of businesses in the 2001 Nisqually earthquake.
including location of consumers, changed demo- Glob Environ Change Part B Environ Hazard
graphics, insurance payouts, geographic disloca- 4(2–3):59–71
tions, and other changes. Then there is the Chang SE, Taylor JE, Elwood KJ, Seville E, Brunsdon D,
Gartner M (2014) Urban disaster recovery in Christ-
inevitable reconstruction boom, regardless of church: the central business district cordon and other
whether it is fuelled by insurance payouts, per- critical decisions. Earthquake Spectra. doi:10.1193/
sonal savings, or direct government intervention. 022413EQS050M
There will be both disruptions and opportunities. Chang-Richard AY, Wilkinson S, Seville E, Brunsdon
D (2012) Resourcing issues in past disaster recoveries:
Tracking this unfolding and complex picture is an some perspectives. Resilient Organisations Research
important feature of economic resilience as pol- Report 2012/07. http://www.resorgs.org.nz/Publica-
icy makers and business owners review the met- tions/research-reports.html. Retrieved 12 May 2013
rics that can warn of troubled economic sectors Chang-Richard AY, Seville E, Wilkinson S, Walker
B (2013) Building natural disaster response capacity:
needing help as well as highlight opportunities sound workforce strategies for recovery and recon-
for growth. struction in APEC economies. APEC Human
Resources Development Working Group, Asia Pacific
Economic Cooperation. http://publications.apec.org/
Cross-References publication-detail.php?pub_id=1534. Retrieved 15
August 2014
De Mel S, McKenzie D, Woodruff C (2012) Enterprise
▶ Community Recovery Following Earthquake
recovery following natural disasters. Econ J 122(559):
Disasters 64–91
▶ Earthquake Disaster Recovery: Leadership and Doyle L, Noy L (in press) The short-run nationwide mac-
Governance roeconomic effects of the Canterbury earthquakes.
New Zealand Economic Papers, Taylor and Francis
▶ Economic Impact of Seismic Events:
Hatton T, Seville E, Vargo J (2012) Improving the resil-
Modeling ience of SMEs: policy and practice in New Zealand.
▶ Reconstruction Following Earthquake Resilient Organisations Research Report 2012/12.
Disasters http://www.resorgs.org.nz/Publications/research-reports.
html. Retrieved 12 May 2013
▶ Resilience to Earthquake Disasters
Kachali H, Stevenson JR, Whitman Z, Seville E, Vargo J,
▶ Resourcing Issues Following Earthquake Wilson T (2012) Organisational resilience and recov-
Disaster ery for Canterbury organisations after the 4 September
2010 earthquake. Australas J Disaster Trauma Stud
2012(1):11–19
Katz J (2013) The big truck that went by. Palgrave Mac-
References millan, New York
Olshansky RB, Hopkins LD, Johnson LA (2012) Disaster
Asian Development Bank (2008) Earthquake insurance: and recovery: processes compressed in time. Natl Haz-
lessons from international experience and key issues ard Rev (ASCE) 13(3):173–178
for developing earthquake insurance in the PRC, Sawada Y, Shimizutani S (2008) How do people cope with
observations and suggestions, 24 June 2008. http:// natural disasters? Evidence from the great Hanshin‐
reliefweb.int/sites/reliefweb.int/files/resources/F_R. Awaji (Kobe) earthquake in 1995. J Money Credit
pdf.pdf. Accessed 5 Dec 2013 Bank 40(2–3):463–488
932 EEE Catalogue: A Global Database of Earthquake Environmental Effects
SPUR (2012) Safe enough to stay. Spur Report 01/2012. Effects (EEEs; i.e., any phenomena generated by
www.spur.org. Retrieved 12 May 2013 a seismic event in the natural environment;
Stevenson JR, Kachali H, Whitman Z, Seville E, Vargo
J (2011) Preliminary observations of the impacts the Michetti et al. 2004, 2007; Guerrieri et al. 2007)
22 February Christchurch earthquake on organisations is available for a very large number of earthquakes
and the economy: a report from the field that occurred not only in the instrumental period
(22 February – 22 March 2011). N Z Soc Earthq Eng and in historical time but also in the prehistorical
Bull 44(2):65–76
Tierney KJ (1997) Business impacts of the Northridge period (paleoearthquakes, e.g., Mc Calpin 2009;
earthquake. J Crisis Conting Manag 5(2):87–97 Reicherter et al. 2009). However, available infor-
Venn D (2012) Helping displaced workers back into jobs mation located in different sources (scientific
after a natural disaster: recent experiences in OECD papers, historical documents, professional reports)
countries, OECD social, employment and migration
working papers, vol 142 OECD Publishing, Paris can often be difficult to access.
The EEE Catalogue is a free accessible data-
base designed at global level to collect informa-
tion about the characteristics, size, and spatial
EEE Catalogue: A Global Database of distribution of Earthquake Environmental Effects
Earthquake Environmental Effects in a standard way from modern, historical, and
paleoearthquakes. This database aims to properly
Luca Guerrieri1, Alessandro Maria Michetti2 and retrieve the available information about EEEs at a
Pablo G. Silva3 global level and archive it into a unique database,
1
Department of Geological Survey, ISPRA, in order to facilitate its use for seismic hazard
Istituto Superiore per la Protezione e la Ricerca purposes. The database implementation has been
Ambientale, Rome, Italy endorsed by the INQUA TERPRO Project #0811
2
Dipartimento di Scienza e Alta Tecnologia, (“A global catalogue and mapping of earthquake
Università dell’Insubria, Como, Italy environmental effects”; 2008–2011), through a
3
Departamento de Geologı́a, Universidad de working group coordinated by ISPRA – Geolog-
Salamanca, Escuela Politécnica Superior de ical Survey of Italy. The first official release of
Ávila, Avila, Spain the EEE Catalogue was done under the frame-
work of the XVIII INQUA Congress, held in
Bern in July 2011.
Synonyms The information included in the EEE Cata-
logue relates to Earthquake Environmental
Earthquake ground effects; Environmental seismic Effects triggered by individual seismic events,
intensity; Macroseismic scale; Paleoseismology; classified according to the EEE types considered
Seismic hazard in the Environmental Seismic Intensity scale
ESI-07. For each event, information is collected
at three levels of increasing detail: earthquake,
Introduction locality, and site. Also available imagery docu-
mentation (photographs, videos, sketch maps,
The recent destructive earthquakes that occurred stratigraphic logs) can be stored in the database.
in Japan (2011; e.g., Lay et al. 2013) and New Moreover, within the EEE Catalogue, the epicen-
Zealand (2010–2011 earthquake sequence; e.g., tral and local intensity values based on the ESI-07
Quigley et al. 2012) have clearly pointed out that scale are recorded. This allows an objective com-
traditional seismic hazard assessment based only parison of earthquake intensity for events that
on vibratory ground motion data needs to be inte- occurred in different areas and/or in different
grated with information about the local vulnerabil- periods.
ity of the territory to earthquake occurrence. The quality of the database in terms of com-
Nowadays, a huge amount of information about pleteness, reliability, and location resolution is
the characteristics of Earthquake Environmental obviously time dependent and, therefore, is
EEE Catalogue: A Global Database of Earthquake Environmental Effects 933
generally less reliable with increasingly more each locality and site, it is possible to better
ancient events. Nevertheless, even where the locate each individual element (red box in detail
information is less accurate (historical earth- in C). Some general information about the local-
quakes, archeoseismological evidence), the ity (e.g., local intensity) and site (e.g., EEE
documented effects are typically the most rele- classification, description, reference, and photo-
vant (i.e., most diagnostic for intensity assess- graph) are visible on a Google Balloon, by
ment). Similarly, the information from clicking on the locality (frame D) and site
paleoseismic investigations, although poorly rep- (frame E). The entire database and photogallery
resentative of the entire scenario, still includes (including maps, video, and other .pdf docu-
significant data (i.e., local coseismic fault dis- ments) related to a specific earthquake can be
E
placements), which are very helpful in estimating consulted by clicking on the earthquake string in
a minimum size for the earthquake. At present the table of contents (frame F).
about one hundred earthquakes are included in The EEE Catalogue is also available on the
the EEE Catalogue. IAEA-ISSC (International Atomic Energy
The implementation of the EEE Catalogue is Agency- International Seismic Safety Centre)
always in progress through a Web Implementa- portal (https://issc.iaea.org/home.php) for use
tion Interface developed for remote data collec- and dissemination within the worldwide nuclear
tion at the following URL: http://www. engineering community. To this end, a new data-
eeecatalog.sinanet.apat.it/login base infrastructure has been developed by the
Earthquake records can be edited only by adaptation and customization of the EEE Cata-
authorized compilers. Data compilation must be logue to the specific IAEA-ISSC standard
done at three different levels of generalization requirements (e.g., content, reliability, magni-
(earthquake, locality, and site). Moreover, it is tude thresholds, geographic priorities, IPR
currently possible to upload (i) the pattern of (intellectual property rights), security policies).
coseismic ruptures as linear elements For a quick field survey of EEE data consis-
(previously edited as .kml file), at earthquake tent with the EEE Catalogue standards, an appli-
level; (ii) photographs, sketches, or videos cation has been developed named “Earthquake
documented a single effect, at site level. Geo Survey” that can be used by IOS and
In the next section the structure of the database android mobile devices (e.g., smartphones and
including mandatory fields at earthquake, local- tablets). Via this application, the user is able to
ity, and site level is described. document on-fault and off-fault effects, collect
Data can be explored on a public interface GPS waypoints, and indirectly geo-tag the col-
(Fig. 1) based on Google Earth; the web page lected photos. The types of effects that can be
address is available at http://www.eeecatalog. reported and described in detail are surface
sinanet.apat.it/terremoti/index.php faulting (on-fault effects) and off-fault effects
From this URL, it is possible to view (Fig. 2) including slope failures, liquefaction, tsunami,
some general information about the earthquake ground cracks, hydrological anomalies, and
visible on a Google Balloon, by clicking on the other effects like trees shaking. For example,
epicenter (yellow star, frame A) including the regarding the earthquake-induced liquefaction
pattern of coseismic ruptures (primary or sec- phenomena, the user can select a subtype of
ondary), surveyed or interpreted, by clicking on liquefaction from a list including the three most
the gray triangle corresponding to rupture zone characteristic ones (ejection of sandy material,
and activating the box rupture zone (frame B). subsidence, and lateral spreading), describe in
The geographic distribution of localities (green detail the failure within an extra field, report the
drops) and sites (classified in 7 EEE categories) maximum diameter of sand boils, select either
can be viewed by clicking on the gray triangle water or sand ejection, and finally locate the site
corresponding to locality level and activating the by activating the GPS and take a photo of the
box locality (frame C). By zooming in detail on liquefaction manifestation. In addition, the
934
EEE Catalogue: A Global Database of Earthquake Environmental Effects, Fig. 1 The public interface of the EEE Catalogue, developed on Google Earth. http://www.
eeecatalog.sinanet.apat.it/terremoti/index.php
EEE Catalogue: A Global Database of Earthquake Environmental Effects
EEE Catalogue: A Global Database of Earthquake Environmental Effects
EEE Catalogue: A Global Database of Earthquake Environmental Effects, Fig. 2 How to explore data into the EEE Catalogue (see text for details)
935
E
936 EEE Catalogue: A Global Database of Earthquake Environmental Effects
EEE Catalogue: A Global Database of Earthquake terminology (surface faulting, slope movements, ground
Environmental Effects, Fig. 4 The site table provides settlements, ground cracks, hydrological anomalies,
a description of the single effect as available in the orig- anomalous waves, other effects)
inal source and a first classification according to a standard
historical earthquakes, the following general induced by past earthquakes and therefore to
mandatory fields are considered: site name, type identify the areas where the human settlements
of effects, site latitude, site longitude, EEE and infrastructures are more exposed to this haz-
description, and reference for description. Then, ard. To this end, a high accuracy for the location
according to the type of effect, the following of EEEs becomes crucial. Typically, EEEs from
fields are highlighted: recent earthquakes are mapped with good accu-
racy immediately after the event. Nevertheless,
(a) Hydrological anomalies: surface water even for some historical earthquakes, it is possi-
effects and/or groundwater effects ble to retrieve with very good detail this informa-
(b) Anomalous waves and tsunamis: max wave tion (cfr. Serva et al. 2007).
height, anomalous wave width, and length of It is the case of the December 28, 1908
affected coast (M 7.1), Messina-Reggio Calabria earthquake
(c) Ground cracks: strike, dip, max opening, and and consequent tsunami (Fig. 5), where the EEE
length Catalogue allows to locate the earthquake/tsu-
(d) Liquefactions: max diameter and max nami effects over the present urban texture with
lowering a spatial resolution of a few meters, pointing out
(e) Slope movements: type, total volume, and the areas characterized by the highest risk.
max size of blocks (only for rockfalls) Comerci et al. (2008) have provided a careful
review of all available contemporary docu-
Where local surface faulting has been ments, i.e., technical and photographic reports,
observed, strike, dip, type of movement, and newspapers, and other archive material related
local offset (vertical and/or horizontal) are also to the seismic and tsunami event, allowing
recorded. For paleoearthquakes, at site level only to recognize and classify a large number
general fields and fields related to local surface (365) of independent environmental effects.
faulting (expected to be found by paleoseismic These have been categorized as slope move-
investigation) must be filled. ment, ground settlement, ground vertical move-
ment and coastal retreat, liquefaction, ground
crack, hydrological anomaly, gas emission,
Case Studies light, and rumble. A large part of the achieved
information has allowed to evaluate the epicen-
The major added value of the EEE Catalogue for tral intensity, besides local intensities at 67 sites,
seismic risk assessment is the possibility to by applying the Environmental Seismic Inten-
explore the scenarios of environmental effects sity scale ESI 2007.
EEE Catalogue: A Global Database of Earthquake Environmental Effects
EEE Catalogue: A Global Database of Earthquake Environmental Effects, Fig. 5 EEEs induced by the December 28, 1908, Messina Straits, Italy, earthquake in the area of
the Messina harbor. If information from contemporary sources is very precise, it is possible to use the EEE Catalogue also for local seismic microzonation
939
E
940 EEE Catalogue: A Global Database of Earthquake Environmental Effects
This study has led to the most comprehensive A similar result is shown by the spatial distri-
picture of the distribution and characteristics of bution of EEEs induced by the 1811–1812 New
coseismic environmental effects of that earth- Madrid, Missouri, earthquakes, mapped in Fig. 7.
quake, a crucial knowledge to better estimate Indeed, the most relevant primary and secondary
the future impact of a similar earthquake on a effects are located along the Mississippi Valley
region that has seen a broad, and often poorly near New Madrid, consistently with the surface
concerned, urban, and infrastructural develop- projection of the causative faults, and unques-
ment since then. tionably provide diagnostic elements for
Furthermore, the EEE Catalogue can reveal assessing an epicentral intensity equal to XI
possible trends in the spatial distribution of pri- (Guerrieri et al. 2011).
mary and secondary effects. The spatial distribu- In the winter of 1811 and 1812, the New
tion of EEEs induced by the October 8, 2005, Madrid Seismic Zone generated an earthquake
Muzaffarabad, Pakistan, earthquake shows that sequence that included three very large to great
the location and amount of surface faulting are earthquakes. The three largest earthquakes in this
consistent with the spatial distribution of sequence are estimated to be of moment magni-
coseismic slope movements, in terms of both tude M 7.8–8.1 and were felt as far away as
areal density and size (Fig. 6). Hartford, Connecticut, Charleston, South Caro-
Ali et al. (2009) provide a detailed field survey lina, and New Orleans, Louisiana. In the northern
of EEEs in the epicentral area of the part of the Mississippi embayment, these earth-
Muzaffarabad event. The displacement and sur- quakes caused widespread liquefaction (Fuller
face expression of the causative fault and accom- 1912; Saucier 1977; Obermeier 1989) and severe
panying secondary environmental effects were ground failure (Tuttle and Barstow 1996). Lique-
observed at a number of places along a capable faction was reported about 200 km northeast of
thrust fault structure (Kashmir Thrust). the New Madrid Seismic Zone in White County,
A complex, clearly segmented, at least 112-km- Illinois; 240 km to the north-northwest near
long surface rupture was mapped with maximum St. Louis, Missouri; and 250 km to the south
values of vertical displacement on the order of near the mouth of the Arkansas River. Although
4–7 m mainly along the central segment of the the exact magnitudes of the earthquakes are
rupture (52 km between Muzaffarabad and debatable, there is no doubt that a repeat of an
Balakot) associated with maximum slip at depth 1811–1812 New Madrid sequence today would
and a major portion of the energy release. The cause a great deal of damage in the central United
adjacent segments are instead characterized by States (Tuttle 1999).
scattered minor surface ruptures with a few cen-
timeters of displacement, accompanied by exten-
sive surface cracking, landslides, and severe Future Developments
damage, concentrated in a narrow belt along the
fault trace. At June 2014, 101 earthquake records have been
A maximum intensity of XI on the Modified collected into the EEE Catalogue: among them,
Mercalli Intensity (MMI) scale and on the Envi- 44 are modern (after 1900), 39 are historical, and
ronmental Seismic Intensity scale (ESI 2007) was 18 are paleoearthquakes. Other 31 earthquake
recorded in the epicentral area between records are currently under implementation.
Muzaffarabad and Balakot. Extremely severe Most of the compiled earthquake records have
damage and very important secondary environ- occurred in Europe (62), then in South America
mental effects in the hanging wall adjacent to the (19), in Asia (16), and in North America (4).
trace of the causative fault plane are mainly due Nevertheless, the compilation of the EEE
to near-fault strong motion and rupture directivity Catalogue is still ongoing with the aim to collect
effects. any available information resulting from
EEE Catalogue: A Global Database of Earthquake Environmental Effects
EEE Catalogue: A Global Database of Earthquake Environmental Effects, Fig. 6 Surface faulting and slope movements induced by the October 8, 2005, Muzaffarabad,
Pakistan, earthquake
941
E
942
EEE Catalogue: A Global Database of Earthquake Environmental Effects, Fig. 7 EEE induced by the December 16, 1811, New Madrid, Missouri, USA, earthquake.
Primary and secondary effects indicative of intensity XI in the ESI 2007 scale are located in the epicentral area along the Mississippi Valley
EEE Catalogue: A Global Database of Earthquake Environmental Effects
EEE Catalogue: A Global Database of Earthquake Environmental Effects 943
specific field surveys after recent earthquakes, as identify the areas where the human settlements
well as from historical and paleoseismic studies. and infrastructures are more exposed to this
For the next future, the goal will be to improve source of potential hazard.
the completeness of the EEE Catalogue with the
aim to properly use collected data for an
improved assessment of seismic hazard based
Cross-References
on a credible scenario of EEEs.
▶ Archeoseismology
▶ Earthquake Location
Summary
▶ European Structural Design Codes: Seismic E
Actions
The EEE Catalogue is a free accessible database
▶ Intensity Scale ESI 2007 for Assessing Earth-
(http://www.eeecatalog.sinanet.apat.it/terremoti/
quake Intensities
index.php) designed at global level to collect
▶ Paleoseismic Trenching
information about the characteristics, size, and
▶ Paleoseismology
spatial distribution of Earthquake Environmental
▶ Paleoseismology and Landslides
Effects in a standard way from modern, histori-
▶ Paleoseismology: Integration with Seismic
cal, and paleoearthquakes.
Hazard
This database has been promoted with the aim
▶ Probabilistic Seismic Hazard Models
to properly retrieve the available information
▶ Seismic Actions Due to Near-Fault Ground
about EEEs at a global level and archive it into
Motion
a unique database, in order to facilitate their use
▶ Seismic Risk Assessment, Cascading Effects
for seismic hazard purposes. The information
▶ Time History Seismic Analysis
included in the EEE Catalogue pertains to Earth-
▶ Tsunamis as Paleoseismic Indicators
quake Environmental Effects triggered by indi-
vidual seismic events, classified according to the
EEE types considered in the Environmental References
Seismic Intensity scale ESI-07. For each event
the information is collected at three levels of Ali Z, Qaisar M, Mahmood T, Shah MA, Iqbal T, Serva L,
Michetti AM, Burton PW (2009) The Muzaffarabad,
increasing detail (earthquake, locality, and Pakistan, earthquake of 8 October 2005: surface
site). Also, available imagery documentation faulting, environmental effects and macroseismic
(photographs, videos, sketch maps, stratigraphic intensity. Geol Soc Lond Spec Publ 316:155–172
logs) can be stored in the database. Moreover, Comerci V, Blumetti AM, Brustia E, Di Manna P,
Fiorenza D, Guerrieri L, Lucarini M, Serva L, Vittori
within the EEE Catalogue the epicentral and E (2008) The effects on the environment of the 1908
local intensity values based on the ESI-07 scale Southern Calabria – Messina earthquake (Southern
are recorded. This allows an objective compari- Italy). In: 1908–2008 Scienza e Società a cento anni
son of earthquake intensity for events which dal Grande Terremoto, Reggio Calabria, 10–12
dicembre 2008. Miscellanea INGV, n. 03
occurred in different areas and/or in different Fuller ML (1912) The New Madrid earthquake. US Geol
periods. Surv Bull 494:119
The implementation of the EEE catalogue is Guerrieri L, Tatevossian R, Vittori E, Comerci V,
always in progress through a Web Implementa- Esposito E, Michetti AM, Porfido S, Serva L (2007)
Earthquake environmental effects (EEE) and intensity
tion Interface developed for remote data collec- assessment: the INQUA scale project. Boll Soc Geol It
tion at the following permanent URL: http:// (Ital J Geosci) 126(2):375–386
www.eeecatalog.sinanet.apat.it/login Guerrieri L, Blumetti AM, Brustia E, Esposito E,
The major added value of the EEE Catalogue Lucarini M, Michetti AM, Porfido S, Serva L,
Vittori E, The INQUA TERPRO Project #0811 Work-
for seismic risk assessment is the possibility to ing Group (2011) Earthquake environmental effects,
explore the scenarios of environmental effects intensity and seismic hazard assessment: the EEE cat-
induced by past earthquakes and therefore to alogue (INQUA Project #0418). In: Proceedings of the
944 Electrochemical Seismometers of Linear and Angular Motion
Electrochemical
Seismometers of Linear
and Angular Motion,
Fig. 1 Schematic of the
basic MET sensing element
sensing technology “Solion.” Solion technology tolerance, wide operating frequency range, high
was first implemented into a practical device in sensitivity, and low noise, as well as low depen-
the 1950s by a US Navy-sponsored project. The dence or complete independence of the response
early applications of Solion devices were to on installation angle. These features are widely
detect low-frequency acoustic waves, either in used in practice especially in field and ocean-
the form of an infrasonic microphone or bottom applications, giving important benefits
limited-band seismometer (Hurd and Lane in using a relatively non-expensive, easy to
1957; Wittenborn 1958; Collins et al. 1964). install, compact and at the same time sensitive
The first theoretical model describing the physi- and low-noise device.
cal principles of operation of such devices was The main part of the current paper includes the
developed by C. W. Larcam in 1965 (Larcam following divisions: operating principles, descrip-
1965) who studied a simple one-dimensional tion of the method, how the transfer function is
model of four electrodes (two anodes and two built, and examples of technical parameters for a
cathodes) placed along a channel filled with an commercial MET, including velocimeters, seismic
electrolyte moving under the action of external accelerometers, and unique highly sensitive com-
inertial forces. pact rotational seismometers.
Significant work on Solion motion detectors
was continued in Russia, where the term “molec-
ular electronic transfer” was firstly introduced to Molecular Electronic Transducer:
describe such a device (Lidorenko et al. 1984). Principle of Operation
Inspired by the exceptionally high rate of
mechanical signal conversion to electric current Figure 1 illustrates the basic concept of a MET
in MET involving mass and charge transport, sensing element, consisting of two transforming
pioneer studies of MET (Kozlov et al. 1991; electrochemical cells (Lidorenko et al. 1984).
Kozlov and Safonov 2003; Panferov and Four electrodes configured as anode–cathode–-
Kharlamov 2001; Reznikova et al. 2001; cathode–anode (ACCA), separated by dielectric
Agafonov and Krishtop 2004; Volgin spacers, which protect adjacent electrodes from
et al. 2003) provided an alternative paradigm in contact, span the channels filled with an electro-
the development of seismic sensors. The advan- lyte containing iodide ions. Holes through the
tages of MET motion sensors include small size, electrodes and dielectric spacers allow fluid to
lack of fragile moving parts, high shock flow through the channels.
946 Electrochemical Seismometers of Linear and Angular Motion
The sensing mechanism of the MET is based Fig. 2a are symmetric about the central plane of
on reversible chemical reactions transferring the cell at x = 250 mm, left and right parts of the
charge between anode and cathode via the elec- sensing element are equivalent, and there is no
trolyte ions in the solution (Sun and Agafonov electric current difference between the two pairs
2010). A solution with a high concentration of a of electrodes.
neutral electrolyte (taking no participation in the The ion distributions in the moving fluid are
electrode reactions) and a small fraction of the shown in Fig. 2b. From Fig. 2b, the maximum of
active component that is responsible for the the triiodide concentration is found at the
charge transport over the fluid–electrode metal upstream anode, while the minimum is at the
interface is usually used. downstream cathode. For iodide, as shown in
Most of the practical devices use Fig. 2b, the maximum concentration occurs at
iodine–iodide electrochemical systems with plat- the upstream cathode, while the minimum con-
inum electrodes. The electrolyte consists of a centration is found at the upstream anode. The
high-concentration aqueous solution of potas- maximums and minimums found in Fig. 2b are
sium iodide KI (lower temperature range bound- due to the fact that the triiodide generated at the
ary at 15 C) or lithium iodide LiI (lower upstream anode is consumed at the downstream
temperature range boundary at 55 C) as neu- cathode. By comparing Fig. 2a, b, it is found that
tral electrolyte and a small quantity of molecular both the maximum and the minimum concentra-
iodine I2. If there is an excess supply of iodide, tions of triiodide are greater when the fluid flows;
iodine enters a freely soluble complex compound for iodide the maximum concentration is greater
(triiodide) according to the following scheme: when the fluid flows, while the minimum is
greater in the stationary electrolyte. In turn,
I2 þ I ! I
3 (1) changes in the ion distribution result in the
changes of the electrode current. Specifically,
When the current flows through the MET, the increase of the triiodide concentration in the
following electrochemical reactions proceed at vicinity of a cathode means acceleration of the
the electrodes: the reduction of iodine at the electrochemical reaction according to Eq. 2.
cathode Inversely, decrease of the triiodide concentration
results in the deceleration of the same reaction.
I 3 þ 2e ! 3I (2) So there is a difference between electrical cur-
rents for the two electrode pairs in the moving
and the oxidation of iodine at the anode electrolyte. Usually the difference between cath-
ode currents is used as an output signal of the
3I 2e ! I 3 : (3) transducer.
As the interelectrode potential difference is
K and Li ions do not participate in the reactions. increased, the rate of electrochemical reactions
The sensing element operation is based on the increases gradually until such a condition occurs
fact that rates of the electrochemical reaction at that all the triiodide ions reaching the electrode
the electrodes are significantly higher than the enter the electrochemical reaction instantly.
reactant supply rate. The electrode reactions in This condition corresponds to the saturation cur-
this case lead to the emergence of a reactant rent regime, and further increase in the potential
concentration gradient. Figure 2a shows the dis- difference does not alter the current. Under this
tribution of I
3 and I in the stationary electrolyte. condition the conversion coefficient of the sens-
Because of the electrode reactions, triiodide ing cell achieves its maximum value. That is
concentration maximizes at the anodes and min- why the saturation regime is typically used as a
imizes at the cathodes, while iodide concentra- working point for the transducer. The current is
tion maximizes at the cathodes and minimizes at limited by the volume rate of supply of the active
the anodes. Since the ion concentrations shown in component to the electrodes. In turn, the supply
Electrochemical Seismometers of Linear and Angular Motion 947
A C C A
−5
x10 Max: 42.158
5 40
35
y [m]
2.5 a 30
25
20
15
10
0 5
Min: 0.0564
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x [m] x10−4 E
x10−5 Max: 4022.704
4020
5 4010
y [m]
4000
2.5 a 3990
3980
3970
0 3960
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Min: 3956.616
x [m] x10−4
Liquid motion
Max: 57.141
x10−5
5 50
40
y [m]
2.5 b 30
20
0 10
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Min: 0.0803
x [m] x10−4
x10−5
Max: 4033.545
4030
5 4020
4010
y [m]
4000
2.5 b 3990
3980
3970
3960
0 3950
3940
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Min: 3938.688
x [m] x10−4
Electrochemical Seismometers of Linear and Angu- and (b) moving electrolyte. Both in (a) and (b), I3 con-
lar Motion, Fig. 2 Schematic diagram of a four- centration is shown on the upper graph, I on the
electrode conduit (the electrodes are denoted by the lower one
shaded portions) and ion distributions in (a) stationary
rate is determined by the ion diffusion in the In the general case, the process is characterized
stationary electrolyte. The flow of the solution by a high conversion factor. This is reflected in a
introduces an additional convective ion trans- strong electric response that exceeds the noise of
port that changes the saturation current at each accompanying electronic circuitry significantly,
pair of electrodes (see the volt–ampere charac- even when the input mechanical influence level is
teristics (VACs) of the electrode system in the low. A high signal-to-noise ratio is ultimately
stationary and flowing electrolyte shown in achieved throughout the entire measurement chan-
Fig. 3). The convective flow is directed from nel. Numerous experimental and theoretical stud-
cathode to anode for the first pair of electrodes ies revealed a considerable dependence of the
and in opposite direction for the second one. output parameters of the device on the conversion
That’s why changes of the concentration are system geometry and the broad possibilities for
not symmetrical, and cathode current difference optimizing the conversion system response
(output signal) is generated. depending on the problems being solved.
948 Electrochemical Seismometers of Linear and Angular Motion
Electrochemical
Seismometers of Linear
and Angular Motion,
Fig. 3 VACs of the
transducer electrode system
in a stationary electrolyte
(1) and in an electrolyte
flowing from the anode to
the cathode (2) and cathode
to anode (3)
Electrochemical Seismometers of Linear and Angular Motion, Fig. 4 Schematic of different type of MET motion
sensor
displacement of the liquid inertial mass m relative QðoÞ rL
jH mech ðoÞj ¼ ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi;
to the ground. There are two real forces acting on aðoÞ 2 ðo2 o 2 Þ
rL 0
the liquid mass: Sch þ Rh 2
o2
Restoring force from the flexible membrane, (8)
kV(t), is negative since it opposes the flow qffiffiffiffi
motion, where V(t) is the volume of fluid passing where o0 ¼ rL k
is the mechanical resonance
through the channel and k is the flexibility coef- frequency of the device.
ficient of the membrane and depends only on the An approximation of the electrochemical
characteristics of the membrane. transfer function in frequency domain is
Hydrodynamic damping force, Rh SchdtdV ðtÞ , is
E
linearly proportional to the volumetric flow rate 1
of the liquid associated with the hydrodynamic jH ec ðoÞj ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ffi ; (9)
resistance Rh and is negative since it also opposes 1 þ ooD
the flow motion; Sch is the cross-section area of
the channel. In this consideration we suppose the where oD ¼ dD2 is the diffusion frequency and d is
cross-section area is constant. the interelectrode distance (Kozlov et al. 1991).
The acceleration of the mass relative to The overall frequency-dependent transfer
an inertial reference frame will be the sum of function of a MET seismometer is a superposition
the acceleration d dtxð2tÞ with respect to the ground
2
Electrochemical
Seismometers of Linear
and Angular Motion,
Fig. 5 Responses of a
typical MET transducer to
applied acceleration (blue)
and velocity (red). Units on
vertical axis are mA/m/s2
and mA/m/s for the blue and
red curves,
correspondingly
Electrochemical Seismometers of Linear and Angular Motion, Fig. 6 Functional diagram of a seismometer with
the filter equalization
display a noticeable spread, so the task of achiev- H dev ðoÞ ¼ H sens ðoÞ H corr ðoÞ H out ðoÞ: (11)
ing the desired accuracy can only be performed
by means of the measuring the cell’s transfer The adjustment procedure is the following:
function directly. Shake tables of various types First the seismometer transfer function is deter-
are necessary equipment to implement this mined by measuring the seismometer response
procedure. with some initial parameters of the correcting
The approach is based on the directly mea- circuit. After that, the transducer transfer function
sured response of a particular sensing cell in the is calculated by dividing the seismometer transfer
frequency range of interest and adjustment of the function by a product of the electronic filter trans-
corrective filters to get the desired response in the fer functions:
specified operating frequency band.
Figure 6 presents the functional diagram for H dev ðoÞ
H sens ðoÞ ¼ : (12)
the seismometer where the transfer function was H corr, initial ðoÞ H out ðoÞ
built using the filter equalization procedure.
Mathematically, the transfer function of a seis- After that the parameters of the correcting
mometer Hdev(o) is presented as a product of the filters are remade using adjustable electronic
transfer function of the transducer by the transfer parts. The mathematical modeling is used to
functions of the successive electronic filters: find the required modifications. The result of the
Electrochemical Seismometers of Linear and Angular Motion 951
Electrochemical Seismometers of Linear and Angu- functions Hcorr(o) • Hout(o) of equalization filters, (c)
lar Motion, Fig. 7 (a) Transfer function Hsens(o) of a resulting transfer function Hdev(o) of a horizontal MET
horizontal MET transducer, (b) product of the transfer seismometer
correction is checked on the shake table. within 1 dB for sensitivity and 30 in the
If necessary, the adjustment procedure is passband. What’s more, the filter equalization is
repeated successively until the measured unable to correct any changes in the transducer’s
response function Hdev(o) meets the standard transfer function that may occur in the long term
curve within a tolerance predetermined as accept- after the factory adjustment. Now this type of
able. The iterative adjustment process eliminates equalization is used only in a few models of
the influence of natural spread of electronic com- electrochemical seismometers.
ponent values. The equalization procedure and More precise devices use the feedback design.
the final result of the adjustment are illustrated The commonly used technique for a practical
on Fig. 7a–d, where real data are used as an feedback implementation is a moving coil sys-
example. The amplitude vs. frequency character- tem. The flexible membrane (see Fig. 4a) is
istics are shown, where Fig. 7a is the measured equipped with a coil placed in a magnetic field
transfer function to applied velocity of the trans- of a firmly installed strong magnet. Applying a
ducer, Fig. 7b is the transfer function for the current into the coil gives rise to a force acting on
electronic filters, Fig. 7c is the final amplitude the moving liquid inside the sensor. Usually, neg-
vs. frequency curve, and Fig. 7d is the final ative feedback is implemented, so that the force
phase vs. frequency curve for the seismometer. produced by the coil acts oppositely to the inertial
The accuracy of the filter equalization proce- force. It is important to note that for this
dure is limited by the tolerance of the electronic approach, due to the high original conversion
components and the accuracy of the shake table coefficient of the transducer, the decrease in the
used to determine the transfer function. The usual conversion factor due to the negative feedback,
variation in the transfer function for a broadband practically, does not affect the instrument
seismometer with filter equalization stays resolution.
952 Electrochemical Seismometers of Linear and Angular Motion
The block diagram of the seismometer The most important feature of a feedback is that,
equipped with a feedback mechanism is provided the system remains stable and the amount
presented in Fig. 8, where Hcorr(o) represents of feedback is sufficient, the output function
the transfer function of filters and amplifiers in Hdev(o) depends mostly on the feedback transfer
direct circuit, Hfb(o) is the same of the feedback function H fb ðoÞ Hcoil ðoÞ rather than on the trans-
circuit, and Hout(o) is the transfer function of the ducer transfer function H sens ðoÞ H corr ðoÞ.
output filter. Indeed
In case the amplification inside the closed- A velocimeter is more common for broad-
loop system is much higher than 1 in band seismometers. In this case, an additional
absolute value jH sens ðoÞ H corr ðoÞ H fb ðoÞ circuit, an integrator as a part of the output
H coil ðoÞj >> 1 and Eq. 13 is approximated by filter (H out ðoÞ o1 in the passband), or a
the following formula: differentiator at the feedback amplifier
(Hfb ðoÞ o in the passband) should be
1 added (both shown in dashed lines in Fig. 8).
H dev ðoÞ ¼ H out ðoÞ: (14)
H fb ðoÞ H coil ðoÞ Whenever an integrator or differentiator is
used, the transfer function of the instrument
The feedback circuit results in a counterforce would give an output voltage proportional to
(i.e., a force compensating effect of the acting the ground velocity.
acceleration) proportional to the current flow. The introduction of an integrator into cir-
So Hcoil(o) can be expressed as a function that cuits at lower frequencies is limited by elec-
converts voltage to acceleration. tronic parts leakage and op-amps offset. Also,
The practical implementation of the electro- the substantial variation of capacitors’ imped-
chemical seismometer equipped with feedback is: ance over the passband in frequency-
dependent circuits may impair the noise
Task 1: Providing sufficient amplification to have performance of a device. Therefore special
a “deep” feedback schematic techniques and careful amplifier
Actually, this is the simplest task among selection should be made in order to maintain
others listed below. First, the transducer itself the optimal noise performance of the integra-
has a high original conversion factor. Addi- tor or differentiator stage.
tionally, the electronic industry produces a Another aspect of introducing a
variety of low-noise operational amplifiers differentiator into the feedback circuit is that
which can be used for additional amplification it brings a leading p/2 phase shift in the loop
of a signal. amplification. Typically, this phase shift
Task 2: Producing an output proportional to improves feedback stability, and velocimeters
either acceleration or velocity, depending on with a differentiator in the feedback have bet-
the instrument type ter performance at higher frequencies in com-
In case the electronics in the feedback has parison with devices equipped with an output
frequency-independent response, the transfer integrator.
function of the instrument would be propor- Task 3: Ensuring the stability of the closed-loop
tional to acceleration. The output filter transfer system
function Hout(o) is used to shape the low- and Maintaining the feedback loop stable
high-cutoff frequencies. is a comparatively sophisticated task.
Electrochemical Seismometers of Linear and Angular Motion 953
Electrochemical Seismometers of Linear and Angular Motion, Fig. 8 Block diagram of a closed-loop electro-
chemical seismometer
Being negative in the middle of the frequency Very strong feedback depth (dark red curve on
band, the feedback may even change its sign the Fig. 9) could result in the feedback loop
and become positive at some frequency out- becoming unstable during the transition pro-
side the high- or low-frequency band limits cesses or under variations in temperature.
due to the phase response variation as a func- Most electrochemical seismometers have
tion of frequency. To avoid self-oscillation, it maximum feedback depth in the range from
is important to keep the amplification inside 10 to 50. The feedback for the electrochemical
the closed-loop system at such a frequency accelerometers could be as big as few
less than 1 in absolute value by all means. hundred.
This task not only requires considerations of Task 4: Sufficiency in feedback dynamic range
stability when the system is being adjusted, For the sake of keeping the dynamic range
but it’s also essential to maintain the closed- at a maximum, the feedback amplifier should
loop stable over all temperature ranges and provide sufficient current into the coil.
under transitions processes after the device is The main limitation comes from the fact that
just powered on or overloaded by a strong modern systems are continuously using a
input signal. lower and lower voltage supply, while coil
Changes of an electrochemical seismome- size and weight of the moving body cannot
ter frequency response with variation of the be correspondingly reduced. Moreover, a
feedback depth are illustrated in Fig. 9. The heavier moving body of a sensor gives less
device response at a low feedback depth is thermal self-noise. A low-noise seismometer
close to the curve of an uncorrected sensor usually needs a larger current for the
with no visible flat region at all. The deeper feedback coil.
the feedback, the wider is the flat region in the The procedure of adjusting a sensor with
device response, but at the same time the feedback equalization consists of two stages.
lower is the overall sensitivity of the device. During the first step, a proper frequency
954 Electrochemical Seismometers of Linear and Angular Motion
Electrochemical
Seismometers of Linear
and Angular Motion,
Fig. 9 Device response
vs. feedback depth (depths
1, 5, 40, 120)
response is made paying no attention to actual The last step in manufacturing the device is
device sensitivity. In the second step, the the adjustment of its sensitivity. The operation
required device sensitivity is adjusted. can be carried out by adjusting the overall gain
The open-loop characteristics can be of the output filter’s Hout(o).
obtained by means of an external digital-to- After this final adjustment, the resulting
analog converter (DAC) and analog-to-digital frequency response characteristic of a closed-
converter (ADC) with a harmonic response loop device is measured and verified. The
scan. The frequency response of the open main point of this method is in simulating
loop outside the operational range is essential ground motion by adding of an external signal
to ensure the system stability. After the initial into the closed-loop circuit. This signal addi-
open-loop response is obtained, the filter cor- tion works in the same way as a real input
rection is made to adjust the feedback ampli- signal. An external sine wave generator can
fication, to make a flat response in the be used to provide such a signal at various
passband and predetermined behavior at low- frequencies and thereby get a full calibration
and high-cutoff frequencies. The process is function.
iterative in the same way as it was for the The block diagram of the closed-loop cali-
non-feedback devices. bration procedure is illustrated in Fig. 10
The main difference compared to the where the sine wave test signal from a gener-
non-feedback case is that the adjusting proce- ator is added to theP closed-loop signal in the
dure gives a much more accurate response of a summation unit . The signal disturbance
seismometer within the passband than for the introduced into the closed-loop and the output
non-feedback sensor. Generally the high- and response are measured by an external 2 chan-
low-cutoff frequencies are extended by 2–3 nel analog-to-digital converter.
octaves. In parallel with the feedback adjust- The examples of the resulting calibration
ment, a fine-tuning of the output filter is done. curves are shown in Figs. 11 and 12.
In case the closed-loop response is actually flat The closed-loop calibration procedure
over the passband, the device characteristics makes the subsequent maintenance and vali-
coincide with the output filter characteristics. dation of a device possible with a common
This decreases the complexity of further DAC–ADC and a reference sensor with ver-
device response presentation in terms of ified sensitivity and noise performance. The
poles and zeroes. test is carried out only by means of the
Electrochemical Seismometers of Linear and Angular Motion 955
Electrochemical
Seismometers of Linear
and Angular Motion,
Fig. 11 Seismometer
response obtained by the
self-calibration test. The
actual CME-6011 sensor
was used to generate an
example
standard device connector and does not two sensors placed close to each other.
require any intervention into the sensor or One of the approaches is to calculate
disassembling the device. The presence of a non-correlating components in the signal of
shake table becomes unnecessary. The two sensors of the same type with identical
device sensitivity can be determined using parameters (sensitivity, phase response,
either factory-determined calibration con- passband). Another approach is to make
stants or by analyzing the simultaneous simultaneous recording with a sensor with
recording made with a sensor with known far better noise performance. In the latter
sensitivity. The noise performance of a case the noise can be calculated by
device can only be obtained by analyzing subtracting the low-noise sensor’s signal
the simultaneous record made by at least from the tested sensor’s signal.
956 Electrochemical Seismometers of Linear and Angular Motion
Electrochemical Seismometers of Linear and Angular Motion, Table 1 The comparison chart for MET sensors
for permanent seismic stations (sensors are sorted in alphabetical order)
Sensor type, BB603 PMD Scientific, Inc., CME-6211, R-sensors LLC, Russia EP-300, Eentec, USA
producer, USA http://www.pmdsci. http://www.r-sensors.ru/1_products/ http://www.eentec.
country com/pdfs/BB603.pdf Descriptions/CME-6211.pdf com/EP-300_data.htm
Operating Electrochemical transducer Electrochemical transducer Electrochemical
principle transducer
Equalization Force-balancing feedback Force-balancing feedback Force-balancing
type feedback
Output Triaxial Triaxial Triaxial
signals
Sensitivity 2,000 V/(m/s) 2,000 V/(m/s) 2,000 V/m/s E
Frequency Standard: 0.0167–50 Hz Standard: 0.0167–40 Hz Standard:
bandwidth 0.0167–50 Hz
Optional: 0.0083–50 Hz Optional: 0.0083–40 Hz Optional: 0.0083–50 Hz
Maximum 10 V 20 V 10 V
output swing (20 Vp–p differential) (40 Vp–p differential) (20 Vp–p differential)
Self-noise Below USGS NLNM in Below NLNM in 0.1–7 Hz band Below NLNM in
0.05–5Hz range 0.9–12 Hz band
Temperature Standard 12 C to +55 C Standard 12 C to +55 C Standard 12 C to
range Extended 40 C to +55 C +55 C
Nominal Standard: 12Vdc, 28 mA Standard: 12Vdc, 55 mA Standard: 12Vdc,
supply 30 mA
voltage, Low power: 12Vdc, 12 mA Low power: 12Vdc, 27 mA Low power: 12Vdc,
consumption 12 mA
Mass lock, None required None required None required
centering
Maximum 10 15 10
installation
tilt
Case 200 mm/220 mm 272 mm/240 mm 200 mm/220 mm
diameter/
height
Weight 11 kg 12 kg 11 kg
operating characteristics are attained when world’s smallest tri-axis 1-Hz seismometer.
MET-based sensors are equipped with an electro- MTSS-2003 is designed primarily for measuring
dynamic negative feedback mechanism. With a small low-frequency signals. It combines high
feedback, the instrument scale factor of 300 V/m/ gain, small size, and low power consumption.
s is achieved without using any additional ampli- The 3-component high-performance acceler-
fication, which is equal to or greater than that ometer MTSS-1033A and its 1-component ver-
achieved by other known low-frequency geo- sion MTSS-1031A are designed for strong
phones. At the same time, for a frequency range motion seismic measurements, industrial vibra-
starting at 1 Hz, the MET geophones are much tion monitoring, vibration isolation analysis,
smaller than typical electrodynamic geophones. etc. MET accelerometers combine very small
The external view of sensors is shown in size and high sensitivity over wide frequency
Fig. 12. MTSS-1001 is a single-axis, band. The wide dynamic range, low distortion,
ultracompact, high-gain, 1-Hz seismic sensor. high temperature stability, and competitive price
MTSS-2003 contains three orthogonal oriented make it ideal and cost-effective solution for var-
MTSS-1001 seismic sensors. It is probably the ious applications such as earthquake, strong
958 Electrochemical Seismometers of Linear and Angular Motion
Electrochemical Seismometers of Linear and Angular Motion, Table 2 The comparison chart for seismic MET
sensors for field work (sensors are sorted in alphabetical order)
BB313 CME-4211 CME-6011 MP403
PMD Scientific, R-sensors, Russia R-sensors, Russia http:// PMD Scientific,
USA http://www. http://www.r-sensors. www.r-sensors.ru/ USA http://www.
Sensor type, pmdsci.com/pdfs/ ru/1_products/ 1_products/Descriptions/ pmdsci.com/pdfs/
manufacturer BB313.pdf Leaflets/4211.pdf CME-6011.pdf MP403.pdf
Operating Electrochemical Electrochemical Electrochemical motion Electrochemical
principle motion transducer motion transducer transducer motion transducer
Equalization Force-balancing Vertical: force- Force-balancing feedback Force-balancing
type feedback balancing feedback feedback
Horizontal: filter
equalization
Output Triaxial Triaxial Triaxial Triaxial
signals
Sensitivity Standard: 2,000 V/ Standard: 2,000 V/ Standard: 2,000 V/(m/s) Standard: 2,000 V/
(m/s) (m/s) (m/s)
Frequency Standard: Standard: 0.033–50 Hz Standard: 0.033–40 Hz Standard:
bandwidth 0.033–50 Hz Optional: 0.1–100 Hz Optional: 0.1–50 Hz 0.1–50 Hz
Maximum 10 V (20 Vp–p 15 V (30 Vp–p 20 V (40 Vp–p diff.) 10 V (20 Vp–p
output swing diff.) diff.) diff.)
Temperature Standard 12 C to Standard 12 C to Standard 12 C to Standard 12 C to
range +55 C +55 C +55 C +55 C
Extended 40 C to Extended 40 C to
+55 C +55 C
Nominal Standard: 12Vdc, Standard: 12Vdc, Standard: 12Vdc, 26 mA Standard: 12Vdc,
supply 28 mA 27 mA 28 mA
voltage, Low power: 12Vdc, Low power: 12Vdc, Low power: 12Vdc, Low power: 12Vdc,
consumption 12 mA 8 mA 10 mA 12 mA
Mass lock, None required None required None required None required
centering
Maximum 12 15 15 12
installation
tilt
Case 155 mm/185 mm 180 mm/140 mm 204 mm/210 mm 200 mm/220 mm
diameter/
height
Weight 5.5 kg 4.3 kg 6.6 kg 9 kg
Electrochemical Seismometers of Linear and Angular Motion, Table 3 The technical performance parameters
for MET compact low-frequency geophones and accelerometers
Performance MTSS-1001 MTSS-1041A MTSS-1031A
Frequency range 1–300 Hz 0.1–120 Hz 0.1–120 Hz
Sensor noise 150 nm/s, rms over frequency range 110 ng/√Hz 150 ng/√Hz
Conversion factor 250 V/(m/s) 6 V/g 2.4 V/g
Maximum negative feedback depth 20 100 150
Maximum applied acceleration – 2 g 4 g
Dynamic range >110 dB >110 dB >120 dB
Non-linearity <1 % <1 % <1 %
Orientation Any direction Any direction Any direction E
Weight 180 g 180 g 180 g
Temperature “55 +65 C” “55 +65 C” “55 +65 C”
Power <80 mW <80 mW <80 mW
Electrochemical Seismometers of Linear and Angu- et al. 2013). The models QRS-116, SDG-1000, and
lar Motion, Fig. 13 Comparison of the self-noise for CRS-09 are angular motion sensors based on MEMS
some electrochemical angular seismometers (models R1 technology. The TRS-500 and VG-103LN are fiber-optic
and R2) with other angular motion sensors (Lin gyros
input, to measure more accurately the seismic aspects of the seismic rotational measurements
field spatial distribution, to be able separate for a variety of applications. A special issue
between modes of seismic waves based on their “Advances in rotational seismology: instrumen-
polarization, and to determine better the site tation, theory, observations, and engineering”
effect (Lee et al. 2011; Bernauer 2012a; was issued in Journal of seismology,
Kapustian et al. 2013). Detailed information volume 16, # 4 in October 2012.
can be found at web page www.rotational- The mechanical configuration of the rotational
seismology.org, especially devoted to different seismometer, based on the electrochemical
960 Electrochemical Seismometers of Linear and Angular Motion
Electrochemical Seismometers of Linear and Angular Motion, Table 4 Technical parameters of commercially
available rotational seismometers
METR-03/01, LLC R-sensors, http://
www.r-sensors.ru/1_products/ METR-13/11, LLC R-sensors, http://
Seismic_sensors_METR-03_CME- R2, Eentec, www.r-sensors.ru/1_products/
106C.pdf, R1 eentec, http://www. (Bernauer Seismic_sensors_METR-03_CME-
Model eentec.com/R-1_data_new.htm; et al. 1012) 106C.pdf,
Output Angular rate Angular rate Angular rate
Sensor Electrochemical Electrochemical Electrochemical
type
Noise at 107 rad/s/√Hz (for R1, see Fig. 13) 2*107 rad/s/
1 Hz √Hz (see
Fig. 13)
Full-scale 0.1 rad/s 0.3 rad/s 0.1 rad/s
range
Bandwidth 0.033–50 Hz 0.033–50 Hz 1–150 Hz
Size 120 120 102 120 120 120 120 102
102
sensing technology, is presented in Fig. 4 above. electrochemical cell. Among the commercial
The sensor consists of the toroidal channel, filled devices developed on the base of this technology
with an electrolyte, and the sensitive cell placed are broadband seismometers, compact high-
across the channel, converting liquid motion into sensitive MET geophones, accelerometers, and
the electrical response. An expansion volume is unique angular seismometers.
used to compensate for the temperature expan-
sions of the liquid. Electrochemical sensors are
probably the only rotational sensors which are
commercially available at a reasonable cost with Cross-References
a resolution appropriate to seismic applications.
Figure 13 shows the spectral noise density spec- ▶ Downhole Seismometers
trum of rotational electrochemical sensors com- ▶ Ocean-Bottom Seismometer
pared with other commercial devices measuring ▶ Principles of Broadband Seismometry
rotation. The parameters of some of the commer- ▶ Seismic Accelerometers
cially available electrochemical rotational ▶ Sensors, Calibration of
seismometers are shown in Table 4. The indepen-
dent test data can also be found in Bernauer References
et al. (2012b).
Agafonov VM, Krishtop VG (2004) Diffusion sensor of
mechanical signals: frequency response at high fre-
quencies. Russ J Electrochem 40:537–541
Summary Agafonov VM, Egorov EV, Zaitsev DL (2010) Molecular
electronic linear accelerometers. Preliminary test
results. Gyroscopy Navigation 1(4):246–251
The electrochemical seismometers offer an alter- Agafonov VM, Egorov IV, Shabalina AS (2014) Operating
native to regular electromechanical devices. principles and technical characteristics of a small-
They are compact, robust, and easy to install sized molecular–electronic seismic sensor with nega-
and provide good quality data. The operation tive feedback. Seismic Instrum 50(1):1–8
Bernauer M, Fichtner A, Igel H (2012a) Measurements of
principles of the electrochemical seismometers translation, rotation and strain: new approaches to
are based on using the charge transfer variations seismic processing and inversion. J Seismol
due to electrolyte motion in the four-electrode 16(4):669–681
Emergency Response Coordination Within Earthquake Disasters 961
preparedness to respond and recover; respond in a response capabilities; and, of more relevance to
coordinated, adaptive, and effective manner this entry, to develop and deliver effective
including all relevant organizations and effected response and recovery capabilities and outcomes.
communities; and initiate regenerative recovery This entry commences with an overview of
parallel with response so as to further enhance more formal leadership and coordination of
resilience and make the most of the opportunities responses. It then goes on to cover the topics of
that disaster provides to communities. monitoring and analysis of ongoing seismic
The 1980s and 1990s saw the development of activity during response; acquiring, analyzing,
the “all hazards” approach to emergency man- managing, and sharing information during disas-
agement, intended to broaden the application of ter response; capturing and making use of phys-
preparedness for hazards beyond a handful of ical and social impact data; urban search and
locally relevant natural hazards and the then rescue (USAR); medical response; emergency
diminishing threat of nuclear conflagration, with welfare; and emergency food, water, and
the fall of the Berlin Wall and the end of the sanitation services. This entry concludes with
cold war. The all hazards approach tended to discussion of post-quake building management
suggest that the arrangements, capabilities, and and critical infrastructure or lifeline utility coor-
resources put in place by government agencies at dination: essential elements of response and
all levels for the major hazards they faced could recovery.
then be applied any and all hazards that might
then arise.
The world has experienced increasing expo- Leadership and Coordination
sure to natural hazards and their consequences
since the 1960s, as populations have boomed in When significant earthquakes strike, the chal-
climatic and geological risk-prone areas in less lenges for those in positions of leadership or who
developed countries and as communities and are responsible for response coordination are many
industries in developed economies became and varied. Not the least due to the reality that
increasingly dependent on technology, particu- affected communities, ethnic and social networks,
larly electricity and telecommunications. businesses, service clubs, and faith organizations,
Increasing exposure to risks has resulted in emergency services in the area at the time will be
continual improvement and refinement of mounting their own responses, while official
approaches to managing risks internationally, responses will generally still be regrouping and
culminating in the United Nations making the reestablishing communication with their field
1990s the “International Decade for Natural resources and external support. The very nature
Disaster Reduction” and the initiation of of damaging quakes makes them some of the
programs such as the United Nations Office most difficult hazards to manage responses to, as
for International Disaster Risk Reduction earthquakes tend to undermine the ability of
(UNISDR). response organizations to deliver their services
UNISDR projects, including the “Hyogo more than other hazards tend to.
Framework for Action 2005–2015 Building the Leadership and coordination, nonetheless, are
Resilience of Nations and Communities to Disas- crucial to ensuring that the needs of affected
ters,” have seen emergency management, and communities and the services they depend upon
disaster risk management in general take a more are met in as timely, effective and equitable man-
comprehensive turn. Nations and communities, ner as possible. Key to effective emergency lead-
in their own ways, are increasingly attempting ership and coordination is legislation that
to understand the natural and human-induced adequately provides for the situations that are
hazards that they face; to inform risk reduction likely to arise from major hazards.
legislation, regulations, and programs; to build The legislative requirement to identify the
community and organizations’ resilience and major hazards and risks that respective national,
Emergency Response Coordination Within Earthquake Disasters 963
Emergency Response
Coordination Within
Earthquake Disasters,
Fig. 1 Basic response
coordination structure – NZ
CIMS manual (Ministry of
CDEM 2014)
state, regional, and local communities face is an context and, if necessary, that of its neighboring
essential foundation to emergency management – or closely related nations. All incident manage-
E
as are legal responsibilities to plan at national and ment systems provide a common structure, roles,
subnational levels to manage hazards and their processes, etc. for overall response coordination
risks holistically. at various levels of response. Each response orga-
Coordination of responses to the immediate nization applies the standard system to their own
and longer-term consequences of earthquakes is operating procedures to greater or lesser degrees,
considerably more important than command and depending on the requirements of the overall
control-style leadership, as no single individual system and any statutory or regulatory mandate
or organization can have adequate knowledge of attached to it.
impacts and response to be able to comprehen- Incident management systems tend to provide
sively lead or direct responses. More implicit an organizational structure and hierarchy with a
coordination within and between organizations response director or controller in the senior posi-
and communities is more influential than individ- tion, albeit often under the direct influence of
ual or organizational leadership and direction. senior appointed or elected officials. Reporting
The concept of disaster response management to the “controller,” “director,” “commander,” or
that is increasingly seen as most effective in similar, are managers of response management
understanding and delivering responses that actu- functions, including public information manage-
ally work is “network-centric operations” (Von ment, operations, intelligence, planning, and
Lubitz et al. 2008) or, in short, net-centric logistics (Fig. 1).
operations. Note: “PIM” is public information manage-
The impacts of earthquakes tend to not only ment in the New Zealand context and “public
cause buildings and other vertical and horizontal affairs” or similar in other contexts.
infrastructure to fail or be compromised but also
to cause injuries, distress, entrapments, and fatal-
ities. They also disrupt response capabilities, Response Coordination Principles
despite how well planned and resourced those
capabilities are, due to overload and damage to Levels of response coordination tend to include:
telecommunications systems, the initial confu-
sion in the community and official responders, • Site – management closest to the response
and direct impacts on responders and their equip- field activities themselves. In an earthquake
ment and facilities – as well as their families, context, this will tend to be a large building,
their homes, and the services they depend upon. block or buildings, or a small geographic area
Effective coordination depends on a range of encompassing areas of structural damage and
factors. An essential prerequisite is a shared geological damage, including liquefaction and
understanding of and commitment to response rockfall, landslide, etc. or social impact or
coordination structure, roles, terminology, and social support.
processes. Each nation tends to have adopted or • Local – coordination and direction of
developed an incident coordination, command, or response activities and resources in support
management system adapted to suit the national of site-level response, often at suburb or
964 Emergency Response Coordination Within Earthquake Disasters
authorities and infrastructure providers, as well earthquake, it is important that there be as good
as response controllers and senior emergency an understanding of the characteristics of the
service managers actually managing the likely aftershock sequence as possible. The com-
response. Politicians have an important role to munication of this information in a timely way
play in response, with mayors usually taking requires the specific integration of national seis-
spokesperson role at a local level and responsible mological representatives within operational
cabinet members doing so at state or national structures. Information about the possible magni-
level, and prime ministers, governors, or presi- tude, intensity, and location of aftershocks is of
dents doing so in more challenging emergencies particular interest to USAR teams, those involved
that warrant higher-level political engagement. with rapid building assessments, and for public
One of the risks of attempting to assimilate information purposes.
community, agency, media, and social media Through their training, the engineers involved
information on behalf of the entire response and both with USAR (Urban Search and Rescue)
to produce the outputs to affected communities in teams and the rapid building assessment opera-
a timely manner in particular is that the processes tions have an expectation of major aftershocks of
of doing will be slowed by the need to develop typically up to a magnitude less than the main
outputs in a collaborative manner. Bureaucratic shock to occur in the days, weeks, or months
development of outputs combined with following, but possibly up to the same magnitude
compromised means of communication can as the main shock.
result in critical information not reaching key For USAR teams, as the likelihood of rescuing
audiences in a timely enough manner – some- live people from collapsed structures diminishes
thing that needs to be kept in mind and mitigated with time following the main shock, the “risk
prior to and during responses. vs. reward” equation changes, and therefore a
Engaging with affected communities is a cru- sense of the likelihood of significant aftershocks
cial aspect of public information management, is a useful input.
particularly where community responses are not Engineers and others undertaking rapid build-
well integrated into the official response and ing assessments have a longer time frame to con-
recovery activities. Including spontaneous and sider as they assess the suitability of damaged
organized community response and preexisting buildings for continued occupancy. The placards
and post-event community groups as partners to they post on buildings should take into account
public information management is increasingly the additional damage that could occur from
seen as critical to effective response and recovery aftershocks in the days, weeks, and months fol-
efforts. lowing the earthquake and prior to detailed dam-
Significant events will not only receive wide- age evaluations and repairs being undertaken.
spread coverage outside the affected area but may Assessment of the short-, medium-, and long-
result in hundreds and possibly thousands of term seismicity effects after a major earthquake is
reporters and production crews converging on important as the response transitions to recovery.
the affected area and coordination centers and This may, for example, lead to an increase in the
sites of particular interest. Dealing constructively seismic design coefficients used for repairs and
with this number of additional personnel and the rebuilds due to the heightened seismicity, as hap-
activities, some potentially risky or at least intru- pened following the Christchurch earthquake in
sive or distracting, has to be taken into account in February 2011.
planning for and delivery of response and recov- The nature of the expected aftershocks can
ery public information management. broadly be characterized from the fault activity
which caused the earthquake itself. For example,
Event Hazard (Seismicity) Monitoring earthquakes that occur on established and active
While the aftershock sequence cannot be pre- faults close to plate boundaries typically generate
scribed or predicted following a major few significant aftershocks, and these will
Emergency Response Coordination Within Earthquake Disasters 967
generally have the same directionality. In con- critical infrastructure groups prior to and their
trast, earthquakes that occur away from plate efforts to restore services during emergencies
boundaries (intraplate earthquakes) on “new” and after emergencies.
faults that have not previously shown evidence Organizations with roles in response, includ-
of surface expression can generate aftershocks ing critical infrastructure providers, are better
which migrate, with much less predictability able to manage and share information during
about the number of significant aftershocks and emergencies when they have invested in post-
their intensity and directionality. quake accessible and resilient data backup ser-
The post-event seismicity is typically vices, ruggedized or robust and seismically
expressed in probabilistic terms, which can take secured information management hardware, and
E
the form of “the probability of an aftershock of alternate means of communications, such as
magnitude range (e.g., M6.0–6.4) occurring in a voice and data over satellite or radio. Even resil-
given region within a specified period of time.” ient data storage, management and communica-
This form of estimation often uses different seis- tion organizations with response roles also have
micity models, depending whether the time frame to have paper-based systems to fall back to, at
is short term (e.g., months or a year) or longer least temporarily, in the event of their first line of
term (e.g., 20 or 50 years). alternate arrangements being compromised or
Assessment of secondary risks should also be overwhelmed.
undertaken and must be considered within the Effective response management depends on
post-quake context. These include those that are emergency-related information management
consequential in additional to the primary seis- applications and operational procedures that are
mic hazard, such as tsunami and landslide/rock- compatible with partner organizations, fit for pur-
fall, liquefaction, natural and technology-related pose, well trained, and exercised.
environmental contamination, as well as conse- Cloud-based data storage and email systems
quentially unstable slopes and land susceptible to are becoming increasingly popular in not only
subsidence and increased flood risk. emergency response-focused information man-
agement but in business continuity and disaster
recovery processes. Software and equipment that
Emergency Information Management are intended for emergency purposes and person-
nel required to use or manage them are consider-
Managing impact and response information is a ably better able to meet response and recovery
particular challenge during damaging earth- needs if the response roles are integrated as much
quakes, particularly if inadequate thought and as possible into business-as-usual practices.
preparation has gone into this function prior to
the event or key components of information col-
lection, analysis, sharing, management, storage, Impact Overview: Rapid Impact
or dissemination are compromised by the event Assessment
itself. The benefits of a well-developed incident
management system and supporting information One of the first tasks of affected communities and
management systems are vital to effective quake organizations, and the emergency response orga-
response activities. nizations supporting them, after a hazard such as
Contemporary information management relies an earthquake causes damage is to carry out a
on networks and equipment which when, unless rapid impact assessment. These usually occur
designed, constructed, supported, and used within 8–48 h on the incident occurring to:
appropriately can be particularly vulnerable to
direct or indirect earthquake interruption. • Assist in determining what ongoing or new
Telecommunication providers are an essential risks may be present
element in the resilience-building work of such • Planning initial response actions
968 Emergency Response Coordination Within Earthquake Disasters
• Acquiring, activating, and tasking necessary In the first few hours and days after a signifi-
initial resources cant earthquake, telecommunications are likely
• Provide the basis for more detailed needs and to be compromised, and personnel available to
structural assessments carry out physical and social impact assessments
will be in short supply. Therefore, for impact
Rapid impact assessment is much less about assessment to be conducted efficiently and effec-
building-by-building, household-by-household, tively, it may be necessary for the coordinating
and facility-by-facility assessment generally ini- emergency management agencies and emergency
tiated several days into the response, but about services to plan and prioritize the deployment of
quickly gaining a sufficiently detailed and reli- resources in pre- or response-determined sectors.
able intelligence picture of the known and likely Impact data collection has traditionally been
direct and indirect impacts of the hazard. The carried out in a paper-based environment, but
process of planning, tasking, collecting, analyz- electronic data collection, often with digital pho-
ing, and disseminating rapid impact assessment tographic records and GPS location data, is
data should follow the intelligence cycle process becoming increasingly affordably feasible –
used in all aspects of emergency response to even in the immediate wake of earthquakes and
inform initial and later action planning similar hazards. Handheld devices, including
processes. smartphones, tablets, iPads, etc., and purpose-
Analysis of earthquake risk, including esti- designed or business-as-usual applications are
mates of likely impact and, particularly in rela- revolutionizing the collection, management, and
tion to critical infrastructure and vulnerabilities application of data in virtually all emergency,
of particular and interrelated infrastructure, can incident, and disaster management contexts.
greatly assist rapid impact assessment. Knowl- Although standard telecommunications ser-
edge of quake likelihoods and consequences can vices will often be compromised or at least
change relatively rapidly as aspects of earthquake overloaded in the first few days after an earth-
science and knowledge improve. quake, data can be readily transferred to recipient
It is essential for those preplanning or actually databases by way of direct data download at the
managing rapid impact assessment to use quake impact assessment base of operations or respec-
risk studies to inform their work where any such tive emergency operations center.
studies are available. An example of such a study
carried out recently is that for the Wellington
region, New Zealand (Cousins 2013), which pro- Rapid Physical Impact Assessment
vides estimates of between $NZ1 billion and
$NZ16.5 billion in damage in varying scenarios In the emergency response coordination context,
generated by the seven major fault lines known in post-earthquake, rapid impact assessment tends
the city and in the nation’s capital, with a popu- to focus first on identifying sites and areas when
lation of approximately 460,000 and 195,000 search and rescue may need to occur – usually
buildings where they are built on or adjacent multistory buildings that have failed or been seri-
to. The damage would include up to almost ously compromised and buildings, roads, bridges,
7,000 building failures caused by shaking, geo- ports, airports, and other geological features that
logical failure, or tsunami, up to approximately may be required for evacuation and emergency
2,000 fatalities and 9,300 serious injuries from response access and egress.
earthquake, and 3,500 fatalities and 7,000 inju- Identification of actual and potential environ-
ries from tsunami. Pre-event impact estimations mental contamination, from analysis of
such as these are of considerable assistance in preexisting records of contamination risks, con-
determining impact assessment, response, and tamination identified or reported in the immedi-
recovery resources that are likely to be necessary ate response or in subsequent managed
when real disasters strike. reconnaissance tasks.
Emergency Response Coordination Within Earthquake Disasters 969
Early assessment also includes determining safe place to sleep, and adequate sanitation.
whether buildings, other infrastructure, or geo- Immediate needs then extend to social
logical features potentially damaged in early reconnection, particularly for family members
shocks may fail later, either of their own accord separated as a consequence of the emergency
or due to later seismic activity. and identification and support of the vulnerable,
Rapid impact assessment informs immediate whether due to pre-emergency circumstances or
planning and management of tactical functions as a result of the emergency.
such as search and rescue, evacuation orders and Most social impact data is collected by per-
support, public information messaging, etc., as sonnel assigned that responsibility in teams
well as more planning for more strategic activi- tasked to carry out survey of individuals, fami-
E
ties and resourcing. lies, or groups sheltering in place, seeking sup-
Rapid physical and social impact assessment port at relief centers, or evacuees in transit or at
tasking, data collection, analysis, and dissemina- reception centers or group lodging. However,
tion are usually managed by the emergency man- rapid social need assessments can also be aug-
agement arm of the local or regional/county mented by data collected by those carrying out
government involved, in conjunction with local rapid physical impact assessment, particularly if
emergency services, particularly fire and police those gathering data on physical impacts have
resources working in or supporting immediate been prepared to observe and gather social
response in the affected areas. impact information.
Residential impact assessment can often be
most effectively carried out in conjunction with
Rapid Social Impact Assessment physical impact assessment, through the estab-
lishment of composite teams of building asses-
The process of rapid social impact assessment is sors and social services workers. In areas where
activated alongside rapid physical impact there are larger buildings, impact assessors and
assessment – informally by communities and engineers are likely to be deployed on behalf of
more formally by government and nongovernment building owners and insurance companies rela-
relief, social service, and emergency management tively early in the response. Opportunities to
organizations. Social impact assessment and emer- coordinate and share at least outline impact data
gency social/welfare services are usually coordi- between public and private assessors should be
nated by the emergency management arm of the pursued. Private assessors should at least work
respective local or regional government, supported under the safety and access requirements
by state, national, and international entities. imposed by emergency management agencies.
Even in post-quake international disaster relief Access to damaged areas is likely to be con-
situations, the responsibility for authorizing the trolled by emergency services or emergency
cross-border entry of search and rescue, impact management agencies early in the response. If
assessment, and relief personnel and their deploy- that is the case, impact assessors will need to be
ment and activities within the affected areas lies required to carry pre-authorized means of iden-
within the government of the recipient country tification or identification and access documen-
and, where delegated authority exists, within the tation or printed cards developed for the
government of the local jurisdiction involved. particular response. Assessors will also be
Social impact assessment includes gathering required to use personal protective equipment
data on known needs and making estimates of and regular reporting procedures as prescribed
likely needs based on known data and knowledge by the coordination entity or at standard occupa-
of needs generated in similar events in compara- tional health and safety practice for the respec-
ble communities elsewhere. tive jurisdiction.
Emergency social needs include the basic Impact and needs information is of little use if
hierarchy of needs: safe air, water, and food, a it is not made available to those delivering
970 Emergency Response Coordination Within Earthquake Disasters
immediate support services or planning for structural collapse where the most complex res-
larger-scale longer-term search and rescue, cues may be required, having due regard to both
safety, relief, and eventual recovery programs. the current stability of the structure and potential
It is essential that all impact assessment data impacts from following aftershocks. Ideally
captured during response are captured and col- USAR engineering teams should contain a geo-
lated at a local level and then made available to technical engineer, as the threats of further move-
response and relief organizations in a timely ment posed by unstable hillsides (landslip and
manner. Some users of impact assessment data, rockfall) add considerable complexity to
such as rescue teams and risk managers, will earthquake-affected areas.
need detailed reports and data feeds, while The principal source of reference in relation to
others, such as social service providers and gov- USAR practice and arrangements is the UN-based
ernance bodies, will require more aggregated, International Search and Rescue Advisory Group
area-wide data. (INSARAG) www.insarag.org. The focus of
Impact assessment is most effective when it INSARAG is the development of standardized
has been planned and resourced in sufficient guidelines and procedures and sharing of best
detail prior to an emergency and has ideally practices among national and international USAR
been tested through exercising both data collec- teams and defining standards for minimum
tion and analysis in particular. requirements of international USAR teams. The
INSARAG Guidelines provide frameworks for
Urban Search and Rescue the response operation for the country affected by
Urban search and rescue (USAR) involves the a sudden-onset disaster causing large-scale struc-
location, rescue, and initial medical stabilization tural collapse, as well as for international USAR
of victims trapped in confined spaces following a teams responding to the affected country. They
structural collapse. USAR is an integrated also outline the role of the UN assisting affected
multiagency response beyond the capability of countries in on-site coordination.
normal rescue arrangements in most jurisdic- Irrespective of the scale of the USAR response
tions. In addition to rescue technicians, USAR to a significant earthquake event, the Local Emer-
teams include medical and canine search and gency Management Authority (LEMA) is the
rescue engineering capabilities. ultimate responsible authority for the overall
The canine search component of a USAR team command, coordination, and management of the
focuses on carrying out rapid searches over and response operation. The LEMA can refer to local,
within collapsed or seriously damaged structures. regional, or national authorities or combinations
Being trained to operate remotely from their han- thereof, which are collectively responsible for the
dlers, the search dogs are able to go into areas that disaster response operation.
may be considered too dangerous for rescue tech- In the rescue response following a major
nicians. By alerting when they encounter trapped earthquake, the operations focal point is the
live people, they provide a focal point for search On-Site Operations Coordination Centre
operations and the other equipment that they (OSOCC). The OSOCC is established close to
have, thereby assisting with prioritization and the LEMA and as close to the disaster site as is
saving significant time in a search and rescue safely possible. It provides a platform for the
operation that covers a large area or numbers of coordination of international responders and
buildings. LEMA. The OSOCC is established by the
The engineering component of a USAR team UNDAC team or by the first arriving interna-
assists with the identification, monitoring, and tional USAR team, who will then hand over the
reduction of hazards (within site) and provides OSOCC to the UNDAC team when they arrive.
detailed/specific structural engineering advice An outline representation of the relationships
where required. This includes advice on the between these aspects of response is provided as
least dangerous points of access to areas of Fig. 2.
Emergency Response Coordination Within Earthquake Disasters 971
spontaneous formal and informal responders con- being required, will include not only triaging
verging to sites of structural collapse. and treating the injured but also dealing with
Significant medical support to entrapped, damage to their own facilities and impacts of
ambulatory, and evacuated casualties will be staff capabilities, coordination of service delivery
required following earthquake and may extend and patient care with partner health and social
as far as limb amputation to enable timely extrac- service providers, decanting of less-urgent
tion and specialist support to sufferers of crush patients from the service, and transfer of higher
syndrome or occasionally prolonged dehydration acuity patients to unaffected or less affected facil-
following protracted entrapment. ities elsewhere.
Earthquakes will often have serious implica- Retirement homes and associated hospital and
tions for health services and supporting infra- support services are a growing part of health
structure in the vicinity, as utilities services and services in many communities, due to aging
access for ambulances are compromised; populations and decreasing family size. The con-
community-based health services, general prac- centration of vulnerable people in elderly care
tices, and hospitals, particularly multi-floor facil- facilities or receiving support in the community
ities are damaged; the homes, places of learning becomes a priority for local health services, both
or care, and work places of health workers and in the area affected and, potentially, recipient
their families are also damaged or isolated; and communities in the event of a significant self- or
health workers or family or friends become casu- managed evacuations occurring.
alties themselves. Even in earthquake-resilient
health facilities, damage will be experienced,
not the least to supporting infrastructure and Emergency Welfare
often older neighboring facilities that are less
resilient. Direct and indirect impacts on humans are the
Following significant earthquakes and simi- main focus of response and recovery activities
larly damaging emergencies, the scale of injured following earthquake. Emergency welfare or
who self-report or are transported to existing and social services are the formal and informal pro-
possibly damaged on-site or makeshift triage cesses put in place to assist those affected to
and treatment centers setup on-site will be meet their basic needs following emergencies.
beyond normal mass-casualty planning and These needs are created when people no longer
experience. Innovative approaches will be have access to the essentials of life, including
required to meet needs generated. Preplanning, accommodation or shelter; personal safety and
training, and exercising for earthquake-scale security; food, water, medication, sanitation,
medical responses will substantially enhance and hygiene; care for dependents and vulnerable
site response, field triage, medical center and people; clothing; means of income; and psycho-
hospital resilience and response, and local, social support.
regional, and higher-level health service emer- The majority of emergency welfare services
gency coordination. tend to be provided informally by the impacted
Engineering support is critical to health-care communities themselves, along with their
providers, initially in terms of damage assess- preexisting social networks; extended families;
ment and operational advice, in the immediate neighborhoods; faith, ethnic, and community ser-
aftermath of earthquakes, as decisions are made vice groups; business associations; and spontane-
on the least risk-prone facilities in which to con- ous volunteer groups formed in the wake of the
duct emergency medical care and whether evac- emergency. These efforts are augmented and,
uations are required from possibly damaged where possible, supported by more formal service
facilities. delivery arrangements.
Hospital emergency responses, ideally Initial emergency welfare response tends
planned for and resourced well prior to them to focus on making emergency shelter, food,
Emergency Response Coordination Within Earthquake Disasters 973
water, and relatively hygienic toileting services agencies sometimes give them credit, with alter-
available to affected resident, transient, or visi- native food supply logistics arrangements seldom
tor communities. These services are provided in being required to be established as part of official
community emergency centers, evacuations responses. It is essential, however, that food sup-
centers, or other facilities established to central- pliers are included in emergency preplanning,
ize services and physical resources in relatively relationship-building, and exercises to ensure
safe and secure locations or dispersed within that their needs and capabilities are realistically
the affected communities if residents are shel- taken into account of and included in responses.
tering in place but with access to normal services Distribution of food within affected commu-
compromised by the effects of the earthquake. nities tends to be more difficult to reestablish as
E
Public facilities, such as community centers, retail premises and transport routes within the
sports facilities, halls, or schools, will often be quake zone may well be out of commission for
used in formal and informal welfare support – at least several days. Community engagement,
assuming these facilities remain relatively intact. preparedness, or resilience-building programs
People who have evacuated from their homes, that encourage community members, groups,
workplaces, or visitor accommodation will tend and businesses to have food and water supplies
to congregate open spaces such as parks and for several days will enable them to support
playing fields, particularly where built facilities themselves more readily, while formal and
are unavailable or insufficient to meet local needs spontaneous arrangements are put in place.
or community members are reluctant to enter
buildings following earthquake. Temporary tent
communities will often be established by people Post-earthquake Building Management
who are reluctant or unable to remain in their In broad terms, a post-earthquake building man-
homes but who want to remain in their agement program is comprised of three phases:
communities.
• Overall damage survey
• Rapid building assessments
Emergency Food and Water Supplies • Detailed damage evaluations
The images of food products broken and strewn The overall damage survey is typically carried
across supermarket floors, collapsed shelving in out as part of the immediate response by emer-
warehouses, damaged roads and bridges, burst gency services personnel and local authority
water pipes, and lines of vehicles at petrol sta- staff, as referred to under Rapid Physical Impact
tions are common scenes in news coverage after Assessment above.
earthquakes. The realities of this for the affected Rapid building assessments are generally
communities run much deeper. undertaken by a number of small teams compris-
Earthquakes cause damage and contamination ing volunteering engineers and other building
to water reticulation systems and to food supplies professionals working with local authority build-
in people’s homes, retail and wholesale food ing officials.
facilities, as well as food processing plants. The objectives of a rapid building assessment
Transport routes, including road, rail, water- process are to:
based, and airport infrastructure into the affected
areas, will be damaged to varying degrees, reduc- • Confirm where damage to buildings is concen-
ing supplies for at least days and often weeks or trated to assist response and recovery
months. decision-making
Regional, state, or national food supply whole- • Indicate whether physical action is to be taken
salers are better able to reestablish supply chains to enable, restrict, or prevent access to indi-
more readily than emergency management vidual buildings
974 Emergency Response Coordination Within Earthquake Disasters
• Commence the process of systematically gath- considerable amount of planning and preparation
ering data on damaged buildings that will facil- is required prior to an event.
itate the planning and monitoring of longer- The following key roles are suggested for
term recovery actions and reoccupation inclusion within the rapid building assessment
operational arrangements and leadership
The outputs from a rapid building assessment structure:
process are posted placards on the buildings and
completed forms summarizing the damage • Building assessment manager
observed and placards posted that are returned • Technical and process adviser
for entry into the database of the respective • Central business district/commercial technical
local authority or other responsible agency. lead
The focus of the rapid building safety evalu- • Residential technical lead
ation process is on immediate public safety, not • Engineering resource coordinators
the provision of an engineering assessment ser- • Information management coordinator
vice to building owners. It is an initial triaging • Regulatory process support (transition
process, akin to initial medical assessments at management)
the scene of a large emergency or in a hospital
emergency department. Buildings that have The scale of the event (and hence rapid build-
sustained visible and significant structural dam- ing assessment operation) will dictate whether all
age are marked via a placarding operation as not of these roles are required to be filled.
being suitable for reoccupancy. Other buildings Detailed damage evaluations are generally
that do not show signs of visible damage or carried out by structural and geotechnical engi-
movement may be suitable for occupancy, but neers specifically engaged by building owners.
they require further attention subsequently from They involve accessing all available information,
building owners and their engineers when detailed interior inspections, and performing spe-
resources become available. cific calculations where required. They can take
Rapid building assessments are just anywhere from one day to a week or more,
that – undertaken quickly in the face of poten- depending on the size of the building and the
tially large numbers of buildings to be appraised type of damage. Detailed damage evaluations
within an affected area. Rapid assessments of for critical facilities, particularly hospitals,
only the exterior are expected to take of the should however be undertaken in the immediate
order of 10–20 min, while rapid assessments aftermath of an earthquake, and this requires
involving exterior and interior inspections can appropriate priority response agreements to be
take anywhere from 1 to 4 h, depending on the established with engineers familiar with the facil-
size of the building. These assessments are ities prior to an earthquake.
almost always undertaken without recourse to
structural drawings. Lifeline Utility Coordination
The coordination of a rapid building assess- The utility services sector is invariably complex
ment operation requires an operational structure in a day-to-day context, much less in a post-
that links between the local jurisdictions’ build- earthquake situation. Within the key sectors
ing control unit and the emergency management including water, energy, telecommunications,
leadership arrangements. As any such operation and transport, there are typically a range of
requires considerable follow-up activity by the providers.
affected council(s), it is important that the oper- The coordination of the information and
ation is integrated as much as possible with build- actions across the utility sector following a
ing control officials, particularly the leadership of major earthquake is accordingly complex and
the operation and management of data. Hence, a requires specific planning. The specific
Emergency Response Coordination Within Earthquake Disasters 975
coordination arrangements will vary depending utility coordination is most likely to be activated
on the legislative structures and operating at both regional and national levels, rather than
arrangements in place. The key common feature local levels.
is for such arrangements to enable a strong link- The Lifeline Utility Coordinator (LUC) tends
age between the utility sector (collectively) and to be a non-statutory regional and national emer-
the emergency management sector. This linkage gency management position responsible for coor-
needs to be active in nonemergency times, given dinating lifeline utilities on behalf of the response
the emphasis that should be given to risk reduc- controller or director during emergency response
tion and preparedness. and the recovery manager throughout recovery
The objectives of post-earthquake utility coor- activities.
E
dination are to: The roles are assigned before an emergency
occurs and activated if required during an
• Coordinate the actions of lifeline utilities and emergency.
emergency/disaster management organiza-
tions to provide a safe and effective response
• Support the restoration of lifeline services as Summary
quickly as possible
In summary, responses to significant earthquakes
Utility coordination functions include: are challenging and complex. This entry has
explored the key aspects of earthquake response
• Managing information flows that are particularly pertinent to the context of
• Monitoring impacts earthquake engineering. Although responses to
• Communicating lifeline utility status informa- virtually all emergencies and disaster start and
tion within the respective coordination center finish with the affected communities, formal
(s) leadership and coordination of responses are
• Facilitating any support required by utilities vitally important for effective response to an
for emergency/disaster management or vice earthquake event. When significant earthquakes
versa strike, the challenges for those in positions of
leadership or who are responsible for response
All countries have emergency management coordination are many and varied. This is due in
legislation and related regulations and plans large part to the reality that affected communities
which, to varying degrees of detail and compul- will be mounting their own responses while offi-
sion, define key infrastructure providers as “life- cial responses are still regrouping and
line utilities,” “critical Infrastructure,” or similar reestablishing communications. The very nature
and require organizations falling under this defi- of damaging quakes makes them some of the
nition to be able to continue to operate during and most difficult hazards to manage responses to,
after an emergency. This has been spelled out as the sudden and widespread impacts signifi-
more clearly in the New Zealand context than cantly affect the ability of response organizations
some other, where a wide range of life utilities to deliver their services more than other hazards
are required to have planned and resourced their tend to.
organizations, facilities, and services to be able to
continue to operate to the fullest extent possible
during and after emergencies. Cross-References
Lifeline utility coordination is a function
defined usually carried out in response coordina- ▶ Resilience to Earthquake Disasters
tion centers under the respective incident man- ▶ Resiliency of Water, Wastewater, and Inunda-
agement system’s operations function. Lifeline tion Protection Systems
976 Empirical Fragility
Empirical Fragility, Fig. 1 Illustration of (a) a column of a DPM for intensity im, (b) fragility curves corresponding to
n = 3 damage states for the same building class
Empirical fragility is typically expressed lifeline inventory. Many of these studies include
either in terms of damage probability matrices fragility functions for multiple classes of assets.
(DPMs) that express the probability of a class of Figure 2 shows a steady increase in the number
assets with similar attributes suffering given of studies on empirical fragility of buildings over
damage states under specified ground motion or the past 40 years. Recently, the Global Earthquake
a continuous relationship between a measure of Project (GEM) (Rossetto et al. 2013) compiled a
damage (damage state) and ground motion or database of empirical fragility functions for the
ground failure intensity (see Fig. 1) for a specified global building inventory. GEM’s database con-
asset class. The continuous functions are over- sists of 175 empirical fragility functions
whelmingly expressed in terms of fragility constructed for the global building inventory. Fig-
curves. These curves are obtained by fitting a ure 3 shows that almost half of existing functions
statistical model to the post-earthquake damage have been constructed for the European building
observations. The two expressions of empirical stock. Most empirical functions are based on post-
fragility are related as depicted in Fig. 1b, where earthquake observations collected from a single
it can be noted that the distances between two earthquake event (e.g., Braga et al. 1982). Few
successive fragility curves for a given intensity exceptions exist of functions based on databases
measure level relate to the corresponding values from multiple global (e.g., Spence et al. 1991;
in the damage probability matrix. Rossetto and Elnashai 2003) or national (e.g.,
Empirical fragility assessment was firstly Rota et al. 2008) events. Recently, GEM
introduced by Whitman et al. (1973), who (Rossetto et al. 2014) built on the state-of-the-art
constructed DPMs for tall buildings by relating research and proposed new guidelines for the con-
an 8-state discrete damage scale to loss data col- struction of empirical fragility curves from single
lected from 1,240 questionnaires in the aftermath or multiple earthquake event databases. Within
of the 1971 San Fernando earthquake, USA. Since these guidelines procedures are proposed for the
then, approximately 89 studies are seen in the assessment of the quality of the post-earthquake
published literature (i.e., academic journals, con- observation databases and for fragility relation-
ference proceedings, reports, and book chapters) ships to be derived using statistical approaches of
to have focused on the development of empirical increasing complexity according to the character-
fragility functions for the global building and istics of the adopted observational databases.
978 Empirical Fragility
Empirical Fragility,
Buildings
Fig. 2 Studies focused on
25 Lifelines
empirical fragility
assessment of buildings and
lifeline networks published
20
No of Studies
in the last four decades
15
(Data sourced from the
compendium of empirical
10
fragility functions of
buildings in GEM
5
(Rossetto et al. 2013) and
Syner-G (Pitilakis et al.
0
2014))
1973−1980 1981−1990 1991−2000 2001−2010 2011−2014
Year of publication
Montenegro
Spain
Europe
Iran
Armenia
Taiwan
Worldwide
Mexico
Japan USA
Fragility
Function
Empirical Fragility, Table 1 Main survey types and their main properties
Typical
Typical no. of Reliability
Typical building damage of
Survey method sample sizes classes states observations Typical issues
Rapid surveys Large All 2–3 Low Safety not damage evaluations.
buildings Misclassification errors
Detailed Large to Detailed 4–6 High Possibility of under-coverage errors
“engineering” small Classes
surveys
Surveys by Very small Detailed 4–6 High Possibility under-coverage errors or
reconnaissance Classes very small samples
teams
Remotely Very large All 2–3 Low Only collapsed or very heavy
sensed buildings damaged states may be reliable.
Misclassification errors
commonly found to suffer from two main errors, “engineering” surveys cover large number of
namely, misclassification and nonresponse assets, they tend to concentrate on the damaged
errors. The former type of errors is introduced assets, ignoring the undamaged ones. This practice
when the safety level of a building is different introduces an error, termed under-coverage, in the
from the actual damage suffered by the assets. database. This error introduces a bias in the data-
For example, a building sustained substantial base, which can be reduced by the use of data from
damage and should be classified as “unsafe” but alternative sources when the source of under-
the survey form indicates it is “safe.” This error coverage error is known. For example, if it is
can potentially have a significant impact on fra- known that only damaged buildings have been
gility functions (see Ioannou and Rossetto 2013). surveyed, it may be reasonable to estimate the
Nonresponse errors are instead introduced when number of undamaged buildings from census data.
the forms used for collecting the survey data are Reconnaissance surveys are conducted by
incomplete. For example, the location or address small teams of highly experienced engineers.
of surveyed assets is missing from the form. The These surveys are therefore of high quality but
mechanism of the missing data determines how often suffer from small sample sizes, given the
this error is treated. For example, due to the speed limited time the reconnaissance teams spend in
of “rapid;” surveys, some boxes in the survey the field and the small areas surveyed.
forms are left blank completely at random. In Finally, surveys based on remote sensing tech-
this case, the forms with incomplete data can be niques have been recently introduced for the
removed from the database. This leads to, some- empirical fragility assessment of buildings. Of
times substantially, smaller sample sizes. By con- note is the recent establishment of the Virtual
trast, the missingness mechanism could depend Disaster Viewer (Bevington et al. 2010),
on the missing observations. For example, the supported by several international institutions
safety assessment on the form is left blank, if that provide images and a forum for experts to
the building is classified as safe. In this case, the interpret damage from satellite images. The sub-
knowledge of this mechanism is essential in order sequent GEO-CAN initiative (Ghosh et al. 2011)
to effectively reduce the bias introduced by this also brought together experts from all over the
nonresponse error. world to interpret damage in the 2010 Haiti and
“Rapid” surveys can be followed by detailed 2011 Christchurch, New Zealand earthquakes.
“engineering” surveys. These surveys collect These surveys have the ability to gather data for
detailed data concerning the characteristics of large sample sizes of the assets of interest. How-
assets and damage levels. Although the ever, as they are based on observations of
Empirical Fragility 981
changes to building roofs and plans from com- 6 damage states. Fragility functions are
parison of pre- and post-earthquake aerial constructed for each damage state, and thus
images, they are unable to distinguish between there needs to be a significant number of obser-
building classes and are only able to assess the vations for each damage state in order to ensure
collapse (and to lesser extent heavy damage) state the reliability of the corresponding empirical
in buildings. Consequently, the inaccuracies function. In practice, a balance needs to be
involved in the assessment of the damage lead reached between the number of damage states
to substantial misclassification errors for the and the number of observations. In other words,
non-collapse damage states (Booth et al. 2011). for the same sample size, the fewer the damage
states, the more data in each, but the wider the
E
Damage Assessment damage states, the less informative the fragility
The seismic damage suffered by assets is typi- functions in the context of loss estimation. Fur-
cally expressed in terms of a damage scale with thermore, due to the rarity of large earthquake
n number of discrete qualitative states. The dam- events near urban areas, higher damage states
age states can be defined in terms of structural (e.g., collapse) tend to contain few observations,
and nonstructural damage levels. They vary from and hence fragility functions constructed for
no damage to collapse and are often considered to these states may be less reliable than the func-
be mutually exclusive and collectively exhaus- tions corresponding to less severe damage states.
tive. The scales, adopted in the literature, vary in A notable exception to the above discrete def-
the number of damage states and the level of inition concerns the pipes, whose levels of dam-
detail in the description of damage for each dam- age are measured in terms of the number of
age state according to the needs of the survey. repairs per mile.
Damage scales with n = 4–6 states are mostly
adopted for the classification of damage observed Characteristics of Assets
in buildings and lifeline components. Some of the Assets affected by similar intensity measure levels
most commonly used damage scales for post- often perform very differently. These differences
earthquake damage surveys for buildings are the can be, at least partially, attributed to their unique
European Macroseismic Scale (Gr€unthal 1998) structural characteristics. Empirical fragility
and its predecessor the Medvedev-Sponheuer- assessment requires the availability of statistically
Karnik scale (Medvedev 1977). For lifelines, significant sample sizes in order to construct reli-
HAZUS proposes qualitative damage scales for able fragility functions. Thus, classes of assets are
most components, which are observed to be sim- often determined based on the characteristics that
ilar to those for buildings. Syner-G (2014), significantly influence their seismic performance.
instead, recently proposed damage scales In theory, the more detailed the class, the more
expressed in terms of operational states, which homogenous the group of assets and the smaller
describe whether or not the component functions. the variation in seismic response. However, in
Despite the presence of well-established damage practice narrowly defined classes often result in
scales, many studies adopt national damage small sample sizes. A careful balance is, therefore,
assessment scales (e.g., AeDES in Italy) or needed between level of detail in the class defini-
bespoke scales. This wealth of damage scales is tion and sample size.
problematic when databases or fragility functions For example, the structural characteristics of
need to be combined or compared. This problem buildings, which influence their seismic perfor-
has been recently addressed by Rossetto (2014) mance, are well documented (e.g., FEMA -
who, building on the work of Hill and Rossetto 547 2007) and include (but are not limited to):
(2008), has drawn equivalence tables between the
various damage scales found in the literature. • Construction material
In the construction of empirical fragility func- • Lateral load resisting system (e.g., moment-
tions, ideally a damage scale should contain 3 to resisting frame, shear wall, etc.)
982 Empirical Fragility
to the dynamic loads imposed on very stiff Empirical Fragility, Table 2 Example of a damage
structures and, therefore, may not be adequate probability matrix for a theoretical construction class
for the characterization of seismic demand in Bins of
medium or high rise building populations. intensity (e.g.,
PGA in g) 0.02–0.04 0.04–0.06
Peak ground velocity (PGV) and peak ground
Damage Counts DPM Counts DPM
displacement (PGD) are more representative of (%) (%)
the seismic demand on structures of intermedi- No damage 100 50 60 30
ate and long natural period of vibration Slight 80 40 90 45
than PGA, respectively. Both PGV and PGD Moderate 18 9 45 23
are generally calculated through the integration Collapse 2 1 5 3 E
of accelerograms and may be sensitive to both Total 200 100 200 100
the record noise content and the filtering
process.
The peak ordinates of a ground motion record, the construction of fragility functions for classes
however, are unable to capture features of the of assets. Empirical fragility functions
earthquake record, such as frequency content, are essentially stochastic relationships, which
duration, or number of cycles, which affect the relate the damage to an asset class with the
response of a structure and its consequent damage corresponding ground motion or ground failure
and loss. Hence, in an attempt to better capture intensity. These relationships can be discrete
the influence of accelerogram frequency content, (i.e., DPMs) or continuous (e.g., fragility
more recent empirical fragility studies have curves).
favored the use of response spectrum-based
parameters as measures of ground motion inten- Damage Probability Matrices
sity. These parameters include the spectral accel- In its simplest form, empirical fragility is
eration for the fundamental period and less often expressed in terms of a discrete function. Its
the spectral displacement. construction requires the aggregation of the dam-
Ground failure (particularly liquefaction, lat- age data for assets of a given class into bins of
eral spreading, and fault rupture) is expressed in similar ground motion or ground failure intensity
terms of permanent ground deformation, which levels. The counts of buildings suffering given
can be measured laterally, vertically (i.e., settle- levels of damage for an examined bin, normal-
ment), or volumetrically. Vertical settlement can ized by the total number of buildings in this bin,
be measured in terms of total settlement or dif- provide the values of the damage probability
ferential settlement, which has greater implica- matrix (see Table 2).
tions for damage to structures and lifelines.
There are currently no standardized parameters Continuous Fragility Functions
for describing permanent ground deformation, Fragility functions are typically constructed by
though Bird et al. (2006) summarize some fitting a statistical model to the available post-
more common parameters, which vary in terms earthquake data. Practically, a statistical model
of input requirements and methodological consists of two components: a random and a
approach. There have been few empirical dam- systematic component. The random component
age observation studies related to ground failure expresses the probability distribution of the
incidents and therefore fewer studies on fragility response variable (e.g., the counts of buildings
estimation to these hazards (e.g., Tokimatsu reaching or exceeding a damage state) given the
et al. 1994). explanatory variable (i.e., the ground motion or
ground failure intensity). The systematic compo-
Statistical Model Fitting Procedures nent expresses the mean response as a function of
Having collected post-earthquake data, the the explanatory variable, in other words the mean
empirical fragility assessment concludes with empirical fragility function.
984 Empirical Fragility
There are two main types of continuous fra- where Ni are the counts of assets that have been
gility functions, described below, depending on damaged to a level equal or exceeding a
whether the examined assets can be considered predetermined damage state, DS dsi; n is the
point-like or line-like. The difference between total number of assets in a bin associated with
the two types is that point-like assets occupy a intensity level im. This distribution is essentially
relatively small area where the intensity is the fragility curve and is characterized by the
assumed constant for each asset. Point-like assets mean response for given intensity, mi, as:
include standard buildings and all components of
lifeline networks presented in Fig. 4, with the y1i lnðimÞ þ y0i ð:1Þ
gðmi Þ ¼ (2)
exception of pipes which are considered line- y1i im þ y0i ð:2Þ
like assets.
where g(.) is termed the link function and relates
Fragility Functions for Point-Like the mean response with the intensity or the natu-
Assets Fragility functions for point-like assets ral logarithm of the intensity. The link function
are expressed in terms of fragility curves. In the can be expressed in three forms:
literature, fragility curves are typically
constructed by fitting parametric statistical 8 1
models to the available data. A parametric >
> F ðm Þ probit ð:1Þ
< i
mi
model represents a family of probability distribu- gðmi Þ ¼ log logit ð:2Þ
>
> 1 mi
tions of the response conditioned on the explan- :
log½logð1 mi Þ complementary loglog ð:3Þ
atory variable defined by a small number of
(3)
parameters. The most commonly used approach,
which is described below, is to fit a generalized
linear model to the data (e.g., Shinozuka The first two functions presented in Eq. 3 are
et al. 2000). These models are based on strong symmetrical functions, and the last one is
assumptions, which, when they are satisfied, lead asymmetrical.
to reliable predictions. Nonparametric models The parameters of the generalized linear
have also been adopted for the construction of models can be estimated by two procedures. In
empirical fragility curves, but they suffer from the first procedure, the parameters are estimated
the danger of over-fitting, especially if the by maximizing the likelihood function through an
data are aggregated in only small number of iterative least squares algorithm, as:
bins of ground motion or ground failure intensi-
ties. These models include the generalized addi-
yi opt ¼ maxarg½Lðyi Þ
tive models (see Rossetto et al. 2014) and " #
the Gaussian Kernel Smoothers (see Noh mbins nj
nj nij
¼ maxarg ∏ mij 1 mij
nij
et al. 2013). j¼1 nij
The generalized linear model is based on the
(4)
underlying assumption that the seismic response
of each asset is a Bernoulli trial, i.e., the response
is equal to 1 if the sustained damage is DS dsi In the second approach, a Bayesian regression
and 0 otherwise. Given that the damage data analysis (e.g., Straub and Der Kiureghian 2008;
typically are aggregated in bins of similar ground Ioannou and Rossetto 2013) is adopted in order to
motion or ground failure intensity, the response is take into account prior information regarding the
considered to follow a binomial distribution for model’s parameters, u, especially when the avail-
intensity level, im: able number of observations is small. Prior infor-
mation is obtained from existing fragility
n functions or independent post-earthquake data
N i jim mi ni ½1 mi nni (1)
ni of similar groups of assets. In addition, this
Empirical Fragility 985
analysis allows for the increase in the model’s where f(u) is the prior distribution of the regres-
complexity by explicitly accounting for the mea- sion parameters accounting for existing prior
surement error in the intensity measure levels knowledge.
(e.g., Straub and Der Kiureghian 2008) or the
misclassification of the damage (e.g., Ioannou Fragility Functions for Line-Like Assets The
and Rossetto 2013). The aim of the Bayesian continuous fragility functions, constructed for
analysis is the estimation of the posterior distri- groups of line-like assets, measure the damage
bution of the model parameters, f(u|y, x), from in terms of the number of repairs per unit length
Bayes’ theorem: (termed repair rate) of these assets for given
levels of intensity. The repair rate (RR) is related
to the intensity through the use of a linear (i.e., the E
Lðy; y, xÞf ðyÞ Lðy; y, xÞf ðyÞ
f ðyjy, xÞ ¼ ¼ð parameters of the model are linearly combined)
f ðy Þ
Lðy; y, xÞf ðyÞdy or nonlinear (i.e., the parameters of the model are
nonlinearly combined) statistical model. These
(5)
models can be expressed in different forms:
where s is the standard deviation, which is con- observations on damage to assets, their charac-
stant for each intensity level; m(im) is the mean teristics, and the ground motion or ground failure
repair rate or the lognormal repair rate or the they are subjected to. This typically includes the
natural logarithm of this rate, which is expressed fitting of a parametric statistical model to the
as a function of the intensity. The model is fit to observational data. The reliability of empirical
the available data using a least squares method. fragility functions depends on two main condi-
The number of breaks or leaks for the total tions: firstly, on the quality of the post-earthquake
length (L) of the line-like asset is considered to observational data and whether the database
follow a Poisson distribution. Therefore, if we includes statistically representative samples of
assume that the examined asset fails if there is the exposed assets and, secondly, on whether
at least one break, then the probability of failure the assumptions on which the selected statistical
for the total length of the asset is obtained, as: model is based upon are satisfied. To date, studies
struggle to satisfy both of these conditions.
Pðline-like asset segment failsÞ
¼ 1 PðNo failureÞ ¼ 1 eRRL (7)
Cross-References
Summary
▶ Analytic Fragility and Limit States [P(EDP|
The empirical seismic fragility of an asset (i.e., IM)]: Nonlinear Dynamic Procedures
buildings or components of lifeline networks) ▶ Analytic Fragility and Limit States [P(EDP|
expresses the potential for its damage in future IM)]: Nonlinear Static Procedures
earthquake events. Empirical fragility is assessed ▶ Seismic Risk Assessment, Cascading Effects
by the statistical analysis of post-earthquake ▶ Seismic Vulnerability Assessment: Lifelines
986 Engineering Characterization of Earthquake Ground Motions
procedures to compute acceleration, velocity, and that were colocated. In this particular example,
displacement time series for use for seismic haz- the analog instrument was triggered by the
ard studies and engineering applications; and S-waves.
(iv) definition of the main intensity measures
computed from the accelerometric time series.
Strong-Motion Data Processing
Engineering Characterization of Earthquake Ground UTC): acceleration time series at the analog station NCR
Motions, Fig. 2 Aftershock of the Umbria-Marche seis- (top); acceleration time series at the digital station NCR2
mic sequence, central Italy (event 1997-10-0705:09:56 (bottom) (Source http://itaca.mi.ingv.it)
Engineering
Characterization of
Earthquake Ground
Motions, Fig. 3 Example
of spurious spikes (raw
record of event 1998-01-
1301:49:16 UTC at SELW)
(Source http://itaca.mi.
ingv.it)
sources of noise and they can often be visually an approach to compensate for these problems
identified and removed prior to further consisting in the use of baseline adjustments
processing. (Graizer 1979; Iwan et al. 1985; Boore 2001;
As an example, spurious “spikes” in the digi- Boore et al. 2002; Paolucci et al. 2008). In par-
tized record can be replaced with the mean of the ticular, the use of baseline-fitting techniques can
accelerations of the data points on both sides of sometimes recover permanent ground displace-
the spike. Another problem encountered with ments from accelerograms.
both analog and digital accelerograms are distor-
tions and shifts in the baseline (Fig. 4), which Filtering
result in unphysical velocities and displacements The process of filtering generally involves the
(Boore 2003). Boore and Bommer (2005) detail removal of most of the signal at frequencies
990 Engineering Characterization of Earthquake Ground Motions
Engineering 80
Acceleration [cm/s2]
Characterization of NCR
Earthquake Ground
Motions, Fig. 4 Example
of baseline shift and its 0
effect on the velocity time
series (event 1979-05-
2112:36:40 UTC recorded
at NCR) (Source http://
itaca.mi.ingv.it) −80
0 2 4 6 8 10 12
Time [s]
8
Velocity [cm/s]
Χ8
0 2 4 6 8 10 12
time [s]
where the Fourier amplitude spectrum (FAS) Low Frequency Noise Removal
shows a low signal-to-noise ratio. Therefore, the The impact of low-frequency noise (<1 Hz) can
critical issue for end users is to appreciate the have strong influence on strong-motion intensity
record limitations after certain frequencies have measures such as ground velocities, displace-
been removed. ments, and response spectral ordinates, and there-
Filters can be applied in the frequency or time fore most of the studies have investigated the
domains; exactly the same results should be effects of low-cut (high-pass) filters.
obtained in both cases. While the choice of filter The low-frequency cutoff, the lowest fre-
type is not crucial, the way in which the filter is quency for which the data are reliable in terms
applied to the accelerograms is very important. of signal-to-noise ratio, can be determined in
The fundamental choice is between causal and several ways: (i) comparison of the FAS of the
acausal filtering, where the latter does not pro- record with that of a model of the noise obtained
duce any phase distortion in the signal, whereas from the pre-event memory for digital records,
causal filters do result in phase shifts in the for example; (ii) judgment of the frequency where
record. The influence of causal and acausal filters the record FAS deviates from the tendency to
on both elastic and inelastic response spectra has decay in proportion to the reciprocal of the fre-
been investigated by Boore and Akkar (2003), quency squared, based on a seismological model
who found that both elastic and inelastic response of the radiated energy (omega-square model,
spectra computed from causally filtered acceler- Brune 1970); or (iii) visual inspection of the
ations can be sensitive to the choice of filter velocity and displacement time series obtained
corner periods even for oscillator periods much by double integration of the filtered acceleration.
shorter than the filter’s corner periods. Finally, Figure 5 shows the effect of low-cut filtering.
Bazzurro et al. (2005) performed a statistical A digital record is filtered using two different
study on the influence of various signal- low-cut frequencies. The lowest frequency gen-
processing techniques used for low-cut filtering erates a low signal-to-noise ratio and leads to
ground motion records on both elastic and inelas- unrealistic velocity and displacement time series
tic response spectra. (gray boxes highlight the differences).
Engineering Characterization of Earthquake Ground Motions 991
Engineering Characterization of Earthquake Ground low-cut equal to 0.01 Hz; right: low-cut = 0.2 Hz (ACC,
Motions, Fig. 5 Example of low-cut filtering (event acceleration; VEL, velocity; DIS, displacement)
2010-01-1213:35:45UTC recorded at AQA). Left:
Acausal low-cut filters are generally preferred frequencies greater than the Nyquist will have no
(Boore 2005) and, in order to achieve the zero effect on the record.
phase shift, they need to start prior to the begin- Douglas and Boore (2011) give an exhaustive
ning of the record and finish after the end of the overview of the approaches and effects of high-
accelerogram, which can be accomplished by cut filtering. While low-cut filters have large
adding points of zero amplitude, known as pads, influence on ground motion intensity measures,
to the two ends of the record. The length of the high-cut filtering is sometimes even not necessary
pads depends on the filter frequency and its order and accelerometric response spectra can often be
(Boore 2005). used up to frequencies much larger that the
After filtering, the processed accelerograms applied high-cut corner used to remove the
are integrated once to obtain velocities and noise. As in the case of low-cut filters, acausal
again to obtain displacement time series. They high-cut filters are generally preferred.
can also be used to obtain other intensity mea-
sures (see below). Applicability of Processed Data
The user should be aware that different corrected
High-Cut Filters versions of the same record may be distributed in
Ground motion at periods of less than roughly different strong-motion databases. Foti and
0.05 s is generally only relevant to particular Paolucci (2012) studied the influence of correc-
engineering problems, such as the design and tion procedures by performing numerical simula-
analysis of nonstructural elements, equipment, tions (seismic behavior of a diaphragm wall). The
and pipework (e.g., in nuclear power plants). results show large differences, especially in terms
Before applying high-cut (low-pass) filters of displacements, which is a parameter of partic-
one should consider that the sampling rate deter- ular relevance when dealing with performance
mines the upper usable frequency, termed the based design.
Nyquist frequency, equal to (1/2Dt) where Dt is Sometimes database providers distribute
the sampling interval. High-cut filter applied at processed accelerograms that are incompatible
992 Engineering Characterization of Earthquake Ground Motions
with the provided velocity and displacement time Differently from PGA, PGV and PGD can be
series because they apply additional processing to read only after processing the accelerometric
reduce noise that is still present in the record. In record, and their values are strictly related to the
other cases, time series are not compatible band-pass filtering, especially in case of PGD,
because the pads added for the application of which strongly depends on the low-cut corner
the filter are removed. The consequence of their frequency.
removal is to weaken the effect of the filter and Although the PGA is widely used in earth-
this can result in offsets and trends in the base- quake engineering, as it is directly related to the
lines of the velocity and displacements obtained inertia force, it poorly characterizes the ground
by integration. motion, as demonstrated in Fig. 6. Records with
In order to avoid the latter effects the correct the same PGA can have completely different
initial conditions in the pad-stripped time series features in the other ground motion parameters,
have to be used when computing displacements, depending on the magnitude of the event, the
velocities, and linear oscillator response. Another source-to-site distance and the geological condi-
way of ensuring compatibility is to post-process tion of the site where the recording station is
the pad-stripped acceleration time series. located.
Boore et al. (2012) examined the post- A preliminary estimation of the frequency
processing adopted for two databases, ITACA content of the records may be inferred by the
(Italian Accelerometric Archive, Paolucci PGA/PGV ratio (Zhu et al. 1988; Sawada
et al. 2011; Pacor et al. 2011) and PEER NGA et al. 1992; Kwon and Elnashai 2006). Low
(Pacific Earthquake Engineering Research PGA/PGV ratios generally indicate earthquakes
Center–Next Generation Attenuation), and dem- with low predominant frequencies, broader
onstrated that any biases and distortions are small response spectra, and longer duration.
for the vast majority of the records of the two The duration of the accelerogram is a param-
databases, so data can be used with confidence. eter used to identify the portion of record in
which ground motion amplitude can potentially
cause damage to engineering and geotechnical
Strong-Motion Intensity Measures structures.
Several definitions are proposed (Bommer and
For many engineering and scientific applications, Martinez-Pereira 1999), as many different mean-
it is useful to describe the seismic ground motion ings can be attached to “duration.” The most
in terms of a few simple parameters instead of the commonly used are as follows:
entire waveform. These parameters are evaluated
both in the time and frequency domains, each of (a) Bracketed duration, introduced by Bolt
them capturing some features of the recorded (1973): a threshold is fixed, typically 0.05 g
accelerograms. or 0.1 g, above which it is deemed that the
motion has relevance for engineering pur-
Time-Domain Parameters poses; the duration is the time interval
The most common parameter is the peak ground between the first and the last exceedance of
acceleration, PGA, which denotes the maximum this value.
ground acceleration (absolute value) observed on (b) Significant duration, introduced by Trifunac
the accelerometric time series, usually evaluated and Brady (1975) on the basis of the study by
for each component of the ground motion. It can Husid (1967), is defined as the time interval
be evaluated directly either from raw or corrected over which the integral of the square of the
signals, since PGAs change slightly with respect ground acceleration (Husid plot) is within a
to the processing procedure. given range of its total value. Usually this
Similarly, the peak ground velocity, PGV, and range is between 5 % and 95 % or between
peak ground displacement, PGD, are defined. 5 % and 75 %. An example of the two
Engineering Characterization of Earthquake Ground Motions 993
Engineering 1
Characterization of
0.8 T90 =7.6s
Earthquake Ground
Husid [%]
Motions, 0.6
Fig. 7 Accelerometric 0.4
waveform (bottom) and
corresponding Husid plot 0.2
(top) at station AQV (event 0
2009-04-06 01:32:32 UTC, 0 5 10 15 20 25 30 35
Italy). The values of
bracketed duration
(bottom) and significant 400
<T> = 12.1s
[cm/s2]
−400
0 5 10 15 20 25 30 35
Time [s]
definitions of duration is presented in Fig. 7 The Arias intensity is directly related to the
for the L’Aquila, Italy, earthquake (Mw 6.3) definition of the significant duration. It is an inte-
recorded at AQV. As marked in this figure, gral parameter adopted to measure the severity of
the time interval between 5 % and 95 % of the ground motion and it has been found to be a
Husid plot represents the significant duration useful measure for thresholds of shaking that
of the record. trigger landslides (Jibson et al. 1998).
994 Engineering Characterization of Earthquake Ground Motions
Engineering
Characterization of Mw 5.1
Earthquake Ground 100 Mw 6.3
Motions, Fig. 8 Fourier
amplitude spectra (FAS)
for the accelerometric 10
FAS [cm/s]
waveforms recorded at
AQV station for event
2009-04-07, M 5.1 (Italy) at 1
epicentral distance R =
5.8 km and event 2009-04-
06, M 6.3 (Italy) at
epicentral distance R = 0.1 E
5.1 km
0.01
0.1 1 10
Frequency [Hz]
Engineering Characterization of Earthquake Ground accelerometric time series of Fig. 6. Different axis scales
Motions, Fig. 9 Acceleration response spectra are adopted to highlight the different spectral shapes asso-
(SA) (left), velocity response spectra (SV) (center) and ciated to the signals
displacement response spectra (SD) (right) for the
different damping ratios, x (usually a value equal natural circular frequency of the oscillator. When
to 5 % of the critical damping is assumed, which acceleration, velocity, or displacement is consid-
is roughly applicable to most structures). ered, the following definitions are introduced:
The maximum amplitude of the response is
obtained by integrating the equation of motion Displacement spectrum SDðT n , xÞ ¼ maxt jyðtÞj
(relative)
of the SDOF system:
Velocity spectrum (relative) SV ðT n , xÞ ¼ maxt jy_ðtÞj
Acceleration spectrum SAðT n , xÞ ¼ maxt jx€ðtÞj
x€ðtÞ ¼ o2n yðtÞ 2xo2n y_ðtÞ (6) (absolute)
Engineering Characterization of Earthquake Ground Motions, Table 1 Strong-motion parameters inferred from
accelerometric records plotted in Fig. 6
PGA [cm/s2] PGV [cm/s] PGD [cm] AI [cm/s] HI [cm] T90 [s]
U301 113 13.0 2.60 17.0 43.7 16.6
NCR2 103 2.1 0.06 2.5 2.1 1.5
T0800 104 5.1 0.35 5.3 11.1 11.6
SRT 106 7.0 1.20 5.8 23.1 12.7
Figure 9 shows the acceleration, SA, velocity, (SI) defined as the area under the pseudo-velocity
SV, and displacement, SD, response spectra response spectrum (PSV) over the period range
(5 % damping) for the acceleration time series 0.1–2.0 s (Housner 1952):
shown in Fig. 6.
At period T = 0 s, the spectral displacement ð
2:5
and velocity are zero and the spectral acceleration SI ðxÞ ¼ PSV ðT, xÞdT (9)
equals the PGA; this is the reason for which the 0:1
PGA is largely adopted in seismic codes to
anchor design spectra. On the other hand, when This parameter measures the severity of seismic
T approaches infinity, SA approaches zero and motion in terms of potential damage, since the
SD approaches PGD. majority of structures have a fundamental period
The pseudo-acceleration and pseudo-velocity of vibration in the range 0.1–2.5 s. Dimension-
spectra are also widely used in practice. They are ally, the Housner intensity is a displacement.
defined as a function of the displacement spec- Table 1 lists the main strong-motion parame-
trum as follows: ters relative to the records plotted in Fig. 6.
pseudo-acceleration spectrum
2 Public Sources of Strong-Motion
2p Records
: PSAðT n , xÞ ¼ SDðT n , xÞ (7)
Tn
Various governmental and local agencies, as well
pseudo-velocity spectrum: as many universities and research institutes, oper-
2p (8) ate strong-motion networks. Depending on the
PSV ðT n , xÞ ¼ SDðT n , xÞ
Tn agency, accelerometric data are distributed in
various ways (real time or triggered by an event,
The response spectra (SA or PSA) of the raw or processed, with basic metadata or revised
processed ground accelerations are usually metadata). The main sources of strong-motion
published alongside the accelerograms by vari- data (worldwide, European, national) for engi-
ous recording agencies. neering application are discussed in detail in this
The pseudo-spectra well approximate the paragraph. Other sources are listed in Table 2.
absolute response spectra, in case of small
damping values and intermediate and high fre- Worldwide Databases
quencies, while they exactly match the absolute The major source of US strong-motion data is the
response spectra in case of an undamped oscilla- California Strong Motion Instrumentation Pro-
tor. Pseudo-spectra and spectra differ at very low gram (CSMIP). The Centre for Engineering
frequencies, since SV tends to PGV and PSV Strong Motion Data (CESMD http://www.
tends to zero (Hudson 1979). strongmotioncenter.org/) distributes processed
A ground motion parameter related to PSV is and unprocessed strong-motion data from US
the Housner intensity (HI) or spectral intensity and major international earthquakes.
Engineering Characterization of Earthquake Ground Motions 997
Engineering Characterization of Earthquake Ground Motions, Table 2 List of the main public sources for
archiving and distributing strong-motion data
Database name Database code URL
Centre for Engineering Strong Motion Data CESMD http://www.strongmotioncenter.org/
COSMOS Virtual Data Center COSMOS http://www.cosmos-eq.org/VDC/index.html
Pacific Earthquake Engineering Research PEER http://peer.berkeley.edu/
Center database peer_ground_motion_database
European Strong-Motion Database ESMD http://www.isesd.hi.is/ESD_Local/frameset.htm
European Integrated Data Archive EIDA http://eida.gfz-potsdam.de/webdc3/
Italian Accelerometric Archive ITACA http://itaca.mi.ingv.it/
Real-time INGV strong- motion data ISMD http://ismd.mi.ingv.it/ E
Italian Strong Motion Network – RAN MOT1 http://www.mot1.it/randownload/EN/index.php
Download
Strong-motion database of Turkey http://kyhdata.deprem.gov.tr/2K/kyhdata_v4.php
HEllenic Accelerogram Database HEAD http://www.itsak.gr/en/head
Kyoshin network and Kiban Kyoshin K-NET and http://www.kyoshin.bosai.go.jp/
network Kik-NET
Swiss Seismological Service SED http://www.seismo.ethz.ch/index
French accelerometric network RAP http://www-rap.obs.ujf-grenoble.fr/
The Guerrero Accelerograph Network http://crack.seismo.unr.edu/guerrero/description.
(Mexico) html
GeoNet (New Zealand) http://info.geonet.org.nz/display/appdata/
Applications+and+Data
Euroseistest database (Greece) http://euroseisdb.civil.auth.gr/
Iran Strong Motion Network ISNM http://www.bhrc.ac.ir/Portal/Default.aspx?
tabid=635
USGS National Strong- Motion Project NSMP http://nsmp.wr.usgs.gov/
The COSMOS Virtual Data Center (http:// 104 shallow crustal earthquakes worldwide
www.cosmos-eq.org/VDC/index.html) is a web compiled from over 1,000 stations, extensive
portal to access strong ground motion records. It metadata (including different distance mea-
distributes records from the United States and sures, various site characterizations and earth-
14 contributing countries for use by the engi- quake source data) with corrections to the
neering and scientific communities. It contains information contained in the original databases.
flexible search methods, including map-based, The web interface provides tools for searching,
parameter-entry, and earthquake- and station- selecting, and downloading ground motion data.
based searches. Display and download options
allow users to view data in multiple contexts, European/Pan-European Databases
extract and download data parameters, and The European Strong-Motion Database (http://
download data files in convenient formats. www.isesd.hi.is/ESD_Local/frameset.htm) con-
The Pacific Earthquake Engineering tains more than 3,000 uniformly processed and
Research Center (PEER) database (http://peer. formatted waveforms. Each waveform is associ-
berkeley.edu/peer_ground_motion_database) ated to earthquake and station parameters. The
has been created for engineering applications user can search the database and databank inter-
(e.g., development of ground motion prediction actively and download selected strong-motion
equations in the framework of the Next Genera- records and associated parameters.
tion Attenuation, NGA, program). The latest WebDc (http://eida.gfz-potsdam.de/webdc3/)
version of the PEER Ground Motion Database is a portal to the European Integrated Data
contains 3,182 three-component recordings of Archive (EIDA) infrastructure, which is a
998 Engineering Characterization of Earthquake Ground Motions
distributed data center that aims to securely 677 three-component records associated to
archive seismic waveform data gathered by Euro- 319 earthquakes, obtained from 117 strong-
pean research infrastructures. Through the EIDA motion stations. For the major earthquakes
portal, the user can request real-time data streams after 2000, an additional web page has been
from permanent networks that are mostly created (http://www.itsak.gr/en/db/data/sm/
acquired over the internet and are immediately after2000). Waveforms can be searched spec-
accessible as continuous data streams by ifying event date and time and station code.
SeedLink in real time. Raw seismological data Corrected and uncorrected time series and
(waveform) and station lists (Dataless SEED, response spectra are available from the down-
Inventory XML) can be obtained from different load pages.
data centers spread across Europe. 4. Japan. Japanese records are available at http://
www.kyoshin.bosai.go.jp/. K-NET (Kyoshin
network) is a strong-motion seismograph net-
National Databases work that consists of more than 1,000 obser-
1. Italy. The Italian Accelerometric Archive vation stations distributed roughly every
(ITACA, http://itaca.mi.ingv.it) contains 20 km. K-NET is operated by the National
about 7,500 three-component processed Research Institute for Earth Science and
waveforms generated by about 1,200 earth- Disaster Prevention (NIED) since 1996. At
quakes with magnitudes greater than 3, in the each K-NET station, a seismograph is
time span 1972–2013. Corrected and installed on the ground surface. KiK-net
uncorrected time series and response spectra (Kiban Kyoshin network) is a strong-motion
are available from the download pages, where seismograph network that consists of pairs of
the parameters of interest can be set and spe- seismographs installed in a borehole and on
cific events, stations, waveforms, and their the ground surface deployed at roughly
metadata can be retrieved. Quasi real-time 700 locations. The strong-motion data
Italian strong-motion data recorded by the recorded by K-NET and KiK-net are transmit-
two national accelerometric networks (RSN, ted in real-time to the NIED data management
managed by the National Institute of Geo- center and can be downloaded by users. The
physics and Volcanology, INGV and RAN, soil conditions at K-NET stations and the geo-
operated by Department of Civil Protection, logical and geophysical data derived from
DPC) can be accessed at ISMD (http://ismd. drilling boreholes at KiK-net stations are
mi.ingv.it/) and MOT1 (http://www.mot1.it/ available.
randownload/EN/index.php).
2. Turkey. The strong-motion database of Turkey
(http://kyhdata.deprem.gov.tr/2K/kyhdata_v4. Codes to Select Spectrum-Compatible Strong-
php) contains about 10,000 (processed and Motion Data
unprocessed) waveforms recorded, from 1976 The strong-motion databases are also employed
to the present, by the network operated by the to select records that match a target spectrum and
Prime Ministry for Disaster and Emergency that can represent a response spectrum for a sce-
Management. Corrected and uncorrected time nario event computed from a ground motion pre-
series and response spectra are available from diction equation or which may be derived from
the download pages, where events, stations, probabilistic seismic hazard analysis and
and waveforms can be searched with respect presented as a uniform hazard spectrum (UHS)
to basic parameters. or which may be developed from building code
3. Greece. The Greek strong-motion database provisions. The selected records, conveniently
from 1973 to 1999 is contained in the HEllenic scaled to fit the target spectrum, are then used as
Accelerogram Database, HEAD (http://www. input into numerical simulations for structural
itsak.gr/en/head). The dataset includes analysis and response site study. Several methods
Engineering Characterization of Earthquake Ground Motions 999
Engineering Characterization of Earthquake Ground Motions, Table 3 List of some computer codes to select
spectrum-compatible records
Name Reference Main characteristics
SigmaSpectra Kottke and Rathje (2008) It selects suites of earthquake ground motions from a
http://nees.org/resources/ library of ground motions such that the median of the
sigmaspectrav suite matches a target response spectrum at all defined
periods, then scales the suite such that the standard
deviation agrees with the target standard deviation
The library of ground motions and the spectrum should
be supplied by users
REXEL/REXEL- Iervolino et al. (2009); Smerzini REXEL allows to define the design spectra according
DISP and Paolocci (2013); Smerzini to the Eurocode 8, the new Italian Building Code E
et al. (2013) (NTC08) or user defined. The strong-motion datasets
http://wpage.unina.it/iuniervo/ contained in REXEL come from ESMD, ITACA and
SIMBAD (Selected Input Motions for Displacement-
Based Assessment and Design, Smerzini and Paolucci
2013)
REXEL-DISP allows to select suites of natural
accelerograms compatible with displacement spectra
of NTC08, EC8, or user defined. The records library is
SIMBAD
CMS – Conditional Baker (2011); Jayaram et al. (2011) CSM allows one to compute the expected response
Mean Spectrum http://www.stanford.edu/ spectrum associated with a target spectral acceleration
~bakerjw/research/ (Sa) value at a single period; the records selection is
conditional_spectrum.html based on PEER database; records compatible with
predicted response spectra from NGA model (also as a
conditional spectrum) can be directly selected from the
web-based tool in the PEER database
ASCONA Corigliano et al. (2012) It allows to select records which are compatible both
with acceleration and displacement response spectra
The strong-motion dataset is extracted from NGA
dataset, ITACA, ESMD, and K-net databases, and it
only includes waveforms recorded on rock site
condition
RSPMatch09 Al Atik and Abrahamson (2010) It performs time-domain spectral matching by adding
http://nees.org/resources/ adjustment wavelets to an initial acceleration time
rpsmatch09 series to generate a modified time series whose
response spectrum is compatible with a specified target
response spectrum
The library of ground motions and the spectrum should
be supplied by users
and computer codes are proposed in literature to seismologists since they are immediately usable
this aim, some of them together their characteris- to estimate damage. Furthermore, from
tics are reported in Table 3; other information can accelerometric waveforms, several strong-
be found in the section devoted to this topic. motion parameters can be inferred, such as peak
values, spectral ordinates, measures of the fre-
quency content, and the shaking duration, which
Summary are fundamental elements for seismic design and
evaluation of the seismic response of structural
Accelerometric time series recorded during mod- and geotechnical systems.
erate and strong seismic events are fundamental This chapter presents engineering characteris-
for earthquake engineers and engineering tics of strong ground motion. In section “Analog
1000 Engineering Characterization of Earthquake Ground Motions
and Digital Recording Instruments,” issues ▶ Site Response for Seismic Hazard Assessment
related to the instrumental characteristics of the ▶ Seismic Noise
analog and digital accelerometers are described. ▶ Time History Seismic Analysis
In section “Strong-Motion Data Processing,”
an overview on the processing of strong-motion
data is presented focusing on the evaluation References
of data quality and on the procedures that are
commonly adopted to remove high- and Al Atik L, Abrahamson NA (2010) Nonlinear site
low-frequency noise in the signals. In section response effects on the standard deviations of
predicted ground motions. Bull Seismol Soc Am
“Strong-Motion Intensity Measures,” the main
100(3):1,288–1,292
strong-motion parameters in the time and fre- Ambraseys NN, Smit P, Douglas J, Margaris B,
quency domains and their characteristics and Sigbjörnsson R, Olafsson S, Suhadolc P, Costa
applications are illustrated. These parameters G (2004) Internet site for European strong-motion
data. Boll Geofis Teorica Appl 45(3):113–129
include peak values (acceleration, velocity, and
Baker JW (2011) Conditional mean spectrum: tool
displacement), integral parameters (Arias and for ground motion selection. J Struct Eng 137(3):
Housner intensities) and Fourier and elastic 322–331
response spectra. Finally, in section “Public Bazzurro P, Sjoberg B, Luco N, Silva W, Darragh
R (2005) Effects of strong motion processing proce-
Sources of Strong-Motion Records,” the main
dures on time histories, elastic and inelastic spectra.
worldwide, European and national public In: Proceedings of the workshop on strong-motion
strong-motion databases are listed, together record processing, Richmond, 26–27 May 2004,
with a short description of their content in pp 1–39
Bindi D, Pacor F, Luzi L, Massa M, Ameri G (2009) The
terms of strong-motion data and associated
Mw 6.3, 2009 L’Aquila earthquake: source, path and
information. site effects from spectral analysis of strong motion
data. Geophys J Int 179:1573–1579. doi:10.1111/
j.1365-246X.2009.04392.x
Bolt BA (1973) Duration of strong ground motion. In:
Cross-References Proceedings, 5th world conference on earthquake
engineering, Rome, pp 1304–1313
▶ Conditional Spectra Bommer JJ, Martinez-Pereira A (1999) The effective
▶ Earthquake Location duration of earthquake strong motion. J Earthq Eng
3:127–172
▶ Earthquake Magnitude Estimation
Boore DM (2001) Effect of baseline corrections on dis-
▶ Earthquake Response Spectra and Design placements and response spectra for several recordings
Spectra of the 1999 Chi-Chi, Taiwan, earthquake. Bull Seismol
▶ Engineering Characterization of Earthquake Soc Am 91(5):1199–1211
Boore DM (2003) Analog-to-Digital conversion as a
Ground Motions
source of drifts in displacements derived from digital
▶ European Structural Design Codes: Seismic recordings of ground acceleration. Bull Seismol Soc
Actions Am 93(5):2017–2024
▶ Probabilistic Seismic Hazard Models Boore DM (2005) On pads and filters: processing
strong-motion data. Bull Seismol Soc Am 95(2):
▶ Recording Seismic Signals
745–750
▶ Response-Spectrum-Compatible Ground Boore DM, Akkar S (2003) Effect of causal and acausal
Motion Processes filters on elastic and inelastic response spectra. Earthq
▶ Response Spectrum Analysis of Structures Eng Struct Dyn 32(11):1729–1748. doi:10.1002/
eqe.299
Subjected to Seismic Actions
Boore DM, Bommer JJ (2005) Processing of strong-
▶ Review and Implications of Inputs for Seismic motion accelerograms: needs, options and conse-
Hazard Analysis quences. Soil Dyn Earthq Eng 25:93–115.
▶ Seismic Actions due to Near-Fault Ground doi:10.1016/j.soildyn.2004.10.007
Boore DM, Stephens C, Joyner WB (2002) Comments on
Motion
baseline correction of digital strong-motion data:
▶ Seismic Instrument Response, Correction for examples from the 1999 Hector Mine, California,
▶ Seismic Network and Data Quality earthquake. Bull Seismol Soc Am 92(4):1543–1560
Engineering Characterization of Earthquake Ground Motions 1001
Boore DM, Azarisisi A, Akkar S (2012) Using Jibson RW, Harp EL, Michael JA (1998) A method for
pad-stripped acausally filtered strong-motion data. producing digital probabilistic seismic landslide
Bull Seismol Soc Am 102(2):751–760. doi:10.1785/ hazard maps: an example from the Los Angeles,
0120110222 California, area. US Geological Survey O-F report
Brune JN (1970) Tectonic stress and the spectra of seismic 98–113
shear waves from earthquakes. J Geophys Res Kottke A, Rathje EM (2008) Earthq Spectra 24:911.
75(26):4997–5009 doi:10.1193/1.2985772
Castro RR, Anderson JG, Singh SK (1990) Site response Kwon O, Elnashai AS (2006) The effect of material and
attenuation and source spectra of S waves along the ground motion uncertainty on the seismic vulnerability
Guerrero, Mexico, subduction zone. Bull Seismol Soc curves of a RC structure. Eng Struct 28:289–303
Am 80:1481–1503 Lermo J, Chávez-Garcı́a FJ (1993) Site effect evaluation
Corigliano M, Lai CG, Rota M, Strobbia CL using spectral ratios with only one station. Bull
(2012) ASCONA: automated selection of compatible Seismol Soc Am 83:1574–1594 E
natural accelerograms. Earthq Spectra 28(3):965–987 Pacor F, Paolucci R, Ameri G, Massa M, Puglia R (2011)
Douglas J (2003) What is a poor quality strong-motion Italian strong motion records in ITACA: overview
record? Bull Earthq Eng 1:141–156 and record processing. Bull EarthqEng 9(6):
Douglas J, Boore DM (2011) High-frequency filtering of 1741–1759
strong-motion records. Bull Earth Eng 9(2):395–409. Paolucci R, Rovelli A, Faccioli E, Cauzzi C, Finazzi D,
doi:10.1007/s10518-010-9208-4 Vanini M, Di Alessandro C, Calderoni G (2008) On
Foti S, Paolucci R (2012) Influence of strong motion the reliability of long-period response spectral ordi-
processing on numerical simulations of soil-structure nates from digital accelerograms. Earthq Eng Struct
interaction. In: Second international conference on Dyn 37(5):697–710. doi:10.1002/eqe.781
performance-based design in earthquake geotechnical Paolucci R, Pacor F, Puglia R, Ameri G, Cauzzi C, Massa
engineering, Taormina, 28–30 May 2012 M (2011) Record processing in ITACA, the new Ital-
Graizer VM (1979) Determination of the true ground ian strong-motion database, Chapter 8. In: Akkar S,
displacement by using strong motion records. Gulkan P, Van Eck T (eds) Earthquake data in engi-
Izvestiya Acad Sci USSR, Phys Solid Earth neering seismology. Predictive models, data manage-
15(12):875–885 ment and networks, vol 14, Geotechnical, geological
Graves RW (2004) Processing issues for near source and earthquake engineering series. Springer, Dor-
strong motion recordings. In: COSMOS workshop on drecht, pp 99–113
archiving and web dissemination of geotechnical Sawada T, Hirao K, Yamamoto H, Tsujihara O (1992)
data – 2004, pp 1–9. http://www.cosmos-eq.org/ Relation between maximum amplitude ratio and spec-
events/wkshop_records_processing/index.html tral parameters of earthquake ground motion. In: Pro-
Housner GW (1952) Spectrum intensities of strong- ceedings of 10th world conference on earthquake
motion earthquakes. In: Duke CM, Feigen M (eds) engineering, vol 2. Madrid, Spain, pp 617–22
Proceedings, symposium on earthquake and blast Smerzini C, Paolucci R (2013) SIMBAD: a database with
effects on structures, Los Angeles, pp 21–36 selected input motions for displacement-based assess-
Housner GW (1970) Design spectrum. In: Wiegel RL ment and design – 3rd release. Report of the research
(ed) Earthquake engineering. Prentice-Hall, Engle- project DPC – RELUIS (2010–2013). http://wpage.
wood Cliffs unina.it/iuniervo/SIMBAD_Database_Polimi.pdf
Hudson DE (1979) Reading and interpreting strong Smerzini C, Galasso C, Iervolino I, Paolucci R (2013)
motion accelerograms. Earthquake Engineering Ground motion record selection based on broadband
Research Institute, Berkeley (Monograph) spectral compatibility. Earthquake spectra, in Press,
Husid R (1967) Gravity effects on the earthquake response doi:10.1193/052312EQS197M
of yielding structures. Earthquake Engineering Trifunac MD, Brady AG (1975) A study of the duration of
Research Laboratory, California Institute of Technol- strong earthquake ground motion. Bull Seismol Soc
ogy, Pasadena Am 1975(65):581–626
Iervolino I, Galasso C, Cosenza E (2009) REXEL: com- Trifunac MD, Todorovska M (2001) Evolution of
puter aided record selection for code-based seismic accelerographs, data processing, strong motion arrays
structural analysis. Bull Earthq Eng 8:339–362. and amplitude and spatial resolution in recording
doi:10.1007/s10518-009-9146-1 strong earthquake motion. Soil Dyn Earthq Eng
Iwan WD, Moser MA, Peng C-Y (1985) Some observa- 21:537–555
tions on strong-motion earthquake measurements Tucker B, King J (1984) Dependence of sediment-filled
using a digital accelerograph. Bull Seismol Soc Am valley response on input amplitude and valley proper-
75(5):1225–1246 ties. Bull Seismol Soc Am 74:153–165
Jayaram N, Lin T, Baker JW (2011) A computationally Zhu TJ, Heidebrecht AC, Tso WK (1988) Effect of peak
efficient ground-motion selection algorithm for ground acceleration to velocity ratio on ductility
matching a target response spectrum mean and vari- demand of inelastic systems. Earth Eng Struct Dyn
ance. Earthq Spectra 27(3):797–815 16:63–79
1002 Equivalent Static Analysis of Structures Subjected to Seismic Actions
Equivalent Static Analysis of Structures Subjected to Seismic Actions, Fig. 1 Equivalent lateral shear force
along two orthogonal axes
substantially in excess to the one at the levels seismic weight W and a response coefficient CS.
immediately above or below it. It can be written as follows:
• In-plane discontinuity irregularity – This
occurs when the vertical elements of a struc- V ¼ CS W (1)
ture’s seismic-force-resisting system such as
walls or braced frames do not align vertically where
within a given line of framing or the frame or W is the effective seismic weight. It includes
the wall has a significant setback. the total dead load above the foundation or above
the top of a rigid basement and applicable por-
The classification of a structure as “regular” or tions of other loads (live load and snow). For
“irregular” affects several aspects of the design instance, in areas used for permanent storage
and analysis such as the permission to use the such as library shelves, fixed equipments, etc.,
equivalent lateral force, the behavior factor, and the total floor live load should be applicable. In
the type of models for analysis. areas used for housing or administrative pur-
poses, only a percentage (20–30 %) of the floor
live load should be applicable.
Equivalent Lateral Static Force Procedure
CS is the seismic response coefficient
The equivalent lateral force procedure provides a
expressed as a spectral pseudoacceleration, in
simple way to incorporate the effects of inelastic
g units, which depends on several factors to
dynamic response into a linear static analysis. This
account for the seismic zone, the structure period,
procedure is useful in preliminary design of all
the local soil conditions, the inelastic perfor-
structures and is allowed for final design of the
mance of the lateral resisting system, and the
vast majority of structures. As stated before, the
nature of the structure’s occupancy and use.
procedure is valid only for structures without sig-
It might help to understand this formula to go
nificant discontinuities in mass and stiffness along
back to the basic relationship between the force
the height, where the dominant response to ground
and the acceleration, which is F = Ma.
motions is in the horizontal direction without sig-
nificant torsion (FEMA 750 2007). The equivalent
Seismic Response Coefficient CS
static analysis procedure involves the following
As noted previously, the seismic response coeffi-
steps:
cient CS depends on several factors and can be
written in different manner according to the seis-
• Calculation of the total lateral seismic force V
mic codes. In all cases the controlling parameters
• Vertical distribution of seismic forces along
are the same:
the height of the structure
• Horizontal distribution of the level forces
Cs ¼ f ðSa , I, R, T, T S , T L Þ (2)
across the width and breadth of the structure
• Calculation of the additional forces from the
Sa: design ground acceleration
inherent and accidental torsion
I: importance factor
• Determination of the drift, overturning
R: lateral resisting behavior factor (reduction
moment, and P-Delta effect
or modification factor)
T: structure fundamental period
TS and TL: soil/seismicity depending factors
Calculation of the Total Lateral Seismic (periods)
Force V Before going into the details of each parame-
ter, it is useful to know what data is necessary to
Although the equivalent lateral static force is a obtain the seismic response coefficient. The fol-
code-dependent relationship, the general formula lowing items refer to the information needed to
is expressed as the product of the effective obtain the seismic response coefficient:
Equivalent Static Analysis of Structures Subjected to Seismic Actions 1005
Constant
velocity
E
C
Constant
displacement
TS Period, T TL
– The location where the building will be fundamental period of the structure. Recently, the
constructed to determine the seismic zone design ground motion is defined by an accelera-
(seismic hazard). tion response spectrum characterized by three
– The soil conditions must be known to deter- ranges as shown on Fig. 2.
mine the Site Class (TS and TL), and this AB represents the maximum constant acceler-
requires an investigation that includes boring ation part of the spectrum and lays in the range
and removing soil samples at various depths. 0.0 < T < TS. TS is a function of seismicity and
– The structure fundamental period T, calcu- site (in some codes it depends only on the site
lated or estimated using empirical formulae. conditions). Depending on the seismic code, TS
– The lateral resisting system type, to determine may be as low as 0.15 s for low-hazard regions on
the modification factor R. very hard soil conditions or as high as 0.9 s in
– The structure use, including consideration of high-hazard regions on very soft soil conditions.
the number of people who would be affected BC represents the constant velocity part of the
by the structure failure and the need to use the spectrum in the range TS < T < TL. In this region,
structure for its intended purpose after an the seismic response coefficient is inversely pro-
earthquake. These information are necessary portional to the period, and the pseudovelocity
to determine the importance or occupancy (pseudoacceleration divided by circular fre-
class (I) of the building. quency, o) is constant. TL the long-period transi-
tion ranges from 3 to 4 s in some codes.
Design Ground Acceleration (Sa) CD represents the constant displacement part
The design ground motion is one of the primary of the spectrum for T > TL. In this region CS is
factors used to determine the required seismic inversely proportional to T 2. This part only
resistance (strength) of structures and supported affects tall and flexible structures.
nonstructural components. In many seismic
codes the design ground acceleration is used as Importance Factor
a product of an acceleration factor which depends The importance class or factor depends on the
on the seismic zone and a dynamic amplification occupancy category of the building. Hence,
factor which depends on the soil class and the essential facilities such as hospitals and fire and
1006 Equivalent Static Analysis of Structures Subjected to Seismic Actions
police stations are designed for seismic forces extent to which its nonlinear response is to be
greater than normal. In this way, such structures exploited. In operational terms, such balance
are expected to resist an earthquake and be able to between resistance and energy-dissipation
be functional in the post-earthquake period. capacity is characterized by the values of the
Buildings containing certain highly toxic sub- behavior factor R and the associated ductility
stances must also be designed to meet these classification. As a limiting case, for the design
higher standards. Similarly, high-occupancy of structures classified as non-dissipative, no
buildings that represent a substantial hazard to account is taken of any hysteretic energy dissi-
human life (schools, health care facilities, and pation and the value of the behavior factor may
stadiums) must be designed for seismic forces not be taken, in general, as being greater than the
less than vital category and greater than normal. value of 1.5. For dissipative structures, the value
Finally, for constructions which are not subject to of the behavior factor can be taken greater than
human occupancy with insignificant economic this limiting value in order to account for the
loss can be designed for seismic forces less than hysteretic energy dissipation that mainly occurs
normal. The importance factor is used either as an in specifically designed zones, called dissipative
independent factor or combined with the seismic- zones or critical regions (EN 1998–1 2004). In
ity and site parameters to categorize structures general, seismic codes provide tables listing dif-
according to the seismic risk they could pose ferent lateral resisting systems with associated
(seismic design category concept) (IBC 2012; values of R.
FEMA 750 2007). In certain codes the impor- In terms of calculation, instead of performing
tance factor may increase the seismic forces for an explicit inelastic structural analysis in
vital facilities up to 1.5 times the normal seismic design, the capacity of the structure to dissipate
forces. energy, through mainly ductile behavior of its
elements, is taken into account by performing
Site Class (Ground Conditions) an elastic analysis based on a lateral force
To take into account the site effect on the estima- reduced by a factor R (Fig. 3). Thus, the behav-
tion of the equivalent lateral static force, the ior factor R is an approximation of the ratio of
concept of Site Class is used to categorize com- the seismic forces that the structure would expe-
mon soil conditions into broad classes to which rience if its response was completely elastic to
typical ground motion attenuation and amplifica- the seismic forces that may be used in the
tion effects are assigned. Site Class is determined design, with a conventional elastic analysis
based on the average properties of the soil within model.
a certain depth (30 m) from the ground surface.
Most of the seismic codes specify 4–6 categories Fundamental Period Determination
of soils ranging from hard, competent rock mate- The fundamental period of the structure, T, is
rials to very soft, loose materials. For each cate- used to determine the design ground accelera-
gory are assigned the parameters TS and TL which tion and in some codes to establish the distribu-
define the transition points on the acceleration tion of the shear along the height of the structure.
design spectrum. The fundamental period T can be determined
from a rational analysis using the structural
Behavior (Modification) Factor R properties and deformational characteristics of
The behavior factor or the response modification the resisting elements or, alternatively, using an
factor R, which is determined by the type of approximate empirical relationship. In most
lateral load resisting system used, is a measure seismic codes, the calculated period by analysis
of the system’s ability to accommodate earth- is limited by a coefficient times the approximate
quake loads and absorb energy without collapse. (empirical) period. This period limit prevents
The resistance and energy-dissipation capacity the use of unusually low base shear when using
to be assigned to the structure are related to the the equivalent static lateral force for design of
Equivalent Static Analysis of Structures Subjected to Seismic Actions 1007
δE δP Displacement, δ
buildings (or computational models) that are h times the lumped mass W at that height
overly flexible (FEMA 750 2007). according to Eq. 4. The sum of the loads at each
The most common empirical expression used level equals the total base shear. Also note that
to estimate the fundamental period is given by: while the greatest force is at the top of the build-
ing, the shear increases from zero at the top to its
T ¼ C t hx (3) maximum at the base. Each floor shear is succes-
sively added to the sum from above.
where Ct and x are coefficients which depend on To account for higher modes participation in
the type of lateral resisting system and h is the tall buildings that do not have a uniform triangu-
height of the building, in m, from the foundation lar distribution, some codes require that an addi-
or from the top of a rigid basement. tional force Ft be applied at the top of the
building. The lateral force induced at any level
hi can be determined from the following
Vertical Distribution of the Equivalent equation:
Static Force
W i hi
Once the total base shear is known, it is used to Fi ¼ X
n ðV F t Þ (4)
determine the forces on the various building ele- W j hi
j¼1
ments. For multistory buildings, the forces must
be distributed according to the fundamental mode
which can be simulated by an inverted force where
triangle varying from zero at the base to a maxi- Fi: lateral force applied at level i
mum at the top (Fig. 4). With this simplified V: total lateral static force (Eq. 1)
approach, it is assumed that there is uniform Ft: additional force to be applied at the top of
mass distribution and equal floor heights, but tall buildings to account for higher mode
the codes do provide for variations, where the participation
force at a given level is proportional to the height Wi, Wj: lumped story masses at level i and j
1008 Equivalent Static Analysis of Structures Subjected to Seismic Actions
Equivalent Static Analysis of Structures Subjected to Seismic Actions, Fig. 4 Vertical distribution of the lateral
static force
hi, hj: heights of the masses WiWj above the interpolated from these values for structures
level of application of the seismic action with a fundamental period that falls between
(foundation or top of a rigid basement) these values (FEMA 750 2007).
n: number of lumped masses along the height
of the building (the number of floors in multistory
buildings)
Alternatively, to take account of higher
Horizontal Distribution of the
modes, some codes prescribed the following dis-
Equivalent Static Force
tribution formula:
Within the context of an elastic analysis, the
distribution of lateral forces to various seismic-
W i hki
Fi ¼ V (5) force-resisting elements depends on the type,
X
n
W j hi k geometric arrangement, and vertical extents of
j¼1 the resisting elements and on the shape and flex-
ibility of the floor diaphragms. In a simplified
The superscript k has a value of unity for manner, the forces Fi applied at each level shall
structures with a fundamental period (T) less be distributed to the various elements of the ver-
than or equal to 0.5 s, has a value of 2 for struc- tical lateral force-resisting system in proportion
tures with a fundamental period greater than or to their rigidities, considering that the diaphragm
equal to 2.5 s, and has a value that is linearly is rigid and the lateral force-resisting system is
Equivalent Static Analysis of Structures Subjected to Seismic Actions 1009
Equivalent Static
Analysis of Structures
Subjected to Seismic
Actions,
Fig. 5 Horizontal
distribution of the lateral
force at floor (level) i
composed of several lines parallel to the direction vertical axis) and will be resisted by the lateral
of the seismic force (Fig. 5). The distribution force-resisting elements (lines) in proportion to
expression can be written as follows: their rigidities and their distances from the center
of rigidity (Fig. 6).
K ij The additional force resulting from the tor-
Fij ¼ Fi (6)
X
nk sional moment is given by the following
K ik expression:
k¼1
K ij r j
Fij: force acting on the lateral force-resisting Fyj ¼ Fi e (7)
X
nk
line j at a floor level i K ik r 2k
nk: number of lateral force-resisting elements k¼1
(lines)
Kij, Kik: story stiffness of the lateral force- Fyj: force resulting from the torsional
resisting element (line) k and j at level i moment acting on the lateral force-resisting
Fi: seismic force at floor (level) i element (line) j
The design shall include the torsional moment, Fi: seismic force at floor (level) i
resulting from the eccentricities of the location of e: total eccentricity
the masses and the centers of rigidities. In addi- KijKik: story stiffness of the lateral force-
tion to the torsional moment, the design also resisting element (line) k and j at level i
should include accidental torsional moments. rkrj: distance from the axes of the force-
For rigid diagrams, the mass at each level is resisting element (line) k and j to the rigidity
assumed to be displaced from the calculated cen- center Cr
ter of mass in each direction a distance equal to nk: number of lateral force-resisting elements
5 % or 10 % of the building dimension at that (lines)
level perpendicular to the direction of the force It should be noted that in seismic design when
under consideration. The torsional moment cre- the force on a vertical element caused by the
ates torsional displacement (twisting about the eccentricity acts in the same direction as that
1010 Equivalent Static Analysis of Structures Subjected to Seismic Actions
ithFloor (level) rj Fi
Cm
Fθj
Fij Cr
ea ec
e
Fi
Equivalent Static Analysis of Structures Subjected to Seismic Actions, Fig. 6 Torsional moment effect due to
eccentric lateral force for rigid diaphragm
caused by the lateral load directly (lateral force- earthquake shaking does not exceed acceptable
resisting elements located on the flexible side), levels. The lateral deflection is expressed in
they must be added. However, when the torsional terms of the inter-story drift which is a measure
force acts in the opposite direction (rigid side), it of how much one floor or roof level displaces
will not be subtracted. under the lateral force relative to the floor level
immediately below. It is typically presented as
a ratio of the difference in deflection between
Minimum Base Shear Limit two adjacent floors divided by the height of the
story that separates the floors. The determina-
In many seismic codes, the equivalent static lat- tion of the story drifts has to be based on the
eral force V is used as a reference value of the application of the design seismic forces to a
total seismic design base shear. For instance, the mathematical model of the physical structure.
base shear calculated using the modal response The model shall include the stiffness and
spectrum analysis should not be less than 80 % or strength of all elements that are significant to
90 % of the equivalent lateral base shear. This the distribution of forces and deformations in
minimum base shear is provided because the the structure.
computed period of vibration may be the result Since the forces used for design include the
of an overly flexible (incorrect) analytical model. response modification coefficient, R, the
resulting displacements are too small and must
be amplified. Most of the seismic codes account
Drift Story for this difference, by multiplying the elastic
displacements by the modification coefficient
Seismic design of structures must ensure that the R or a deflection amplification factor which
anticipated lateral deflection in response to is typically similar to R, but a little less than,
Equivalent Static Analysis of Structures Subjected to Seismic Actions 1011
the R coefficient. Therefore, the deflection at story drift determined using Eq. 9, Vi is the sum of
a level I is given by the lateral seismic design forces above the story,
and hi is the story height. If the calculated value
dI ¼ R:15emdei (8) of y at each story is less than or equal to 0.1, the
structure is considered to have adequate stiffness
The inter-story deflection becomes and strength to provide stability. If the value of y
exceeds 0.1, the lateral force analysis must
DI ¼ di di1 (9) include explicit consideration of P-Delta effects.
These effects are an amplification of forces that
The drift is also an important consideration for occurs in structures when they undergo large
E
structures constructed in close proximity to one lateral deflection. A limiting value for y (ymax)
another. In response to strong ground shaking, varying between 0.2 and 0.3 is generally pre-
adjacent structures can hit one another, an effect scribed. If the structure exceeds this limiting
known as pounding. Pounding can induce very value, it is considered potentially unstable and
high forces in a structure at the area of impact and must be redesigned unless nonlinear response
has been known to be the cause of collapse of some history analysis is used to demonstrate that the
structures. Therefore, the seismic codes require that structure is adequate.
structures be set far enough away from one another
and from property lines to avoid pounding.
Summary
Overturning Moment and P-Delta The equivalent static lateral force method is a
Effects simplified technique to substitute the effect of
dynamic loading of an expected earthquake by a
Because the inertial force created by an earth- static force distributed laterally on a structure for
quake represented by the equivalent lateral force design purposes. The total applied seismic force
acts through the center of mass of a building, V is generally evaluated in two horizontal direc-
there is a tendency for the moment created by tions parallel to the main axes of the building.
this force acting above the base to overturn the The equivalent static analysis procedure
structure. This overturning force must be involves the following steps:
counteracted by stabilizing load. Normally, the
dead weight of the building, also acting through • Calculation of the total lateral seismic force
the center of mass, is sufficient to resist the V which depends on the seismic zone (design
overturning force, but it must always be checked. ground acceleration), the building importance,
P-Delta or second-order effect is an instability the soil conditions, the fundamental period of
phenomenon that results from the simultaneous the structure, the lateral force-resisting sys-
actions of the lateral force and dead load. Seismic tem, and the effective seismic weight (dead
provisions require an evaluation of the structure and part or entire of live load)
stability under the anticipated lateral deflection • Determination of the vertical distribution of
by calculating a “stability coefficient” y for each seismic forces along the height of the struc-
story. The value of y is given by the following ture, where the force at a given level is pro-
expression: portional to the height h times the lumped
mass m at that height (simplified method)
Pi Di • Considering a horizontal distribution of the
y¼ (10)
V i hi level forces to the various elements of the
vertical lateral force-resisting system in pro-
In this formula, Pi is the weight of the structure portion to their rigidities, considering that the
above the story being evaluated, Di is the design diaphragm is rigid
1012 Estimation of Potential Seismic Damage in Urban Areas
Estimation of Potential Seismic Damage in Urban Areas, Table 1 A typical damage probability matrix
Damage state Central damage ratio
(DS) (%) MMI = V MMI = VI MMI = VII MMI = VIII MMI = IX
None 0 Damage state probabilities, Pr(DS,I)
Light 5
Moderate 30
Heavy 70
Collapse 100
E
express the same information in terms of dis- Method
crete damage probabilities for a particular inten-
sity level. In this section, empirical and analytical methods
Empirical method is based on the fundamental for estimating damage state probabilities will be
idea that similar type of structures experiences presented. Before that, damage probability matri-
comparable damage rates under earthquakes. It ces are explained briefly.
involves analysis of empirical data collected in
post-earthquake surveys. In the analytical
method, in order to estimate the limit states of Damage Probability Matrices
the structure, the seismic performance under a Damage probability matrices (DPMs) were first
given ground motion level is assessed through introduced by Whitman (1973) for multistory
detailed time history analysis or other simplified buildings in the aftermath of the 1971 San
methods. Subjective method includes expert Fernando earthquake. An element in this matrix,
opinions for the assessment of the damage prob- Pr(DS,I), gives the probability of occurrence of a
abilities under various levels of shaking intensi- damage state (DS) under an earthquake intensity,
ties. Experts are requested to give their opinions I. The sum of any column in this matrix is equal to
on the level of the damage based on their relevant unity. Table 1 shows an example damage proba-
experience on seismic damage assessment. It is bility matrix.
also possible to generate fragility curves based on Damage state is the verbal or quantitative
the experimental data that has been obtained representation of the damage that would result
through a series of tests. Finally, hybrid method in a structure under an earthquake of specific
involves a combination of the previously intensity. Verbal definitions are generally used
described approaches for assessment of damage in qualitative assessment of damage.
probabilities. A quantitative expression is used to represent
An important problem is the comparison of the damage if the original cost of the structure
damage probabilities obtained with these alter- is known. The ratio of the cost of repair to the
native methods. In this study, empirical and cost of replacement of the structure is defined as
analytical approaches are compared in terms of damage ratio (DR). As the name implies, this
damage data from a recent earthquake that ratio takes values that range from 0 % to
occurred in Turkey. For this purpose, first, 100 %. However, representative damage ratios
empirical damage probabilities are obtained are required in order to associate with the dam-
from the damage database of the 17 August age states and make the calculations easier.
1999 Kocaeli earthquake. Then, the results are Therefore, the term central damage ratio (CDR)
compared with the damage rates from the ana- is defined for a representative value of each
lytical method. Results of this study and similar damage ratio. The CDR values in Table 1 are
studies are important in terms of accurate esti- taken from G€urpınar et al. (1978).
mation of seismic risks in earthquake-prone Using a single damage ratio is preferred to
regions in the world. summarize a full DPM. This ratio is called
1014 Estimation of Potential Seismic Damage in Urban Areas
Estimation of Potential Seismic Damage in Urban Areas, Fig. 1 Observed intensity map of the 17 August 1999
Kocaeli (Mw = 7.4) earthquake (Adapted from Özmen 1999)
mean damage ratio (MDR) and is obtained by state probabilities given in Table 1 are calcu-
the following equation: lated as follows:
X
MDR ðI Þ ¼ PrðDS, I Þ CDRðDSÞ (1) N ðDS, I Þ
DS PrðDS, I Þ ¼ (2)
N ðI Þ
where MDR(I) is the mean damage ratio
corresponding to the intensity level I, Pr(DS,I) is where N (DS,I) is the number of buildings that are
the damage state probability of the building in damage state DS due to an earthquake with
type of interest subjected to the earthquake of intensity I and N(I) is the total number of build-
intensity I, and CDR(DS) is the central damage ings subjected to the same event.
ratio corresponding to damage state DS. In this study, regional empirical DPMs are
In most of the previous studies, Modified formed using the detailed damage database of
Mercalli Intensity (MMI) is used as the seismic 17 August 1999 Kocaeli (Mw = 7.4) earthquake.
intensity scale in DPMs. Similarly, in this study This destructive earthquake caused significant
MMI scale is employed to quantify the intensity structural damage resulting in almost 20,000
of the ground shaking. casualties and considerable economic losses.
While computing empirical damage probabili-
ties, initially the isoseismal (intensity) map of
Empirical Method for Damage Probability the earthquake is investigated (Fig. 1). Isoseismal
Matrices maps demonstrate spatial distribution of seismic
In the empirical approach, damage state proba- intensities in the form of contour lines of equal
bilities are obtained with simple frequency ana- intensity values. These maps are mostly gener-
lyses of different damage states in the damage ated in the field in the aftermath of an earthquake.
database of interest. In other words, the damage Based on the fact that different intensities are
Estimation of Potential Seismic Damage in Urban Areas 1015
Estimation of Potential Seismic Damage in Urban Estimation of Potential Seismic Damage in Urban
Areas, Table 2 Empirical damage probability matrix Areas, Table 3 Empirical damage probability matrix
for Kocaeli city center (MMI = IX) for Sakarya city center (MMI = X)
Reinforced Reinforced
CDR Masonry concrete CDR Masonry concrete
Damage state (%) buildings buildings Damage state (%) buildings buildings
No damage 0 0.70 0.16 No damage 0 0.80 0.12
Light damage 5 0.12 0.30 Light damage 5 0.08 0.22
Moderate 30 0.08 0.29 Moderate 30 0.04 0.25
damage damage
Severe 85 0.11 0.25 Severe 85 0.07 0.41
damage and damage and E
collapse collapse
MDR (%) 11.91 31.75 MDR (%) 7.51 43.28
Total number of flats 14.086 25.351 Total number of flats 78.444 22.463
Estimation of Potential Seismic Damage in Urban Areas, Fig. 2 Fragility curves of mid-rise reinforced concrete
buildings (Erberik 2008b)
Analytical Method for Damage Probability past major earthquakes. The reinforced concrete
Matrices buildings are classified as “low rise” (1–3
The first step of the analytical method involves stories) and “mid rise” (4–9 stories). These
construction of fragility curves based on the buildings are also classified with respect to
structural analyses of the building group of their compliance to current seismic codes and
interest. Then, the continuous information on their performance levels as “high,” “moderate,”
fragility curves is converted to discrete damage and “low.” Thus, there are six subclasses for the
state probabilities in DPMs. There are different buildings. Typical fragility curves for the
analysis approaches for deriving fragility reinforced concrete buildings are shown in
curves: Static methods are simpler but yield Fig. 2. The MR abbreviation in the figure indicates
only approximate results. Dynamic analyses “mid-rise” buildings while the letters A, B, and
are mostly more complex but they generate C represent the high, moderate, and low seismic
more accurate results in terms of structural performance levels. The masonry buildings, on the
response. In addition, nonlinear analyses should other hand, are classified as “urban engineered,”
also be used in fragility analyses due to the fact “urban non-engineered,” and “rural.” The
that damage is directly related to nonlinear corresponding abbreviations, respectively, are
behavior. Currently, two common analytical “UE,” “UN,” and “RN.” In addition to these, num-
methods for generating fragility curves are the ber of stories is another structural parameter used
nonlinear pushover method and nonlinear in the classification of masonry buildings. Figure 3
dynamic methods. In addition, it is possible to displays an example fragility curve for two-story
perform analyses on idealized models of build- urban-engineered and urban-non-engineered
ings as single- or multi-degree-of-freedom masonry buildings (UE2, UN2). For harmoniza-
structures. tion purposes, two limit states are selected: Lower
In this study, in order to compare the empir- limit state (LS1) is between no damage and mod-
ical and analytical DPMs, previously available erate damage states, while the upper limit state
fragility curves for masonry and reinforced con- (LS2) is between moderate damage and heavy
crete buildings are used (Erberik 2008a, b). damage/collapse states.
These curves are derived using the analytical The information on fragility curves can be
method described previously, and they present easily converted to DPMs. In the example
the local characteristics of existing reinforced shown in Fig. 4, the fragility curves indicate the
concrete and masonry buildings in Turkey. exceedance probability of two different limit
They are also validated against real data from states (LS1, LS2) in terms of a selected ground
Estimation of Potential Seismic Damage in Urban Areas 1017
Estimation of Potential Seismic Damage in Urban Areas, Fig. 3 Fragility curves of 2-story urban masonry
buildings (Erberik 2008a)
Estimation of Potential Seismic Damage in Urban Areas, Fig. 4 Transformation of the information in a fragility
curve into a damage probability matrix
motion intensity parameter (peak ground accel- procedure is repeated for other intensity levels,
eration, PGA in this case). The regions surround- it is possible to form a full DPM.
ing the fragility curves correspond to three
different damage states (DS1, DS2, and DS3).
To obtain the discrete probability of a damage Comparison of Results from Alternative
state given an intensity level (PGA = 0.6 g in Methods
this case), a vertical line is drawn which inter-
sects the fragility curves as shown in Fig. 4. This As discussed previously, it is possible to obtain
vertical line yields the probability of the build- damage state probabilities with alternative
ing type of interest to experience different dam- approaches. In this study, in order to make a
age states under the corresponding PGA. In the comparison with the empirical results, the analyt-
example shown in Fig. 4, when PGA = 0.6 g, the ical fragility curves complying with the charac-
probability of no damage (or light damage) is teristics of the existing building stock in Kocaeli,
7 %, the probability of moderate damage is 57 %, Sakarya, and Bolu are employed. Considering the
and the probability of severe damage local and structural-deficient characteristics of
(or collapse) is 36 %. This information consti- buildings in these city centers, fragility curves
tutes one column of the DPM. If the same for “non-engineered” buildings are used.
1018 Estimation of Potential Seismic Damage in Urban Areas
Estimation of Potential Seismic Damage in Urban Estimation of Potential Seismic Damage in Urban
Areas, Table 5 Comparison of damage probabilities Areas, Table 7 Comparison of damage probabilities
for masonry buildings in Kocaeli (MMI = IX) for masonry buildings in Sakarya (MMI = X)
CDR Analytical Empirical CDR Analytical Empirical
Damage state (%) approach approach Damage state (%) approach approach
No damage + 2.5 0.90 0.81 No damage + 2.5 0.75 0.89
light damage light damage
Moderate 30 0.08 0.08 Moderate 30 0.16 0.04
damage damage
Severe damage 85 0.03 0.11 Severe damage 85 0.09 0.07
+ collapse + collapse
MDR (%) 6.89 13.78 14.60 9.38
MDR (%)
Estimation of Potential Seismic Damage in Urban Estimation of Potential Seismic Damage in Urban
Areas, Table 6 Comparison of damage probabilities Areas, Table 8 Comparison of damage probabilities
for reinforced concrete buildings in Kocaeli (MMI = IX) for reinforced concrete buildings in Sakarya (MMI = X)
CDR Analytical Empirical CDR Analytical Empirical
Damage state (%) approach approach Damage state (%) approach approach
No damage + 2.5 0.15 0.46 No damage + 2.5 0.03 0.34
light damage light damage
Moderate 30 0.85 0.29 Moderate 30 0.58 0.25
damage damage
Severe damage 85 0.00 0.25 Severe damage 85 0.39 0.41
+ collapse + collapse
MDR (%) 25.9 31.1 MDR (%) 50.8 43.2
Estimation of Potential Seismic Damage in Urban probabilities obtained from analytical fragility
Areas, Table 9 Comparison of damage probabilities curves. The damage data is collected in three
for masonry buildings in Bolu (MMI = VII)
city centers, which are all located in the most
CDR Analytical Empirical active seismic zone of Turkey, after the 17 August
Damage state (%) approach approach
1999 Kocaeli earthquake.
No damage + 2.5 0.91 0.92
light damage As a result of the comparisons in these city
Moderate 30 0.06 0.03 centers, in some cases very close results are
damage obtained despite some noticeable differences in
Severe damage 85 0.02 0.05 other cases. In particular, the use of equivalent
+ collapse simplified models of reinforced concrete build-
MDR (%) 6.15 7.45 E
ings resulted in deviations from the empirical
results. Such differences are expected as both
approaches involve inherent uncertainties due to
Estimation of Potential Seismic Damage in Urban
Areas, Table 10 Comparison of damage probabilities assumptions and simplifications made. However,
for reinforced concrete buildings in Bolu (MMI = VII) in all cases, for all building types, the MDR
CDR Analytical Empirical values are found to be in the same order of mag-
Damage state (%) approach approach nitude. This observation is encouraging to use
No damage + 2.5 0.46 0.36 both approaches in the future.
light damage According to the results of this case study, it is
Moderate 30 0.53 0.36 suggested that alternative methods are employed
damage
for estimating potential seismic damage in urban
Severe damage 85 0.01 0.27
+ collapse
regions. Hybrid damage models are also
MDR (%) 17.9 34.65 recommended as they yield more accurate
results. In the long run, results of this study and
similar studies can be used in seismic loss esti-
structural analysis approach taken or the ideali- mation models, disaster management, emergency
zations of the exact buildings. For instance, the planning, as well as rapid response purposes in
fragility curves for masonry buildings employed urban regions.
in this study are derived using three-dimensional
building models; however, the curves for
reinforced concrete buildings are based on equiv- Cross-References
alent single-degree-of-freedom models. Due to
the equivalent idealized models, the analyses ▶ Damage to Buildings: Modeling
are simplified but on the other hand some accu- ▶ Empirical Fragility
racy is lost. This point is believed to be the main ▶ Seismic Loss Assessment
cause of difference between empirical and ana-
lytical results for the reinforced concrete build-
ings. This trade-off between cost of computing References
and cost of modeling real buildings more accu-
rately is always a choice that a researcher Erberik MA (2008a) Generation of fragility curves for
must make. Turkish masonry buildings considering in-plane fail-
ure modes. Earthq Eng Struct Dyn 37:387–405
Erberik MA (2008b) Fragility-based assessment of typical
mid-rise and low-rise RC buildings in Turkey. Eng
Summary Struct 30:1360–1374
G€
urpınar A, Abalı M, Y€ucemen MS, Yeşilçay Y (1978)
Compliance with compulsory earthquake insurance.
In this study, observed damage state probabilities
Earthquake Engineering Research Institute Report,
of masonry and reinforced concrete buildings in 78–05. Middle East Technical University, Ankara
Turkey are compared with the damage state (in Turkish)
1020 European Structural Design Codes: Seismic Actions
Kafalı C, Grigouri M (2004) Seismic fragility analysis. In: European Structural Design Codes: Seismic Actions,
Proceedings of the 9th ASCE specialty conference on Table 1 Eurocode 8 parts
probabilistic mechanics and structural reliability,
Part EN Title
Rome, 19–23 June 2005
Özmen B (1999) Iso-intensity map of İzmit earthquake. 1 EN 1998- Design of structures for earthquake
Earthquake Research Department, General Directorate 1:2004 resistance – general rules, seismic
of Disaster Affairs, Ankara actions, rules for buildings
Whitman RV (1973) Damage probability matrices for 2 EN 1998- Design of structures for earthquake
prototype buildings, vol 380, Structures Publication. 2:2005 resistance – bridges
MIT, Boston 3 EN 1998- Design of structures for earthquake
3:2005 resistance – assessment and
retrofitting of buildings
4 EN 1998- Design of structures for earthquake
4:2006 resistance – silos, tanks, pipelines
European Structural Design Codes:
5 EN 1998- Design of structures for earthquake
Seismic Actions 5:2004 resistance – foundations, retaining
structures, geotechnical aspects
Michael N. Fardis 6 EN 1998- Design of structures for earthquake
Department of Civil Engineering, School of 6:2005 resistance – towers, masts,
Engineering, University of Patras, Patras, Greece chimneys
Synonyms
loadings, materials, or geotechnical design, all
Design response spectra; Design seismic actions; aspects of seismic design are covered by
Eurocode for seismic design, Seismic actions, EN-Eurocode 8: “EN1998: Design of Structures
Seismic response spectra for Earthquake Resistance.” This is for the con-
venience of countries with very low seismicity,
as it gives them the option not to apply
Introduction EN-Eurocode 8 at all.
EN-Eurocode 8 has six Parts, listed in Table 1.
Between 1992 and 1998, close to 60 Eurocodes The EN-Eurocodes are the recommended
were published by the European Committee for European structural design codes for civil engi-
Standardization (CEN) as European neering works and their parts (including
pre-standards for structural design (ENVs). prefabricated structural components), in order to
Their conversion to the first generation of Euro- facilitate the integration of the construction mar-
pean Standards (EN) for structural design started ket in the European Union. The European Com-
at the end of the 1990s and was concluded after mission has published a “Recommendation on
the mid-2000s. The conversion entailed the implementation and use of Eurocodes for
a thorough revision of the ENVs on the basis of construction works and structural construction
recent scientific/technical developments and of products” (European Commission 2003),
comments from National Standards Bodies according to which Member States should adopt
which are members of CEN. the Eurocodes as a suitable tool for the structural
Among the ten EN-Eurocodes, two cover the design of construction works and refer to them in
basis of structural design and the loadings their national provisions on structural construc-
(“actions”), one deals with geotechnical and tion products. More specifically, the Eurocodes
foundation design, while five cover aspects spe- should be used as the basis for the technical
cific to concrete, steel, composite (steel- specifications in the contracts for works and
concrete), timber, masonry, or aluminum con- related engineering services in the public sector,
struction. Instead of distributing the seismic as well as in the water, energy, transport, or
design aspects to the EN-Eurocodes on telecommunications sectors.
European Structural Design Codes: Seismic Actions 1021
According to (European Commission 2003), it EU. Member States should not diverge from the
is up to a Member State to select the level of recommended choice, unless geographical, cli-
safety and protection – which, according to matic, and geological differences or different
(European Commission 2001), may include ser- levels of protection make it necessary.
viceability and durability – offered by civil engi- A National Annex may also contain country-
neering works on its national territory. To allow specific data that are also considered by the
Member States to exercise this authority and EN-Eurocode as NDPs (e.g., a seismic zoning
accommodate geographical, climatic, and geo- map, spectral shapes for the various types of soil
logical (including seismotectonic) differences profiles provided in Eurocode 8, etc.). A decision
without sacrificing harmonization of structural to adopt nationally or not an informative Annex
E
design codes at the European level, the notion of the EN-Eurocode is also stated there.
of “Nationally Determined Parameters” (NDPs) If the National Annex does not exercise
has been adopted as the means to provide the national choice for some NDPs, the choice will
necessary flexibility in the application of be the responsibility of the designer, taking into
Eurocodes across (and outside) Europe. account the conditions of the project and other
According to this approach, the EN-Eurocodes national provisions (European Commission 2001).
allow national choice of all key parameters or A National Annex may provide supplemen-
aspects that control the safety, durability, service- tary information, noncontradictory to any of the
ability, and economy of civil engineering works rules of the EN-Eurocode. This may include ref-
designed and constructed to the EN-Eurocodes. erences to other national documents, to assist the
The “Nationally Determined Parameters” in user in the application of the EN (European Com-
EN-Eurocodes include: mission 2001). What a National Annex cannot do
is modify any EN-Eurocode provisions or replace
– Symbols (e.g., safety factors, the mean return them with other – e.g., national – rules. Although
period of the design seismic action, etc.) not encouraged, such deviations from the
– Technical classes (e.g., ductility classes) EN-Eurocode are allowed in national regulations
– Procedures or methods (e.g., alternative other than the National Annex. However, when
models of calculation) use is made of national regulations allowing devi-
ation from certain Eurocode rules, the design will
Alternative classes and procedures/methods not be called “design according to
considered as NDPs are identified and fully EN-Eurocodes,” as by definition this term
described in the normative text of the means compliance with all EN-Eurocode provi-
EN-Eurocode. For NDP-symbols, the sions, including the national choices for the
EN-Eurocode may give a range of acceptable NDPs (European Commission 2001).
values and normally recommends in If an EN-Eurocode is not considered relevant
a nonnormative note a value for the symbol. It to a particular Member State, then a National
normally also recommends a class or a procedure/ Annex is not required. This may be the case for
method among the alternatives identified and EN-Eurocode 8 in countries of very low seismic-
described in the EN-Eurocode text as NDPs ity (European Commission 2001).
(European Commission 2001). CEN Member States have been given few years
Each Member State publishes its chosen after the publication of an EN-Eurocode by CEN
values for the NDPs in a National Annex which in English, French, and German to publish the
accompanies the national version of the corresponding National Annex (including the
EN-Eurocode. According to European Commis- national choice for the NDPs). During that period,
sion (2003), Member States should adopt for the Member States have calibrated the NDPs so that,
NDPs the choices recommended in the notes of for the target safety level, structures designed
the EN-Eurocode, in order to achieve the maxi- according to the National version of the
mum feasible harmonization across the EN-Eurocodes will not cost significantly more
1022 European Structural Design Codes: Seismic Actions
than those designed according to current National of the US codes. Nonetheless, collapse may be
Standards (European Commission 2001). imminent after such an earthquake (e.g., due to
CEN Member States have withdrawn those strong aftershocks) and repair may be unfeasible
National Standards which conflict in some way or economically prohibitive. Tolerance to
with any EN-Eurocode, so that the EN-Eurocodes extremely strong seismic actions is pursued
become the exclusive structural design standards through systematic and across-the-board applica-
in the European Union. tion of the capacity design concept, which allows
The revision of the EN-Eurocodes toward control of the inelastic response mechanism.
their second generation is expected to be com- The two explicit performance levels – (local-)
plete by 2020, after starting in 2014. collapse prevention and damage limitation – are
pursued under two different seismic actions. The
seismic action under which (local-)collapse
Performance Requirements in should be prevented is termed “design seismic
EN-Eurocode 8 and Corresponding action.” Given that issues of safety and economy
Seismic Hazard Levels are within the competence of individual Member
States, the hazard levels for the seismic
Seismic Design of New Buildings actions associated with these two performance
Part 1 of EN-Eurocode 8 (CEN 2004a) provides levels – (local-)collapse prevention and damage
for two-tier seismic design of new buildings, with limitation – are left open for national determina-
the following explicit performance objective tion as NDPs. For buildings of ordinary impor-
(Fardis et al. 2005): tance, the recommendation in Part 1 of
EN-Eurocode 8 is:
– Under a rare seismic action: to protect life, by
preventing collapse of the structure or its parts – A “design seismic action” (for local-collapse
and by maintaining structural integrity and prevention), with recommended 10 % proba-
residual load capacity bility of been exceeded in 50 years
– Under a frequent event: to reduce property (recommended mean return period: 475 years)
loss by limiting structural and nonstructural – A seismic action for damage limitation, with
damage or – for certain types of structures, recommended 10 % probability of been
such as tanks, pipelines or chimneys – to exceeded (recommended mean return period:
ensure integrity or a minimum operating level 95 years).
– “Significant damage” (SD), corresponding to “passive” ones, while in “passive” programs trig-
“life safety” in the USA and to the (local-) gered by remodelling the building, performance
collapse prevention performance level to requirements for the assessment and the
which new buildings are designed according retrofitting may gradate with the extent and cost
to Part 1 of EN-Eurocode 8 (CEN 2004a). The of the remodelling work undertaken. With this
structure is significantly damaged, may have guidance, National Authorities will hopefully set
moderate permanent drifts, but retains some the performance requirements for existing build-
residual lateral strength and stiffness and its ings so that, in the long run, the percentage of the
full vertical load-bearing capacity. Repair existing building stock to be retrofitted is optimal
may be uneconomic. The verifications should for the society and the national economy as
E
provide a margin against member ultimate a whole.
capacities.
– “Damage limitation” (DL), corresponding to Seismic Design of New Bridges
“immediate occupancy” in the USA. The Part 2 of EN-Eurocode 8 (CEN 2005b)
structure has no permanent drifts; its elements requires single-tier seismic design of new bridges
have no permanent deformations, retain fully with the following explicit performance
their strength and stiffness and do not need objective:
repair. Members should be verified to remain
elastic. • The bridge must retain its structural integrity
and have sufficient residual resistance to be
The “seismic hazard” levels for which the three used for emergency traffic without any repair
limit states are required are meant to be decided after a rare seismic event: the “design seismic
nationally as NDPs, or by the owner, if not fixed in action”; any damage due to this event must be
the National Annex. Part 3 of EN-Eurocode 8 itself easily repairable.
gives no recommendation, but mentions that the
performance objective suitable to ordinary new Although called “non-collapse requirement,”
buildings is a 225-year earthquake (20 % proba- in reality this corresponds to life safety, rather
bility of been exceeded in 50 years), a 475-year than to near collapse, because sufficient residual
event (10 % probability in 50 years), or a 2,475- resistance has to be available after the design
year one (2 % in 50 years), for the DL, the SD, or seismic event for immediate use by emergency
the NC “limit state,” respectively. National traffic (Kolias et al. 2012).
Authorities also decide whether all three limit Part 2 of EN-Eurocode 8 (CEN 2005b) calls
states need to be verified or whether checking also for limitation of damage under a loosely
one or two limit states at the corresponding seismic defined seismic action with high probability of
hazard level suffices. Part 3 of EN-Eurocode been exceeded in the bridge’s lifetime; such dam-
8 guides National Authorities to choose the set of age must be minor and limited only to secondary
limit states to be checked, as well as the return components and to the parts of the bridge intended
period of the corresponding seismic actions, for controlled damage under the “design seismic
depending on whether assessment and retrofitting action.” However, this requirement is of no prac-
takes place as part of an “active” program tical consequence for design: it is presumed to be
(in which owners of certain categories of buildings implicitly met, if all the criteria for compliance
are required to meet specific deadlines for comple- with the “non-collapse requirement” are checked
tion of the assessment and – depending on its and met under the “design seismic action” (Kolias
outcome – of the retrofitting) or a “passive” one et al. 2012). This should be contrasted with new
(that associates seismic assessment – possibly buildings, for which Part 1 of EN-Eurocode
leading to retrofitting – with discontinued use of 8 (CEN 2004a) provides explicit checks under
the building). In “active” programs the relevant a well-defined “damage limitation” seismic action.
requirements may be less stringent than in However, the damage checks (inter-story drifts)
1024 European Structural Design Codes: Seismic Actions
normally refer to nonstructural elements, which damping ratio up to 10 % and for natural period
are not present in bridges. T between 0.2 and 1.0 s, the pseudo-velocity
spectrum approximates well the actual relative
velocity spectrum.
Elastic Response Spectra The EN-Eurocode 8 spectra include ranges of:
Pseudo-values do not correspond to the real (see below; for structures of ordinary impor-
peak spectral velocity or acceleration. For tance, gI = 1.0).
European Structural Design Codes: Seismic Actions 1025
inverse of the annual rate, la(ag), at which this years, which is different from the reference prob-
acceleration level is exceeded: ability, PLR, in the same TLR years, agR should be
multiplied by
T R ag ¼ 1=la ag (5)
gI ¼ ðPLR =PL Þ1=k (8b)
A functional form commonly used for la(ag) is
Ground Acceleration of Seismic Actions Other
k
la ag ¼ K o ag (6) Than the “Design Seismic Action”
Design for damage limitation under the
If the exponent k (: slope of the “hazard curve” corresponding seismic action is explicitly
la(ag) in a log-log plot) is about constant, two required for buildings (as well as for tanks, pipe-
peak ground acceleration levels ag1, ag2, lines, and chimneys, but not for bridges). The
corresponding to two different mean return ratio, n, of the PGA of the seismic action for
periods, TR(ag1), TR(ag2), are related as damage limitation to the “design seismic action”
(for local-collapse prevention) reflects the differ-
!1=k ence in hazard levels. It is an NDP, with
ag1 T R ag1
¼ (7) recommended values for buildings, silos, tanks,
ag2 T R ag2 and pipelines equal to 0.5 for Importance Classes
I and II or 0.4 for III and IV. This gives a ground
The value of k characterizes the seismicity of acceleration, (ngI)agR, equal to 40 %, 50 %, 48 %,
the site. In regions where the difference in peak and 56 % of agR, for Importance Classes I, II, III,
ground acceleration of frequent and very rare and IV, respectively; in other words, the property
seismic excitations is very large, k is low protection is about the same for ordinary, large
(around 2); in such regions, full performance- occupancy, or essential facilities, although the
based design or assessment at several perfor- protection they offer against collapse may be
mance levels with widely different hazard levels very different.
is very meaningful. Large values of k (above 4) The approach highlighted at the end of the last
are typical of regions where high ground acceler- section for the estimation of the importance factor,
ation levels are almost as frequent as smaller gI, on the basis of the Poisson assumption of earth-
ones; one performance level (normally the one quake occurrence may be modified to determine
associated with the lowest among the hazard the ratio, n, of the ground acceleration of the dam-
levels) will almost always govern there; age limitation seismic action, nag, to the “design
performance-based design or assessment at the ground acceleration,” ag. If the seismic action for
other levels may be redundant. damage limitation has a probability of been
The value of the importance factor gI may be exceeded in TDL years which is the same as that
determined as follows: let’s suppose that, for of exceeding the “design seismic action” in
“Importance Class II” (ordinary), the “reference a design life of TL years, then ag should be multi-
peak ground acceleration” on rock, agR, has plied by
a probability, PLR, of been exceeded in the refer-
ence design life of TLR = 50 years; then, the n ¼ ðT DL =T L Þ1=k (9)
“design ground acceleration,” ag, which has that
probability, PLR, of been exceeded in TL years is EN-Eurocode 8 recommends in a note
obtained by multiplying agR by a probability of 10 % for exceeding the damage
limitation seismic action in TDL = 10 years. For
gI ¼ ðT L =T LR Þ1=k (8a) these recommended values, Eq. 4 gives a mean
return period of 95 years. The value of n should
Alternatively, to have a probability of exceed- vary with the exponent k; in order for the
ing the “design ground acceleration,” PL, in TLR recommended values of n for buildings (equal to
European Structural Design Codes: Seismic Actions 1027
0.4 or 0.5) to be consistent with the other European Structural Design Codes: Seismic Actions,
recommended values, k should be around 2. Table 2 Values of viscous damping, z, for different
structural materials
If the construction phase is long and in its dura-
tion the structural system is much more vulnerable Material Damping (%)
to earthquakes than after completion (as, e.g., in Reinforced concrete components 5
segmental construction), the owner may specify Prestressed concrete components 2
Welded steel components 2
seismic performance requirements during con-
Bolted steel components 4
struction and the associated compliance criteria.
Equation 4 may then be applied, using the esti-
mated duration of construction, Tc, as TL, to give
T D T 4 sec : Sa ðT Þ E
the mean return period of the seismic action which
has the target probability of been exceeded, P (e.g., TCTD
¼ ag S 2:5 (10d)
P = 0.05). The resulting mean return period may T2
be used as TR(ag1) in Eq. 7 to find the peak ground where
acceleration, ag1, with probability P of been ag is the design ground acceleration on rock,
exceeded during construction. In that case, ag2 = from Eq. 3
agR and TR(ag2) = TNCR (Kolias et al. 2012). S is the “soil factor”
EN-Eurocode 8 adopts the same spectral pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
shape for the different seismic actions to be ¼ 10=ð5 þ zÞ 0:55 (11)
used at different performance levels or purposes.
The difference in the hazard level is reflected is the correction factor per (Bommer and
only through the peak ground acceleration to Elnashai 1999) for values of the viscous damping
which the spectrum is anchored. ratio, z, other than the reference value of 5 %. In
fact, the value z = 5 % is considered as represen-
Elastic Response Spectra of the Horizontal tative of cracked reinforced concrete structures.
Seismic Action Components in EN-Eurocode 8 The values of z specified in Part 2 of Eurocode
The elastic response spectral acceleration, Sa(T), 8 (CEN 2005b) for components of various struc-
due to a single horizontal component of the seis- tural materials are listed in Table 2.
mic action is described in EN-Eurocode 8 by the The terms “constant spectral pseudo-velocity
following expressions: range” and “constant spectral displacement
range” used above in Eq. 10c and 10d, respec-
In the “short-period range” tively, are a direct consequence of Eqs. 1 and 2,
alongside the functional dependence of Sa(T) on
T T in Eq. 10c and 10d.
0 T T B : Sa ðT Þ ¼ ag S 1 þ ð2:5 1Þ
TB By definition, S = 1 over rock (called “type
(10a) A ground”). The value agS plays the role of an
“effective ground acceleration,” as the spectral
In the “constant spectral pseudo-acceleration acceleration at the constant spectral acceleration
range”: plateau is always equal to 2.5agS. The “soil fac-
tor,” S, amplifies uniformly the entire spectrum
T B T T C : Sa ðT Þ ¼ ag S 2:5 (10b) by S over the spectrum for rock. This feature
came out of the studies that provided the back-
In the “constant spectral pseudo-velocity range”: ground for the Eurocode 8 spectra (Rey
et al. 2002).
TC The values of the “corner” periods TB, TC, and
T C T T D : Sa ðT Þ ¼ ag S 2:5 (10c)
T TD (i.e., the extent of the ranges of constant spec-
tral pseudo-acceleration, pseudo-velocity, and
In the “constant spectral displacement range”: displacement) and of the soil factor, S, are taken
1028 European Structural Design Codes: Seismic Actions
European Structural Design Codes: Seismic Actions, Table 3 Ground types in EN-Eurocode 8 for the definition of
the seismic action
vs,30 cu
Description (m/s) NSPT (kPa)
A Rock outcrop, with less than 5 m of cover by weaker material >800 – –
B Very dense sand or gravel, or very stiff clay, with depth of several tens of 360–800 >50 >250
meters and mechanical properties gradually increasing with depth
C Dense to medium-dense sand or gravel, or stiff clay, several tens to many 180–360 15–50 70–250
hundreds meters deep
D Loose-to-medium sand or gravel, or soft-to-firm clay <180 <15 <70
E Surface alluvium layer with vs < 360 m/s and thickness from 5 to 20 m,
underlain by rock (with vs > 800 m/s)
S1 10 m thick soft clay or silt, with plasticity index >40 and high water content <100 – 10–20
S2 Liquefiable soils; sensitive clays or any soil not of type A to E or S1
to depend mainly on “ground type.” In the define the seismic action. For ground-type S1, the
EN-Eurocodes, the term ground includes any special study should take into account the thick-
type of soil, as well as rock. EN-Eurocode 8 rec- ness and the vs-value of the soft clay or silt layer
ognizes five standard ground types, as well as two and the difference with the underlying materials
special ground types, S1, S2, for which there are and should quantify their effects on the elastic
no recommended values for TB, TC, TD, or S. Note response spectrum. Note that soils of type S1 may
that in reality the value of S decreases with the have low internal damping and exhibit linear
intensity of the motion (as measured by ag) owing behavior over a large range of strains, producing
to the increased nonlinearity of the soil, but this peculiar amplification of the bedrock motion and
effect has been neglected in EN-Eurocode 8. unusual or abnormal soil-structure interaction
The five standard ground types, A, B, C, D, effects. The scope of the site-specific study
and E and the two special ones, S1 and S2, per should also address the possibility of soil failure
EN-Eurocode 8 are described in Table 3. The under the design seismic action (especially at
reference approach to characterize the ground in ground type S2 deposits with liquefiable soils or
this table is through the average value of shear sensitive clays).
wave velocity, vs,30, at the top 30 m: The values of TB, TC, TD, and S for the five
standard ground types A to E are meant to be
30 defined by each country in the National Annex to
vs, 30 ¼ X (12)
hi Eurocode 8, depending on the magnitude of
v
i¼1, N i earthquakes contributing the most to the hazard.
The geological conditions at the site may also be
where hi and vi are, respectively, the thickness taken into account, in addition to the properties
(in m) and the shear wave velocity at small shear of the top 30 m of ground, to determine these
strains (less than 106) of the i-th layer in values.
N layers. If the value of vs,30 is not available, the Instead of spectral amplification factors that
Standard Penetration Test (SPT) blow-count decrease – probably through the S factor – with
number may be used for soil types B, C, or D, increasing design acceleration (spectral or
according to the correspondence of SPT to vs,30 in ground) because of the soil nonlinearity effect
(Ohta and Goto 1976). If neither the SPT number (as in the US codes), the non-binding recommen-
nor vs,30 is available, the undrained cohesive dation in a note in Part 1 of EN-Eurocode 8 is to
resistance, cu, may be used to characterize have two types of spectra:
the soil.
Ground belonging to the two special types, S1 – Type 1: for moderate to large magnitude
and S2, deserves a special, site-specific study to earthquakes
European Structural Design Codes: Seismic Actions 1029
European Structural Design Codes: Seismic Actions, Table 4 Recommended parameter values for the standard
horizontal elastic response spectra in EN-Eurocode 8
Spectrum type 1 Spectrum type 2
Ground type S TB (s) TC (s) TD (s) S TB (s) TC (s) TD (s)
A 1.00 0.15 0.40 2.00 1.00 0.05 0.25 1.20
B 1.20 0.15 0.50 2.00 1.35 0.05 0.25 1.20
C 1.15 0.20 0.60 2.00 1.50 0.10 0.25 1.20
D 1.35 0.20 0.80 2.00 1.80 0.10 0.30 1.20
E 1.40 0.15 0.50 2.00 1.60 0.05 0.25 1.20
European Structural Design Codes: Seismic Actions, for the five standard ground types of Table 3 (PGA on rock
Fig. 1 Five percent-damped horizontal elastic response equal to 1 g)
spectra of type 1 (left) and 2 (right) recommended in EC8
– Type 2: for low magnitude ones (e.g., with produced by moderate to large magnitude earth-
surface magnitude less than 5.5) at close dis- quakes. The recommended values of the period
tance, producing motions rich in high frequen- TD at the outset of the constant spectral displace-
cies over soft soils ment region seems rather low. For flexible struc-
tures (e.g., those with seismic isolation) they may
The values of TB, TC, TD, and S recommended not lead to safe-sided designs. A safeguard
in a non-binding note in Part 1 of EN-Eurocode against the rapid decay of the elastic spectrum
8 for the five standard ground types A to E are for T > TD is provided by the lower bound of
given in Table 4, giving the shapes depicted in 0.2agS recommended in Part 1 of EN-Eurocode
Fig. 1. These values are based on (Rey et al. 2002) 8 for the design spectral accelerations (see
and European strong motion data (Ambraseys Eq. 13c, 13d in section “Design Spectrum for
et al. 2000). There are certain regions in Europe Force-Based Design With Linear Analysis per
(e.g., where the hazard is contributed mainly by EN-Eurocode 8”).
strong, intermediate-depth earthquakes, as in the
part of the eastern Balkans affected by the Simplified Design Rules, or Exemption from
Vrancea region) where the two recommended Applying EN-Eurocode 8, for Low Design
spectral shapes may not be suitable. The lower Ground Acceleration
S-values of type 1 spectra are due to the larger soil EN-Eurocode 8 says that its provisions need not
nonlinearity in the stronger ground motions be applied in “cases of very low seismicity,”
1030 European Structural Design Codes: Seismic Actions
leaving to the National Annex to decide which value of the design ground acceleration on type
combination of categories of structures, ground A ground (i.e., on rock), ag, or the corresponding
types, and seismic zones in a country qualify as value, agS, over the ground type of the site.
“cases of very low seismicity.” It recommends Moreover, it recommends a value of 0.08 g for
(in a note) as a criterion for the “cases of low ag, or of 0.10 g for agS, as the threshold for the
seismicity,” either the value of the design low seismicity cases.
ground acceleration on type A ground (i.e., on
rock), ag, per Eq. 3, or the corresponding value,
agS, over the ground type of the site. The Elastic Response Spectra of the Vertical
recommended threshold values for the very Seismic Action Component in EN-Eurocode 8
low seismicity cases are 0.04 g for ag or 0.05 g EN-Eurocode 8 requires designers to take into
for agS. As the value of ag includes the impor- account the vertical component of the seismic
tance factor gI, certain structures in a region may action in design only in certain very well-
be exempted from the application of prescribed situations.
EN-Eurocode 8, while others (those housing
essential or large occupancy facilities) may not • In buildings, only if the design ground accel-
be. This is consistent with the notion that the eration in the vertical direction, avg, exceeds
exemption from the application of Eurocode 8 is 0.25 g and even then only in the following
due to the inherent lateral force resistance of any cases:
structure designed for non-seismic loadings, – In base-isolated buildings or
neglecting any contribution from the ductility – For (nearly) horizontal members (beams,
and energy dissipation capacity. As girders, or slabs) which:
EN-Eurocode 8 considers that, thanks to its • Have a span of at least 20 m or
overstrength, any structure is entitled to • Cantilever over at least 5 m or
a behavior factor, q, of at least 1.5, the value • Support directly columns or
of 0.05 g for agS, recommended as the threshold • Are prestressed
for very low seismicity cases, implies an • In bridges, for the seismic design of:
assumed inherent lateral force capacity of – Prestressed concrete decks (with the verti-
0.05 2.5/1.5 = 0.083 g, which is a reasonable cal component acting upwards)
assumption. If a National Annex states that the – Any bearings
entire national territory is considered as a “case – Any seismic links
of very low seismicity,” then EN-Eurocode – Piers in zones of high seismicity:
8 does not apply at all in the country. • If the pier is already taxed by bending
In cases of “low seismicity,” EN-Eurocode due to permanent actions on the deck or
8 permits simplified seismic design procedures: • The bridge is less than 5 km from an
e.g., to design a concrete, steel, composite (steel- active seismotectonic fault (defined as
concrete), or masonry building to EN-Eurocodes one where the average historic slip rate
2, 3, 4, or 6, respectively, alone, using very few is at least 1 mm/year and there is topo-
rules from Part 1 of EN-Eurocode 8 (CEN graphic evidence of seismic activity in
2004a); to design a bridge for “limited ductile” the past 11,000 years); in that case, site-
behavior, with simplified rules defined in Part specific spectra which account for near-
2 of EN-Eurocode 8 (CEN 2005b); etc. Again it source effects should be used.
leaves to the National Annex to decide which • For silos
combination of categories of structures, ground • For tanks
types, and seismic zones in a country belongs to
the “cases of low seismicity” and recommends Therefore, the practical importance of the ver-
(in a note) employing in the criterion either the tical spectrum is less than that of the horizontal.
European Structural Design Codes: Seismic Actions 1031
Sa(T)/a vg
In the “short-period range”: 2
1.5
T
0 T T B : Sa,vert ðT Þ ¼ avg 1 þ ð3 1Þ
TB 1
(13a) 0.5 E
0
In the “constant spectral pseudo-acceleration 0 0.5 1 1.5 2 2.5 3 3.5 4
range”: T (sec)
TC
T C T T D : Sa,vert ðT Þ ¼ avg 3
T – avg = 0.9ag, if the type 1 spectrum applies to
(13c) the site.
– avg = 0.45ag, if the type 2 spectrum is used.
In the “constant spectral displacement range”:
The vertical response spectrum recommended
TCTD in Part 1 of EN-Eurocode 8 is based on work and
T D T 4s : Sa,vert ðT Þ ¼ avg 3
T2 data specific to Europe (Ambraseys and Simpson
(13d) 1996; Elnashai and Papazoglou 1997).
The ratio avg/ag is known to be higher at short
with from Eq. 11. distances (epicentral or to a causative fault).
However, as distance does not enter as
The main differences between the horizontal a parameter in the definition of the seismic action
and the vertical spectra are in: in EN-Eurocode 8, the type of spectrum has been
chosen as the parameter determining this ratio, on
• The amplification factor in the constant spec- the basis of the finding that avg/ag increases also
tral pseudo-acceleration plateau, set equal to with magnitude (Ambraseys and Simpson 1996;
3 instead of 2.5 Abrahamson and Litehiser 1989), which in turn
• The lack of a uniform amplification of the determines the selection of the type of spectrum.
entire spectrum due to the type of soil
EN-Eurocode 8 for force-based design of new There is no clear, well-known energy dissipa-
structures is: tion mechanism for the response in the vertical
direction. So, the “behavior factor” q in that
In the “short-period range”: direction is attributed to overstrength alone and
is taken equal to 1.5, unless a higher value is
2 T 2:5 2 supported by special studies and analyses.
0 T T B : Sa, d ðT Þ ¼ ag S þ
3 TB q 3
(14a)
EN-Eurocode 8 (CEN 2004a) requires as input distance, and mechanism of rupture at the source
for a response-history analysis an ensemble of at consistent with those of the design seismic action.
least three records (or pairs or triplets of different The travel path and the subsoil conditions should
records, for analysis under two or three concur- preferably resemble those of the site. These
rent components of the action, respectively). It requirements are not only hard to meet but also
accepts for this purpose artificial, historic, or may conflict with conformity (in the mean) to the
simulated records. target spectrum of the design seismic action.
Artificial (or “synthetic”) records can be math- EN-Eurocode 8 also requires designers to scale
ematically produced using random vibration the- individual historic or simulated records so that
ory to match almost perfectly the response their peak ground acceleration (PGA) matches on
spectrum defining the seismic action (Gasparini average the value of agS of the design seismic
and Vanmarcke 1976). It is fairly straightforward action; this requirement may also be considered
to adjust the phases of the various sinusoidal against physical reality. It is more meaningful,
components of the artificial waveform, as well instead, to use individual historic or simulated
as the time-evolution of their amplitudes (“enve- records with PGA values already conforming to
lope function”), so that the artificial record the target value of agS. Note also that the PGA
resembles a specific recorded motion; this type alone may be artificially increased or reduced,
of record is called “modified historic” in Part 2 of without affecting at all the structural response.
EN-Eurocode 8. Figure 4 shows seven pairs of So, it is more meaningful to select the records on
15 s long “modified historic” records, modulated the basis of conformity of the spectral values
to match the type 1 horizontal response spectrum alone, as described below.
over ground type C for a PGA on rock of 0.10 g If pairs or triplets of different records are used
(0.115 g over ground type C); it also compares the as input for analysis under two or three concur-
resulting spectra with the target spectrum per rent seismic action components, conformity to
EN-Eurocode 8. the target 5 %-damped elastic response spectrum
Note, though, that records which are equally may be achieved by scaling the amplitude of the
rich in all frequencies are not realistic. Moreover, individual records as follows:
a motion with a smooth response spectrum, with-
out peaks and troughs, introduces a conservative – For each earthquake consisting of a triplet of
bias in the response, as it does not let the inelastic translational components, the records of hori-
response help the structure escape from a spectral zontal components are checked for conformity
peak to a trough at a longer period. Therefore, separately from the vertical one (CEN 2005b).
historic records are favored in EN-Eurocode – The records of the vertical component, if con-
8. Records simulated from mathematical source sidered, are scaled so that the average 5 %-
models, including rupture, propagation of the damped elastic spectra of their ensemble is at
motion through the bedrock to the site, and finally least 90 % of the 5 %-damped vertical spec-
via the subsoil to the surface, are also preferred trum at all periods between 0.2Tv and 2Tv,
over artificial ones, as the final records resemble where Tv is the period of the lowest mode
natural ones and are physically appealing. Obvi- with a participation factor of the vertical com-
ously an equally good average fitting of the target ponent higher than those of both horizontal
spectrum requires more – appropriately ones (CEN 2005b).
selected – historic or simulated records than arti- – For analysis in 3D under both horizontal com-
ficial ones. Individual recorded or simulated ponents, the 5 %-damped elastic spectra of the
records should be “adequately qualified with two horizontal components in each pair are
regard to the seismogenetic features of the combined by applying the Square-Root-of-
sources and to the soil conditions appropriate to Sum-of-Squares (SRSS) rule at each period
the site” (CEN 2004a). In plain language, they value. The average of the “SRSS spectra” of
should come from events with magnitude, fault the two horizontal components of the
European Structural Design Codes: Seismic Actions 1035
2.0
Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2) Sa (m/s2)
1.0 3.0 EC8
m/s2
2.0 Record
0.0
−1.0 1.0
Montenegro 1979 - Ulcinj (2) - Y
−2.0 0.0
2.0
3.0 EC8
1.0
m/s2
2.0 Record
0.0
−1.0 1.0
Imperial Valley 1940 - El Centro Array #9 - 180
−2.0 0.0
2.0
1.0 3.0 EC8
m/s2
European Structural Design Codes: Seismic Actions, Fig. 4 Historic records modified to match EC8’s type
1 horizontal response spectrum for ground type C
individual earthquakes in the ensemble should lowest natural period of the structure in any
be at least 0.9√2 1.3 times the target 5 %- horizontal direction. If it isn’t, all individual
damped horizontal elastic spectrum at all horizontal components are scaled up, so that
periods from 0.2T1 up to 2T1, where T1 is the their final average “SRSS spectrum” exceeds
1036 European Structural Design Codes: Seismic Actions
by a factor of 1.3 the target 5 %-damped hor- the bridge. So, Part 2 of EN-Eurocode 8 (CEN
izontal elastic spectrum everywhere between 2005b) also gives a simplified method to account
0.2T1 and 2T1. for such effects in design. This may be achieved by
superimposing to the seismic action effects from
For analysis under a single horizontal compo- an analysis that neglects the spatial variability of
nent, Part 1 of Eurocode 8 (CEN 2004a) restricts the motion the effects of postulated static relative
the mean 5 %-damped elastic spectrum of the displacements of the supports in the same or in the
applied motions not to fall below 90 % of that opposite direction, which are a function of the peak
of the design seismic action at any period from ground displacement from Eq. 15 and the afore-
0.2T1 to 2T1. mentioned distance beyond which the ground
If the response is obtained from at least seven motion may be considered as uncorrelated.
nonlinear time-history analyses with (triplets or
pairs of) ground motions chosen in accordance
with the above, the relevant verifications may use Summary
the average of the response quantities from all
these analyses as action effect. The set of seven The six parts of the first European Standard for
pairs in Fig. 4 meets this condition for analysis in seismic design, assessment, and retrofitting of
3D under two simultaneous horizontal compo- structures were published from 2004 to 2006 as
nents of ground motion. If less records are used EN-Eurocode 8. In this European structural
(but not less than three), the most unfavorable design standard, the seismic action for which
value of the response quantity among the ana- new structures are designed for the Ultimate
lyses should be used. Limit State (life safety performance) is the
“design seismic action”; it is determined as the
product of the reference seismic action times the
Spatial Variability of the Seismic Action: importance factor of the structure. The return
Asynchronous Ground Motions period of the reference seismic action is chosen
by the country, which retains the authority to set
Parts 2 and 4 of EN-Eurocode 8 (CEN 2005b, the safety level in its territory, taking into account
2006) require designers to consider the spatial economic aspects. Seismic zoning is expressed in
variability of the seismic action in the design of terms of the peak ground acceleration on rock, to
extended structures, namely, of pipelines and which the elastic acceleration response spectrum
multi-support bridges with a continuous deck is anchored. This spectrum has regions of con-
longer than two-thirds of the distance beyond stant spectral pseudo-acceleration, pseudo-
which the ground motion is taken as uncorrelated. velocity, and displacement, the extent of which
This distance is an NDP with recommended value is decided by the country depending on the earth-
ranging from 600 m to 300 m for ground type quake magnitude, as a function of the type of soil
A to D, respectively. Informative Annex D in in the top 30 m and the geologic conditions
(CEN 2005b) provides models of the spatial var- underneath. Five standard types of ground are
iability of the ground motion and appropriate defined; for each one of them, two types of indic-
methods of analysis, including guidance for the ative spectral shapes are given: one for a surface
generation of seismic motion samples reflecting magnitude above 5.5 and the other for below that
the probable spatial variability between supports. limit. The indicative spectral shapes have been
However, this theory is complex and requires developed on the basis of strong motion data from
special tools for its numerical implementation the south of Europe. The peak ground displace-
and special site data for the determination of ment is given as a function of the peak ground
the model parameters. So, the code allows acceleration, along with guidance for the construc-
using a simplified approach to take into account tion of the displacement response spectrum on the
the spatial variability of seismic motions along basis of the peak ground displacement and the
European Structural Design Codes: Seismic Actions 1037
acceleration response spectrum. Vertical response ▶ Selection of Ground Motions for Response
spectra are introduced and indicative shapes are History Analysis
given. Elastic response spectra are utilized to ▶ Site Response for Seismic Hazard Assessment
derive reduced design spectra for conventional ▶ Spatial Variability of Ground Motion: Seismic
force-based design or to generate/adopt accelera- Analysis
tion time-histories for nonlinear dynamic analysis. ▶ Stochastic Ground Motion Simulation
For the design of new buildings, towers, chimneys, ▶ Structural Design Codes of Australia and New
tanks, silos, or pipelines, a second seismic action is Zealand: Seismic Actions
introduced for the damage limitation state (for ▶ Time History Seismic Analysis
protection of property or operation of the facility
E
under a frequent or occasional event). Three limit
states (performance levels) are introduced for the
assessment and retrofitting of existing buildings: References
damage limitation, significant damage, and near
Abrahamson NA, Litehiser JJ (1989) Attenuation of peak
collapse. The country may set the levels of seismic vertical acceleration. Bull Seismol Soc Am 79:549–580
action for which these limit states are checked or Ambraseys N, Simpson K (1996) Prediction of vertical
may leave the decision to the owner. response spectra in Europe. J Earthq Eng Struct Dyn
25:401–412
Ambraseys N, Simpson K, Bommer JJ (1996) Prediction
of horizontal response spectra in Europe. J Earthq Eng
Cross-References Struct Dyn 25:371–400
Ambraseys N, Smit P, Berardi R, Rinaldis D, Cotton F,
Berge C (2000) Dissemination of European strong
▶ Behavior Factor and Ductility
motion data. CD-ROM collection, European Council
▶ Building Codes and Standards Environment and Climate Program
▶ Earthquake Recurrence Bommer JJ, Elnashai AS (1999) Displacement spectra for
▶ Earthquake Response Spectra and Design seismic design. J Earthq Eng 3(1):1–32
Bommer JJ, Scott SG (2000) The feasibility of using real
Spectra
accelerograms for seismic design. In: Elnashai A,
▶ Earthquake Return Period and Its Antoniou S (eds) Implications of recent earthquakes
Incorporation into Seismic Actions on seismic risk. IC Press, London, pp 115–126
▶ EEE Catalogue: A Global Database of Bommer JJ, Elnashai AS, Chlimintzas C, Lee D (1998)
Review and development of spectra for displacement-
Earthquake Environmental Effects
based seismic design. Research report ESEE/98-3,
▶ Engineering Characterization of Earthquake engineering seismology and earthquake engineering
Ground Motions section, Imperial College, London
▶ Nonlinear Dynamic Seismic Analysis Bozorgnia Y, Campbell KW (2004) The vertical-to-
horizontal spectral ratio and tentative procedures for
▶ Performance-Based Design Procedure for
developing simplified V/H and vertical design spectra.
Structures with Magneto-Rheological J Earthq Eng 8(2):175–207
Dampers CEN (2004a) European standard EN 1998-1:2004.
▶ Physics-Based Ground-Motion Simulation Eurocode 8: design of structures for earthquake resis-
tance. Part 1: General rules, seismic actions and rules
▶ Probabilistic Seismic Hazard Models
for buildings. Comite Europeen de Normalisation,
▶ Random Process as Earthquake Motions Brussels
▶ Response-Spectrum-Compatible Ground CEN (2004b) European standard EN 1998-5:2004.
Motion Processes Eurocode 8: design of structures for earthquake resis-
tance. Part 5: Foundations, retaining structures, geo-
▶ Response Spectrum Analysis of Structures
technical aspects. Comite Europeen de Normalisation,
Subjected to Seismic Actions Brussels
▶ Review and Implications of Inputs for Seismic CEN (2005a) European standard EN 1998-3:2005.
Hazard Analysis Eurocode 8: design of structures for earthquake resis-
tance. Part 3: Assessment and retrofitting of buildings.
▶ Seismic Actions due to Near-Fault Ground
Comite Europeen de Normalisation, Brussels
Motion CEN (2005b) European standard EN 1998-2:2005.
▶ Seismic Sources from Landslides and Glaciers Eurocode 8: design of structures for earthquake
1038 European Structural Design Codes: Seismic Actions
resistance. Part 2: Bridges. Comite Europeen de EN 1998-5: Eurocode 8: design of structures for earth-
Normalisation, Brussels quake resistance. General rules, seismic actions,
CEN (2005c) European Standard EN 1998-6:2005. design rules for buildings, foundations and retaining
Eurocode 8: Design of structures for earthquake resis- structures. Thomas Telford Publishers, London
tance. Part 6: Towers, masts, chimneys. Comite Gasparini DA, Vanmarcke EH (1976) Simulated earth-
Europeen de Normalisation, Brussels quake motions compatible with prescribed response
CEN (2006) European standard EN 1998-4:2006. spectra. Res. Report R76-4, Department of Civil Engi-
Eurocode 8: design of structures for earthquake resis- neering, Massachusetts Institute of Technology, Cam-
tance. Part 4: Silos, tanks, pipelines. Comite Europeen bridge, MA
de Normalisation, Brussels Kolias B, Fardis MN, Pecker A (2012) Designers’ guide to
Elnashai AS, Papazoglou AJ (1997) Procedure and spectra Eurocode 8: design of bridges for earthquake resis-
for analysis of RC structures subjected to strong verti- tance, EN 198–2. Institution of Civil Engineers (ICE)
cal earthquake loads. J Earthq Eng 1(1):121–155 Publishing, London
European Commission (2001) Guidance paper L: applica- Ohta Y, Goto N (1976) Estimation of S-wave velocity in
tion and use of Eurocodes. CONSTRUCT 01/483 terms of characteristic indices of soil. Butsuri-Tanko
Rev.1, Brussels 29(4):34–41
European Commission (2003) Commission recommenda- Paolucci R (2002) Amplification of earthquake ground
tion on the implementation and use of Eurocodes for motion by steep topographic irregularities. Earthq
construction works and structural construction prod- Eng Struct Dyn 31:1831–1853
ucts. Document no. C(2003)4639, Brussels Paolucci R (2006) Numerical investigation of 3D seismic
Faccioli E, Vanini M, Frassine L (2002) Complex site amplification by steep topographic profiles and check
effects in earthquake ground motion, including topog- of the EC8 topographic amplification coefficients. In:
raphy. In: The 12th European conference on earth- Bouckovalas G (ed) General report proceedings of the
quake engineering, London, paper 844 athens workshop, ETC12. Geotechnical evaluation
Faccioli E, Paolucci R, Rey J (2004) Displacement spectra and application of the seismic Eurocode EC8, National
for long periods. Earthq Spectra 20(2):347–376 Technical University of Athens, Athens, pp 187–191
Fardis MN (2009) Seismic design, assessment and Rey J, Faccioli E, Bommer JJ (2002) Derivation of design
retrofitting of concrete buildings (based on soil coefficients (S) and response spectral shapes for
EN-Eurocode 8). Springer Science+Business Media Eurocode 8 using the European strong-motion data-
BV, Dordrecht base. J Seismol 6(4):547–555
Fardis MN, Carvalho EC, Elnashai A, Faccioli E, Pinto P, Tolis SV, Faccioli E (1999) Displacement design spectra.
Plumier A (2005) Designers’ guide to EN 1998-1 and J Earthq Eng 3(1):107–125
F
processing and, more in general, of pattern rec- can be computed from the elements of the GLCM
ognition: dimensionality reduction. In fact, which describe quantitatively the textural behav-
dimensionality reduction can be a fundamental ior of the subregion considered. Another
step in the implementation of many classifica- approach for textural characterization relies on
tion or retrieval problems. If we are forced to probabilistic modeling. Markov random field
work with a limited quantity of data, as we (MRF) models fall under this category and can
often are, then increasing the dimensionality be quite successful for texture modeling (Cross
of the space rapidly leads to the point at and Jain 1983). They capture local characteristics
which the data is very sparse, therefore provid- of an image considering regional patterns and by
ing a very poor representation of the mapping. assuming a local conditional probability distribu-
To describe the extraordinarily rapid growth in tion. Spatial feature extraction can be based on
the difficulty of problems as the number of multichannel filtering as well. The image is
variables (or dimension) increases, the phrase decomposed into a set of sub-bands which are
“the curse of dimensionality” has been coined calculated by convolving it with a bank of linear
by Richard Bellman (1957). Reducing the filters, nearly uniformly covering the spatial-
number of input variables can lead to improved frequency domain, followed by some nonlinear
performance for a given data set, even though procedures. Gabor filters can be an example for
information is being discarded. The mapping is this type of techniques (Jain and Farrokhnia
better specified in the lower-dimensional space, 1991). Mathematical morphological filters
which more than compensates the loss of should also be mentioned for the detection of
information. textural features. In this case, the extraction is
obtained by performing shape-oriented opera-
tions, like openings and closures, based on set
An Overview of Feature Extraction theory (Soille 2003).
Techniques
Transform Domain Features
According to the type of satellite data to be ana- An image transform is a mathematical operation
lyzed (SAR, optical/multispectral, time series) that re-expresses in a different, and possibly more
and the considered application, different meaningful, form all, or part of, the information
approaches can be used for feature extraction: content of a multiband or gray scale image. One
of the most used transform methods for feature
Spatial and Contextual Features extraction is the principal component analysis
Spatial and contextual features try to synthesize (PCA) (Singh and Harrison 1985) where
the information contained in neighboring pixels a multispectral set of m images is re-projected in
with respect to a pixel under examination. Differ- terms of a set of m principal components. Two
ent contextual descriptors can be defined. One of particular properties exist: the components are
the earliest and most important techniques used extracted for decreasing variance, and the princi-
to express textural and tonal variations in pal components are statistically unrelated or
a remotely sensed image data is the one based orthogonal. For multiband images, the first PCA
on the gray-level co-occurrence matrix (GLCM) components can represent effective features to be
(Haralick et al. 1973). The GLCM is the core of used in the operational scenarios to be consid-
a mathematical framework capable of ered, as they reveal the structure behind the cor-
representing the distance and angular relation- relation of the variables. Indeed no linear
ships among the pixels over an image subregion combination (i.e., weighted sum) of the original
of a specified size. Each element of the GLCM is bands can contain more information than is pre-
a measurement of the probability of occurrence sent in the first principal components.
of two gray scale values separated by a given An alternative approach considers harmonic
distance in a given direction. Several quantities analysis, a field of mathematics that studies the
Features Extraction from Satellite Data 1041
Features Extraction from Satellite Data, Fig. 1 1,500 1,500 pixels synthetic interferogram’s example (left) and
reconstructed one by means of first 500 DWT components (right)
I2 I2
I3 I3
. .
. .
. .
. .
. .
IN IN
INPUT OUTPUT
LAYER LAYER
BOTTLE-NECK
LAYER
MAPPING DE-MAPPING
LAYER LAYER
Features Extraction from Satellite Data, Fig. 3 Original Landsat image (left) and two computed nonlinear principal
components (center and right)
only provide a sparse covering, because the areas valuable items of information in this respect,
of coseismic ground displacement are not known thanks to its synoptic capability, in particular
a priori. Satellite imagery is naturally suited for when the seismic event is located in remote
such measurements, because it regularly provides regions or a failure affected the main communi-
comprehensive and detailed images of the ground cation systems. Both optical and radar sensors
with high radiometric and geometric quality. The can be exploited for the extraction of this type
displacement field can be measured by the com- of feature. Despite the great potential provided by
parison of images acquired before and after the Earth Observation data for supporting opera-
event. Both InSAR (interferometric SAR) and tional activities in the field seismology, it has
optical data can be exploited for this type of been pointed out that the level of automatic data
investigation. Other effective uses of Earth processing is mostly inadequate with “still too
Observation data for seismology are connected much manual labour and author arbitrariness”
with disaster management. In the management of (Tronin 2010). This is not the case when
a seismic event to immediately draw up a draft, considering the estimation of tectonic parameters
damage map of the stricken urban areas is of great from SAR interferometry where feature extrac-
importance in order to organize civil protection tion, according to innovative technologies based
interventions. Remote sensing can provide on neural networks, can be fully automated.
1044 Features Extraction from Satellite Data
Dimensionality Reduction of the Data to all neurons of the successive layer but has no
Data dimensionality reduction is one of the key feedback to neurons in the previous layers.
issues in retrieval and classification problems. Compared to other methods, the use of NNs is
In this specific case, it is certainly not appropriate often effective because they can simultaneously
to feed a classificator or a parameter estimator address nonlinear dependences and complex
with all the DN numbers forming the interfero- physical behavior with reduced computational
metric image. In such a context, a standard efforts and without requiring any a priori infor-
technique for performing dimensionality reduc- mation (Bishop 1995). Moreover, NNs learn the
tion is to sample the interferogram with an ade- input–output relationships exclusively from the
quate frequency. A better performance can be training patterns; therefore, no explicit expres-
obtained considering the combination of two sion addressing the interactions between the
techniques for feature extraction. In particular, parameters of interest is needed with such an F
the first processing step relies on the harmonic approach.
analysis, and in the second, the AANN is applied. In the training phase, the weights (internal
Such an approach leads to the determination of parameters of the network) are progressively
a considerably reduced number of components adjusted to enable the approximation of the
for the interferogram representation without needed functional relationships. The perfor-
involving the training of huge AANN topologies. mance of the algorithm is monitored by an error
An interferogram with a dimension of 1,500 (or cost) function which has a nonlinear depen-
1,500 pixels can effectively be represented by dence on weights. Once the error is minimized,
a vector of 50 components if the chaining of the the trained MLP performs the desired mapping of
AANN and the DWT is applied (Picchiani input space onto output space.
et al. 2012). As far as the source mechanism classification
is concerned, the training and validation sets can
Classification and Retrieval Using Neural be created through the generation of synthetic
Networks differential interferograms. Such a generation
Once the data dimensionality reduction has been can be performed using a forward model comput-
performed, and the interferogram is represented ing the surface deformation (the displacement
by means of a vector expressing a limited number vector) associated to specific values of the given
of significant feature values, an algorithm is nec- fault parameters. For this purpose, the Okada
essary to transform such a concentrated informa- formulation can be considered as it explains, in
tion content into a set of values to be associated a close analytic form, the surface deformation
with the desired quantities. Such desired quanti- due to a seismic event by a dislocation model in
ties may either be values of tectonic parameters an elastic half space. The data set is obtained by
or probabilities of belonging to a specific fault running the forward model varying the values of
class. In both cases, the problem is rather com- its parameters appropriately all along a certain
plex due to its inherent nonlinearities. For this range (see Fig. 5). In the classification case, the
reason, the use of neural networks can be consid- output layer of the network consists of three units
ered as they are recognized as a powerful tool for each associated with one of the three considered
inversion procedure in remote sensing applica- fault mechanisms. The network is trained in order
tions (Stramondo et al. 2011). to give the value one to the unit corresponding to
Among various topologies, multilayer the desired type of mechanism and the value zero
perceptrons (MLPs) have been found to have to the other two units. In the parameter retrieval
the best suited topology for classification and case, the output of the network is formed by the
inversion problems and can also be used in this desired quantities. Once the network is trained,
case. These are feedforward networks where the the technique performs in a fully automatic mode,
input only flows in one direction, towards the providing the fault classification and/or the esti-
output, and each neuron of a layer is connected mation of the fault parameters in real time.
1046 Features Extraction from Satellite Data
Features Extraction from Satellite Data, from the simulated data set generated by means of
Fig. 5 Different examples for the three considered RNGCHN software. Top: normal fault; middle: strike
types of faults of differential interferograms extracted slip fault; bottom: reverse fault
▶ Earthquake Damage Assessment from VHR Picchiani M, Del Frate F, Schiavon G, Stramondo S
Data: Case Studies (2012) Features extraction from SAR interferograms
for tectonic applications. EURASIP J Adv Signal
▶ Earthquake Magnitude Estimation Process 2012:155
▶ Earthquake Mechanism Description and Schölkopf B, Mika S, Burges CJC, Knirsch P, Muller K,
Inversion Ratsch G, Smola AJ (1999) Input space versus feature
▶ Earthquake Mechanisms and Tectonics space in kernel-based methods. IEEE Trans Neural
Netw 10(5):1000–1017
▶ Estimation of Potential Seismic Damage in Singh A, Harrison A (1985) Standardized principal com-
Urban Areas ponents. Int J Remote Sens 6(6):883–896
▶ Hyperspectral Data in Urban Areas Soille P (2003) Morphological image analysis: principles
▶ InSAR and A-InSAR: Theory and applications. Springer, New York
Stramondo S, Del Frate F, Picchiani P, Schiavon G (2011)
▶ Land Use Planning Following an Earthquake Seismic source quantitative parameters retrieval from
Disaster InSAR data and neural networks. IEEE Trans Geosci F
▶ Noise-Based Seismic Imaging and Monitoring Remote Sens 49(1):96–104
of Volcanoes Tronin AA (2010) Satellite remote sensing in seismology.
A review. Remote Sens 2010(2):124–150
▶ Remote Sensing in Seismology: An
Overview
▶ SAR Images, Interpretation of
▶ Seismic Event Detection
▶ Spatial Variability of Ground Motion: Seismic
Frequency-Magnitude Distribution
Analysis
of Seismicity in Volcanic Regions
▶ Urban Change Monitoring: Multi-temporal
SAR Images
John J. Sanchez Aguilar1, Hernán G. Alvarado2
and Santiago López2
1
Departamento de Geociencias, Universidad
Nacional de Colombia, Bogotá, Colombia
References 2
Departamento de Geociencias, Universidad
Bellman R (1957) Dymanic programming. Princetown Nacional de Colombia-Sede Bogotá, Bogotá,
University Press, Princetown Colombia
Bishop C (1995) Neural networks for pattern recognition.
Oxford University Press, New York
Cross GR, Jain AK (1983) Markov random field texture
models. IEEE Trans Pattern Anal Mach Intell PAMI- Synonyms
5(1):25–39
Feigl KL, Duprè E (1999) RNGCHN: a program to calcu-
late displacement components from dislocations in an b-value anomalies; b-values; Earthquake size dis-
elastic half-space with applications for modeling geo- tribution; Seismic b-values around volcanoes;
detic measurements of crustal deformation. Comput Volcanic seismicity; Volcano-tectonic
Geosci 25(6):695–704
earthquakes
Haralick RM, Shanmugam K, Dinstein I’H (1973) Tex-
tural features for image classification. IEEE Trans Syst
Man Cybern SMC 3(6):610–621
Jain AK, Farrokhnia F (1991) Unsupervised texture seg- Introduction
mentation using Gabor filters. Pattern Recogn
24(12):1167–1186
Massonnet D, Rossi M, Carmona C, Adragna F, Peltzer G, The spatial and temporal distribution of earth-
Feigl K, Rabaute T (1993) The displacement field of quakes sizes near and around volcanoes can be
the Landers earthquake mapped by radar interferome- characterized through the well-known frequency-
try. Nature 364(6433):138–142
magnitude distribution:
Okada Y (1985) Surface deformation due to shear and
tensile faults in a half space. Bull Seismol Soc Am
75(4):1135–1154 logN ¼ a bM (1)
1048 Frequency-Magnitude Distribution of Seismicity in Volcanic Regions
Frequency-Magnitude Distribution of Seismicity in Total number of earthquakes used is 5,858. The solid
Volcanic Regions, Fig. 1 Average magnitude of com- line shows a least squares best fit for the linear part of
pleteness and b-value for the data compiled by the frequency-magnitude distribution (FMD). The insets
Observatorio Vulcanológico y Sismológico de Pasto show histograms of magnitudes, with bins equal to 0.1.
(OVSP, Servicio Geológico Colombiano) at Galeras Vol- Right: Data for the period 2008.3–2010.6. Total number of
cano, Colombia. Left: Data for the period 2004–2013. earthquakes is 1,384. Conventions as in left panel
where N is the cumulative number of earthquakes magnitudes shows a slightly skewed distribution
with magnitude M and a and b are constants with a right tail and a maximum that occurs in the
representing the seismic productivity of a region range 0.7 ML 0.9, a first indication of the
and the slope of the linear part for the distribu- probable value of MC in the catalog. Another
tion, respectively. This is a well-known power estimate of completeness may be obtained by
law (Gutenberg and Richter 1944) that is applied finding the maximum curvature of the FMD
herein. using a combination of 90–95 % confidence in
Earthquake catalogs compiled by volcano the curvature (derivative) estimation. Indeed, this
observatories around the world are commonly procedure gives Mc = 0.8 for the whole catalog
used to estimate b values and map their spatial implying that earthquakes with ML < 0.8 are
distribution. From Eq. 1, it is evident that earth- probably not being always detected. Inspection
quake magnitude plays a fundamental role and is of a curve of cumulative number of earthquakes
important to mention that its computation may versus time revealed that during 2004–2013 rates
vary from one catalog to another. Firstly, differ- of seismicity at Galeras varied, most likely
ent magnitudes have been traditionally calculated because of occurrence of swarms, groups of
at different observatories (duration magnitudes, earthquakes with similar magnitudes that occur
MD, and local magnitudes, ML, are most com- closely separated in time and not associated to
mon) and secondly, detection capabilities vary a mainshock. The period 2008.3–2010.6, how-
because of the different network configurations ever, exhibited a fairly stable behavior and the
and equipment. The magnitude of complete corresponding FMD did not vary (Fig. 1, right
detection or magnitude of completeness (Mc) is panel). The data for both time periods follows
the value above which the catalog is considered fairly well a Gutenberg-Richter law and the dif-
to be approximately complete (Fig. 1). ference in average b-values between both
In a practical way Mc can be estimated by datasets is small. Earthquake catalogs usually
visual inspection of any of the forms of the are culled at the Mc value to ensure that the
FMD. For example, in Fig. 1 the FMD for earth- dataset is mostly complete. The linear part of
quakes at Galeras Volcano, Colombia, is shown. the FMD can be fit by a least squares procedure
The histogram of the number of events versus to obtain a straight line with negative slope.
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions 1049
The slope is called the b-value and it represents body of accepted knowledge on seismic b-values
the average size distribution of earthquakes. near volcanoes is presented. Finally, a short case
Because the b-value is related to the mean study on one of the most active volcanoes is
magnitude in a given region, it is important to shown to exemplify applications of the technique
spatially map its distribution in both volcanic and and to describe some possible implications for
tectonic environments because of the possible volcano and earthquake hazards and risk.
relationships with the state of stress in the crust.
When calculating the spatial distribution of b, one
must consider (1) the calculation method, which Seismic b-Values at Volcanoes
can either be maximum likelihood or weighted Worldwide
least squares, and (2) the parameters of the
gridding scheme and the number of events to The distributions of seismic b-values at depth F
sample around each node of the grid (Wiemer have been studied (by spatial mapping mostly)
and McNutt 1997). at a number of volcanoes around the world using
To calculate the b-values and estimate their a variety of catalogs. A compilation of results of
spatial distribution, it is necessary to create a grid 2D and 3D mapping at 16 volcanoes is presented
in two or three dimensions. As a general rule the (Table 1).
node separation should be proportional to the Table 1 highlights that b-value anomalies are
formal errors in the spatial parameters of the common regardless of the type of volcano, con-
earthquake locations. Regarding the sampling figuration of the seismograph network, magni-
mode around each node, there are two possibili- tude scale used, or amount of earthquakes in the
ties: either selecting a constant number of events dataset. The pattern of the anomalies varies from
within circles of variable radii or selecting place to place, however (Wyss et al. 1997), and it
a variable number of events inside circles of is established that, with no exception, volcanic
constant radii. In both methods the sampling vol- areas exhibit irregularly shaped volumes of
ume has the shape of a cylinder, with its base anomalously high b-values surrounded by
centered on a particular node of the grid. If the regions of normal-to-low b-values (Wiemer and
calculation takes place in map view, the cylinder McNutt 1997; Wiemer and Wyss 2002). From
will be vertical, but if the calculation is on Table 1, it is observed that there is notable spatial
a vertical cross section, the cylinder will be hor- variability of b, with several volcanoes having
izontal, with variable height and radius more than one region of increased b-values.
depending on the sampling method. In this way There is variability in the computed b-values as
the cylinders can be coin shaped or pipe shaped. well, with b-anomalies in the range 1.0 < b < 3.0,
If the grid is for a 3D region, the sampling vol- and although it would seem appealing to compare
umes will be spheres. these values among volcanoes, conclusions
Within each cylinder the b-value is calculated drawn from such task would not be solid, because
from the FMD obtained from the sample and in of the different magnitude scales used. For
this case one can either assume a constant Mc instance, the b-values of 1.55 reported for shal-
(usually equal to the minimum magnitude within low anomalies at Mt. St. Helens and Nevado del
the sample) or calculate the FMD locally. It is Ruiz are not necessarily comparable.
usually a good practice to calculate b-values It is reasonable to assume that mineral stabili-
using both methods and trying several sampling zation (formation of crystals) in magmas takes
schemes to test the stability of the results. place at depth within the magma chamber and
The estimation of b-values has been an active that exolution of gas and vesiculation might
research area over the last two decades and here occur at shallower levels. The mineral and vola-
a description of results for volcanoes around the tile species and types of magmas studied vary and
world is presented, then the possible implications thus the calculated depths for stabilization of
for magma processes are summarized and the minerals or gas exolution and formation of
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions, Table 1 Main features of b-value anomalies and data on mineral stabilization and gas processes at
1050
volcanoes around the world. Reported and calculated stabilization pressure values from various authors are converted to pascals (Pa, N/m2 = m1 kg s2) or megapascals
(MPa = 106 Pa) as this is the SI derived unit for pressure or stress
Number of Anomaly
earthquakes Magnitude b-value centroid Pressure
Volcano Type used scale Mc (anomaly)a depth (km) References (MPa) References
St. Helens (USA) Stratovolcano 1,674 ML 0.4 1.5 0.3 3.25 0.55 Wiemer and 175 Pallister
McNutt et al. (2008)
(1997)
1.6 8.35 1.65
Mt. Spurr (USA) Stratovolcano 643 ML 0.1 1.7 3.4 1.1 Wiemer and
McNutt
(1997)
1.2 >10.0
Mammoth Mtn. (Long Caldera dome ML 1.3 1.8 5.0 Wiemer 20 Burkett
Valley Caldera) et al. (1998) (2007)
1.73 8.0
0.06
Resurgent Dome (Long Active dome ML 1.3 1.9? 4.0
Valley Caldera)
1.36 10.0
0.05
Teishi Knoll (Izu-Tobu Submarine Crater, 10,000 ML? 1.3 1.00 8.5 Wyss 800 Kushiro
group), located off Ito within pyroclastic 0.03 et al. (1997) (1991)
(Japan) cones field
1.54 13.0
0.05
Soufriere Hills (West Stratovolcano 1,900 MD 1.7 3.07 2.0 Power 150 Barclay
Indies) 0.07 et al. (1998) et al. (1998)
Mt. Etna (Italy) Stratovolcano 450 MD 2.5 3.0 0.5 7.5 2.0 Murru 155 Trigila
et al. (1999) et al. (1990)
14.5 2.5
Martin (Alaska, USA) Stratovolcano 2,500 ML 0.7 1.6 6.0 1.0 Jolly and 62.5 Coombs and
0.05b McNutt Gardner
(1999) (2001)
Mageik (Alaska, USA) Stratovolcano ML 0.7 1.8 6.0 1.5
0.24c
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions
Katmai (Alaska, USA) Stratovolcano, Caldera ML 0.7 1.30 8.0 1.0 37.5 Coombs and
0.25 Gardner
(2001)
12.0 3.0
Kilauea-East Rift Zone Shield Volcano 16,963 Mpd 1.5 1.73 5 2.0 Wyss
et al. (2001)
Mt. Redoubt (USA) Stratovolcano 1.7 0.1 4.0 Wiemer and 180.9 Gerlach
Wyss (2002) et al. (1994)
2.2 0.2 9.0
Nevado del Ruiz Stratovolcano 2,769 MD 0.7 1.55 1.0 Rodrı́guez 50 Stix
(Colombia) 0.25 (2003) et al. (2003)
1.65 1.0 0.5
0.05
1.55 5.35 0.35
0.15
1.55 2.0 0.2
0.15
Pinatubo (Philippines) Stratovolcano 1,406 MD 0.73 1.7 0.1 2.0 1.0 Sánchez 220 Bernard
et al. (2004) et al. (1996)
11.0 1.0
Galeras (Colombia) Stratovolcano 1,918 MLe 1.2 1.3 0.05 2.5 0.5 Sánchez 60 Stix
et al. (2005) et al. (1997)
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions
Miyake-jima-Tokushima Stratovolcano 10,000 ML? 2.5 1.86 0.1 8.0 0.5 Rierola 224 Amma-
(Izu Islands, Japan) (2005) Miyasaka
et al. (2003)
1.3 0.1 12.5 2.5
1.3 0.1 19 2.0
Makushin (Alaska, USA) Stratovolcano 491 ML 1.25 1.9 0.1 4.0 1.7 Bridges and 169 Goldberg
Gao (2006)f (2010)
a
Values for b-anomalies and their locations are taken from results of 3D mapping and vertical cross sections, where reported
b
During July 27, 1995, to December 31, 1997, the Alaska Volcano Observatory reported over 2,500 earthquakes located in the general area Martin-Mageik-Trident-Katmai
Caldera. The b-values reported in the table were estimated from Fig. 6 (Jolly and McNutt 1999)
c
b-values at Martin-Mageik region may include data from swarms in Sept. 1996–April 1997. Values calculated pre- and post swarm were in the range 0.92 < b < 0.98
d
Mp is a “preferred magnitude” chosen hierarchically and defined in the data description prepared by F. Klein for distribution of the earthquake catalog at Hawaii. Mostly equal to
MD and in few cases equal to ML
e
Magnitudes were recalculated by regression equation from MD to ML using a sample of earthquakes simultaneously recorded by broad band and short period subnetworks
f
Bridges and Gao report a second, deeper b-value anomaly (b 1.6) detected at roughly 12 km depth southeast of Makushin, but it was deemed statistically insignificant
1051
F
1052 Frequency-Magnitude Distribution of Seismicity in Volcanic Regions
Frequency-Magnitude 6
Distribution of
Seismicity in Volcanic
Regions, 5
Fig. 2 Histogram of
Number of anomalies
b-value anomaly centroid
depths beneath volcanoes 4
0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Depth of centroid of b-value anomaly
vesicles: For Mt. Etna, Italy, Trigila et al. (1990) in a range of 1.26–2.12 kbar has been suggested.
report a range of 85–225 Mpa pressure range for Evidence from Geology, Petrology, and Geo-
Plagioclase stabilization in basaltic lavas and it chemistry thus indicates that several processes
has been suggested that the feeding system for take place at different depths beneath volcanoes.
magmas at Etna may be composed of different From Table 1 it is also evident that seismicity
reservoirs; at the Katmai cluster, USA, Coombs studies have also revealed that beneath many
and Gardner (2001) estimated a range of volcanoes, b is anomalously high at different
40–100 Mpa for Plagioclase stabilization in depths. All volcanoes listed (17) exhibit b-value
andesite-dacite-rhyolite magmas erupted during anomalies embedded in an otherwise region of
the 1912 eruption from Novarupta; at Redoubt, normal-to-low b-values. With respect to the dis-
USA, Gerlach et al. (1994) suggested that SO2 is tribution of b-values at depth, most volcanoes
stored as vapor at 6–10 km depth in the magma studied (16, b-values at Yellowstone were not
before the 1990 eruptions and perhaps at mapped at depth) have at least one anomaly,
shallower levels (4–6 km) prior to the 2009 erup- and 11 out of 16 volcanoes exhibit more than
tions (Coombs and Bacon 2012); at Nevado del one anomalous region. The centroid depths for
Ruiz, Colombia (Stix et al. 2003), it has been b-value anomalies vary in the range 2–19 km and
suggested that ascent of volatile-bearing magma their distribution are shown in Fig. 2.
took place from 9 to 15 km depth and remained at From Fig. 2 it is evident that, in many cases,
shallow residence depth of <3 km; at volcanoes exhibit two anomalies at different
Mt. Pinatubo, Philippines, estimation of volatile depths: one shallow anomaly is common in the
contents in the range 5.1–6.4 % suggested a vesic- range 2–4 km and a deeper one appears in the
ulation depth of 6 km for pre-eruptive magmas range 7–8 km. The first anomaly range includes
and stabilization of olivine at 2.2 kbar or a depth not only the depth at which volatiles in magma
of magma storage in the range 7–11 km was also with 4 % weight fraction may start to exsolve but
inferred (Bernard et al. 1996); at Galeras, Colom- also the region where the existence of open
bia (Stix et al. 1997), gas exolution might have cracks is plausible. At depths similar to the
taken place at 20–100 Mpa or 0.8–3.9 km depth; range 7–8 km, alternate geophysical evidence
and at Makushin, USA, scoriae within mafic and geology studies support the nearby presence
ignimbrite layers were analyzed (Goldberg of magma bodies. At Galeras Volcano, Colom-
2010) and a stabilization pressure of pyroxenes bia, the shallow plumbing magmatic system has
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions 1053
been conceived as a vertically elongated feature complex configurations in the magma and
such as a reservoir extending down to 5–6 km plumbing systems beneath volcanoes is gen-
depth or a fractured plug beneath de active cone erally accepted.
that overpressurizes and acts as the source of – That factors that may contribute to variations
material during Vulcanian explosions (Stix in b are:
et al. 2003) and a deeper magma reservoir (a) To increase b:
below 4 km depth. 1. Increase in material heterogeneity. At
Regarding the cause of anomalous b-values, volcanoes this can be associated to
there is general agreement that several factors increase in cracks, caused by vesicula-
may be involved; these are summarized as follows: tion and fragmentation of ascending
magma, processes that likely take
place at relatively shallow levels F
Accepted Knowledge on Seismic (roughly 4 km depth for magmas with
b-Values at Volcanoes 4 wt. % H2O).
2. Increase in pore pressure caused by
– That the frequency-magnitude distribution is magma nearing groundwater (which
related to the heterogeneity (degree of crack- would imply decreasing the effective
ing) in the upper crust and can be appropri- stress). This process may also likely
ately measured by a power law expressed in take place at relatively shallow levels.
the Gutenberg-Richter law. The parameters of 3. Increase in the thermal gradient. Quite
this law are understood as the seismic produc- possible near magma systems or intru-
tivity of the volume (“a-value”) and the mean sions, because magmas may reach tem-
magnitude of the linear part (“b-value”). peratures near 1,200 C.
– Estimating the b-value is equivalent to esti- (b) To decrease b:
mating the mean magnitude and it is reason- 1. Increase in applied shear stresses.
able to assume the mean magnitude as Difficult to model this near volcanoes,
proportional to mean crack length. but plausible along the faults that are
– Heterogeneity in volcanic areas can be perva- common in volcanic areas. Zones of
sive and may be related not only to the nature lower-than-normal b-values have been
and amount of cracking that the magma sys- associated to regions of high stress load
tem or volcano edifice may have but also to the and therefore areas where large earth-
nature of rocks and deposits that form volca- quake nucleation could take place.
noes, usually a combination of hydrothermally In this perspective, seeking for regions
altered lava flows and pyroclastic deposits. of lower-than-normal b-values is just
– Although volcanic areas were usually reported as important and may have hazard
as having higher-than-normal b-values, with implications.
respect to a “global” average of b = 1.0, it is 2. Increase in effective pressure or decrease
now known that instead, volcanoes exhibit in pore pressure. Could be conceivable
regions of anomalously high b-values within during or after volcanic eruptions.
a “normal” crust. (c) Either to increase or decrease b:
– That magma (molten rock inside the earth) 1. Changes in the configuration of the
cannot support shear stresses and therefore detection network or changes in ana-
earthquakes cannot occur there. b values give lyses procedures. These may introduce
information on faults and fractures in the artifacts in the catalog. For example,
vicinities of magma chambers. The resolution a change in the way magnitudes are
of earthquake locations may not be still determined may cause a “stretching”
enough to reveal details on the configuration of the magnitude scale, offsetting
of magma systems, although the concept of b toward lower values.
1054 Frequency-Magnitude Distribution of Seismicity in Volcanic Regions
It has been suggested that processes such as processes do represent a significant hazard for
vesiculation, fragmentation, and magma intru- the region (INGEOMINAS 1997), but also earth-
sions can affect the crack distribution beneath quakes possibly associated to volcanic-tectonic
a volcano. Gas exolution and subsequent bubble interactions have caused damage and fatalities
nucleation and growth inside magmas take place in recent times. Thus, seismically characterizing
under colossal pressure conditions. Bubbles can the region near Galeras is of relevance and map-
increase in numbers and accumulate near the top ping of b-values offers the opportunity to identify
of magma reservoirs increasing pressure on the zones of smaller-than-average earthquakes as
host rock and facilitating cracking. Also, magmas well as regions where relatively large earth-
intrude through fractures and weak zones of the quakes are typically produced.
crust and this process may cause new cracks and Using preliminary earthquake locations
fracture propagation thus changing the homoge- between September 1995 and June 2002, a first
neity of the rock masses. approach to spatially map the earthquake size
From Table 1 a scatter plot of the centroid distribution around Galeras was taken (Sánchez
depths for b-value anomalies versus the pressure et al. 2005). Their results of 3D mapping readily
at which mineral stabilization and/or gas allowed to identify a prominent high b-value
exolution might take place or the estimated anomaly beneath the active crater and vertically
depth for magma storage (from Petrology, Geo- elongated down to 5 km depth which was inferred
chemistry, and Geology data) was built. Errors in to be the region adjacent either to the volcano’s
the centroids of b-value anomalies were taken as conduit, the shallow magma reservoir, or alterna-
the maximum and minimum observable extents tively the remnants of a semi-brittle intrusion. In
at depth from published and authoritative mate- map view, this anomalous feature was basically
rial. Results are shown in Fig. 3. surrounded by regions of normal-to-low
Based on Fig. 3 (right panel), a positive rela- b-values. Of particular interest here is the cluster
tion between the depth of deeper b-value anoma- of earthquakes located northeast of Galeras,
lies and the magma process beneath volcanoes is where the source of three swarms that included
mildly revealed for a group of eight volcanoes: damaging earthquakes (ML 4.7) was identified
Mammoth Mt. (Long Valley Caldera), Nevado during 1993–1995 (April 1993, November–
del Ruiz, Mt. St. Helens, Katmai-Novarupta, December 1993, and February 1995). This source
Redoubt, Mt. Pinatubo, Mt. Etna, and Miyake- falls within a region of low b-values detected
jima. These include diverse-sized volcanic sys- preliminarily in the early 2005 study (Fig. 4).
tems from a variety of geological environments: Recent data for Galeras has been analyzed and
six continental arc volcanoes (formed above the catalog of earthquakes for the period
oceanic-continental subduction zones), a hot 2004–2013 was used to spatially map b-values
spot volcano, and an island arc volcano (formed using both the entire catalog data and a subset of
above oceanic-oceanic subduction zone). it (2008–2010), during which seismicity rates
were observed to be fairly stable. Results are
presented in Fig. 5. In map view the region
b-Values at Galeras Volcano, Colombia, north-northeast of Galeras still appears as
and Implications for Earthquake a region of low b-values and remains as
Hazard and Risk a source of moderate-sized earthquakes
(a sequence of four 4.6 < ML < 4.7 earthquakes
Galeras Volcano, one of the most active volca- occurred there during mid 2005). The active cra-
noes in Colombia, looms over the city of San Juan ter region remains highly productive, but it does
de Pasto (pop. 400,000) which lies just a few not show as a zone of smaller-than-average earth-
kilometers east of the active crater. Not only quakes, on the contrary, falls within an area of
have repeated eruptions of Galeras affected life low b-values and was the source of two separate
and property over the years and volcanic 4.6 < ML < 4.7 earthquakes in 2006 and 2010.
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions 1055
Frequency-Magnitude
Distribution of
Seismicity in Volcanic
Regions, Fig. 3 Scatter
plots of the depths of
b-value anomalies
observed at volcanoes
versus pressure conditions
calculated for gas exolution
or mineral stabilization
(from Geology, Petrology
and Geochemistry data).
Many volcanoes from
Table 1 exhibit more than
one b-value anomaly F
usually at different depths.
Left: Centroid depth of
“shallower” anomalies.
Right: Centroid depth for
“deeper” anomalies. Thick
black lines in both panels
show a basic linear fit
(R = 0.0 for the data on the
left panel, R = 0.5 for data
on the right panel)
Frequency-Magnitude Distribution of Seismicity in black lines: faults. Thick red contour: limits of San Juan
Volcanic Regions, Fig. 4 Map of b-values in the de Pasto. Swarms of damaging earthquakes were pro-
Galeras region for the period 1995–2002. High b-values duced between 1993 and 1995 in the area of low b-values
are mapped as regions of hot colors (red-orange), regions northeast of the volcano (Modified from Sánchez
of low b-vales appear as areas of cold colors (blue). Dots: et al. 2005)
earthquake epicenters. Triangle: active vent. Continuous
Frequency-Magnitude Distribution of Seismicity in seismicity rate was fairly constant. The different b-values
Volcanic Regions, Fig. 5 Maps of b-values in the in the western portions of Galeras between the different
Galeras region. Left: period 2004–2013, using all earth- datasets may be an artifact caused by the different sam-
quakes with ML > 0.8. Relatively large earthquakes pling around each node, necessary because of the lower
(4.6 < ML < 4.7) occurred in 2006.16 (near the crater), number of earthquakes in the shorter period 2008–2010.
2010.64 (just east of the crater), and during The region of interest, northeast of the active crater
2005.64–2005.65 (north-northeast of the crater, four remains mostly unchanged. Conventions as in Fig. 4
earthquakes). Right: period 2008–2010, during which the
itself is spatially related to these structures, par- this area a particularly complex one in terms of
ticularly to the splay of the two faults near the stresses and earthquakes.
active crater. It is conceivable that magma has Interestingly, the area southeast of the vol-
used faults as pathways to the surface, making cano which shows persistently as a region of
Frequency-Magnitude Distribution of Seismicity in Volcanic Regions 1057
Gutenberg B, Richter CF (1944) Frequency of earth- areas: Mount St. Helens, Washington, and Mt. Spurr,
quakes in California. Bull Seismol Soc Am Alaska. Geophys Res Lett 24:189–192
34:185–188 Wiemer S, Wyss M (2002) Mapping spatial variability of
INGEOMINAS (1997) Mapa de amenazas volcánicas del the frequency–magnitude distribution of earthquakes.
Galeras-tercera versión. Instituto de Investigaciones Adv Geophys 45:259–302
en Geociencias, Minerı́a y Quı́mica. http://www. Wiemer S, McNutt SR, Wyss M (1998) Temporal and
ingeominas.gov.co/Pasto/Volcanes/Volcan-Galeras/ three-dimensional spatial analyses of the frequency-
Mapa-de-amenazas.aspx. Accessed 4 June 2013 magnitude distribution near Long Valley Caldera,
Jolly AD, McNutt SR (1999) Seismicity at the volcanoes California. Geophys J Int 134:409–421
of Katmai National Park, Alaska; July 1995–Decem- Wyss M, Shimazaki K, Wiemer S (1997) Mapping active
ber 1997. J Volcanol Geotherm Res 93:173–190 magma chambers by b values beneath the off-Ito
Kushiro I (1991) Origin of volcanic rocks in Japanese volcano, Japan. J Geophys Res 102:20,413–20,422
island arcs. Episodes 14:258–263 Wyss M, Klein F, Nagamine K, Wiemer S (2001) Anom-
Murru M, Montuori C, Wyss M, Privitera E (1999) The alously high b-values in the south flank of Kilauea
location of magma chambers at Mt. oEtna, Volcano, Hawaii: evidence for the distribution of
Italy, mapped by b-values. Geophys Res Lett magma below Kilauea’s east rift zone. J Volcanol
26:2553–2556 Geoth Res 106:23–37
Pallister JS, Thornber CR, Cashman KV, Clynne MA,
Lowers HA, Mandeville CW, Brownfield IK, Meeker
GP (2008) Petrology of the 2004–2006 Mount
St. Helens lava dome – implications for magmatic
plumbing and eruption triggering. In: Sherrod DR,
Friction Dampers for Seismic
Scott WE, Stauffer PH (eds) A volcano rekindled: the Protections of Steel Buildings
renewed eruption of Mount St. Helens, 2004–2006, Subjected to Earthquakes: Emphasis
U.S. Geological Survey professional paper on Structural Design
2008–1750. U.S. Geological Survey, Vancouver,
pp 647–702
Power JA, Wyss M, Latchman JL (1998) Spatial varia- Lucia Tirca
tions in the frequency-magnitude distribution of earth- Department of Building, Civil and
quakes at Soufriere Hills Volcano, Montserrat, West Environmental Engineering, Concordia
Indies. Geophys Res Lett 25:3653–3656
University, Montreal, QC, Canada
Rierola M (2005) Temporal and spatial transients in
b-values beneath volcanoes. Diploma thesis, ETH
Zurich, Institute of Geophysics, Zurich, 77 pp
Rodrı́guez SP (2003) Mapeo tridimensional del valor b en Synonyms
el Volcán Nevado del Ruiz. Trabajo de Grado,
Universidad de Caldas, Manizales, Colombia, 72 pp
Sánchez JJ, McNutt SR, Power JA, Wyss M (2004) Spatial Damping; Earthquake; Energy dissipation; Fric-
variations in the frequency-magnitude distribution of tion dampers; Seismic design; Steel buildings
earthquakes at Mount Pinatubo Volcano. Bull Seismol
Soc Am 94:430–438
Sánchez JJ, Gómez DM, Torres RA, Calvache ML,
Ortega A, Ponce AP, Acevedo AP, Gil-Cruz F, Introduction
Londoño JM, Rodrı́guez SP, Patiño J de J, Bohórquez
OP (2005) Spatial mapping of the b-value at Galeras To reduce the seismic demand, researchers have
Volcano, Colombia, using earthquakes recorded from
proposed to incorporate supplemental energy dis-
1995 to 2002. Earth Sci Res J 9:30–36
Stix J, Torres RA, Narváez L, Cortés GP, Raigosa J, sipation devices into the structural system of
Gómez DM, Castonguay R (1997) A model of vulca- buildings. According to the primary dissipation
nian eruptions at Galeras volcano, Colombia. mechanism, supplemental energy dissipation
J Volcanol Geotherm Res 77:285–303
devices are grouped into two categories: hyster-
Stix J, Layne GD, Williams SN (2003) Mechanisms of
degassing at Nevado del Ruiz Volcano, Colombia. etic and viscoelastic. Hysteretic devices rely on
J Geol Soc 160:507–521 the relative displacements of components within
Trigila R, Spera FJ, Aurisicchio C (1990) The 1983 Mount the device and are typically based on either
Etna eruption: thermochemical and dynamical infer-
metallic yielding or frictional sliding, while vis-
ences. Contrib Miner Petrol 104:594–608
Wiemer S, McNutt SR (1997) Variations in the frequency- coelastic devices are velocity dependent. More
magnitude distribution with depth in two volcanic specifically, friction devices dissipate energy
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes 1059
Ev = % EI
allel with the beam located at the top of chevron 40%
bracing system. The activation of slip forces that
30%
characterize the designed friction dampers occur
simultaneously with the maximum internal forces 20%
allowable to develop in the system during the 10%
ground motion excitation. The building reaches
0%
the peak interstorey drift when the available slip 0% 20% 40% 60% 80% F
distance provided by friction damper devices was Damping
consumed. “The forces generated by these
devices installed in the structural members are Friction Dampers for Seismic Protections of Steel
usually in phase with the internal forces resulting Buildings Subjected to Earthquakes: Emphasis on
Structural Design, Fig. 1 Vibrational energy versus
from ground motion shaking” (Christopoullos supplemental damping
and Filiatrault 2006).
The total input energy, EI, induced by a seis-
mic event into a structural system can be members are protected from damage associated
expressed as a summation of kinetic energy, Ek, with permanent deformations. In this light, the
cumulative strain energy, ES, inherent damping, main design objective is to minimize the differ-
ED, and the hysteretic damping Eh of the seismic ence between the seismic input energy and that
force resisting system (SFRS). In this study, Eh is dissipated by the dampers (Christopoullos and
the damping induced by friction devices. The Filiatrault 2006). The variation of Ev as a function
energy balance equation is: of supplemental damping is depicted in Fig. 1.
Pioneering work on friction devices was
EI ¼ Ek þ ES þ ED þ E h (1) conducted by Pall (1979) and Pall and Marsh
(1981). Since then, several types of friction
The kinetic and cumulative strain energy are dampers have been developed and studied in the
accumulated into the primary structural system literature. These devices differ in their mechani-
and rely on structural damage (Akiyama 2000; cal shape and materials used for the sliding
Tirca 2009), while the system is damped by both surfaces.
ED and Eh, which are amplitude-dependent. In Pall friction dampers dissipate energy through
general, the contribution of ED and Eh is related friction developed by the relative sliding within
to the amount of post-yielding response and Eq. 1 two surfaces in contact which are clamped by
can be rearranged as follows: posttensioned bolts. In order to obtain stable rect-
angular hysteresis loops similar to that of Cou-
Ek þ ES ¼ EI ð ED þ E h Þ (2) lomb friction, different types of surface treatment
and lining material were studied experimentally.
In Eq. 1, the term (Ek + ES) expresses the vibra- After investigating the response of slip bolted
tional energy, Ev, or more specifically the poten- joints under monotonic loading, Pall reported
tial damage energy, while (ED + Eh) is the energy that the most stable behavior was obtained when
dissipated by viscous damping in the structural brake lining pads in contact with mill scale sur-
members and supplemental devices. Thus, by face on plate was chosen (Fig. 2a). Nevertheless,
adding damping into a structural system the elas- minor differences between the static and dynamic
tic vibration energy diminishes, while structural friction coefficient were observed. However,
1060 Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes
Friction Dampers for Seismic Protections of Steel joint: (a) monotonic test, (b) back-bone curve, and (c)
Buildings Subjected to Earthquakes: Emphasis on hysteretic behavior (After Pall 1979)
Structural Design, Fig. 2 Response of slip-bolted
under large seismic excitations, the posttensioned United States, and India (Christopoullos and
bolts of friction dampers may impact into the end Filiatrault 2006). Examples of Pall friction
of slotted hole and undergo bearing or even bolt damper installed in an X-bracing system and
shear failure. The hysteretic behavior of friction single-diagonal brace are showed in Fig. 3.
damper shown in Fig. 2b is similar to that of an In the last two decades, the following types of
elasto-perfectly plastic system. Herein, the back- friction devices have been developed: slotted
bone curve is composed of four segments: elastic, bolted connections (Fitzgerald et al. 1989;
slipping, bearing, and bolt shear failure. When Grigorian et al. 1993; Tremblay 1993),
the demand is higher than the available slip Sumitomo devices developed by Sumitomo
length which is equal to the length of slotted Metal Industries Ltd. of Japan and reported by
hole, a sudden increment in storey shear forces Aiken and Kelly (1990), energy dissipating
accompanied by decreasing of Coulomb damping restraint damper (EDR) developed by Flour Dan-
is encountered. The hysteresis behavior depicted iel Inc. (Nims et al. 1993), friction variable
in Fig. 2c in terms of slip load versus the slip damper developed based on the EDR damper
length, D, shows rectangular symmetrical loops (Zhou and Peng 2009), friction damper devel-
which are largely influenced by the fluctuation of oped by Mualla (2000), etc.
friction coefficient during the slipping stage. The The energy dissipated by the Sumitomo fric-
elastic stiffness, ko, shown in Fig. 2c, is the stiff- tion damper is due to friction generated when
ness of the attached brace member. friction pads, made of cooper alloy with graphite
Due to their efficiency (Pall and Pall 2004), plug inserts, slide directly on the inner surface of
Pall friction dampers were employed in more the outer cylinder. The precompressed internal
than 40 new and retrofit buildings in Canada, spring, incorporated into the device, applies a
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes 1061
F
Friction Dampers for Seismic Protections of Steel installed in X-bracing at Concordia Library, Montreal,
Buildings Subjected to Earthquakes: Emphasis on and (b) installed in single-diagonal brace at Boeing Com-
Structural Design, Fig. 3 Pall friction dampers mercial Airplane Factory Everrett, USA (Courtesy
manufactured by Pall Dynamics Lmt. Montreal: (a) Dr. Pall)
force that is converted through the action of inner The friction damper proposed by Mualla
and outer wedges into a normal force on the (2000) was added to connect the top part of chev-
friction pads. Sumitomo device can be installed ron braces to the mid-span of a moment resisting
in parallel to the beam located on top of chevron frame beam. This friction damper, patented by
bracing system or in-line with steel braces. As DAMPTECH Denmark, consists of three steel
reported by Aiken and Kelly (1990), Sumitomo plates that are able to rotate against each other
dampers were used in two high-rise buildings in around a pre-stressed bolt which passes through.
Japan in order to resist small intensity earth- Two circular friction pad discs inserted between
quakes and ground floor vibrations. the aforementioned steel plates provide dry fric-
The EDR damper is similar to Sumitomo tion lubrication and ensure stable friction forces.
device. It includes the following components: an As reported by Liao et al. (2004), a full scale test
internal spring, steel compression wedges, conducted on a three-storey moment resisting
bronze friction wedges, stops at both spring’s frame with added chevron bracing system and
ends, and the outer steel cylinder. Its functioning Mualla’s dampers was investigated. The total
comprising the generating slip force depends on weight of the frame was about 34 t. The bracing
the length of the internal spring. In addition, the system consisted of bar members that were
normal force acting on the cylinder wall and the pre-tensioned in order to avoid buckling. Due to
friction force developed between the bronze fric- the application of lower scaling simulated ground
tion wedges and the inner surface of the outer motions, no damage of structural members was
cylinder determine the slip force in the device. As reported during the conducted experimental tests.
noted by Nims et al. (1993), the EDR device can
produce a wide range of hysteretic behavior such
as double-flag shape and triangular lobed shape. Design Provisions
Both types of hysteresis loops are self-centering.
By using the same principles, Zhou and Peng Since 1980, seismic response control techniques
(2009) proposed a new friction variable damper have been used as complementary solutions to the
where a sliding shaft and a friction ring were used existing SFRSs. Despite of their wide-spread
to replace the spring and wedges. In addition, two applications, design guidelines for structures
zones with high and low friction coefficient have equipped with supplemental energy dissipation
been defined in the internal walls of the outer devices (hysteretic and viscoelastic) are still in
cylinder. evolving phases. This study refers to the design
1062 Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes
philosophy of structures with incorporated fric- testing program, etc. A brief review of these
tion damper devices (FDD). As mentioned above, guidelines with highlights on friction damper
FDDs are hysteretic devices. For example, devices design is presented below.
buckling-restrained braces are also known as hys-
teretic damper devices and are designed to dissi-
pate part of vibration energy through yielding of FEMA P-750 and ASCE/SEI 7-10
the core plate. Provisions
In general, hysteretic devices are installed in
braces and are able to undergo large inelastic Design Philosophy
response, while they dissipate most of the hyster- General requirements:
etic energy. The purpose of installing these
devices into the structural system is to maintain – Structures with a damping system have a seis-
the main structure either elastic or within low mic force resisting system that provides a
inelastic deformations. Both, the main frame complete load path. The base shear used to
and the supplemental energy dissipation system, design the employed SFRS shall not be less
share the same deformation, which in turn is that than Vmin, where Vmin is the grater of:
of the entire system. It is important to assure
stable response of these devices under dynamic V min ¼ V=Bvþ1 and V min ¼ 0:75 V:
loading. Thus, the added dampers installed in
new or retrofit buildings should yield or slip Herein, V is the seismic base shear in the
before the shear resistance of the main structure direction of calculation determined in accor-
is reached. dance with the code procedure corresponding
In North America, the first design guidelines to the selected SFRS and Bv+1 is the numerical
addressing provisions for passive energy dissi- coefficient set for effective damping reduction
pated devices were introduced in FEMA factor that is equal to the damping provided by
356 (2000) and FEMA 450 (2003) that refer to supplemental dampers in addition to the inher-
the seismic rehabilitation of buildings and seis- ent damping (2 % for steel structures). For
mic design for new buildings, respectively. Later irregular building structures and for damping
on, FEMA 450 was replaced by FEMA-P system that has less than four damper devices
750 (2009). The aforementioned design guide- per floor disposed in the direction of loading,
lines were incorporated in Chapter 14 of ASCE/ the minimum base shear will not be less than
SEI 41-13 (Seismic evaluation and retrofit of 1.0 V.
existing buildings), as well as in Chapter 18 of – Damping system may be used in addition to
ASCE/SEI 7-10 standard. It is noted that Com- the SFRS in order to meet interstorey drift
mentary J of NBCC 2010 (National Building limits and it may be located external or inter-
Code of Canada) includes two clauses in regard nal to the structure, while it shares or not the
to seismic design with supplemental energy dis- members of SFRS. In Fig. 4 is illustrated a
sipated devices. However, both CSA/S16-2009 SFRS consisting of one bay moment resisting
and ANSI/AISC 360-2010 standard provide frame (MRF) and a damping system com-
design regulations for the “ductile buckling- posed of diagonal braces with installed FDs.
restrained braced frames” system. The FD system is located independently to the
In Europe, the Eurocode 8 part 1 (CEN 2004) MRF system.
contains Chapter 10 entitled Base isolation, but – Energy dissipating devices must be tested
this chapter does not cover regulations for passive before installation for displacement and slip
energy dissipation devices installed into several force corresponding to seismic demand that is
storeys of building’s structure. required by code.
The aforementioned guidelines specify gen- – Linear static and response spectrum analysis
eral requirements, analysis procedures, required methods for design are accepted for regular
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes 1063
structures that have at least four damper consists of changing the yield load corresponding F
devices per each floor located in one principal to metallic dampers with the slip load of friction
direction of the building. Like MRF or braced dampers. As mentioned above, friction dampers
frame structures, damped structures yield dur- are installed in braces and are designed not to slip
ing the design ground motions, while they under the wind forces. On the other hand, in order
need to behave elastically under the wind to increase the building performance under earth-
loads. quake loads, systems composed of friction-
– For detailed design and seismic response damped braces are added to the MRF structures,
investigations, nonlinear dynamic time- while the MRF system should posses enough
history analyses are recommended. strength, stiffness, and ductility. Furthermore,
– Dampers are modeled to perform in the plastic all members of the SFRS and those attached to
range. The computer model should be cali- FDs should be designed to carry the internal
brated based on experimental test results. For forces developed when slip forces are activated.
designs in which the SFRS is expected to Thus, the application of capacity design checks to
yield, the postyield behavior of the structural structural members is required. According to
elements must be modeled explicitly. Christopoullos and Filiatrault (2006) the design
of structures equipped with friction dampers can
In the ASCE/SEI 7-10 provisions, damping be divided in four steps:
system is defined as being “the collection of
structural elements that includes all the individ- • Calculate the demanded slip force and size the
ual damping devices, all structural elements or in-line brace by employing simplified
bracing required to transfer forces from damping methods (e.g., the equivalent static force pro-
devices to the base of the structure, and the struc- cedure). Then, the demanded peak interstorey
tural elements required to transfer forces from drift (e.g. linear dynamic analysis methods)
damping devices to the seismic force-resisting should be evaluated.
system.” For details of design procedure, the • Optimize the design of friction dampers and
reader is referred to the ASCE/SEI 7-10 docu- adjacent members.
ment. It is noted that other design methodologies • Apply the capacity design and size all struc-
can be applied to design structures equipped with tural members to carry the designed slip
friction dampers. forces.
• Use the nonlinear time-history dynamic anal-
Design of Structures Equipped With Pall ysis to check the design performance of build-
Friction Dampers ing structure under an ensemble of selected
As noted above, metallic dampers and friction ground motions scaled to match the design
dampers are hysteretic dampers. The design of response spectrum corresponding to 2 % prob-
structures equipped with metallic dampers or ability of exceedance in 50 years (NBCC
friction dampers is similar. The difference 2010).
1064 Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes
A review of different design procedures from ðEf =ðV br ¼ V t =ku 2V br =ku ¼ 0 (6)
the literature is discussed below. However, in all
design procedures, the building structure is com- From Eq. 6, the shear force exerted by braces is:
posed of a bare frame system and a supplemental
damping system, whilst its response is analyzed V br ¼ 0:5V t (7)
before and after the friction dampers are
activated. Thus, to maximize the energy dissipated by fric-
tion devices, the total shear force must be equally
Design Procedures from the Literature shared between friction devices and the bare
The first design procedure for MRF system frame (MRF). In addition, it was reported that
equipped with friction damped braces was pro- the optimum slip force of friction devices
posed by Baktash and Marsh (1986). They con- installed in structure depends on the structural
sidered a single storey MRF structure with characteristics only and not on the ground motion
friction damped braced bays. Based on this pro- intensity (Baktash and Marsh 1987).
cedure, the slip force is activated when yielding Filiatrault and Cherry (1988) proposed to
in the moment resisting frame brace is reached. determine the optimum activation of friction
For a single bay MRF frame that was designed devices based on minimizing the Relative Perfor-
based on the principle “strong column weak mance Index, RPI, devised from the application
beam”, the shear force resisted by the frame of the energy concept.
action, Vs is related to the plastic moment capac-
ity of the MRF beam, Mp. Herein, Mp = Vsh/2, RPI ¼ 1⁄2½SEA=SEAo þ Umax =U max0
(8)
where h is the storey height. Thus, the proposed
design method is based on the assumption that the where SEA is the strain energy area of all struc-
lateral shear force leading the damper in-line with tural members of a friction damped system, SEA0
brace to slip must be equal to the shear force is the strain energy area corresponding to a zero
causing the MRF to yield and is expressed by activation force, Umax is the maximum strain
the following equation: energy stored in all structural members of a fric-
tion damped system, and Umax0 is the maximum
strain energy for a zero slip force. The values
V s ¼ 2Mp =h (3)
resulted for the RPI yield to three cases:
(i) RPI = 1 corresponds to the response of a
The corresponding shear deflection of the storey, bare frame structure; (ii) RPI < 1 corresponds
Ds is: to the response of a damped structure which is
smaller than the response of the bare frame struc-
Ds ¼ ðV t V br Þ=ku (4) ture; (iii) RPI > 1 corresponds to the response of
a damped structure which is larger than the
where Vt is the total shear force exerted by the response of the bare frame structure. Authors
frame and braces, Vbr is the total shear force recommended the selection of diagonal braces
exerted by braces alone, and ku is the lateral to comply to Tb/Tu < 0.4, where Tb and Tu are
stiffness of the bare MRF. the fundamental period of the braced frame struc-
The hysteretic energy dissipated by friction ture and bare frame structure, respectively.
devices Ef is: Herein, the fundamental periods Tb and Tu can
be expressed as Tb = 2p/ob and Tu = 2p/ou,
Ef ¼ V br Ds ¼ V br ðV t V br Þ=ku (5) where ob is the natural circular frequency of the
fully braced frame before dampers are activated,
By differentiating Eq. 5 with respect to Vbr and ob = (kb/m)0.5 and ou is the natural circular
setting it equal to zero, the maximum energy frequency of bare frame, ou = (ku/m)0.5. In addi-
dissipated by friction devices is obtained: tion, kb and ku are the lateral stiffness of the
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes 1065
braced frame structure and bare frame structure, • l = kb/kf which is the ratio between the stiff-
respectively. In these expressions, m is the total ness of bracing system and the stiffness of the
seismic mass of the system. Later on, Filiatrault bare moment resisting frame.
and Cherry (1990) proposed the following equa- • ß = dby/dfy which is the ratio between the
tion to calculate the total shear force, Vo that is displacement that cause yielding of bracing
required to activate all friction damper devices in system and that that causes yielding of bare
a structure: MRF system.
• f = Ffy/mag which is the yield strength of the
frame normalized to the mass of the structure
V o =W ¼ ag =g Q T b =T g , T b =T u , N f (9)
multiplied by the design peak ground acceler-
ation. Similarly, it can be defined
where Nf is the number of floors, ag is the design
b = Fby/mag, where Fby is the yield force of F
peak ground acceleration, g is the gravity accel-
bracing system. The total normalized force is
eration, Tg is the predominant period of the
II = f + b.
selected design ground motion, and Q is
an unknown single valued function that is
The selection of the above parameters sug-
given in Eq. 10a for 0 Tg/Tu 1 and Eq. 10b
gests the variation effect of ß and l on the build-
for Tg/Tu > 1.
ing response. When ß = 1, bracing system and
MRF system yield for the same lateral displace-
Q ¼ T g =T u 1:24N f 0:31 T b =T u
ment. When ß = 0, the structural system corre-
þ 1:04N f þ 0:43 sponds to the MRF system only. They
(10a) recommended an optimal value of ß around 0.5.
The application of this design methodology to
Q ¼ ðT b =T u Þ 0:01N f þ 0:02 T g =T u 1:25N f 0:32 MDOF systems requires uniform distribution of
þ T g =T u 0:002 0:002N f þ 1:04N f þ 0:42 stiffness and slip forces along the building height,
(10b) aiming at uniform activation of friction devices in
order to avoid damage concentration within a
Equations 9 and 10a or 10b can be used directly to specific floor. In addition, they proposed to main-
calculate the total base shear force V0. It is noted tain at each storey the stiffness of braces propor-
that for a single diagonal brace located at floor i, tional to the bare frame stiffness. Also, they
the slip force is Pi and the activation shear recommended keeping proportionality between
at that floor is Vi = Pi cosyi, where yi is the the horizontal projection of slip forces per floor
angle of brace inclination reported to a and the corresponding static or dynamic distrib-
horizontal line. uted storey shear over the building height. In
In the meantime, studies on hysteretic bracing addition, two design assumptions were made:
systems were carried out by Ciampi et al. (1990, (i) the MRF system is designed to respond elasti-
1992, 1995, etc.) who considered that both cally while the friction damper devices behave in
MRF and bracing system are arranged in parallel. the non-linear range and (ii) the MRF system may
They concluded that the activation of friction yield when dfy was reached, therefore after fric-
damper devices should occur before yielding of tion damper devices were activated. The tri-linear
moment resisting frame. In addition, Ciampi force-displacement constitutive low is depicted
et al. (1995) referred to four key parameters in Fig. 5.
such as: Fu and Cherry (1999, 2000) proposed a sim-
plified design method that leads to a code com-
• Tf, which is the fundamental period of the bare patible procedure for a friction-damped steel
moment resisting frame. A deductable param- braced frame system. First, they developed the
eter is a = Tb/Tf where Tb is the fundamental method for a SDOF system based on establishing
period of the bracing system. an equivalent ductility-related force modification
1066 Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes: Emphasis on Struc-
tural Design, Fig. 6 Mechanical model of a SDOF system
stiffness of friction-damped braces, kbd, as shown simulate the highly nonlinear Coulomb friction F
in Fig. 2c. As noted above, the stiffness of the and has the ability to represent different hystere-
frame and the stiffness of friction-damped braces sis shapes according to the values of the parame-
are in parallel (kf + kbd), whereas Dy = Fs/(kf + ters involved (Morales Ramirez 2011). Since the
kbd) and Fs is the slip restoring force. In this desired shape of the Coulomb dry friction law is
study, the seismic response of retrofitted symmetric and strength and stiffness degradation
building equipped with friction-damped braces is neglected, the Bouc-Wen model is reduced to a
was verified to remain within the interstorey nonlinear restoring force (Eq. 11) of a SDOF
drift code limits and forces developed in struc- system shown in Fig. 6. The evolutionary vari-
tural members to be below the members’ resis- able z, given in Eq. 11, is defined in Eq. 12.
tance. It was concluded that friction-damped
braces installed in moment frame buildings are fsðu,_ zÞ ¼ ako u þ ð1 aÞko z (11)
able to control the interstorey drift and floor
acceleration. _ Þ
v
A jzjn ½g þ bsgnðuz
z_ ¼ u_ (12)
a
fs
b Ps + fd
kbr
BW Slipping Ps Bearing No
Brace slipping
–ua
EPPGap No u
uy ua + uy
slipping
–Ps Slipping
EPPGap
Bearing
–(Ps + fd)
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes: Emphasis on Struc-
tural Design, Fig. 7 Friction-damped brace: (a) Schematic model and (b) Hysteresis model plus the bearing state
analysis requires the calculation of the evolution- Fig. 7b (Morales Ramirez and Tirca 2012).
ary variable z at each step of time for a single Herein, uy is the elastic axial deformation of the
element. By using n = 10, it gives an acceptable brace. To simulate the slip-lock phase, one equiv-
level of prediction because the difference is alent EPPGap spring made of three bilinear gap
quickly reduced throughout the evolution of the springs arrange in parallel each other are defined
slipping stage. to act in tension and another set in compression.
To simulate the slip and slip-lock phases of Each uniaxial ElasticPPGap material (OpenSees
friction-damped brace device in OpenSees notation) has a defined stress–strain or force-
(McKenna and Fenves 2004), Morales Ramirez deformation relationship either in tension or in
(2011) proposed a novel equivalent uniaxial compression. Thus, one equivalent EPPGap
material composed of BoucWen material spring is activated when the displacement
(BW) in parallel with gap springs made of demand exceeds the available slip length, +ua,
Elastic-Perfectly Plastic Gap material and the other is activated when the -ua slip length
(EPPGap), which are already available in the is reached. Once activated, this component is able
OpenSees library (Mazzoni et al. 2006). The fric- to limit the displacement and to increase the force
tion damper was arranged in series with the brace experienced by the friction damper. The thresh-
member made of steel material. The brace was old force of these gap elements is related to the
modeled with a truss element. The schematic maximum force that the device is able to with-
representation of friction-damped brace device stand (Ps + fd) without reaching failure (Fig. 7b).
is shown in Fig. 7a. In the illustrated model, To control the failure phase, the MinMax material
BoucWen material was selected to replicate the was set to decouple the friction device when the
smooth hysteresis behavior of friction damper strain of the ElasticPPGap material exceeds the
(stick–slip and slip phase), which was activated predefined bounds either in tension or compres-
when the axial stress ss generated by the slip sion. In addition, when the MinMax material is
force Ps was reached (ss = Ps/Abr and Abr is the activated (e.g., t = ti), the device is decoupled
gross area of the attached brace). Each friction from t = ti until the end of analysis.
damper is characterized by its slip force and slip Detailed design examples of 4-, 8-, and
length that are resulted from calculation. How- 12-storey MRF building equipped with Pall fric-
ever, under strong ground motions, the provided tion dampers installed in-line with braces are
slip length, ua, may be limited and the given in Morales Ramirez (2011) reference. The
pretensioned edge bolt may drive the friction building is located in Montreal, Canada, and was
damper into the slip-lock phase as depicted in subjected to ten simulated and historical ground
Friction Dampers for Seismic Protections of Steel Buildings Subjected to Earthquakes 1069
Filiatrault A, Cherry S (1990) Seismic design spectra for Moreschi LM, Singh MP (2003) Design of yielding metal-
friction damped structures. ASCE J Struct Eng lic and friction dampers for optimal seismic perfor-
116(5):1334–1355 mance. Earthq Eng Struct Dyn 32:1291–1311
Fitzgerald TF, Anagnos T, Goodson M, Zsutty T (1989) Mualla IH (2000) Experimental and computational eval-
Slotted bolted connections in a seismic design of con- uation of a novel friction damper device. PhD thesis,
centrically braced connections. Earthquake Spectra, Department of Structural Engineering and Materials,
EERI 5(2):383–91 Technical University of Denmark
Fu Y, Cherry S (1999) Simplified seismic code design National Research Council of Canada (2010) National
procedure for friction damped steel frames. Canadian building code of Canada. National Research Council
J Civil Eng 26:55–71 of Canada, Ottawa
Fu Y, Cherry S (2000) Design of friction damped struc- Nims D, Ritcher R, Bachman R (1993) The use of the
tures using lateral force procedure. Earthq Eng Struct energy dissipating restraint for seismic hazard mitiga-
Dyn 29:989–1010 tion. Earthq Spectra 9(3):467–489
Grigorian CE, Yang TS, Popov EP (1993) Slotted bolted Pall A (1979) Limited slip bolted joints, a device to control
connection energy dissipators. Earthq Spectra the seismic response of large panel structures. PhD
9(3):491–504 thesis, Building, Civil and Environmental Engineer-
Liao WI, Mualla I, Loh CH (2004) Shaking table test of a ing, Concordia University, Montreal
friction-damped frame structure in. Struct Des Tall Pall A, Marsh C (1981) Response of friction damped
Spec 13:45–54 braced frames. J Struct Div ASCE 108(ST6):
Mazzoni S, McKenna F, Scott M, Fenves G (2006) 1313–1323
Opensees command language manual. PEER, Univer- Pall A, Pall R (2004) Performance based design using Pall
sity of California, Berkeley friction dampers – An economical design solution. In:
McKenna F, Fenves GL (2004) Open system for earth- 13th world conference on earthquake engineering,
quake engineering simulation (OpenSees), PEER, Vancouver, paper #1955
University of California, Berkeley Tirca L (2009) A simple approach for the seismic retrofit
Morales Ramirez JD (2011) Numerical simulations of of buildings with staggered friction dampers. In:
steel frames equipped with friction-damped diagonal- Mazzolani FM (ed) Proceedings of the first interna-
bracing devices. MASc thesis, Building, Civil and tional conference on: protection of historical build-
Environmental Engineering, Concordia University, ings. CRC Press, Italy, pp 761–768
Montreal Tremblay R (1993) Seismic behavior and design of friction
Morales Ramirez JD, Tirca L (2012) Numerical simula- concentrically braced frames for steel buildings. PhD
tion and design of friction-damped steel frame struc- thesis, University of British Columbia, Vancouver
tures. In: The 15th world conference in earthquake Zhou X, Peng L (2009) A new type of damper with
engineering, Lisbon, paper #2538 friction. Earthq Eng Eng Vibrat 8(4):507–520
G
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 1 Seismic
collapse of RC building
(2001 Gujarat earthquake
of Mw = 7.7 in India)
resisted to a reasonable extent by this margin, Amplification of Ground Motion in Soft Soil
whereas there is no significant safety margin The intensity of ground motion is increased by
against the horizontal action produced by earth- the surface soft soil deposits and this phenome-
quake excitations. Consequently, many efforts non is called amplification. It is known that the
have been made in the recent times to improve extent of amplification changes basically with the
horizontal resistance in structures. thickness of the surface soil (H) and the velocity
It has been widely recognized that earthquake of S-wave propagation (Vs). The S(econdary)-
damage is more substantial upon soft soil wave means a propagation of shear deformation
deposits than on firm soil or rock outcrop. This and the theory of elasticity defines its velocity by
belief is supported by an experience during the sffiffiffiffi
1923 Kanto earthquake of Mw = 7.9 in Tokyo. G
Vs ¼ (1)
Figure 2a illustrates the geomorphology of the r
present Tokyo at the end of the sixteenth century
when the urban development was started. The in which G and r stand for the shear modulus and
eastern part of the area used to be a mouth of a mass density of soil, respectively. More pre-
big river surrounded by swamps and rice paddies cisely, the extent of amplification is a function
underlain by soft and loose soils. In contrast, the of 2pfH/Vs where f designates the number of
western part of the area is composed of higher shaking cycles per second (frequency). The
terraces covered by volcanic ash materials, asso- most significant amplification, which is called
ciated with occasional small valleys that were resonance in subsoil, occurs when 2pfH/Vs = p/2,
produced by erosion and filled with soft soils. and therefore the frequency is equal to Vs/(4H).
Figure 2b illustrates the distribution of earth- This implies that the surface ground motion is
quake damage percentage of traditional wooden dominated by a component of this special “reso-
houses. Note that this damage implies the damage nant” frequency although the motion in the base
by shaking and does not include those by big fires rock at a deep elevation includes many frequency
that finally destroyed most parts of the city. In this components in a random manner.
figure, a good consistency can be found between Resonance occurs also in structures upon the
the distribution of soft soils in the eastern part and ground surface. Generally, soft wooden houses
in valleys and the location of high damage rate. with heavy roofs have low resonant frequency.
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1073
Geotechnical Earthquake Engineering: Damage (Drawn after Murai 1994). (a) Original geomorphology
Mechanism Observed, Fig. 2 Correlation between of Tokyo (Drawn after Murai 1994). (b) Distribution of
subsoil conditions and earthquake damage of wooden house damage in 1923 (Okamoto 1973)
houses in Tokyo during the 1923 Kanto earthquake
Therefore, if the surface ground motion is rich Seismic inertial force for design
with (dominated by) low-frequency components, ¼ ðWeight of structureÞ PGA=g (2)
the seismic response of such house is significant
and heavy damage is likely. In contrast, rigid where g stands for the gravitational acceleration
structures have higher resonant frequencies and (=9.8 m/s2) and PGA is affected by the afore-
are prone to damage if located upon stiff soil mentioned amplification of motion. Noteworthy
where H is small, Vs is high, and Vs/(4H) is is that this inertial force is considered to be a
close to the structure’s resonant frequency. static force that lasts permanently despite that
Figure 3a, b illustrates two wooden houses, the real earthquake action occurs for a much
one crushed house on soft alluvium and shorter time. Hence, Eq. 2 overestimates the real
survival of the other on a firm gravelly fun seismic effect. To compensate for this, PGA in
deposit. On the contrary, Fig. 3c shows a rigid Eq. 2 has to be made smaller than the real max-
warehouse that was located on a firm gravelly fun imum acceleration.
and suffered a significant damage during the Equation 2 implies that the maximum accel-
earthquake. eration or its equivalent inertial force controls the
seismic failure of structures. This type of failure
Seismic Inertial Force is the case of a brittle structure which collapses if
The earthquake resistant design in the modern the external load exceeds the resistance; see
times was established by the use of seismic iner- Fig. 4. Because traditional masonry buildings
tial force. Theoretically, this force is related with and wooden houses were of brittle nature, seismic
the maximum acceleration (PGA) by design by means of this static seismic design
1074 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 3 Different damage
extents of houses because
of different resonant
frequency and ground
conditions (after the 1995
Kobe earthquake of
Mjma = 7.3). (a) Soft house
on soft soil. (b) Soft house
on firm soil. (c) Rigid house
on firm soil
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1075
Brittle structure
ECLPD.ppt 01
Deformation
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 6 Translation of
retaining wall triggered by
failure of unstable fill in its
foundation (in Takamachi
residential area, Nagaoka,
during the 2004 Niigata-
Chuetsu earthquake of
Mw = 6.6)
a b
Seismic
force
Pile-soil
interaction
ECLPD.ppt 02
Geotechnical Earthquake Engineering: Damage Mechanism Observed, Fig. 7 Schematic illustration of pile-soil
interaction during earthquakes. (a) Soil displacement > pile displacement. (b) Soil displacement < pile displacement
mitigation is to enlarge the size of foundation and and exerted load on the station. Central columns
increase the number of piles (Fig. 8). in the station were destroyed, the ceiling fell
down, and the ground surface caved
Underground Structure and Earthquake in. Noteworthy is that soil becomes softer during
Motion strong shaking because shear strain in soil during
The deformability of subsoil as illustrated in strong shaking is greater and shear modulus of
Fig. 7a is important because this is one of the soil decreases as shear strain amplitude increases.
main reasons why underground structures are Such a situation was not fully considered in the
destroyed during strong earthquakes. Figure 9 past when subsoil was optimistically assumed to
illustrates depression of ground surface that was reduce the deformation of underground struc-
caused by the collapse of a metro station in Kobe tures. After this lesson was learned, central col-
in 1995. This station collapsed because the sur- umns in many subway stations were reinforced to
rounding backfill soil was not sufficiently firm avoid future problems (Fig. 10). However, the
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1077
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 8 Increasing the
number of piles to resist
stronger earthquake force
(Kobe)
Geotechnical Earthquake Engineering: Damage Mechanism Observed, Fig. 9 Depression at ground surface
above the collapsed subway station (Kobe 1995, Photo by Prof. Nozomu Yoshida)
success of reinforcement depends on the intensity Note that tunnels used to be considered stable
of earthquake shaking during future earthquakes. during earthquakes because surrounding “hard”
Similar situation is found in the significant rock was expected to reduce the deformation of
deformation of a tunnel in “soft” rock (Fig. 11). tunnel during earthquakes.
1078 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 10 Seismic
retrofitting of central
columns in Oshiage
subway station in Tokyo
Geotechnical Earthquake Engineering: Damage not have an invert at the bottom). (a) Bending and com-
Mechanism Observed, Fig. 11 Significant seismic pression in the top of tunnel (crown). (b) Uplift at the
deformation of tunnel in soft rock (Haguro Tunnel, 2004 bottom of tunnel
Niigata-Chuetsu earthquake of Mw = 6.6) (this tunnel did
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 12 Former site of
Yungay City (see a hill in
the back side and
Mt. Huascarán with snow)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 13 Site of
former Valdez, Alaska
(Photograph taken by
Dr. K. Horikoshi in 2000)
from the glacier through the river in the figure Consequently, this submarine slope easily lique-
into the sea bottom. The fine grain size of the fied and failed during the gigantic earthquake
material made the sedimentation rate very low in and triggered a tsunami.
the seawater, resulting in loose deposit, while
the lack of cohesion made the material highly Natural Dam
prone to liquefaction. Even worse was that the The 1999 Chi-Chi earthquake in Taiwan regis-
fine grain size reduced the permeability and tered Mw = 7.7. The rock mass in the geologi-
hence the dissipation of excess pore water pres- cally young Taiwan island was not fully lithified.
sure during shaking was very slow. The complicated tectonic environment around
1080 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 14 Tsao-
Ling slope failure in
Taiwan caused by the 1999
Chi-Chi earthquake
the Taiwan island disturbed the rock mass. The travel distance of a failed landslide mass
The high extent of annual rainfall made steep (runoff distance) is an important issue because
slopes in the mountainous regions. These situa- the longer runoff results in damage in a bigger
tions were combined to develop many slope fail- area. Hs€u (1975) proposed to study the H/L ratio
ures during the earthquake. Among them the of the soil flow geometry in which H stands for
biggest slope failure occurred in Tsao-Ling the height of fall of the soil mass, while L the
(Fig. 14), and 120 million m3 of materials flowed horizontal travel distance, H/L suggesting tan
down into the river to form a natural dam (Hung (friction angle) where the friction angle is in
2000). This slope had failed several times in the terms of the total stress. Figure 16 was drawn by
past because of heavy rains or earthquakes in collecting data from many literatures. It is impor-
1862, 1941, 1942, 1979, and 1999 (Hung 2000). tant herein, as has been pointed out by many
Noteworthy was that the 1999 event produced a authors, that the greater volume of soil mass is
big natural dam, similar to the previous event in associated with the smaller H/L and, hence, the
1941 (Kawata 1943), and a reservoir which was friction angle. Submarine slope failures are
later filled with river sediments. accompanied by even smaller friction angle.
The 2008 Wenchuan earthquake of Mw = 7.9
in the Sichuan Province of China affected moun- Instability of Slope Surface
tainous area where body of mountains has been It is often the case that a mountain slope is cov-
deteriorated by the tectonic action. Slope failures ered by a weathered unstable material. This weak
produced natural dams (Fig. 15). The problem of surface material may fall down during strong
the natural dams is the possibility of breaching shaking. Although the size is small, this type of
and flooding in the downstream area. Therefore, slope failure occurs at many places and often
it is important after big earthquakes in mountain close mountain roads, making emergency activi-
regions to find the formation of natural dams and ties very difficult.
if a dam is subject to overtopping, erosion, and Figure 17 reveals a typical seismic failure of a
breaching to drain water as soon as possible. mountain slope. While rainfall-induced slope
Empirically it is known that seepage and piping failures often start from the shoulder of a slope,
failure is rather rare in natural dams (Tabata earthquake-induced failures start from the moun-
et al. 2002). taintop. It should be stressed that slopes of
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1081
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 15 Site of
natural dam of Qing zhu
Jiang River near
Qingchuan, Sichuan
Province, China
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 16 Empirical
relationship between height
of fall (H) and horizontal
travel distance (L) of
landslide mass
(Reproduced from p. 309 of
Towhata (2008))
volcanoes are more vulnerable to failure because, lands have been constructed in a less costly man-
first, generally the volcanic slopes are made of ner. As a consequence, unfortunately, fill parts of
ashes and pumices without rigid rock mass and, land became unstable during strong earthquakes.
second, these materials are pervious and include a See Fig. 19. Because residents do not have engi-
big amount of water. The volcanic slope in neering knowledge, quality assessment of resi-
Fig. 18 started its failure from the top. dential lands is recommended as an urgent
Another problem of slope instability has necessity.
emerged in residential land developments in
hilly areas where cut and fill is a common way Long-Term Effects of Earthquake Shaking
of construction. Being different from public and on Slope Instability
industrial projects, individuals cannot always It is noteworthy that strong earthquake shaking in
afford reliable but expensive safety measures. mountainous regions may initiate long-term
As a business practice, therefore, residential instability of slopes and debris-flow hazards.
1082 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 17 Seismic failure in
the surface of mountain
slope (Sichuan Province in
China)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 18 Failure of slope of
volcanic deposits, Costa
Rica, 2008, after the
Cinchona earthquake of
Mw = 6.1
The slope in Fig. 20 seismically collapsed in was relieved. Similarly, the slope behind the city
1707 (120 million m3) and, thereinafter, has pro- of Muzaffarabad in northern Pakistan became
duced debris flows during many heavy rainfalls. unstable after the 2005 earthquake of
It seems that the strong shaking disturbed the Mw = 7.6; see Fig. 21.
entire mountain body and the rock mass The second type of long-term effect is caused
became subject to disintegration more easily by a huge deposit of debris in mountain valleys.
after a huge mass of rock fell down and stress Figure 22 shows a huge mass of debris in a valley.
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1083
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 19 Seismic instability
of residential land in Sendai
City after the 2011 gigantic
earthquake of Mw = 9 in
Japan
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 20 Ohya
slide site in Shizuoka,
Japan, which became
unstable after gigantic
collapse caused by the 1707
earthquake of
magnitude = 8 or more
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 21 Seismically
disturbed slope behind
Muzaffarabad, Pakistan
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 22 Huge
debris deposit in a valley as
a consequence of slope
failures during an
earthquake in Sichuan
Province of China
is classified into normal, reverse, and strike-slip power of fault action overwhelmed the resistance
types. Those types of fault depend on the local or of a massive concrete structure. As such, fault is
regional tectonic stress conditions. Traditionally generally perceived as a formidable natural disas-
fault has been a target of scientific study and has ter. In contrast to this idea, Fig. 24 shows the
not been investigated for engineering purposes. situation around the fault displacement in Taiwan
Recently, however, it attracts more engineering as well. Noteworthy is that the damage was lim-
concern because of the increasing demands for ited within the area of ground distortion, and the
safety. houses out of this distortion were not affected.
This implies that the essence of fault-induced
Damage Examples damage lies in the substantial distortion of
Figure 23 illustrates the destroyed shape of a ground, while the intensity of shaking is not so
concrete gravity dam in Taiwan. The energy and significant as may be imagined. Further, Fig. 25
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1085
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 24 Limited damage
in a fault area (1999
Chi-Chi earthquake)
1086 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 25 Survival of rigid
house resting on strike-slip
type of ground distortion
(1995 Kobe earthquake)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 26 Prohibition of
land use along fault
(Yokosuka, Japan)
foundation in Fig. 5 and this soft structure. concrete key block is installed between them in
The shield tunnel for the subway of Osaka order to prevent possible leakage of water after
City is made of more ductile iron segments at fault rupture (Hatton et al. 1991).
the crossing of the Uemachi fault (Azetori
et al. 2006). The Shin-Kobe Station is situated Coseismic Subsidence and Uplift
on a normal fault, and the station building is Because earthquakes are induced by rupture
divided into three parts without structural con- and distortion in the earth crust, vertical dis-
nection among them so that a possible fault placement often remains and affects the soci-
movement may not destroy the building. The ety. Figure 28 is the result of the subsidence
gravity-type Clyde Dam in New Zealand is of the earth crust (coseismic subsidence)
divided into two parts across the fault line, and a after the 2011 gigantic earthquake in Japan.
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1087
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 28 Coseismic
subsidence in Mangoku-
Ura, Japan
1088 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 29 Coseismic uplift
in Kisagata, Japan (former
lagoon is rice field today)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 30 Deposit of sand
ejecta (Urayasu City on
March 13)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 31 Subsidence and
tilting of building with
eccentric center of gravity
(Niigata, 1964)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 32 Falling of Showa
Bridge in Niigata, 1964
buildings, and hence subsoil liquefaction easily structures from liquefaction problems by densifi-
induced this damage. Figure 32 depicts the falling cation, installing drains and grouting.
down of Showa Bridge. Although the exact cause Earthfill dams with insufficient compaction
of this damage is not yet known, liquefaction in are subject to liquefaction as exemplified by the
the riverbed somehow affected the stability of the collapse of the Lower San Fernando Dam in
bridge foundation. After 1964, much effort was California in 1971 caused by the San Fernando
made to understand the mechanism of liquefac- earthquake of Mw = 6.6 (Seed et al. 1975); see
tion damage and to develop preventive measures Fig. 33. Moreover, tailings, which are a waste
as well as assessment of liquefaction risk. By from mining industries, are prone to liquefaction
1980, it had become possible to protect important because it is disposed into reservoir water, its
1090 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 33 Liquefaction-
induced slope failure of
Lower San Fernando Dam,
California, USA, in 1971
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 34 Liquefaction of
mine tailings in
Mochikoshi, Japan, caused
by the 1978 Izu-Ohshima-
Kinkai earthquake of
MJMA = 7.0
grain size is small, being of silty size, and it consolidation settlement in the ground around a
deposits softly in water to form loose water- pile-supported building (Fig. 35). Accordingly,
saturated cohesionless subsoil. Figure 34 shows lifeline connections are damaged.
an example of liquefaction in a tailings dam.
Although end-bearing pile is an effective pro- Horizontal Displacement of Liquefied Subsoil
tection measure of buildings from subsurface liq- Those preventive measures that were developed
uefaction, it is not perfect. Because water and before 1980 were suitable for important struc-
sand are ejected after liquefaction (Fig. 30), the tures for which soil improvement was financially
volume of subsoil contracts, leading to feasible. Since the 1980s, liquefaction problem of
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1091
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 35 Elevation
difference between pile-
supported building and
surrounding liquefied
ground (Urayasu 2011)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 36 Buckling and
floating of water pipeline in
Dagupan, the Philippines
less expensive structures has attracted concern. Figure 36 indicates a water pipeline after
For example, a lifeline network cannot improve extensive liquefaction caused by the 1990
soil conditions although its possible liquefaction Luzon earthquake of surface-wave magni-
damage affects the entire community signifi- tude = 7.8. The buckling of this pipeline was
cantly. Accordingly, an idea emerged in which caused by compressional deformation of the
the liquefaction-induced damage or deformation local subsoil. Whether compression or extension,
should be reduced to an acceptable extent. This the horizontal deformation and displacement of
idea has developed to the seismic performance- liquefied ground exert serious effects on under-
based design in more recent times. ground structures. Hamada et al. (1986) found a
1092 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 37 Devastated
gravity quay wall in
Kobe, 1995
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 38 Cracks parallel to
quay wall in Nishinomiya
Harbor after 1995 Kobe
earthquake
liquefied gentle slope to have moved downward many cracks parallel to the quay wall line in
and destroyed gas pipelines. Kawamura Fig. 38.
et al. (1985) reported bending failure of a con-
crete pile foundation in Niigata where subsoil Liquefaction Problems in the Twenty-First
translated toward the river and the associating Century
lateral displacement of a surface building gener- The gigantic earthquake in 2011 in Japan (Mw =
ated significant bending moment in the pile. Fur- 9.0) and the sequence of seismic events in New
thermore, in 1995, gravity quay walls moved Zealand in 2010 and 2011 caused many liquefac-
toward the sea in Kobe and other harbors where tion problems. They are characterized by damage
backfill soils liquefied (Figs. 37 and 38). The in relatively inexpensive structures such as per-
lateral expansion of backfill is evidenced by sonal houses, lifelines, and river levees. It is the
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1093
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 39 House
damage in Christchurch
in 2011
case therein that the presently available mitiga- improvement under an existing structure is
tion technologies have reduced the liquefaction expensive and time-consuming.
problems of important structures for which finan- River levee is prone to liquefaction because it
cial resource is available for soil improvement. In is often situated upon water-saturated loose sand.
contrast, individuals cannot afford the cost of Formerly it was supposed that earthquake resis-
perfect soil improvement, which may be too con- tant design is not always necessary in levees and
servative for a small house. Moreover, lifelines that seismic damage of levees should be restored
and river levees do not have sufficient financial within a short time prior to the next flooding.
resources either for soil improvement over their Probabilistically earthquake and flooding are
long length. Thus, liquefaction problem is still unlikely to occur at the same time. Importantly,
important today. levees allow a certain extent of seismic deforma-
Figure 39 illustrates a house that tilted because tion and do not require a strict control of safety.
of subsoil liquefaction during the 2011 Christ- Such an idea started to change in the 1990s.
church earthquake of Mw = 6.2 in New Zealand. Figure 42 illustrates the damaged shape of Yodo
Although the damage may not appear serious, River levee in Osaka City after the 1995 Kobe
minor tilting can cause headache and dizziness earthquake. See sand ejecta on the ground surface
to residents. The Christchurch officials declared as an evidence of liquefaction. Because this site is
liquefaction-prone area where all the residents close to the sea where high tide occurs twice a
were advised to move to safer places. Figure 40 day, a significant subsidence of this levee could
shows liquefaction-induced tilting and subsi- have led to overtopping of water and “flooding”
dence of a low building resting on a shallow in the back area that was lower than the average
foundation without pile. Figure 41 demonstrates sea level. Since then, the importance of seismic
liquefaction in a recent artificial land in Itako, design of river levees has been recognized under
Japan, in 2011. Being different from Christ- certain circumstances.
church, residents in Japan wish to restore their During the gigantic earthquake in 2011 in
houses in order to continue to live, while improv- Japan, liquefaction damage occurred in many
ing the subsoil condition. Apparently, soil river levees that were located on clayey subsoil
1094 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 40 Liquefaction in
building foundation in
Christchurch
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 41 Liquefaction in
residential area of
Itako City
which was unlikely to liquefy (Sasaki et al. 2012). From the engineering viewpoint, this kind of
It was interpreted that the sandy body of the levee liquefaction revealed such problems as identifi-
caused consolidation settlement in the clayey cation of liquefaction-prone sites, improvement
subsoil, got saturated with water, and developed of vulnerable soils, and assessment of seismic
liquefaction (Fig. 43). Figure 44 illustrates an performance (assessment of seismically induced
example of this type of liquefaction in which the deformation), all at reasonable costs.
original ground surface has no liquefaction in Buried lifelines are subject to liquefaction
contrast to the significant distortion of the levee. problems as well. In particular, liquefaction in
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1095
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 42 Liquefaction
effects on Yodo river levee
at Torishima of
Osaka, 1995
Geotechnical Earthquake Engineering: Damage Mechanism Observed, Fig. 44 Example of liquefaction dam-
age in a river levee resting on clayey subsoil (Naruse River, Japan, after the 2011 earthquake)
1096 Geotechnical Earthquake Engineering: Damage Mechanism Observed
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 45 Floating of water
pipe caused by liquefaction
(Itako, Japan, 2011; by
Prof. J. Koseki)
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed, Fig. 46 Two-
meter floating of manhole
caused by liquefaction
(Urayasu City, 2012)
the backfill soil frequently occurs because com- problem of sewerage pipes is more important
paction of the backfill soil around a pipe in a than that of water supply pipes because the for-
small trench is not easy. Figures 45 and 46 show mer is embedded at deeper elevation below the
floating of a pipe and a manhole. Liquefaction groundwater table than the latter and also because
Geotechnical Earthquake Engineering: Damage Mechanism Observed 1097
Geotechnical
Earthquake Engineering:
Damage Mechanism
Observed,
Fig. 47 Disconnection at
sewerage pipe joint and
flow-in of liquefied sand
(Urayasu City, 2011)
the sewerage water is not pressurized and flows in structures as lifelines, river levees, and residential
accordance with the carefully designed slope of houses. Because people are demanding a higher
pipes; minor change of pipe slope makes the flow level of seismic safety, engineers have to develop
difficult. Figure 47 demonstrates one type of liq- methodologies by which their lives and proper-
uefaction damage in which a pipe joint was dis- ties are protected from earthquake effects at
connected by large deformation, and the invasion affordable costs.
of liquefied sand clogged the water flow.
Cross-References
Summary
▶ Analysis and Design Issues of Geotechnical
This section reviewed the history of earthquake-
Systems: Rigid Walls
induced damage in ground and geotechnical
▶ Liquefaction: Performance of Building Foun-
structures, focusing on local soil conditions,
dation Systems
intensity of ground shaking, seismic earth pres-
▶ Seismic Actions Due to Near-Fault Ground
sure, underground structures, slope failures, fault
Motion
effects, tectonic subsidence, and liquefaction. It
is therein seen that the types of damage have been
changing with the change of our culture and life- References
styles. In particular, liquefaction and lifeline
problems were not important 50 years ago and Azetori R, Nagataki M, Izumiotani T, Kitada N (2006)
Protection of railway shield tunnel from fault displace-
fault problems are now seen as concerns. Another ment. Proc 61st Annu Conv JSCE 3:187–188
example is the ductile behavior of structures for (in Japanese)
which the conventional seismic inertial force for Hamada M, Yasuda S, Isoyama R, Emoto K (1986) Gen-
design is not good enough. It is therefore impor- eration of permanent ground displacements induced by
soil liquefaction. Proc JSCE 376(III-6): 211–220
tant for engineers to pay attention to the changing Hatton JW, Foster PF, Thomson R (1991) The influence
situation and, when a big earthquake occurs, con- of foundation conditions on the design of Clyde
duct damage reconnaissance survey in order to Dam. Transactions of 17th ICOLD, Question 66,
detect new types of damage and develop new pp 157–178
Hs€
u KJ (1975) Catastrophic debris streams (Sturzstroms)
mitigation measures. Recent earthquakes generated by rockfalls. Geol Soc Am Bull 86:129–140
revealed that earthquake engineering has not yet Hung J-J (2000) Chi-Chi earthquake induced landslides in
been helpful for such relatively inexpensive Taiwan. Earthq Eng Eng Seismol 2(2):25–33
1098 Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of
navigation information and the precise time when wavelength gives the information associated
the GNSS satellite emits the signals. As receiving with the distance between the receiver and satel-
the signals, the receiver estimates the travel time lite. Since the wavelength is about 19.0 cm in
by subtracting the broadcast time from the case of L1 carrier wave and the phase can be
received time and calculates the range between determined with an accuracy of 1–0.1 %, the
the receiver and satellite by multiplying the travel range can be estimated less than 2 mm accuracy.
time by the propagation velocity. Since an inex- This phase pseudorange enables us to estimate
pensive crystal clock is employed in the receiver the receiver position with an accuracy of a few
for decreasing its cost, the travel time is contam- millimeters. However, how many cycles exist in
inated with receiver’s clock error. Therefore, the the range is unknown. So there is an ambiguity in
range is called “pseudorange” instead of true phase pseudorange. This ambiguity is called
range. So, in the point positioning, the receiver’s “phase ambiguity” or “integer ambiguity” and
position (x, y, z) and the receiver’s clock error dt treated as unknown parameters should be esti-
are treated as unknown parameters. mated as an integer value. G
The receiver calculates the satellite position The phase pseudorange between receiver i and
from the satellite navigation information. Thus, satellite k is generally described as the following
the receiver has the information about the satel- equation:
lites’ positions and the ranges from the satellites.
So the receiver is able to estimate its position and
clock error with the least squares method if more fki ¼ rki þ lN ki þ cDti þ cDtk þ Dktrop, i Dkion, i
than four GPS satellites are available. It is said þ Dkant, i þ eki
that the accuracy of estimated position is about
a few to several meters.
where fki is the phase in meters, rki is the true
Since the velocity of light is about
range, l is wavelength, Nik is the phase ambiguity
3.0 108 m/s, for instance, 3 m error corre-
(or called integer ambiguity), Dti is clock error of
sponds to 10 ns. So, in this case, the receiver is
receiver, Dtk is clock error of satellite, Dktrop, i is
able to synchronize to the GPS system time with
tropospheric delay, Dkion, i is ionospheric delay,
an accuracy of about 10 ns. The GPS system time
and Dkant, i is error related to the antenna includ-
is derived from the atomic clocks which are
ing antenna phase center variation and multipath
implemented in the GPS ground control
noise. The true range rki is expressed as
stations and the GPS satellites. The GPS Naviga-
tion Message included in the GPS signals con-
tains parameters that allow the receiver to rki ¼ xi Xk
calculate the correction time from the GPS sys- qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2 2
tem time to the UTC (Universal Time Coordi- ¼ xi X k þ yi Y k þ zi Z k
nated) with the accuracy of sub-microsecond
order (Dana 1997). Therefore, the GPS receiver
where xi = (xi, yi, zi) is the GPS antenna position
for time synchronization provides the precise
and Xk = (Xk, Yk, Zk) is the satellite position.
UTC time.
To eliminate the clock errors related to the
receiver and satellite, the double-difference oper-
Interferometry Positioning (Static and
ation is used in the relative positioning. In this
Kinematic Positioning)
method, the relative position from a base station
In order to achieve an accuracy of a few or
is accurately estimated. The DD (double-
sub-centimeters, interferometry positioning tech-
differenced) phase pseudorange is expressed as
nique is needed. In the interferometry position-
the following equation:
ing, phase of the carrier wave is analyzed instead
of the code pseudorange. The carrier phase
(expressed in cycles) multiplied by the fkl
ij ¼ rij þ lN ij þ Dtrop, ij Dion, ij þ Dant, ij þ eij
kl kl kl kl kl kl
1100 Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of
time series. This information is expected to pro- displacements of each sensor. From the estimated
vide effective constraints to the magnitude displacements, the useful information about the
estimation. building collapses and road closures due to
A GPS-based tsunami early warning compo- the debris is investigated for the early stage of
nent, developed by the German Research Centre the rescue action after a large earthquake (Oguni
for Geosciences (GFZ) within the German- et al. 2011).
Indonesian Tsunami Early Warning System
(GITEWS) project, was installed in Indonesia
after the disastrous Indian Ocean tsunami Summary
(December 26, 2004). This component consists
of four subcomponents; the GPS real-time refer- GNSS is a very useful system in earthquake engi-
ence stations (GPS RTR) for ground motion neering because it enables us to determine the
detection and reference station for buoy GPS displacements due to earthquake at any point
receivers; GPS sensor stations at tide gauges for wherever the antenna and receiver are properly G
ground motion detection and tide gauge data cor- deployed. The estimated coseismic, postseismic,
rection; GPS sensors on buoys for sea level mea- and interseismic displacements are used for
surements and direct tsunami detection; and determining the geometry of fault plane and rup-
external GPS stations for providing external ref- ture process, estimating the magnitude of earth-
erence frame. At a large earthquake, GPS data is quake, and estimating the future earthquake
transmitted from GPS stations in near real time, potential. This helps our understanding of earth-
and displacements are determined. The ground quake. GNSS is also used as a time synchroniza-
displacements are used to select the most proba- tion system for seismograph network.
ble tsunami scenario from some thousands of GNSS receiver is employed as a displacement
pre-calculated scenarios, based on a matching sensor in various systems for disaster mitigation.
process. The sea level height is determined by A GPS-based tsunami early warning system is
analyzing the GPS data from offshore buoys to installed in Indonesia. And other tsunami early
detect tsunami directly (Falck et al. 2010). warning systems are studied. Earthquake early
warning system with RTK GPS and postseismic
Postseismic Damage Detection building-wise damage detection system are also
Application of the GPS wireless sensor network studied.
(GWSN) to postseismic building-wise damage As of April 2013, only the United States’ GPS
detection is studied. The details of GWSN are and Russian GLONASS are fully operational
described in the later section. GNSSs. In the near future, European Union Gali-
The GWSN consists of a central server and leo positioning system, Chinese Compass navi-
many sensor nodes. The sensor node has gation system, Indian Regional Navigation
a wireless communication module, microcontrol- Satellite System, and Japanese Quasi-Zenith Sat-
ler, small battery, and inexpensive L1 GPS mod- ellites will join in the fully operational GNSSs.
ule connected to a small patch antenna. At a large Therefore, the accuracy and robustness of the
earthquake, the sensor nodes collect raw GPS system with GNSS are expected to be improved
data according to the command from a central further.
server and share the collected data between the
neighboring sensor nodes. After that, the sensor
nodes determine the relative position from the GPS Wireless Sensor Network
neighboring sensor nodes onboard and transmit
the estimated relative positions to the central System Overview
server through wireless communication network. Wireless sensor network is a system in which
The central server combines the relative positions many wireless sensor nodes automatically orga-
to organize the sensor mesh and detects the nize and maintain a wireless network and
1104 Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of
transmit their data or information to each other. The electric current is 14.6 mA (3.0 V) at receiv-
The sensor node consists of a microcontroller, ing data and 17.4 mA at transmitting data. This
physical sensor, low power radio module, and kind of wireless communication module has
small battery. In the GPS wireless sensor network some sleep modes depending on usable functions,
(GWSN), an inexpensive L1 GPS receiver and the electric current reaches at a few mA in the
connected to a small patch antenna is employed deep sleep mode.
as the physical sensor (Saeki et al. 2006). Figure 1 An inexpensive L1 GPS module is connected
shows a prototype of the sensor node. to the wireless sensor module through a serial
This prototype employs a wireless communi- communication. The electric current is about
cation module having a 32 bit microcomputer, 50 mA (3 V) which is highest among the electric
128 kbyte RAM, several serial communication components of the prototype. In the prototype,
ports, as well as an RFIC (Radio Frequency a small patch antenna is employed, which is gen-
Integrated Circuit) based on IEEE802.15.4. erally never used for the precise displacement
determination because such a small patch
antenna has weak directivity and then the data is
largely affected by the multipath noises. This
noise contamination makes it difficult to resolve
the integer ambiguities.
A schematic view of GWSN is shown in
Fig. 2. This system consists of many sensor
nodes and a central server. The sensor nodes are
ordinarily in sleep mode for saving their battery
energy. They wake up, for example, every 1 min
to ask their task to the server. If the task is
Global Navigation Satellite System (GNSS) in Earth- “sleep,” they go into sleep mode again. If the
quake Engineering, Usage of, Fig. 1 Wireless sensor
node equipped with an inexpensive L1 GPS module
task is “observation,” they turn the GPS module
connected to a small patch antenna on and wait a trigger signal. After receiving the
Global Navigation
Satellite System (GNSS)
in Earthquake
Engineering, Usage of,
Fig. 2 A schematic view
of the GWSN (GPS
wireless sensor network)
Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of 1105
Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of, Fig. 3 A format of com-
pressed data of a single epoch in case of static relative positioning with short baseline
trigger signal, the sensor nodes save the required In the GWSN, quasi-static condition is G
raw GPS data on their RAM during a specified assumed, which means a relative position of sen-
time interval. After finishing data acquisition, sor nodes is invariant during the short observation
they turn the GPS module off and return back to period (e.g., 10 min). The quasi-static assumption
sleep mode. If the received task is “data collec- and dense sensor deployment enable us to omit
tion,” the sensor node sends the saved GPS data the data of code pseudorange and integer part of
to the server through wireless communication. phase pseudorange. As a result, the amount of
The sampling interval and data length can be data can be drastically decreased to 28 byte per
changed according to the command from server. epoch (Saeki et al. 2006).
The integer part of phase pseudorange is very
Data Compression Technique important information especially for kinematic
Data compression is a very important technology positioning. As mentioned above, the phase
for wireless sensor network, because a large pseudorange includes phase ambiguity as
amount of data requires long communication unknown value. The value of phase ambiguity is
time and high energy consumption which results constant as long as the GPS module continuously
in short battery lifetime. For example, if a GPS tracks the satellite’s carrier waves. If the integer
module outputs about 300 bytes per epoch part of phase pseudorange is missed, the phase
(practically the data size of original message ambiguity becomes different value and should be
depends on the type of GPS module) and determined again. This phenomenon is called
a sensor node saves the whole data for 10 min, “cycle slip” and recognized as an important prob-
the amount of data reaches to 180 kbyte. It takes lem to be solved.
about 18 s for a sensor node to transmit the data to On the other hand, in case of static positioning,
the server with a transmit rate of 100 kbps. the missing integer part can be fixed by shifting
If a single server covers 100 sensor nodes, it the corrected DD phase pseudorange f ^kl ðtÞ
ij
needs at least 30 min (=1,800 s) to collect data (appears in the following section) by integer
from all sensor nodes. (GWSN for the application value so that the difference between the succes-
of building-wise damage detection is specially sive two values is minimized. This is because the
customized so that the sensor nodes locally share corrected DD phase pseudorange can be approx-
their data between neighboring sensor nodes, esti- imated as a linear function of time with a small
mate their relative position onboard, and transmit slope in case that the relative position of sensor
only the estimated results to the server to avoid nodes is invariant and the baseline is short
transmitting large amount of data.) Since the orig- enough. Figure 4 shows an example of the
inal raw GPS data is large in amount, truly corrected DD phase pseudorange. The fluctuation
required data should be extracted. Figure 3 shows is less than 0.2 cycles after repairing the
a format of compressed data packet. cycle slips.
1106 Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of
Multi-hop Positioning
The observation equation of DD phase
pseudorange is solved with the assumption of
white noise. This assumption may be valid for
some pairs of sensor nodes existing under similar
surrounding condition because such condition
gives similar multipath noises and the noises are
canceled out in the double-difference operation.
However, for the pairs existing under different
condition, multipath noises are different and not
canceled out in double-difference operation. In Global Navigation Satellite System (GNSS) in Earth-
this case, the accuracy of float solution becomes quake Engineering, Usage of, Fig. 6 Function of suc-
worse, and the DD phase ambiguities are not cess probability with respect to the value of ratio test
(Saeki 2012)
successfully resolved as integer value. Therefore,
it is meaningful to select a good pair of sensor
nodes whose noises are similar to each other. (Fig. 5b). In this case, the sum of vectors ID1 !
Figure 5 shows a simple example of relative ID3 and ID3 ! ID2 is more likely to give better
position determination. Suppose that three sensor solution. This is a basic idea of multi-hop posi-
nodes are there and the noises are different from tioning (Saeki and Oguni 2012).
each other but the noise of ID3 has some similar- To select an optimum path quantitatively
ities to those of ID1 and ID2. This situation often among the candidates (possible sum of vectors),
happens in actual observations when many sensor success probability function P(r) is introduced. It
nodes are deployed densely. In this case, the is empirically known that higher ratio J2/J1 (= r)
relative position from ID1 to ID2 tends to be gives higher success probability. Figure 6 shows
estimated incorrectly, and the ratio J2/J1 becomes the experimental results of the success probabil-
small (as described in Fig. 5a) because the noises ity versus the ratio J2/J1. The marks represent the
are not canceled out in the double-difference results obtained from an experiment, and the lines
operation. On the other hand, the vector ID1 ! are the fitted curves using the function
ID3, and also the vector ID3 ! ID2, may be P(r) = 1.0 – bear where a and b are the fitting
estimated better and the ratio becomes larger parameters. The experiment was conducted
1108 Global Navigation Satellite System (GNSS) in Earthquake Engineering, Usage of
Cross-References
References
under an ideal condition (no obstacles over the
site), but the GPS antennas were differently fixed. Allen RM, Ziv A (2011) Application of real-time GPS to
In the first pair (exp.1) the antennas were both earthquake early warning. Geophys Res Lett
38, L16310
fixed on a flat concrete floor. In the second pair Dana PH (1997) Global positioning system (GPS) time,
(exp.2) one antenna was fixed on a flat concrete dissemination for real-time applications. Real Time
floor, but another one was attached on a concrete Syst 12:9–40
block with a height of 15 cm. This difference Dow JM, Neilan RE, Rizos C (2009) The international
GNSS service in a changing landscape of global nav-
gave slightly different multipath noises. In the igation satellite systems, J Geod 83:191–198
third pair (exp.3) one antenna was fixed on a flat Falck C, Ramatschi M, Subarya C, Bartsch M, Merx A,
concrete floor, but another one was fixed on Hoeberechts J, Schmidt G (2010) Near real-time GPS
a tripod. In this case, the multipath noises were applications for tsunami early warning systems. Nat
Hazard Earth Syst Sci 10:181–189
very different to each other. GSI HP (2013) http://www.gsi.go.jp/ENGLISH/
The success probability of path can be evalu- Hirose H, Obara K (2005) Repeating short- and long-term
ated by multiplying the success probabilities of slow slip events with deep tremor activity around the
the corresponding vectors. Bungo channel region, southwest Japan, Earth Planets
Space 57:961–972
Hofmann-Wellenhof B, Lichtengger H, Collins J (1994)
Pðr Þ ¼ Pðr 1 ÞPðr 2 Þ Pðr n Þ GPS theory and practice, 3rd edn. Springer, New York
Hudnut KW, King NE, Galetzka JE, Stark KF, Behr JA,
The optimal path is searched from numerous Aspitoes A, van Wyk S, Moffitt R, Dockter S, Wyatt
F (2002) Continuous GPS observations of post-seismic
candidates so that the success probability is deformation following the 16 October 1999 Hector
maximized. In the multi-hop positioning, Mine, California, Earthquake (Mw 7.1). Bull Seismol
Dijkstra’s algorithm is used as a search algo- Soc Am 92(4):1403–1422
rithm. This algorithm is widely used in many IGS HP (2013) http://igscb.jpl.nasa.gov/
Kato T, Terada Y, Kinoshita M, Kakimoto H, Isshiki H,
applications such as network routing protocols Matsuishi M, Yokoyama A, Tanno T (2000) Real-time
or mobile navigation systems to find the shortest observation of tsunami by RTK-GPS, Earth Planets
(or lowest const) path efficiently. Space 52:841–845
Figure 7 shows an experimental result of the Nikokadis RM, Bock Y, de Jonge PJ, Shearer P, Agnew
DC, Domselaar MV (2001) Seismic wave observa-
optimum path estimated using the multi-hop tions with the Global Positioning System. J Geophys
positioning method. In the experiment, 53 sensor Res 106(B10):21897–21916
nodes were arranged on a rooftop of a building in Oguni K, Miyazaki T, Saeki M, Yurimoto N (2011) Wire-
a grid form at 2 m spaces. The raw GPS data was less sensor network for post-seismic building-wise
damage detection. In: Proceedings of the 20th IEEE
saved for 4 min with 1 Hz sampling rate and international workshops on enabling technologies:
transmitted to the central server through wireless infrastructure for collaborative enterprises. Paris,
communication. The multi-hop positioning was pp 238–243
Global View of Seismic Code and Building Practice Factors 1109
Okada Y, Kasahara K, Hori S, Obara K, Sekiguchi S, the historical changes of codes from 1900 to
Fujiwara H, Yamamoto A (2004) Recent progress of 2013, the years in which updates were made, the
seismic observation networks in Japan - Hi-net, F-net,
K-NET and kiK-net -, Earth Planets Space 56:15–28 number of buildings these codes influence, and a
Ozawa S, Nishimura T, Suito H, Kobayashi T, Tobita M, review of the code quality itself with respect to
Imakiire T (2011) Coseismic and postseismic slip of the hazard of that country. Over 160 countries
the 2011 magnitude-9 Tohoku-Oki earthquake. Nature and nations have some form of seismic code.
475:373–376. doi:10.1038/nature10227
Saeki M, Oguni K (2012) Multi-hop positioning, relative However, the quality, extent of application, and
positioning method for GPS wireless sensor network. methodologies between seismic-resistant codes
In: PECCS2012 – international conference on perva- differ around the world. An exploration of the
sive and embedded computing and communication location of each of these code changes, in terms
systems, Rome, Italy 24–26
Saeki M, Kosaka T, Kaneko S (2006) Phase ambiguity of when they have been implemented, to what
resolution based on linear modeling of DD carrier extent they have been implemented, and which
phase. In: ION GNSS 19th international technical meet- percentage of buildings they encompass, has
ing of the satellite division, Fort Worth, 26–29 Sept 2006 been undertaken. G
Segall P, Davis JL (1997) GPS applications for
geodynamics and earthquake studies. Annu Rev
Earth Planet Sci 25:301–336
Teunissen PJG, de Jonge PJ, Tiberius CCJM (1997) The The History of Seismic-Resistant Codes
least-squares ambiguity decorrelation adjustment: its Worldwide and Their Components
performance on short GPS baselines and short obser-
vation spans. J Geodesy 71:589–602
Yue H, Lay T (2011) Inversion of high-rate (1 sps) GPS Seismic-resistant building codes traditionally
data for rupture process of the 11 March 2011 Tohoku started being implemented during reconstruction
earthquake (Mw 9.1). Geophys Res Lett 38:L00G09. following major earthquakes of the past, includ-
doi:10.1029/2011GL048700
ing 1755 Lisbon, 1880 Luzon, and 1908 Messina.
Although no country had a formal seismic-
resistant code in 1900, some countries such as
Italy, Portugal, the USA, and other locations
Global View of Seismic Code and had influences from historical earthquakes,
Building Practice Factors which aided the building styles of the time
being more seismic resistant. This was also the
James Edward Daniell
case with colonial building styles in Africa.
Geophysical Institute, Karlsruhe Institute of
Using the components of lateral earthquake
Technology, Karlsruhe, Germany
forces, ductility and drift and building height,
these were slowly implemented into design, as
explained in the previous chapter of G€ulkan and
Synonyms
Reitherman.
The first seismic-resistant codes simply
Building design codes; Construction practice;
applied a seismic coefficient and building height
Earthquake construction; Seismic-resistant codes;
limits, such as those implemented within Italy
Worldwide earthquake standards comparison
(1910, 1917), Costa Rica (1914), Turkey and
Japan (1923), the USA (1906, 1927), and other
locations pre-1930s. It should be noted that,
Introduction
pre-1930, some municipalities and cities applied
different seismic-resistant building codes or prac-
This entry looks at a review of seismic-resistant
tices, such as building from wood after the 1848
codes for each country in the world by examining
Wellington earthquake (Beattie and Thurston
2006) or after the 1770 Haiti earthquake.
The online version includes code changes and code types Lateral loads were often included with
through time. the wind, snow, and gravity loads in analysis.
1110 Global View of Seismic Code and Building Practice Factors
Post-1930, the difference of the response spec- the amount of displacement at different limit
trum to various structural periods was looked at, states, has been determined to be increasingly
with the influence being that the lateral forces important, despite both methods looking at the
would be changed depending on the fundamental location of plastic hinges (shear strength of mem-
period of the building. Elastic analysis was, how- bers being greater than that of the shear due to
ever, still undertaken, as opposed to elastoplastic flexural strength) and hence looking at the dis-
analysis (ductile analysis). placement ductility of the components of the
Only in the 1960s were ductility consider- building or structure. The previous chapter of
ations taken into account in the form of force- G€ulkan and Reitherman provides a useful back-
reduction factors, given that ductile structures ground to the implementation of such parameters.
were able to survive ground shaking and thus These difficulties resulted in displacement-
inertia forces many times greater than elastic based design, rather than force-based design,
structures. These force-reduction factors were being undertaken in order to construct buildings
placed slowly into the seismic-resistant codes in safely (Priestley et al. 2007). Displacement-based
different countries from the 1970s to 1990s; how- design is still generally not used in seismic codes
ever, greater research in the field showed the need but can be applied as a valid alternative where
for analysis into the effects of damage potential wanted. A comment to this effect has been incor-
as a result of displacements rather than strength porated in the NZS code 1170 Part 5 (Standards
alone. New Zealand 2004).
There were also two types of building Thus, there are now two main types of seismic
analysis undertaken: force-based design and codes used worldwide. The first, strength-based
displacement-based design. Force-based design design, relies on a design base shear with force-
uses a seismic coefficient (C) and a seismic based design generally based on historical earth-
weight (W) calculated to give the design seismic quake hazard maps, without any view as to the
forces (V): performance of the structure in different events.
The other, performance-based design, still uses
V ¼ Cs W, where these uniform spectra metrics from seismic zon-
ing and traditional force-based design but adds in
Cs ¼ Sa=ðReduction Factor=Importance FactorÞ
performance criteria that need to be satisfied at
more than one earthquake level. In this way,
The seismic coefficient is made up of the performance levels are also defined differently
design spectra of the earthquake horizontal and for schools versus normal buildings, as a different
vertical components calculated from various return period can be defined for the performance,
parameters detailed in Table 1 below, in combi- thus designating different displacements or
nation with a reduction factor corresponding to interstory drifts required. Better levels of
the global ductility capacity and the inherent performance-based design are creeping slowly
overstrength in the lateral force-resisting system. into revisions of seismic-resistant codes world-
Using this system generally works; however, it wide, with life safety and damage limitation of
relies on the fact that strength is being used to buildings at various corresponding levels of
control damage and that strength and stiffness are earthquake motions being implemented and
independent. In reality, these two are interlinked. even displacement-based design (Priestley
Force-reduction factors assume that the ductility 1993) being included.
demand is the same for each type of structure Table 1 represents the key elements of a mod-
unless changed; however, each building gener- ern seismic-resistant code, which then will be
ally acts differently with respect to ductility. The summarized for the purposes of the seismic
distribution of strength in a building, including code index developed.
Global View of Seismic Code and Building Practice Factors 1111
Global View of Seismic Code and Building Practice Factors, Table 1 The criteria examined for each of the
seismic-resistant codes within the worldwide seismic code index (1900–2012) discussed in this study
Value given out of
Element Sub-element Key components 104
Structural Material loads/ Basic principles of design as denoted by the general 30
design strengths, dead and building code in the country
method live load
Seismic Horizontal Design response spectra (Sa) generally determined by 5
actions components the ground seismic acceleration, importance factor, at
characteristic design periods, damping modification
factor. It has different shapes for design periods,
depending on these cutoffs
Seismic Vertical Generally taken as a percentage of the horizontal factor 3
actions components to create Av
Seismic Ground seismic Acceleration at which the seismic zone map coefficient Included as a
actions acceleration will occur for a return period (475 years generally) percentage of G
Can be intensity or PGA based actuality over 30
Seismic Near fault Looks at whether near-fault effects are taken into 2
actions account
Seismic Soil classification Depending on soil type, extra factors applied to 5
actions account for the increased shaking
Seismic Importance factor Structures generally split into various factors in order 5
actions to define different levels of design in terms of life
safety, e.g., hospitals given a higher level of safety
Seismic Behavior factor Reduction of seismic loads based on post-elastic 5
actions (ductility factor) behavior – depending on the material of the structural
system
Seismic Foundation factor Interaction between the foundation and the soil is 5
actions included in some cases
Existing buildings % of buildings
1–2-story building % of buildings
factor
Design Simplified Uses the fundamental mode of oscillation and looks at 5
method spectrum method equivalent seismic forces. Simplified and does not take
into account so many complexities, just using base
shear multiplied by the design spectra value and
distributing the lateral forces over the height of the
building
Also includes eccentricities, other loading details
Design Drift Drift is used to calculate seismic forces as a boundary 2
method condition
Design Dynamic response Uses dynamic response spectrum method, taking into 5
method method account multiple modes of oscillation, which has
higher quality until the sum of effective modal masses
reaches a certain percentage of the oscillating mass of
the system
Design Nonstructural Are nonstructural elements designed for? 2
criteria
Design Avoidance of Special notes with application to the earthquake load 5
criteria collapse combination, avoidance of soft-story mechanisms,
plastic hinge locations, and other combinations
Design Damage limitation Load-bearing systems, pounding taken into account, 5
criteria drift, appendages such as chimney, facades
(continued)
1112 Global View of Seismic Code and Building Practice Factors
Global View of Seismic Code and Building Practice Factors, Table 1 (continued)
Value given out of
Element Sub-element Key components 104
Foundation Foundations, Basic checks as to slope stability and other foundation 5
design retaining structures, design applications such as liquefaction hazard, shear
slopes settlement, differences for shallow and deep
foundations, etc.
Code Displacement- Displacement-based criteria in terms of the design 15
quality based design/ methodology or overengineered status, additional
assessment overall components and methodologies encompassing many
design levels for all types of structures
Which Countries Have Seismic Codes? 1992, 1996, 2000, 2004, 2008, and 2012
(IAEE 1996, 2000, 2013). In addition, there
Not all countries have seismic codes; however, has been a list of countries which have some
documentation does not exist on a global scale to form of seismic codes in the practice of earth-
see which countries have seismic-resistant codes quake hazard assessment (92 countries)
and whether they are part of law, enforced or (McGuire 1993). The “Seismic Code Evaluation”
ignored. In addition, some countries have differ- work in Central America as well as South Amer-
ences on the province-level enforcement or ica has been collected and implemented in the
adherence to such codes. database (Chin and Association of Caribbean
In order to determine the factor by which States 2003).
damage is changing and the complexity of One hundred and sixty-six nations out of two
codes, a significant review and analysis was hundred and forty-four nations have some form
undertaken to create a harmonized up-to-date of seismic-resistant code, but they are usually not
list of locations where seismic-resistant codes implemented for all styles of buildings, given the
have been considered. There were several early large number of nonengineered building styles
initiatives which attempted to undertake this by worldwide as shown in Fig. 1.
looking at only a list of some of the countries In addition, there are nations which have some
worldwide with seismic-resistant building codes, form of code; however, this has not been ratified
including the World List initiative by the IAEE and has not been implemented in practice. For
which includes details as to some seismic- earthquakes, a review of the various seismic-
resistant codes in various countries – IISEE resistant codes around the world has been
Global View of Seismic Code and Building Practice Factors 1113
Global View of Seismic Code and Building Practice Factors, Fig. 2 The seismic zonation changes for the country
of Macedonia (Salic et al. 2010)
undertaken, examining the level of base shear century. Studies such as Salic et al. (2010), as
relative to hazard, seismic zoning, and other depicted in Fig. 2 for Macedonia, or Romeo
parameters ranked in comparison to the relative (2007) for Italy show the move from seismic
hazard of a particular country versus the seismic hazard based on historic earthquakes to probabi-
hazard actually represented. listic hazard with a smoothing of hazard zona-
Many different countries have had multiple tions used in code through time.
editions, adjustments, and changes to their Shown in Fig. 3 are the number of code
seismic-resistant codes and zonations in the past changes through time for each nation.
1114 Global View of Seismic Code and Building Practice Factors
Global View of Seismic Code and Building Practice seismic-resistant and/or building code (or lack thereof)
Factors, Fig. 4 The worldwide seismic code index (level in each country regardless of enforcement and building
1) as calculated for 2012, showing the quality of the practice factor
The seismic code index ranks the quality of earthquake influences. Bilham (2013) estimates
seismic codes since 1900 in each nation, giving a that there are 1–3 billion family homes in this
score between 0 and 104, normalized to 100 as two category. However, many societies have learned
extra components were added later (Daniell to build with lightweight materials such as thatch,
et al. 2011). The criteria was set that a country thus protecting themselves from fatal roof and
with a unified national code was placed at a value wall collapses. The age distribution of buildings
of 30/100, utilizing the minimum criteria set which thus needs to be explored.
is the use of earthquake loading in some form of
design (not necessarily for all structures). The qual-
ity of this code was then objectively ranked by using Age Distribution of Infrastructure and
the above criteria elements to make an assessment Its Effect on the Number of Buildings
of the code quality at the time of implementation. Under Seismic-Resistant Codes
It can be seen from the 2012 picture of the
seismic code index in Fig. 4 that much work is A global building inventory including age and
needed solely on the writing of seismic code indi- number of stories has been built (Daniell
ces. Despite the adherence or intention to adhere to et al. 2011) using census information. The num-
the new Eurocodes in Africa and through Asia, ber of possible buildings that were built under
many nations still have not signed or enforced some form of code was then reduced, as there
codes with respect to seismic resistance. are codes where not all building typologies are
The current state of the world seismic codes included under the seismic-resistant codes. The
with respect to nonengineered buildings is also a age of buildings has also been used to give a
key concern, given the lack of seismic codes for comparison as to the influence of a seismic code
residential buildings and the age of buildings. on the vulnerable percentage of buildings.
Many of these buildings are largely vulnerable, By using the same census data and methodol-
use heavy or inappropriate materials, and are not ogy, the age distribution of infrastructure was
built based on historical earthquake or future collected, with the average shown in Fig. 5.
Global View of Seismic Code and Building Practice Factors 1115
In countries such as Australia, the average age of 475 years globally (Giardini 1999). Different
infrastructure has stayed fairly constant at around intensity-PGA relations were used in Fig. 7 to
18 years from 1950 to 2005 (ABS 1995–1996), convert PGA to MMI depending on the location
with a dip toward 1971 as building increased, and worldwide. It can be seen that most nations
is now currently at about 18 years. In terms of the should thus have some form of seismic code,
average lifetime of a building, this value is quoted and when looking at longer return periods, even
as approximately 75 years. By comparison, this more of the globe can be exposed to damaging
value is around 30 years in China but in Great earthquake intensities.
Britain is around 132 years. For European build- The percentage of structures in each country
ings, a study named “Housing Statistics in the under seismic code differs greatly when only the
European Union” used census data and studies age percentage of structures is only looked at;
in order to examine the age of buildings however, there are usually clauses that buildings
(EUROSTAT 2013). The large difference in the of one or two stories are not built to the earth-
design life of structures leads to an interesting quake code. By using household size and build-
problem when filling in the remaining countries. ing counts globally, a percentage of buildings
If there is not enough new investment in build- built under code has been estimated. The age of
ings, structures, and infrastructure, the average structures in each country is integrated against
age of buildings will increase, if there is no new the seismic code index in order to create a per-
population. As an example, Germany has had centage of structures that are definitely not built
very low population growth since 1991. The age under code. This is, of course, assuming that the
of their infrastructure increased from 23.7 years age percentage of the buildings is the same over
to 26.9 years in 2009. Their general government the whole country, which is a reasonable assump-
buildings increased from 22.1 to 28.4 years as tion as the natural weighting is toward greater
compared to dwellings from 25.1 to 27.8 years. building in cities, which is represented by the
Figure 6 shows the average building year in each entire country values.
country as a result of the analysis. The locations of seismic code zonation were
Figure 7 depicts the locations in the world that then split from the non-code zones by calculating
should be under seismic code as per GSHAP the percentage of buildings in these zones. It
(1999)-derived intensity exceedances for should be noted that as a proxy, the MMI >6
1116 Global View of Seismic Code and Building Practice Factors
Global View of Seismic Code and Building Practice Factors, Fig. 6 The average building year in each country
Global View of Seismic Code and Building Practice exceedances for 475 years globally. Different intensity-
Factors, Fig. 7 The locations in the world that should be PGA relations were used to convert PGA to MMI,
under seismic code as per GSHAP-derived intensity depending on the location worldwide
contour was used from GSHAP, as the seismic calculate this value, the individual building
zone maps collected have not yet been digitized. codes were looked at. It is difficult to make an
However, these maps were looked at to check the assessment in some respects as to small buildings
assumptions, with major changes being made in (residential 1–2 stories, or other forms) as,
Italy (where historical codes have not covered the although these are included in some codes, there
whole of Italy) and China. are many exemptions. The Chinese code (PRC
Unfortunately, every building type is not cov- 2001/2008/2010), for instance, uses intensity
ered by earthquake-resistant codes. The next pro- zones and story heights. Most rural housing is
vision that is examined is that of small buildings only 1 story and is thus not covered under
or those not covered under code. In order to earthquake-resistant codes. In addition, many
Global View of Seismic Code and Building Practice Factors 1117
Global View of Seismic Code and Building Practice Factors, Fig. 8 The percentage of buildings in each country
G
that should have been built under a seismic-resistant code (without small building provision)
urban buildings under 4 stories in the intensity built to former Soviet codes, have become dilap-
zone 6 and 3 stories in the intensity zone 7 do not idated and thus also should be retrofitted or
need tie columns for brick buildings. There is an rebuilt.
exception for 1-story public buildings, which In general, those countries which have high
need to be built under code (2001 code, percentages in Fig. 8 are smaller countries with
Chaps. 7 and 10). In Australia, all structures are most of their building stock built in the last
included; (Standards Australia (1979, 1993, 30 years. Marked differences occur across nations
2007)) however, for domestic buildings less with different zonations and building ages across
than 3 stories, which are ductile, no provision is nations not described by such diagrams.
required. This code, being an objective-based
code, also provides a good basis for engineers to
check that the code is being adhered to. An Seismic Code Provisions: Germany
approximate percentage of buildings coming
under the provisions have been recorded. In addi- An example is presented below for Germany, in
tion, the number of informal and rural buildings is order to show the analysis procedure. The
also taken into account. DIN4149 was first implemented in Germany in
Existing buildings are also included in some 1981 after the 1978 Albstadt earthquake. The
seismic-resistant codes in terms of requiring update to this was given in 2005. Before this, a
retrofitting, such as in Argentina, Italy, EC8 few other codes had been applied on regional
(European Union), the USA, Korea, Mexico, levels, without any federal application
New Zealand, Nicaragua, and Venezuela; how- (Abrahamczyk et al. 2005). The German code
ever, it is unsure as to what extent this has been has five zones (no zone, 0, 1, 2, and 3).
enforced currently and thus has not been Only 31 % of Germany’s building stock has
included. been produced after 1981. The code actually
By combining the age of structures which applied in zones 1–3 in terms of earthquake loss
could have a seismic code and the number of is for very little of the building stock (15.3 %).
nonengineered structures not built to code struc- This means that under 5 % of possible buildings
tures in those zones, a relative percentage of have been built under some form of earthquake-
buildings in earthquake zones coming under resistant code. In DIN4149, not every building
code can be established. It should be noted that needs to have had the code change applied, due to
through Central Asia many buildings, although the premise of the number of stories (Table 2);
1118 Global View of Seismic Code and Building Practice Factors
Global View of Seismic Code and Building Practice with a high-quality code, yet the builders were
Factors, Table 2 Various limits on the requirements not not made to adhere to that code and used substan-
to adhere to earthquake codes in zones 1–3
dard materials, then the building would likely not
Earthquake Importance Maximum story stand up to the earthquake forces.
zone class height
Corruption has been identified as a key com-
1 I–III 4
ponent of earthquake losses, first through the
2 I–II 3
Bilham and Hough (2006) paper in India and
3 I–II 2
then Ambraseys and Bilham (2011) and Bilham
(2013) exploring the Transparency International
thus, even less than 5 % of buildings are applica- corruption index with respect to earthquakes.
ble to this code. State enforcement ratification of There is no doubt that engineered constructions
codes is necessary in order for the DIN4149 to be have increased the quality of building against
used. The 2005 version has been ratified by most earthquake forces, but much of this has to do
states; however, the new Eurocodes have only with the education of the individuals of that coun-
currently (as of mid-2013) been ratified by try. Education can be used as a proxy for the
Baden-W€ urttemberg, a state in the southwest of quality of understanding of building practice as
Germany (European Union 2006). well as the quality of the mathematics, engineer-
In zone 1, 89 % of buildings do not need to be ing, and drawings that govern building construc-
built to code. In zone 2, around 75 % of buildings tion. A study looking at the number of
do not need to be built to code and in zone universities offering earthquake-related subjects
3 around 48 %. This means that less than 0.9 % is another proxy that has been previously used
of buildings in Germany are built to a seismic- (Santa-Ana et al. 2012). Building error and mis-
resistant code. When focusing on the population takes increase with lack of education.
and value of the 0.9 % of buildings, this increases
to approximately 3 % of capital stock, given the
multi-story buildings, infrastructure, and impor- Summary
tant buildings in importance classes 3 and
4 protected by the earthquake code (hospitals, Seismic-resistant design codes have been
schools, industrial facilities, etc.). However, a implemented in many nations around the world
more important problem is posed by the existing over the past century. A summary has been pro-
hazard, building typology, and standard building vided as to their current application and to the
code. This formulates a similarly bleak extent of their use. Although many countries have
engineered view of the world in terms of loss implemented a seismic-resistant building code,
functions, which presents an additional difficulty the current percentage of buildings covered by
in applying a single building approach. these codes is generally less than 50 % of the
building stock.
The enforcement and implementation of
Building Practice Versus Code seismic-resistant codes is important as even though
Compliance a country may have a very high-quality code, if it
is not implemented or enforced by the government,
Just because a country has a seismic code in law, the code is useless. Given the exponential rise in
this does not mean that the seismic code will be the number of seismic-resistant codes and the qual-
enforced. Thus, building practice plays a major ity of these codes, it can be seen that more work is
role in earthquake-resistant building, and not needed in implementing seismic-resistant codes
only the quality of the seismic code is important for existing buildings and assessing residual risk,
but its enforcement, adherence, and lack of cor- and care must be taken when referring to the use of
ruption in the building. If a building was designed codes without looking at exclusion clauses.
Global View of Seismic Code and Building Practice Factors 1119
CPU GPU
example, matrix multiplication can be easily Repeating earthquakes are occurring on the
processed by a GPU due to its natural parallel similar asperity at the plate boundary. These
characteristics. Ryoo (2008) conducted a similar earthquakes have an important property; the seis-
test for matrix multiplication with a program that mic waveforms observed at the identical obser-
attained an efficiency of 91 GFLOPS on the GPU vation site are very similar regardless of their
of a GeForce 8800 GTX. Similarly, matrix occurrence time. Kawakami et al. (2013) pro-
decomposition, which is the foundation for solv- posed a high-speed digital signal processing
ing a linear equation, can also be easily acceler- method for the detection of repeating earthquakes
ated using a GPU. Volkov and Demmel (2008) using GPGPU-acceleration. They compared the
optimized a matrix decomposition algorithm on execution time between GPU (NVIDIA GeForce
the Geforce 8800 platform, and the decomposi- GTX 580) and CPU (Intel Core i7 960)
tion efficiencies of the LU, QR and Cholesky processing. In the case of a band-limited phase
methods were all greater than 180 GFLOPS. only correlation, the obtained results indicate that
Furthermore, Liu et al. (2008) proposed a gen- a single GPU is approximately 8.0 times faster G
eral finite element method solution that enables than a 4-core CPU (auto-optimization with
dynamically fast simulation of deformation on OpenMP). In the case of the coherence function
the newly available GPU hardware with CUD- using three components, the GPU is 12.7 times as
A. Their test results indicate that the GPU with fast as the CPU. This study examines the high-
CUDA enables an increase speed of approxi- speed signal processing of a huge amount of
mately four times for FEM deformation compu- seismic data using the GPU architecture. It was
tation on an Intel(R) Core 2 Quad 2.0 GHz found that the GPGPU-based acceleration for the
machine with GeForce 8800 GTX. Many com- temporal signal processing is very useful. In the
mercial finite element analysis codes, e.g., future, multi-GPU computing will be employed
ABAQUS, MSC. MARC, and ANSYS, have to expand the GPGPU-based high-speed signal
released their own GPU-based versions. The per- processing framework for the detection of repeat-
formance studies demonstrate that the GPU ing earthquakes in the future.
results in speedups in the range of 3–6 times.
In addition to the finite element simulation, The Visualization of Seismic Data
other structural numerical simulations also can The GPU can also be used for the high-
be greatly accelerated by the use of GPUs. For performance visualization of earthquake data.
example, Durand et al. (2012) simulated rock Hsieh and Yang (2011) presented a method
impact on a concrete slab using the discrete ele- using volume rendering based on a GPU to visu-
ment method (DEM), and the use of a GPU (Tesla alize the spectral information of the 1999 Chi-Chi
C2050) reportedly showed a speedup factor of 30. earthquake and the GPU-based volume-rendering
system achieved a rendering speed of 30 frames
Earthquake Detection per second (FPS). Xie et al. (2010) proposed a
Meng et al. (2012) proposed a GPU-based com- parallel visualization technique for large seismic
putation method to accelerate the detection datasets based on a GPU and CPU cooperation
algorithm for earthquakes during intensive after- platform. Their experimental results showed that
shocks or swarm sequences. This group applied the method exhibited effective real-time perfor-
parallel code to search the Salton Sea geothermal mance for a notably large amount of seismic data
field for missing earthquakes in a 90–day time (10 GB) on standard PC hardware.
window around the occurrence time of the 2010 Chen and Hsieh (2012) proposed a GPU-based
Mw7.2 El Mayor-Cucapah Earthquake. By divid- method using volume rendering to better under-
ing the procedure into several routines and stand and analyze time-varying earthquake
processing them in parallel, this group achieved ground-motion data, including acceleration,
an approximate 40 speedup for one NVIDIA GPU velocity and displacement. This group used
card compared with the sequential CPU code. CUDA to implement volume rendering (ray
1124 Graphics Processing Unit (GPU) Technology in Earthquake Engineering, Application of
CPU
Input Data for non linear
THA
GPU
THA for Building THA for Building THA for Building
#1 #2 #3
THA for Each Building
CPU
Copy Data to Host
Memory
Graphics Processing Unit (GPU) Technology in Earthquake Engineering, Application of, Fig. 2 Flow chart of
the seismic damage analysis using GPU
casting), numerical integration (trapezoidal can be used for the simulation of one build-
quadrature) and numerical differentiation ing, only a few task assignment rounds are
(centered finite difference), and achieved a required to complete the simulation of a city
speedup of approximately 100 versus a CPU. with thousands of buildings using GPU com-
puting. Thus, the computational efficiency is
Seismic Damage Prediction expected to be notably high.
Although thousands of buildings may exist in A novel solution proposed by Han et al. (2012)
an urban area, if each building is treated as a used a GPU to accelerate urban seismic damage
subtask and a proper computing model is prediction based on refined building models and
adopted, the computing workload of each nonlinear time-history analysis (THA). Figure 2
subtask is sufficiently small that it can be shows the flow chart of the Han’s program in
performed on a single GPU core. Furthermore, which the CPU reads the data, copies the data
there is little interaction between different into the GPU, and assigns the computational
buildings during an earthquake, which results tasks to each GPU core, whereas the GPU imple-
in there being little data exchange between ments the nonlinear THA for individual buildings
different GPU cores. Because there are hun- and copies the results back to the host memory for
dreds of cores in a GPU and each GPU core output.
Graphics Processing Unit (GPU) Technology in Earthquake Engineering, Application of 1125
150
CPU Double
100
Damage States
None Slight Moderate Extensive Complete
Compiler: Microsoft Visual C++ 2008 SP1 Durand M, Marin P, Faure F, Raffin B (2012) DEM-based
and CUDA 4.2. simulation of concrete structures on GPU. Eur
J Environ Civil Eng 16:1102–1114
Han B, Lu XZ, Xu Z (2012) Urban regional seismic
The benchmark cases demonstrated that the damage prediction based on GPU-CPU hybrid com-
GPU/CPU cooperative computation time is puting. In: Proceedings of the 15th world conference in
approximately 1/39 times the CPU computational earthquake engineering, Lisbon
Hsieh TJ, Yang YS (2011) Visualizing the seismic spec-
time if single precision is used, as shown in Fig. 3. tral response of the 1999 Chi-Chi earthquake using
This ratio increases to 1/21 if double precision is volume rendering technique. J Comput Civil
used. Seismic damage prediction based on a GPU Eng-ASCE 26:225–235
is shown in Fig. 4, which provides an important Kawakami T, Okubo K, Uchida N, Takeuchi N,
Matsuzawa T (2013) High-speed digital signal
reference for earthquake disaster prevention and processing method for detection of repeating earth-
mitigation. quakes using GPGPU-acceleration. EGU General
In summary, existing studies have demon- Assembly 2013, Vienna
strated that GPUs provide a highly efficient but Komatitsch D, Michéa D, Erlebacher G (2009) Porting a
high-order finite-element earthquake modeling appli-
low-cost environment for computation and visu- cation to NVIDIA graphics cards using CUDA. J Par-
alization in earthquake engineering. Therefore, in allel Distrib Comput 69:451–460
the future, GPU technology will come to be Liu Y, Jiao S, Wu W, De S (2008) GPU accelerated fast
widely used in earthquake engineering. FEM deformation simulation. In: IEEE Asia Pacific
conference on circuits and systems, Macao,
pp 606–609
Meng XF, Yu X, Peng ZG, Hong B (2012) Detecting
Acknowledgements The authors are grateful for the
earthquakes around Salton Sea following the 2010
financial support received from the National Key Tech-
Mw7.2 El Mayor-Cucapah earthquake using GPU par-
nology R&D Program (No. 2013BAJ08B02), the National
allel computing. Procedia Comput Sci 9:937–946
Nature Science Foundation of China (Nos. 51222804,
NVIDIA (2013) CUDA programming guide. http://docs.
51178249, 51378299, 51308321) and China Postdoctoral
nvidia.com/cuda/pdf/CUDA_C_Programming_Guide.
Science Foundation (2013M530632).
pdf
Okamoto T, Takenaka H, Nakamura T, Aoki T (2011)
Accelerating large-scale simulation of seismic wave
Cross-References propagation by multi-GPUs and three-dimensional
domain decomposition. Earth Planets Space 62:939–942
Ryoo S (2008) Program optimization strategies for data-
▶ Earthquake Damage Assessment from VHR parallel many-core processors. University of Illinois at
Data: Case Studies Urbana-Champaign, Urbana
Unat D, Zhou J, Cui Y, Baden SB et al (2012) Accelerat-
ing a 3D finite-difference earthquake simulation with a
C-to-CUDA translator. Comput Sci Eng 14:48–59
References Volkov V, Demmel J (2008) LU, QR and Cholesky fac-
torizations using vector capabilities of GPUs. Techni-
Abdelkhalek R, Calandra H, Couland O, Roman cal report UCB/EECS-2008-49. EECS Department,
J et al (2012) Fast seismic modeling and reverse time UC Berkeley
migration on a GPU cluster. Concurrency Comput Xie K, Wu P, Yang S (2010) GPU and CPU cooperation
Pract Exp 24:739–750 parallel visualization for large seismic data. Electron
Chen MD, Hsieh TJ (2012) Run-tme GPU computing and Lett 46:1196–1197
rendering of earthquake ground-motion data. In: Pro- Zhou J, Unat D, Choi DJ, Guest CC, Cui Y (2012) Hands-
ceedings of the IEEE 9th international conference on on performance tuning of 3D finite difference earth-
high performance computing and communication, quake simulation on GPU Fermi chipset. Procedia
pp 812–817 Comput Sci 9:976–985
H
some insight in the earthquake phenomena. Also not described here will be similar to the ones
it cannot be ignored that rough devices to detect covered here.
ground motion were known and used through
antiquity and medieval times. It is confirmed
that such a device was used on the Exeter siege An Overview of Seismic Recording
in 1549. It was used to determine the location of
enemies digging underground tunnels. The coe- To better understand the descriptions of seismic
val document says: “hearing a noise under- instruments, some basic theoretical aspects will
ground, he takes a pan of water, and by be described here. For more details, see e.g.,
removing it from place to place, came at length Wielandt 2009 or Havskov and Alguacil 2010.
to the very spot where the miners were working;
which he knew for certain by the shaking of the The Nature of Seismic Ground Motion
water in the pan.” But no application of this or Earthquakes release energy as elastic waves. The
other instruments for the recording or study of frequency content of the generated waves
earthquakes is known from the renaissance up to depends on the size (total energy) released at the
the eighteenth century. earthquake source. Larger earthquakes generate
In fact, theoretical and technical issues were larger amplitudes at lower frequencies relative to
not developed enough at that time to allow the small earthquakes. Attenuation of elastic waves
recording of ground motion. For these reasons the propagating on Earth is larger for higher frequen-
history of the development of early instruments cies. Thus, with equal amplitude, lower-
for earthquake recording is one of accumulation frequency waves propagate to longer distances
of empirical knowledge and of trial and error. than higher-frequency waves.
Only at the end of the nineteenth century were For these reasons seismographs adapted to
the key issues concerning ground motion record- record lower frequencies (<0.1 Hz) are needed
ing started to be properly defined and mastered to record large and distant earthquakes while
and the technology of instrumental ground instruments with larger amplification at higher
motion recording started to progress steadily. frequencies (>1 Hz) are needed to record near,
In order to clarify the following discussion, it small earthquakes.
should be remembered that, from a technical
point of view, two main components can be iden- From Ground Motion to Seismic Recording
tified in a ground motion recording instrument. As it will be seen in the following sections, the
The first is the transducer; this is a device that most widely used principle for seismic recording
transforms ground shaking into some other vari- is the inertia principle.
able (e.g., a voltage) that can be recorded. The A device of such type, a mechanical sensor,
second one is the recording system itself, a device has a mass with one degree of freedom coupled to
allowing to visualize and to store the output of the a frame (the ground) through a linear restoring
transducer for further inspection and analysis. force (see Fig. 1). This can be achieved with
The third item is a signal-amplifying and a spring with restoring force k. A sensor will
preprocessing system or, in other words, a signal- also have friction which will damp the swinging
conditioning system. But this was not a key issue system. The equation of motion of this system
in the early development of seismographs. can then be expressed as (Havskov and Alguacil
In the following sections a review will be 2010, p. 15)
made on the evolution of seismic recording up
to the middle of the twentieth century. As it is z€ þ 2ho0 ż þ o20 z ¼ €
u,
impossible to describe all types of seismographs,
the description will concentrate on the instru- where u is the ground displacement (the displace-
ments that defined new characteristics or new ment of the frame); z the relative displacement
pffiffiffiffiffiffiffiffiffi
evolution in the recording concepts. Instruments between the frame and the mass and o0 ¼ k=m,
Historical Seismometer 1129
Historical Seismometer, Fig. 1 Schematic illustration a “dash pot” so that the mass will not swing excessively
of the most simple seismometer, the inertial mechanical near the resonance frequency of the system. A ruler is
seismometer. A mass is held on a spiral spring. The spring mounted on the side to measure the motion of the mass
connects the mass with the ground and introduces a linear relative to the ground (From Havskov and Alguacil 2010) H
restoring force. The motion of the mass is damped using
From this equation it is seen that for high frequency is directly related to the ground motion
frequencies A(o) = 1, so the sensor will measure acceleration (Fig. 3c).
pure displacement, and A(o) diminishes rapidly Such principle can be easily applied to the
for lower frequencies, as represented in Fig. 3a. record of vertical and horizontal ground motion.
Thus, it is possible to obtain a record (z) The simplest sensor for vertical motion consists
reproducing the amplitudes of the ground motion in a mass suspended by a spring (Fig. 1). How-
using a sensor with a natural frequency lower ever, getting a natural frequency of 0.1 Hz would
than the expected frequencies of the ground require an unrealistic soft spring combined with
motion. a large mass so some tricks are needed; see
For low frequencies, A(o) = o2/o02 which is below. In the case of horizontal motion, it is
proportional to acceleration. Thus, the recording possible to avoid a spring by using gravity accel-
for frequencies lower than the sensor’s natural eration as the restoring force. The simplest way is
1130 Historical Seismometer
Historical Seismometer,
Fig. 3 Response curves of
a mechanical seismometer
and electromagnetic
seismometer with respect to
different kinds of input
signals: (a) displacement,
(b) velocity, and (c)
acceleration. The
normalized frequency is the
signal frequency divided by
the natural frequency
(corner frequency) of the
seismometer. Plots are in
bi-logarithmic scale (From
Wielandt 2009, Fig. 5.6)
to use a common pendulum (Fig. 4a). In this case until the second half of the century. Early elec-
pffiffiffiffiffiffiffiffi
o0 ¼ g=L , where g is the gravity acceleration tromagnetic sensors were connected to galva-
and L the length of the pendulum. A problem nometers which record the output current
arises when long periods are required since this generated by the sensor. Galvanometers have
would require a very large L. For a frequency of a natural period and damping as sensors so the
0.1 Hz, L = 25 m, not very practical. This prob- whole response of the electromagnetic sensor-
lem can be solved using the p “garden-gate” galvanometer is the result of combining both
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sensor
(Fig. 4b). In this case o0 ¼ g sin a=L. Theoret- elements (see Fig. 6).
ically any period can be reached with such
a configuration, but in practice, due to friction,
it is hard to get a frequency lower than 0.03 Hz. Seismoscopes: The First Known
Electromagnetic sensors were introduced Instrument
early in the twentieth century (Fig. 5). The output
voltage of such a sensor is proportional to ground The first known devices used to detect ground
velocity for frequencies higher than the natural motion are not seismographs but seismoscopes.
frequency (Fig. 3b). But good voltage recorders These devices make it possible to observe ground
for electromagnetic sensors were not available shaking, but they do not give a time history of the
Historical Seismometer 1131
may indicate the direction of shaking, time of the destroyed Pompeii). But it is also noteworthy
event, or maximum amplitude. from the seismological point of view because all
The first known device built specifically for studies on its possible detecting mechanism point
the recording of earthquakes is a seismoscope of to the use of the inertia principle, the same that
Chinese origin. It is acknowledged that it was led to the construction of most of the early seis-
designed by Zhang Heng (78–139) in 132 A.D mographs. Moreover, it gave an indication not
(Zhang Heng is the name of the Chinese philos- only of the occurrence of an earthquake but also
opher as written in the Pinyin transcription sys- of the direction of the shaking.
tem. In many old articles, books, etc., it is found Many discussions have taken place about its
as Chang Hêng or also as Choko and Tyoko, exact internal mechanism, but a consensus has
Japanese adaptations of his name). It should be not been reached. Models with common or
noted that this instrument has not been preserved, inverted pendulums and a model assuming that
and what can be seen in books and museums are the oscillating element was the external jar
recreations produced from the descriptions of the suspended from the pillar have been proposed.
instrument given by Fan Ye and others in the fifth Assuming the chronicles are correct, the most
century (Hsiao and Yan 2009). puzzling question is that only one ball was deliv-
From the translation of these descriptions ered and that it pointed in the direction of the
given by Dewey and Byerly (1969) and Hsiao epicenter. Thinking on simple mechanisms with
and Yan (2009), “The instrument was made of pendulums, it looks like more than one ball can
bronze and it resembled a wine jar with be delivered and the pointing direction may be
a diameter of eight chi (a Chinese measure equiv- toward or opposite to the epicenter. Anyway, the
alent to 2 m approx.). Inside, there was a pillar in use of more complex designs, already known at
the center and eight transmitting rods near the that time, has shown the feasibility of such
pillar. It carried out a switch ball and started the a mechanism (Fig. 7b–e).
mechanism. On the outside of the vessel there
were eight dragon-heads, facing the eight princi-
pal directions of the compass. Below each of the Seismoscopes in Europe: The Eighteenth
dragon-heads was a toad, with its mouth opened Century
toward the dragon. The mouth of each dragon
held a ball. At the occurrence of an earthquake, After Heng’s seismoscope, further references to
one of the eight dragon-mouths would release such instruments disappeared from Eastern liter-
a ball into the open mouth of the toad situated ature, and it is necessary to jump in time to the
below. The direction of the shaking determined eighteenth century in Europe to find new instru-
which of the dragons released its ball and the ments for earthquake recording. The first mention
instrument made a sound (probably the ball beat- of such an instrument was written by Jean de
ing the toad) to warn the operator.” See Fig. 7a for Hautefeuille (1647–1724) who, in 1703, pro-
a possible reconstruction. It is known that the posed a device to detect earthquakes. The back-
instrument was operational because there is ground idea was very similar to that of Heng. He
record of the detection of, at least, a 400-mile suggested to fill a bowl to the brim with mercury,
distant earthquake which was not felt at the loca- so that the shaking produced by an earthquake
tion of the seismoscope (Hsiao and Yan 2009). would cause some of the liquid to spill out. In
The Heng seismoscope is a noteworthy order to determine the direction of the shock, the
machine in itself. It has a place in the history of mercury spilling out in each of the eight principal
technology and it has been studied by many directions of the compass was to be collected in
scholars (notice its place in time with respect to cavities (see Fig. 8 for a sketch of the instrument).
Western history realizing that it was built at the Jean de Hautefeuille was not using the inertia
time of Roman emperor Hadrian and just half principle. His instrument was designed as an
a century after the Vesuvius eruption that inclinometer because he was assuming that, in
Historical Seismometer 1133
Historical Seismometer, Fig. 7 (a) A possible recon- feasibility of a mechanism delivering a ball pointing in the
struction of the outside and inside parts of the Zhang Heng direction of the epicenter due to the compression and
seismoscope. (b) and (c): the position of the eight rods and dilatation of the incident wave (Adapted from Hsiao and
the detail of one of the rods. (d) and (e): detail showing the Yan (2009))
Historical Seismometer,
Fig. 10 The seismoscope
of Forbes (1844). See
detailed description on
the text
is necessary to wait until the nineteenth century to pendulum (see Fig. 10). Using the same words
find the first improved recording device. The of Forbes (1844), “a vertical metal-rod AB, hav-
most interesting instrument was designed by ing a ball of lead C moveable upon it, is supported
Scottish James D. Forbes (1809–1868). Due to upon a cylindrical steel-wire D, which is capable
a series of small earthquakes felt near Comrie, of being made more or less stiff by pinching it at
Scotland, a special committee was appointed by a shorter or greater length by means of the screw
the British Association for the Advancement of E [. . .] The self-registering part of the apparatus
Science being one of its goals to get instruments [. . .] consists of a spherical segment HIK of cop-
to record the shocks. Forbes built an inverted per lined with paper, against which a pencil L,
1136 Historical Seismometer
C
D
Historical Seismometer, Fig. 11 A sketch of the seismoscope of Mallet. The image of a crosshair in C is reflected
from the surface of mercury in the basin B and viewed through a magnifier, D (From Dewey and Byerly 1969)
inserted in the top of the pendulum-rod, is gently Another interesting point about Forbes pendu-
pressed by a spiral spring. The marks thus traced lum was his deliberate use of “magnification”
on the concave surface indicate at once the direc- through the increase of the distance of the writing
tion and maximum extent of the pendulum’s pencil to the mass. It is the first time a reference to
vibration.” this issue is found. Anyway, the real performance
He used deliberately the inverted pendulum of the instrument was somehow disappointing
because he knew this was an easy way to obtain because it just recorded three of the sixty earth-
long free periods with small dimensions. It is quakes felt. From our present knowledge, it is
worth pointing out that Forbes was probably the likely to assume that friction was too high.
first person to study, from a theoretical point of During the same years, Robert Mallet
view and with the aim to calculate its response to (1810–1881), trying to calculate the velocity of
ground motion, the effect of an earthquake on seismic waves, used a totally different
a pendulum. But Forbes assumed that an earth- seismoscope consisting of a mercury container
quake was just a lateral displacement of the and an eleven-power magnifier. Mallet looked
ground with constant speed, and therefore, the through at the image of a crosshair reflected at
results did not give him valuable information the surface of mercury (see Fig. 11). This is
about the best settings for his instrument. As a sophisticated version of the “pan” used
Dewey and Byerly (1969) pointed out, Forbes at the Exeter siege and specifically designed to
“desired a long period in order that the pendulum measure a propagation time, there being no inter-
remained stationary as the Earth moved beneath est in any estimate of the amplitude of the ground
it. He clearly wanted to measure ground displace- motion. In the following years this instrument
ment in an earthquake. However, [. . .] we can’t was to be improved with the addition of photo-
be sure if Forbes knew that a long-period pendu- graphic recording transforming it into an almost
lum would function as a displacement meter for real seismograph without using the inertia
very-short-period oscillations of the ground.” principle.
Historical Seismometer 1137
Historical Seismometer, Fig. 12 Seismoscope by Cavalleri. He studied the frequency contents of seismic waves by
comparing the amplitudes of oscillation of the different pendulums with different periods (Adapted from Cavalleri 1858)
In Italy, Giovanni M. Cavalleri (1807–1874) recorders of this instrument, not the least because
designed and built several seismoscopes. the existing copies are different. Of specific inter-
Cavalleri constructed an instrument with six est, the seismoscope for detecting vertical motion
short pendulums of different periods, each of consisted of a conical mass hanging on a spiral
which traced the record of its motion in fine spring. The mass was suspended just over a basin
powder (Fig. 12). He used them to study the of mercury. When a slight motion caused the tip H
frequency contents of earthquakes and concluded of the cone to touch the mercury, an electric
(wrongly) that a range of frequencies from two to circuit was completed, which caused a clock to
four cycles per second would be “sufficient to stop, indicating the time of the shock. Horizontal
embrace every undulation occasioned by any motion was detected with common pendulums,
earthquake.” Anyway, the Cavalleri studies whose swinging completed the same circuit as
showed that the ground motion during an earth- that completed by the mass-spring seismoscope.
quake is a composition of oscillatory motions In this case, a circular narrow channel filled with
with different frequencies, and it is the first mercury was placed around the bottom of the
attempt to make this kind of analysis. oscillating mass. It was possible to reduce the
It is also around these years that the first distance between the mass at rest and the mercury
devices to record vertical ground motion are to tens of millimeter.
made and that electrical circuits are used for the These types of seismoscopes had a high sensi-
detection of ground motion. The most conspicu- tivity, and their design was copied in a large
ous of such instruments was the “Sismografo number of further starting/stopping devices for
Elettromagnetico” designed by Luigi Palmieri recording apparatus as it will be mentioned
(1807–1896) in 1856. This instrument, really below.
remarkable at that time, was able to assert time, Already at the end of the century, in 1895,
direction, intensity, and duration of an earth- Giovanni Agamennone (1858–1949) introduced
quake. It was not yet a seismograph, as its name his double-action electric seismoscope (Fig. 14).
may suggest, but a collection of seismoscopes, This instrument can be considered as the final
each intended to record particular parameters of synthesis of the Italian know-how on
an earthquake (Fig. 13). Palmieri’s instrument seismoscopes, closing its evolution. It uses the
seems to have been an effective earthquake oscillatory character of the earthquake ground
detector for local shocks. From the historical motion in an elegant form. Two inverted pendu-
records, it looks like it was more sensitive to lums of different periods are set aside each other.
earthquakes than humans, being the first instru- Their suspensions bars, connected to the poles of
ment to clearly attaint this goal. Copies of this a battery, have an electric contact at the top.
instrument reached Japan and were in use (and When at rest, the contact is open through a gap
are still preserved) in Tokyo, at least from 1875 to of few tens of a millimeter. Any slight motion of
1885. the ground causes the pendulums to oscillate
It is difficult to describe with words all the (at different frequencies) and, thus, to complete
complexities of the amplitude indicators and the electric circuit. A bell rings and a clock, or
1138 Historical Seismometer
Historical Seismometer, Fig. 13 Palmieri seismo- an electric circuit activated electromagnets (C and D),
graph. The spiral spring (E) was used to detect vertical which caused the clock to be stopped, thus recording the
motion. Horizontal motions along different axes were time of the arrival of the seismic waves, and the rotation of
detected and recorded electrically by the movement of the drum (i) was started, with paper being unwound and
mercury in the U-tubes (on the left side). The closing of marked at m (From Palmieri 1873)
Historical Seismometer,
Fig. 14 Agamennone
double-action electric
seismoscope. Two inverted
pendulums, with different
free period, are set aside.
The close proximity of the
two upper extremities of the
pendulums had the effect,
in the case of a tremor, of
making them come into
contact with one another,
completing an electric
circuit (From Ferrari 1992)
Historical Seismometer 1139
any other device, may be started by the electric designed to record rotary motions was also incor-
signal. porated into the seismograph. It consisted of
a crossbar with weights at both ends, much like
a dumbbell, which was pivoted at its center of
The First Seismograph mass so as to rotate in a horizontal plane. Restor-
ing force was applied to the dumbbell by springs
A fact ignored presently is that by the middle of so that it oscillated with a period of one second
the nineteenth century, seismoscopes became (Fig. 15). A more detailed description can be
widely distributed around the world. Its presence found in Ferrari (2006).
and use other than in Italy and the UK is The instrument was not designed for continu-
documented in Cincinnati, USA; on the occasion ous recording and Cecchi arranged a seismoscope
of the New Madrid earthquakes of 1811 and to start a clock and to put into motion the record-
1812; in Tabriz, Iran, for the October 4, 1856, ing surface, a smoked glass plate, at the time of an
Persian earthquake; in Manila, Philippines, for earthquake. The recording surface would move
the June 20, 1857, event; or in several places in under the needles at a speed of one centimeter per
Germany for the June 24, 1877, Herzogenrath second for 20 s. From the time on the clock, an H
earthquake. observer arriving at the seismograph would deter-
But seismoscopes do not give much informa- mine how long before his arrival the earthquake
tion about the nature of earthquakes, and this is had occurred.
the most probable cause for the present low rele- It looks like several similar instruments were
vance of seismoscopic observations. The charac- installed at Italian observatories and, certainly,
teristics of ground shaking during earthquakes one in Manila, Philippines, but the first confirmed
continued to be unknown as well as the charac- record of this instrument was on February
teristics of wave propagation at far distances. 23, 1887, at the occasion of the large earthquake
From our present knowledge about earthquake that occurred near Nice, France (Fig. 16). This
recording, it is possible to infer that sensors points to an insensitive instrument and explains
(seismometers) were becoming adequate for the reason why such an achievement passed
their task, but recorders were not. The lack of almost unnoticed. Our present interpretation is
earthquake recordings (amplitude as a function that the friction of the recording system was too
of time, a seismogram) made it difficult to further high and, most likely, the starting seismoscope
study the earthquake phenomena. was not sensitive enough.
But around 1875 Filippo Cecchi (1822–1877) At the same time, other developments impor-
built what it is now considered the first seismo- tant for the future of seismometry were taking
graph. However at the time it passed quite place; among them were experimental and theo-
unnoticed. Unlike any of the instruments retical studies about horizontal pendulums.
discussed up to this point, the Cecchi seismo- Researchers needing to detect small gravity var-
graph was expected to record the relative motion iations on laboratory and those working in geod-
of a pendulum with respect to the Earth as esy studied the properties and behavior of
a function of time. It was based in the inertia horizontal pendulums (its specific characteristic
principle, and as described by Dewey and Byerly was the way to hold the oscillating mass, as can
(1969), two common pendulums were used to be seen in Fig. 20-3) and he also developed the
record horizontal vibrations, each one oscillating theory of such an instrument. A mirror was
in a unique direction and placed perpendicular to attached to the pendulum thereby magnifying its
each other. Their free period was 1 Hz and the motion using a light beam reflected on it.
ground motion was magnified three times by Another important theoretical development
a thread-and-pulley apparatus. A mass held on was done by Perry and Ayrton (1879), both Brit-
a spiral spring, as shown in Fig. 1, was used for ish professors working at Japan. They published
the recording of the vertical motion. A device a paper entitled On a neglected Principle that
1140 Historical Seismometer
Historical Seismometer,
Fig. 15 Cecchi
seismograph. Restoration
of the original instrument
presented at Genoa in 2000.
In the front, the two
common pendulums are
seen. Among them, the
elongated box is the place
where recording plates
were fixed (four plates for
the three translation
components and the
rotational one). The
rotational sensor is seen just
behind to the right of the
common pendulum
(Photograph by the author)
Historical Seismometer, e
Fig. 17 Drawing of the
Ewing horizontal g b
pendulum (From Ewing
1881)
h
c
a
d
e
1142 Historical Seismometer
the ground motion during an earthquake, and for Earth is elastic and that therefore elastic wave
the first time, seismologists could design their theory can be used to analyze the behavior of
instruments with some knowledge of the phe- seismic waves. Looking at the record,
nomena the instruments were to record. a curvature on the baseline can be seen. This
But an instrument for the vertical motion with curvature corresponds to the deformation caused
a similar long period as those designed for the by earth tides. The instrument was a tiltmeter but
horizontal ones was still elusive. The solution it recorded earthquakes, as gravimeters do.
came when Gray introduced a method of increas- The instrument consisted of a heavy base
ing the period of a mass-spring system by containing, inside, a horizontal pendulum sitting
attaching the spiral spring to the short arm of on a rigid frame, rotating around two bearings at
a lever and attaching the mass to the long arm the top and the bottom, each consisting of a point
of the lever. This increased the period to few pressed into a socket (Figs. 19 and 20-2). He used
seconds by increasing the effective mass of the a photographic recording. It was obtained
pendulum bob (see Figs. 20-8 and 9). Thus, even through a mirror attached to the oscillating
with many drawbacks with respect to our present frame, which reflected light from a lamp through
knowledge (not a damped instrument, large dry a cylindrical lens to a rotating drum which was
frictions), a three-component instrument was covered with photographic paper. The drum
now available. turned 11 mm in an hour. As this system reduces
friction significantly, the pendulum was only
10 cm long and carried a mass of only 42 g. It
Recording Earthquakes Globally: was usually used with free periods from 12 to 17 s
Teleseismic Recording and a static magnification of 100 (Dewey and
Byerly 1969). A time reference was obtained
Another step was needed for the seismograph to with a second fixed light trace which wrote on
be recognized as the key instrument not only for the same photographic paper. Every hour, this
the study of earthquakes but also for the study of second trace was eclipsed for 5 min. The
the Earth’s interior. This was initiated by Ernst magnification and the free period of Rebeur’s
von Rebeur-Paschwitz (1861–1895). His main pendulum were 20 times bigger than Ewing’s
research area was geodesy and, specifically, the pendulum. This instrument can be considered as
lateral deformations of the Earth’s crust. To mea- the first approximation to what was needed to
sure the deformations he built sensitive horizon- record teleseismic events.
tal pendulums to measure slight changes in the The original instruments were, for many
direction of the vertical gravity. On April years, assumed lost, but recently original frames
17, 1889, two of these instruments, located in and other parts have been found by Fréchet and
Potsdam and Wilhelmshaven, recorded simulta- Rivera (2012).
neously an anomalous signal that, from the news-
papers information, he related to a large
earthquake which had occurred in Japan. The Seismograph Evolves
He published a note about this observation in
Nature that has become one of the milestones of It is possible to identify, in the last quarter of the
observational seismology (Rebeur-Paschwitz nineteenth century, three main groups or
1889). The accompanying published seismic “schools” of seismograph designers and even to
records have been reproduced many times (here assign nationality to each one: the Italians, with
also, see Fig. 18). The main result of his discov- their already long tradition of experimentation, in
ery was that, for the first time, the possibility to which Palmieri and Cecchi can be their best her-
record distant earthquakes was confirmed. This alds; the Anglo-Japanese, with the Milne-Ewing-
possibility had already been proposed by others. Gray trio, a rapidly progressing group character-
The observation also confirmed that the whole ized by its good equilibrium between the
Historical Seismometer 1143
Historical Seismometer, Fig. 18 Rebeur-Paschwitz seismogram of the 1889 Japanese earthquake (From Rebeur-
Paschwitz 1889)
Historical Seismometer,
Fig. 19 Photography of
the original Rebeur-
Paschwitz pendulum. The
top bearing is clearly seen.
Also seen is a hole behind
the bearing for the light
beam used for photographic
recording (From Fréchet
and Rivera 2012)
1144 Historical Seismometer
Historical Seismometer,
Fig. 20 Different
pendulum settings used for
earthquake recording. It is
possible to establish
a seismograph
classification according to
the mechanical constitution
of the transducer: (1)
conical or bifilar pendulum,
(2) rigid pendulum, (3)
Zöllner pendulum, (4)
vertical pendulum, (5)
torsion pendulum, (6)
inverted pendulum, (7)
flexure pendulum, (8)
vertical oscillator, (9)
vertical rotational oscillator
(From Batlló 2004)
theoretical and experimental approaches to the Although other principles were tested, it is
topic; and finally, with a really poor experience clear that all designers adopted the inertia princi-
of earthquakes due to the low seismicity of its ple for the design of the main instrument. All
country, the Germans, Rebeur-Paschwitz at the kinds of pendulums were tested (see Fig. 20).
front, with a geodetic background and a large Around that time (probably of Italian inven-
experience in detecting very small ground tion) the classical analog continuous seismic
motions at low frequencies. record consisting in a spiral line plotted on
The approach of each group to the seismic a paper placed on a drum was introduced. The
recording problem was almost independent and lateral translation of a drum was obtained by just
quite different. But when, within few years, the substituting its rotational axis with a screw. It was
three arrived to obtain their first valuable records, now possible to have a 24 h record on a unique
the information among the different groups trav- sheet of paper with manageable dimensions.
eled fast, and the new technical procedures of one In 1895, Giuseppe Vicentini (1860–1944) and
group were really soon implemented by the others. Giulio Pacher (1867–1900) introduced a new
Historical Seismometer 1145
Historical Seismometer,
Fig. 21 Vicentini
seismograph, with
horizontal and vertical
components as installed in
the Alicante Observatory,
Spain (Original photograph
from the IGN archive)
common pendulum specially designed to record Milne, on coming back to England, centered
near earthquakes. Now it was quite clear that his interest on global seismicity. For this objec-
signals of near earthquakes contained higher fre- tive he used the horizontal pendulum he designed
quencies than those from teleseismic events. in Japan in 1894 (Fig. 22). In this instrument he
For this reason they designed an instrument opti- adopted the photographic record and, thus, he
mized for near-earthquake recording. A common combined it with a light oscillating mass (less
pendulum of 100 kg with a free period 2.3 s than half kilograms). The free period was set
recorded on smoked paper with a speed of near 16–18 s and magnification was 6–9. He
15 mm/min. The motion of the pendulum was adopted also continuous recording, but did not
amplified and resolved into perpendicular com- adopt the laterally translating drum (he did in
ponents by levers. Static magnification was set to later versions of the instrument). For this reason
80. The recording was on smoked paper, and the early records were long rolls of photographic
following a trend common to Italian instruments paper. Paper speed was just 1 mm/min and one
at that time, they used large masses to overcome roll of 10 m lasted 1 week. The instrument, thus,
the frictions of the amplifying and inscription was not convenient for immediate analysis of
mechanisms. The good results obtained made seismograms. Even not being an instrument of
them add a vertical component. It was quite orig- outstanding performance, it is really important
inal in design and consisted of a flat spring in the history of seismology because, due to
clamped to a wall from one side and with the a proposal of Milne, it was selected by the British
oscillating mass (50 kg) at the other end Association for the Advancement of Science to
(Fig. 20-7). The free period was 1 s approx. and be deployed around the world (and, effectively, it
magnification was 130. The resulting instrument was) in a first attempt of making a worldwide
was known as Vicentini seismograph and it was network.
used at many places for near-earthquake record- By the end of the nineteenth century, the Brit-
ing up to the middle of the twentieth century ish professors in Japan returned to Europe and the
when it was finally superseded by electromag- Japanese took full responsibility for continuing
netic short-period seismographs (see Fig. 21). the outstanding seismological research in Japan.
The three components were written side by side Fusakichi Omori (1868–1923), a collaborator of
on the same continuous record. Milne, continued the design of horizontal
1146 Historical Seismometer
Historical Seismometer,
Fig. 22 Milne
seismograph installed at the
Toledo Observatory, Spain.
This instrument has a drum
recording instead of a roll,
which was common in the
early instruments (From
Batllo 2004)
Historical Seismometer,
Fig. 23 Bosch-Omori
seismograph single
component (the most
common set was a pair
recording in perpendicular
horizontal directions)
installed at the Toledo
Observatory, Spain (From
Batllo 2004)
pendulums. He turned to continuous recording on elsewhere. Still in the middle of the twentieth
smoked paper and, as needed in this case, heavier century, they were common in secondary seismic
masses (10 or more kg). In its most common stations. The Bosch factory, in Strasbourg, com-
form, Omori’s pendulum consisted of a mass on mercialized a popular version under the name
a rod, pivoting about a socket, with the mass held “Bosch-Omori” pendulums (Fig. 23). The world-
up by a flexible wire. With a free period of about wide distribution, relatively long period, and ade-
20 s, static magnifications about 10–15, and paper quate paper speed make their records still
speed of 5/10 mm/min, they were rough instru- valuable for the study of large earthquakes
ments of easy use and maintenance. These qual- which occurred in the early years of the twentieth
ities make them popular and were copied century (Batlló et al. 2004, 2008).
Historical Seismometer 1147
Historical Seismometer, Fig. 25 Record of the earth- seismograph installed at Munich, Germany, at 2,000 km
quake that occurred on April 23, 1909, near Lisbon, Por- from the epicenter. P and S waves arrival and surface
tugal, as recorded in the N-S component of the Wiechert waves are clearly identified (From Batlló et al. 2008)
reference instruments. Figure 25 shows an exam- for each component) used the Zöllner suspension
ple of a typical record obtained with a Wiechert (Fig. 20-3) and the vertical instrument used the
instrument. P and S waves are clearly identified suspension system introduced by Ewing in Japan
and the record is not different for our present (Fig. 20-9). Their masses were few kilograms. Its
seismograms. design allowed to separate the “sensing” device
from the “recording” device. From then it is
called the seismometer-galvanometer system.
The Electromagnetic Seismograph The new instrument was the fruit of careful
experimentation but also, as in the case of the
Boris B. Galitzin (1862–1916), in charge of the Wiechert instrument, a result of theoretical
Pulkovo seismic station, was, as Wiechert, study. The horizontal pendulum’s free period
deeply interested in the improvement of seismic was set around 24 s for the horizontal and 12 s
recording. In 1905 he came with a totally new for the vertical components. The galvanometer’s
instrument, the electromagnetic seismograph free period was set to the same frequency as the
(Fig. 26). The new seismograph used, instead of seismometers and both the seismometer and gal-
mechanical transducers amplifying the ground vanometer had critical damping. It was then pos-
motion with levers, an electromagnetic trans- sible to get an analytical solution to the equation
ducer. To do so coils attached to the oscillating of motion for the whole coupled system
pendulum were moving through a magnetic field seismometer-galvanometer. The peak magnifica-
created by permanent magnets. Ground motion tion was more than one thousand times and
was thus converted to an electric current and a typical paper speed was 30 mm/min. Another
optically recorded, with the help of advantage was that by using parallel and serial
a galvanometer, in a photographic paper intro- coupled resistances, both the damping and the
ducing, thus, a second oscillating element. Vis- gain of the instrument could be regulated easily.
cous damping was obtained magnetically, with The new instrument was much more sensitive
a sheet of copper attached to the oscillating mass than any previous instrument and superior to
moving in the magnetic field created by a second them in almost all aspects, except complexity of
pair of magnets. The horizontal pendulums (one recording and price. At that time photographic
Historical Seismometer 1149
Historical Seismometer,
Fig. 26 Group of Galitzin
electromagnetic
seismometers installed at
the Toledo Observatory,
Spain. The two
perpendicular horizontal
components are placed to
the left and the vertical
component to the right. The
size of the magnets is
notorious (From Batlló
2004)
recording was not affordable to any seismic sta- performance of damping oil systems as viscous
tion. While a standard mechanical seismograph damping was not satisfactory) lowering the qual-
(if properly damped) has a flat response to dis- ity of the obtained seismograms.
placement for frequencies above the natural fre- Nevertheless, the studies on the structure of
quency, the transfer function of the seismometer- the Earth progressed fast. But recording of small
galvanometer system is no longer flat (see Fig. 6) earthquakes in the near field was still deficient.
making it a bit more difficult to analyze the To get a high amplification at frequencies near
seismograms. The available magnets suffered 1 Hz and higher was not an easy task. Some
from demagnetization, and thus, the overall mag- solutions were proposed. To improve the
nification of the system changed. For these rea- mechanical transducers (the main problem was
sons, their use expanded slowly. to overcome the friction), large masses were
used. Italian designers recognized this problem
already at the end of the nineteenth century.
From Teleseisms to Near Earthquakes Wiechert himself moved this way constructing
a short-period Wiechert-type instrument with
Wiechert astatic seismograph and Galitzin’s elec- 17 t. Later, in 1922, the Quervain-Piccard
tromagnetic instruments gave a good solution mechanical seismograph was built in Switzer-
to the recording of teleseisms. They were for land. A mass of 21 t was “floating” on springs,
many years the standard instruments of the most and separate recording in three directions was
important seismic stations around the world, obtained by levers. The static magnification was
complemented at secondary stations with 1,700 and the free period was 3 s for the horizon-
a variety of instruments, many of them horizontal tal components and 1 s for the vertical
pendulums of the Omori type, as the Mainka component.
pendulums (Fig. 27), with nominal characteris- At the same time a totally different solution
tics similar to the Wiechert instruments but of was adopted in California. The result was the
easier maintenance. But overall the quality of famous Wood-Anderson instrument (Fig. 28).
the mechanical parts of these cheaper instruments It takes advantage of a torsion suspension
was not as good as those of the original Wiechert (Fig. 20-5) for the pendulum (Anderson and
(and, as pointed by McComb (1936), Wood 1925). A small mass is enough to obtain
1150 Historical Seismometer
-T
T
-D
m
m
C
C
M
S
2
1
Historical Seismometer, Fig. 28 General sketch of the Wood-Anderson seismometer. A zoom of the oscillating mass
is shown in the right part (elements T, C, and m) (From Anderson and Wood 1925)
Historical Seismometer,
Fig. 29 WWSSN Benioff
variable reluctance
seismometer for the
horizontal component. The
transducer is placed in the
central part of the
instrument between the two
mass disks (Original
photograph in the IGN
archive)
1152 Historical Seismometer
many examples. Ewing, already in 1888 in the recorded. All this knowledge about expected
UK, used his double pendulum to measure the values was already accumulated from the previ-
vibration of the new Tay Bridge during the pass- ously available observations as well as how to use
ing of railway trains. Guido Alfani (1976–1940), mechanical sensors to record ground
in 1911, with a portable seismograph specially acceleration.
designed for use in buildings, called Once specifications were defined, a review of
“trepidometer,” studied the vibration of the available technology was made. The choice was
famous Pisa inclined tower when bells were ring- to use a Wood-Anderson torsion pendulum-type
ing. Contemporaries Albin Belar (1864–1939) in seismometer (Trifunac 2009). In this case
Ljubljana and Manuel Sánchez Navarro a vertical component was also built. Problems
(1867–1941) in Granada, among others, built with the proper set of the delicate suspension
also “tromometers” to use them in engineering (in this case with four wires, as explained by
works. But all of these instruments lacked enough Trifunac 2009) required a change for a more
sensibility in the needed range of frequencies to robust pivoted and spring-stabilized system
properly analyze the obtained signals (even the already in 1933. Analog recording on photo-
frequency contents of those signals were not well graphic paper forced the choice for a triggered
defined at that time). Even more, most of them instrument. The triggering seismoscope was still
were not damped instruments. As it happened in of the Palmieri type. The resulting accelerometer
many other occasions related to the seismic (Fig. 30) had three components with a free period
recording, proper ideas already existed many of 10 Hz (to properly record accelerations at
years before its proper realization. It is worth to lower frequencies). The amplitude of the trace
point out that already in 1915, Galitzin was 4 cm for an acceleration of 0.2 g (remember
studied the possibility of using the piezoelectric that it was expected for this instrument to have
effect to build accelerometers (Fremd 1997), but a flat response to acceleration in the range of
lacking the electronic technology needed to interest). The recording speed was 1 cm/s and
amplify signals, the device was not feasible at time marks were introduced each 0.5 s.
that time. The instrument worked quite well but was
In the early 1930s several damaging earth- quite expensive for particulars (an indication of
quakes in the USA started a discussion about that is that accelerometers were not commercial-
seismic safety, and the engineers realized that ized before 1963). Thus, and mainly for engineer-
real progress in earthquake engineering ing purposes, well-calibrated seismoscopes
(an engineering branch not yet defined at that continued to be on the first line of earthquake
time) passes through detailed knowledge of engineering up to the advent of electronic sys-
accelerations in buildings and structures, but tems and digital recording, in the mid-1970s
there was no instrument capable of doing these (Fig. 31).
measurements. However, theoretical and techno-
logical studies were properly developed at that
time to describe the problem. So just 2 years Summary
passed since the starting of the Strong Motion
Program, in 1931, up to the recording of the first The development of the early stages of seismom-
accelerograms for the Long Beach earthquake of etry and the evolution of the early seismographs
May 10, 1933. and of seismic recording have been presented.
The appointed committee for the Strong Evolution of the instruments, from the first
Motion Program was very fast in determining known devices of the second century A.D. up to
that it was important to record acceleration the devices of the nineteenth century, was slow
instead of displacement and also in the interesting and with many failures, the main issue being the
range of frequencies and the expected amplitudes lack of knowledge of the real ground motion
(f < 10 Hz and amplitude up to 0.5 g) to be during earthquakes and, thus, of a clear goal.
Historical Seismometer 1153
Cylindrical
Lens
Recorder Drive
Motor
Recorder
Pendulum Starter
Longitudinal Displacement
Meter
Commutator Transverse Accelerometer
and
Circuit creater
Clock
Recorder Shutter Cable
Transverse Solenoid
Vertical Accelerometer
Standard
0 10 20 cm Longitudinal
Accelerometer
Transverse Displacement H
Meter
Recorder Shutter Cable
OF THE
UNITED STATES COAST & GEODETIC SURVEY
Historical Seismometer, Fig. 30 USCGS accelerograph. The three accelerometers are placed at the center of the
instrument. The starter pendulum is clearly seen at the top left of the accelerometers (From Hudson 1963)
Historical Seismometer,
Fig. 31 USCGS
seismoscope for strong
motion recording. This
instrument was designed in
the 1950s and continued in
use up to the time of the
digital accelerograph
(From Hudson 1963)
This lack of knowledge was also due to the lack faster. The tools for the theoretical analysis of
of proper recorders able to record continuously. seismic recording (mainly Newtonian mechan-
This was a common problem to all sciences and it ics) were already available. Electromagnetic the-
was not properly solved until the second part of ory, useful to develop electromagnetic
the nineteenth century. Once the first records transducers, was developed at the same time and
were obtained, seismometry progressed much was soon incorporated to seismometry.
1154 Historical Seismometer
Humans experience near earthquakes (felt Dewey J, Byerly P (1969) The early history of seismom-
earthquakes) and this led to the early attempts of etry (to 1900). Bull Seism Soc Am 59:183–227
Ewing JA (1881) On a new seismograph. Proc R Soc
seismic recording. But the large development of 31:440–446
seismic networks at the beginning of the twenti- Ferrari G (ed) (1990) Gli strumenti sismici storici: Italia
eth century was mainly made with the purpose of e contesto europeo. SGA Storia- Geofisica-Ambiente,
recording teleseisms. Near-field recording was Bologna, p 198
Ferrari G (ed) (1992) Two hundred years of seismic instru-
not properly addressed up to the interwar period ments in Italy 1731–1940. SGA Storia- Geofisica-
of the twentieth century. At first specific instru- Ambiente, Bologna, p 156
ments for the recording of near earthquakes were Ferrari G (2006) Note on the historical rotation seismo-
developed and later, by the demand of engineers, graphs. In: Teisseyre R, Takeo M, Majewski E (eds)
Earthquake source asymmetry, structural media and
accelerometers were designed and deployed. rotation effects. Springer, Berlin, pp 367–376
In the middle of the twentieth century, with Forbes JD (1844) On the theory and construction of
the development of the WWSSN and similar a seismometer, or instrument for measuring earth-
instruments (like the Kirnos seismographs), quake shocks, and other concussions. Trans R Soc
Edinb 15:219–228
seismic instruments gave results similar to our Fréchet F, Rivera L (2012) Horizontal pendulum develop-
present instruments. Nevertheless, sophisticated ment and the legacy of Ernst von Rebeur-Paschwitz.
seismoscopes, cheap and easy to handle, were J Seismol 16:315–343
generally deployed in seismic areas up to the Fremd VM (1997) Historical seismometers as predeces-
sors of modern instruments. Cahiers du Centre
advent of new active seismic sensors and digital Européen de Géodynamique et de Séismologie
recording which solved many of the problems of 13:91–93
the older instrumentation. Havskov J, Alguacil G (2010) Instrumentation in
earthquake seismology. Springer, Dordrecht, xii +
358 pp
Hörnes R (1893) Erdbebenkunde: die Erscheinungen und
Cross-References Ursachen der Erdbeben, die Methoden ihrer
Beobachtung. Veit, Leipzig, vi + 452 p
▶ Passive Seismometers Howell FH (1990) An Introduction to Seismological
Research: History and Development (Chapter 4).
▶ Recording Seismic Signals Cambridge University Press, Cambridge, pp 57–72.
▶ Seismic Instrument Response, Correction for Hsiao KH, Yan HS (2009) The review of reconstruction
▶ Sensors, Calibration of designs of Zhang Heng’s seismoscope. J Jpn Assoc
Earthq Eng 9(4):1–10
Hudson DE (1963) The measurement of ground motion of
destructive earthquakes. Bull Seism Soc Am
References 53:419–437
McComb HE (1936) Selection, installation and
Anderson JA, Wood HO (1925) Description and theory of operation of seismographs. Special publication
the torsion seismometer. Bull Seism Soc Am 15:1–72 no. 206. Coast and Geodetic Survey, Washington,
Batlló J (2004) Catálogo Inventario de sismógrafos iv + 43 p
antiguos españoles. Instituto Geográfico Nacional, Oliver J, Murphy L (1971) WWNSS: seismology’s global
Madrid, 413 network of observing stations. Science 174:254–261
Batlló J, Clemente C, Pérez-Blanco F, Vidal F (2004) Palmieri L (1873) The eruption of Vesuvius in 1872.
Reconstruction of a Bosch-Omori seismograph. Asher, London, p 148
Nuncius 19:701–711 Perry J, Ayrton WE (1879) On a neglected principle that
Batlló J, Stich D, Macià R (2008) Quantitative analysis of may be employed in earthquake measurement. Phil
early seismograph recording. In: Fréchet J, Mag Ser 5(8):30–50
Meghraoui M, Stucchi M (eds) Historical seismology: Rebeur-Paschwitz E (1889) The earthquake of Tokyo.
interdisciplinary studies of past and recent earth- Nature 40:294–295
quakes. Springer, Berlin, pp 379–396 Stein S, Wysession M (2003) An introduction to seismol-
Benioff H (1955) Earthquake seismographs and associ- ogy, earthquakes and earth structure. Blackwell
ated instruments. Adv Geophys 2:219–275 Publishing, Oxford, xi + 498 p
Cavalleri GM (1858) Di un nuovo sismometro collocate Trifunac MD (2009) 75th anniversary of strong motion
nel Collegio di Monza. Atti dell’I R Istituto Lombardo observation – a historical review. Soil Dyn Earthq Eng
di Scienze, Lettere ed Arti 1:34–44 29:591–606
Hyperspectral Data in Urban Areas 1155
Urhammer RA, Collins ER (1990) Synthesis of Wood- the chemical composition of the materials. How-
Anderson seismograms from broadband digital ever, some materials may present strong spectral
records. Bull Seism Soc Am 80:702–716
Wiechert E (1904) Ein astatisches Pendel höher similarities except for a narrow part of the spec-
Empfindlichkeit zur mechanischen Registrierung von trum. From this point of view, hyperspectral sen-
Erdbeben. Beitr Geophys 6:435–450 sors have the ability to detect more materials than
Wielandt E (2009) Seismic sensors and their calibration. multispectral sensors.
In: Bormann P (ed) New manual of seismological
observatory practice (NMSOP). Deutsches In the literature, several studies have analyzed
GeoForschungsZentrum GFZ, Potsdam, p 46 the spectral characteristics of material detected
by hyperspectral instruments, usually by means
of laboratory and field spectroscopic investiga-
tions (Ben-Dor 2001).
Hyperspectral remote sensing demonstrated
Hyperspectral Data in Urban Areas its effectiveness in several fields of application,
such as precision agriculture, security surveil-
Giorgio Antonino Licciardi lance, physics and astronomy, environmental
GIPSA-Lab – INP Grenoble, Grenoble Institute analysis, and mineralogy, as well as in hazard H
of Technology, Saint Martin d’Hères, France monitoring and management.
In urban environment, hyperspectral remote
sensing can be extremely useful, in particular
Synonyms for the management of pre-/post-seismic events.
In this optic, the ability of hyperspectral sensors
Change detection; Data fusion; Hyperspectral to discriminate several materials can be
images; Hyperspectral sensors; Image extremely useful in the characterization of the
processing; Material identification; Remote building type, by identifying the material used
sensing for the construction. Another important use of
hyperspectral imaging is the ability to produce
accurate geological maps of the soil.
Introduction Hyperspectral images provide a great amount
of spectral information, however, when com-
The scope of this entry is to introduce pared with multispectral images, the main limi-
hyperspectral remote sensing data and its appli- tation of this kind of data resides in their
cations in urban environment. dimensionality. Since hyperspectral data cubes
In the last decades imaging spectroscopy are large, their dimension may have a negative
(Goetz et al. 1985), commonly referred to as impact in terms of computational processing load
hyperspectral remote sensing, has become and also on the storage. The dimensionality prob-
a widely used method for identification and quan- lem limited also a proper development of
tification of surface materials in different kinds of satellite-based hyperspectral systems, since
environments, such as urban, rural, and geologi- transmission and storage of acquired data become
cal areas (Heiden et al. 2012). extremely difficult. From this point of view,
A hyperspectral sensor is able to record a high reducing the dimensionality of hyperspectral
number of spectral bands, corresponding to nar- images is one of the most important processes
row contiguous wavelength intervals. This char- for the exploitation of the information provided
acteristic permits the discrimination of different by hyperspectral.
materials based on their unique spectral charac- In the following paragraphs, different aspects
teristics, called spectral signature. In general, two of hyperspectral imaging in urban area, with
materials can be distinguished by broad and nar- a particular attention to those related to seismic
row spectral reflectance features determined by events, will be addressed.
1156 Hyperspectral Data in Urban Areas
half the wavelength of the light, then the output across all 220 channels with high radiometric
will be dark. accuracy.
CHRIS is a high-resolution hyperspectral sen-
Hyperspectral Sensors sor installed on board the PROBA satellite. Dis-
Most of the developed hyperspectral sensors are tinctive feature of CHRIS is its ability to observe
designed for airborne carriers. One of the first the same area under five different angles of view
airborne hyperspectral sensors was the AVIRIS in the VIS (visible)/NIR (near infra-red) bands.
(Airborne Visible/Infrared Imaging Spectrome- Another important feature of CHRIS is the ability
ter) developed by National Aeronautics and to be reprogrammable in terms of the number of
Space Administration (NASA). AVIRIS contains spectral bands and spectral resolution (up to
242 different detectors, each with a bandwidth of 150 spectral bands with a spectral resolution of
approximately 10 nm, allowing it to cover the 1.25 nm). This characteristic allows the acquisi-
entire range between 380 nm and 2,500 nm. tion of 62 narrow and quasi-contiguous spectral
Similar sensors were the Compact Airborne bands with the spatial resolution of 34–40 m or
Spectrographic Imager (CASI), able to acquire only 18 spectral bands with an enhanced spatial
images in 288 bands in the 380 nm–1,050 nm resolution of 18 m. H
spectral range, and the Hyperspectral Mapper PRISMA (PRecursore IperSpettrale della
(HyMap) providing 128 bands in the wavelength Missione Applicativa) is a new earth observation
region of 450–2,500 nm. The Airborne integrating a hyperspectral sensor with
Hyperspectral Scanner (AHS) and the Multispec- a panchromatic camera. The advantage of using
tral Infrared and Visible Imaging Spectrometer both sensors is to integrate the classical geomet-
(MIVIS) extended the spectral range up to ther- ric feature recognition to the capability offered by
mal infrared, allowing a wider range of the hyperspectral sensor to identify the chemical/
applications. physical feature present in the scene. The primary
The development of hyperspectral technology applications will be the environmental monitor-
for the space satellites remains difficult and very ing, geological and agricultural mapping, atmo-
expensive in terms of payload design, mainte- sphere monitoring, and homeland security.
nance, and calibration. However, these difficul- The main goal of EnMAP project is to inves-
ties have not deterred the space agencies from tigate a wide range of ecosystem parameters
finding interesting missions carrying on board encompassing agriculture, forestry, soil and geo-
hyperspectral payloads. This is the case of Hype- logical environments, coastal zones, and inland
rion developed by NASA, CHRIS (Compact waters. Thanks to the chosen sun-synchronous
High Resolution Imaging Spectrometer) orbit, each point on earth can be investigated
PROBA-1 (Project for On-Board Autonomy) and revisited within 4 days.
developed by a European consortium founded
by European Space Agency (ESA), the upcoming Hyperspectral Data Processing
PRISMA developed by ASI (Agenzia Spaziale Since hyperspectral sensors are able to measure
Italiana), and EnMAP (Environmental Mapping the radiance, one of the most important
and Analysis Program) developed by Deutschen processing steps in hyperspectral image
Zentrums f€ ur Luft- und Raumfahrt (DLR). processing is the conversion to reflectance. Usu-
Hyperion instrument, mounted on board of the ally, the techniques for retrieving surface reflec-
EO-1 (Earth Observation 1) satellite, provides tance can be roughly divided into two major
a high-resolution hyperspectral imager capable categories, relative reflectance retrieval tech-
of resolving 220 spectral bands (from 400 to niques and absolute reflectance retrieval tech-
2,500 nm) with a 30 m spatial resolution. The niques. Relative reflectance retrieval is based on
instrument covers a 7.5 km by 100 km land area the knowledge of two or more target reflectances
per image and provides detailed spectral mapping in the image. Knowing the reflectance of
1158 Hyperspectral Data in Urban Areas
a material in the image permits to define by dimensionality reduction may become a key
regression a linear equation between each wave- parameter to obtain a good performance.
length and the measured radiance. However, this These techniques are known under the general
approach is also based on the assumption that the name of feature reduction. Besides enabling an
atmosphere is uniform, resulting in significant easier storage and management of the data,
artifacts in strong atmospheric bands if ranges feature-reduction procedures can be crucial for
of elevations are present in the scene. the implementation of optimum inversion
Absolute approaches, on the other hand, per- algorithms.
mit the retrieval of surface reflectance based on Many methods have been developed to tackle
the physical principles, in which measured radi- the issue of high dimensionality of hyperspectral
ance is typically compared to radiance generated data (Serpico and Bruzzone 1994). In summary,
by an atmospheric radiative transfer model, such we may say that feature-reduction methods can
as the MODerate resolution atmospheric be divided into two classes: “feature-selection”
TRANsmission (MODTRAN) model. Usually, algorithms (which suitably select a suboptimal
the relative approach can be useful in processing subset of the original set of features while
airborne data, where the influence of the atmo- discarding the remaining ones) and “feature
sphere is not relevant and can be considered uni- extraction” by data transformation which projects
form. On the other hand, absolute approaches are the original data space onto a lower-dimensional
dominant in spaceborne images. feature subspace that preserves most of the infor-
Another common preprocessing step is mation, such as nonlinear principal component
georectification, used to relate the acquired data analysis (NLPCA; Licciardi and Del Frate 2011).
to the ground. The first analysis suggests that feature selec-
In order to make maps and relate them to tion is a more simple and direct approach com-
ground reference data, it is critical that aircraft- pared to feature extraction and that the resulting
related and ground-related distortions are reduced set of features is easier to interpret. Nev-
removed. The most common approach is to use ertheless, extraction methods can be expected to
a “rubber-sheet” stretching approach and numer- be more effective in representing the information
ous tie points between a base map and measured content in lower dimensionality domain.
image. Recent improvements in onboard naviga- Feature-selection techniques can be generally
tion information and recent software develop- considered as a combination of both a search
ment have made it possible to georectify images algorithm and a criterion function. The solution
to within a few pixels (~10 m for a 4 m GIFOV to feature-selection problem is offered by the
[Ground instantaneous field of view]) in a near- search algorithm, which generates a subset of
automated fashion. features and compares them on the basis of the
Aside from the common processing tech- criterion function. From a computational view-
niques to obtain images more interpretable, point, an exhaustive search for the optimal solu-
there are other techniques that are more related tion becomes intractable even for moderate
to the dimension of hyperspectral data. If we values of features (Siedlecki and Sklansky
consider the consistency of the hyperspectral 1988). Despite these apparent difficulties, many
data, we can easily understand the importance feature-selection approaches have been devel-
of finding a method which can transform the oped (Serpico and Bruzzone 1994). The sequen-
original data cube into one with reduced dimen- tial forward selection (SFS) and the sequential
sionality and maintain, at the same time, as much backward selection (SBS) techniques are the sim-
information content as possible. In particular, plest suboptimal search strategies: they can iden-
dataset composed of hundreds of narrowband tify the best feature subset achievable by adding
channels, besides storage and transmission, may (to an empty set in SFS) or removing (from the
cause problems in terms of complexity of complete set from SBS) one feature at a time,
processing and inversion phases. Therefore, until the desired number of features is achieved.
Hyperspectral Data in Urban Areas 1159
The sequential forward floating selection function. The related artifacts of spectral
(SFFS) and the sequential backward floating misregistration are denoted as “smile” or
selection (SBFS) methods improve the standard “frown.”
SFS and SBS techniques by dynamically chang- • Spatial nonuniformity: is the nonuniformity of
ing the number of features included (SFFS) or the spatial response within an acquired spec-
removed (SBFS) at each step. trum and is usually imaged on a detector col-
A feature-extraction technique aims at reduc- umn. This nonuniformity is represented by the
ing the data dimensionality by mapping the fea- position and shape of the spatial response
ture space onto a new lower-dimensional space. function in both the along-track and across-
Both supervised and unsupervised methods have track dimensions of a spatial pixel. The related
been developed. Unsupervised feature-extraction artifacts in the across-track dimension are
methods do not require any prior knowledge or denoted as “keystone.”
training data, even though these are not directly
aimed at optimizing the accuracy in a given clas- As for the spatial nonuniformity, the influence
sification task. The class comprises the “principal of the “keystone” effect results in a black pixel in
component analysis” (PCA, Fukunaga 1990), the image that can be easily replaced by meaning H
where a set of uncorrelated transformed features the surrounding pixels with a 3 3 moving win-
is generated; the “independent component analy- dow. On the other hand, the removal of the
sis” (ICA) (Jutten and Herault 1991), “smile” effect is not an easy task. The conse-
a computational method for separating quence of this effect is that the central wave-
a multivariate signal into additive subcompo- length of a band varies with spatial position
nents supposing the mutual statistical indepen- across the width of the image in a smoothly curv-
dence of the non-Gaussian source signal; and ing pattern. Very often the peak of the smooth
the “maximum noise fraction” (MNF, Green curve tends to be in the middle of the image and
et al. 1988), where an operator calculates a set gives it a shape of “smile” or “frown.” The effect
of transformed features according to a signal-to- of the smile is not obvious in the individual
noise ratio optimization criterion. bands. Therefore, an indicator is needed to
make evident whether or not a given image suf-
Challenges in Hyperspectral Data Processing fers from smile effect. A way to check for the
The analysis of hyperspectral images is not an smile effect is to look at the band difference
easy task. Due to the complexity of a images around atmospheric absorption
hyperspectral sensor, hyperspectral images can (760 nm). In fact, the region of red-near infrared
be affected by several problems in terms of data transition has high information content of vege-
uniformity. Any nonuniformity in the system tation spectra. This region is generally called “red
generates degrading artifacts. Since pushbroom edge” (670–780 nm) and identifies the red-edge
scanners mainly characterize hyperspectral sen- position (REP). The smile effect is acute due to
sors, in this entry we will focus on the main sharp absorption at 760 nm, which is within the
problems that afflict this kind of sensors. red-edge region, and for this reason atmospheric
In pushbroom imaging spectroscopy, the correction of smiled data will be incorrect.
image is generated in the spectral and spatial The methodologies developed so far can only
dimension simultaneously. This lead to two reduce the intensity of smile effect but cannot
main kinds of problems: remove it entirely, because during its life
a detector element can change its response; there-
• Spectral nonuniformity: is the nonuniformity fore, the knowledge and the correction of this
of the spectral response within a sensor’s spec- phenomenon became fundamental in the analysis
tral band and can be imaged on a detector row. of hyperspectral images.
This nonuniformity is typically represented by Another problem that affects pushbroom scan-
the position and shape of the spectral response ners is the so-called “striping” noise. The striping
1160 Hyperspectral Data in Urban Areas
effect consists in strong variations in the average different types of surfaces characterized by
value in the column for every band of the image. a continuous spatial change. This results in
Theoretically the elements of the detector array small urban structures mostly dominated by arti-
are identical. Thus, the detectors in a row should ficial (man-made) materials. The identification of
be well calibrated to provide the same output these materials requires an accurate analysis of
along the spatial domain, if illuminated by the the spectral characteristics of these materials.
same source. In the real world, a residual Information about the material used in the roof
nonuniformity of the electro-optical response of the buildings permits to determinate the con-
generates different transfer functions. Thermal struction technique of the building and conse-
fluctuations can cause additive small variations quently its level of vulnerability. The
in the alignment of the detector, producing a identification of the roof spectral characteristics
non-predictable noise not constant during the can be useful also for the derivation of the type of
time. Only the systematic knowledge of the gain urban structure where the building is located. For
function permits an accurate compensation of instance, the historical development of European
these fluctuations. The correction of the striping countries leads the urban structure type to be
can be carried out by means of filtering correlated with a certain time period and con-
approaches. This type of destriping involves two struction style. Thus the knowledge of the urban
filters on the original image creating two new structure type may give information about the
images, a low-pass filtered image and a high- type and age of the buildings. In particular, the
pass filtered image, and combining the results. necessary information for the evaluation of
If both a low-pass and a high-pass filter of the pre-seismic events scenarios can be resumed in
same size are run on an image and then added the detection of the building type (i.e., the mate-
together, the result will be the original image. The rial) and in the detection of the identification of
goal in choosing sizes for the two filters is to the contextual information, such as the position
create an output for each without the unwanted of each building in relation to the surrounding
striping. If the filter sizes are not the same, the ones. Analysis of the geological information of
result will not be exactly the same as the original. the soil can be useful but become extremely dif-
When the filter outputs are added together, the ficult if performed by using remote sensing tech-
striping will have been removed or subdued. niques in urban environment, since the soil tends
If the sizes of the two filters differ too much, to be covered by man-made structures.
artifacts may be introduced into the result.
From this brief overview of the principal Material Identification
preprocessing techniques, it is possible to under- Surface materials can be detected based on their
stand the amount of efforts that are necessary in spectral characteristics. In the literature there
order to extract useful and reliable information exist several works on the extraction of informa-
from hyperspectral data. tion about the material of the roofs by means of
multispectral images. However, these techniques
permit to identify a limited number of elements,
Building Vulnerability Assessment due to the low spectral resolution of multispectral
imager. For instance, multispectral images are
One of the aspects of pre-seismic event manage- not sufficient to differentiate between roofs and
ment is related to the assessment of the vulnera- other surfaces having similar spectral character-
bility of buildings in urban environment. Many istics, such as asphalt and bitumen used to seal
features that can influence the seismic perfor- rooftops. In this context hyperspectral imaging
mance of a building can be extrapolated from can be an extremely useful instrument for an
the outside and then be detected by remote sens- effective estimation of the vulnerability, identi-
ing instruments. From this point of view, urban fying more information about the material used in
areas can be considered as an ensemble of the construction of the investigated buildings.
Hyperspectral Data in Urban Areas 1161
In general, several tools have been developed Tel Aviv, Israel. Data acquired by the Digital
to match reference spectra with those measured Airborne Imaging Spectrometer (DAIS) have
by an imaging spectrometer. The most common been used by Roessner et al. (2001) to obtain
approach is based on the use of standard super- a map of urban materials in the city of Dresden,
vised classification techniques, where known Germany. In this study, a maximum likelihood
spectra are used to determine the statistical prop- classifier has been used to derive a first map of
erties of each class based on spectral characteris- pure spectral features and then used these features
tics. Examples of supervised classification to unmix the other spectra.
approaches applied to hyperspectral data are However, since the spectral variations charac-
described in McKeown et al. (1999) and Roessner terizing a certain material are caused by several
et al. (2001), where the maximum likelihood factors, such as the age and deterioration of the
classifier (MLC) was applied to map urban land material, or the color, and also by the illumination
cover. Other techniques are based on the use of angles, several aspects need to be considered.
support vector machines (SVM) (Melgani and In particular, aside from the spectral heterogene-
Bruzzone 2004) and neural networks (NNs) ity of different materials, it is important to take
(Licciardi et al. 2009, 2012). Other approaches into account the high intra-class variability of H
have been designed explicitly for the analysis of spectra corresponding to the same material.
imaging spectrometry data, such as the Spectral Moreover, in urban environment there are many
Angle Mapper (SAM; Kruse et al. 1993). objects having a size smaller than the pixel. This
In order to extract as much information con- results in mixed pixels, presenting spectra that are
tent as possible from hyperspectral images in a combination of two or more materials that are
urban environment, specific methods for auto- imaged in the IFOV (instantaneous field of view)
mated information extraction have been devel- of the sensor. Moreover, the modeling of the
oped (Roessner et al. 2001). In all these spectral mixture becomes more complex as the
methods, the comprehensive mapping of the dif- pixel size increases. From this point of view,
ferent surfaces present on urban environment can airborne hyperspectral sensors, characterized by
be obtained through a classification technique a high spectral and/or spatial resolution, permit to
(Roessner et al. 2001). obtain reliable quantitative measurements of spe-
For example, Clark et al. (2001), in their envi- cific absorption features of urban materials.
ronmental studies on the World Trade Center In terms of spectral unmixing, a number of
area after the September 11, 2001, attack, suc- approaches in the literature are based on the
cessfully applied hyperspectral remote sensing to assumption that the detected spectra are linear
map material in the surrounding area. In this combination of pure spectra. In particular, in
study, the authors used high-spatial-resolution linear spectral unmixing the mixed spectrum is
AVIRIS data acquired over the World Trade modeled as the sum of pure spectra, each
Center to map thermal sources, asbestiform min- weighted by the fraction of the material within
erals, and dust and debris from the collapse, pro- the field of view. Roberts et al. (1998) have
ducing maps of environmental contaminants that developed a general approach for spectral
are difficult to produce cost-effectively in any unmixing in which the number and types of
other way. McKeown et al. (1999) used data end-members are allowed to vary per pixel. In
acquired by the Hyperspectral Digital Imagery Roessner et al. (2001), the authors used an MLC
Collection Experiment (HYDICE) to classify to detect the pure spectra in the image and then
urban and natural materials. The authors then use them as end-members to produce the abun-
fused the classification results with a 3D model dance maps. An approach based on nonlinear
of the buildings. Ben-Dor (2001) developed an unmixing has been proposed in Licciardi and
urban spectral library using a combination of an Del Frate (2011), where a neural network
existing spectral library developed by Price (NN) has been used to quantify the abundances
(1995) and CASI data acquired over the city of of previously detected end-members.
1162 Hyperspectral Data in Urban Areas
as the independence of the distribution of EDP different levels of intensity, thus fundamentally
given the IM from any other seismological changing the nature of records that one would use
parameters that may characterize the ground at low versus high levels of the IM, then methods
motion, e.g., duration, magnitude, spectral of analysis other than IDA should probably be
shape, or the presence of a pulse indicative of employed.
near-source forward directivity. A sufficient IM
essentially captures all seismological information
needed to determine the effect of a ground motion Structural Model
record on the structure being investigated, thus
allowing unrestricted scaling in theory. Natu- For performing IDA, a (a) realistic, (b) low-to-
rally, any IM for which a seismic hazard curve medium-complexity, (c) robust nonlinear struc-
can be practically estimated will never be fully tural model should be formed. Each of these three
sufficient; thus, excessive scaling may introduce requirements comes with its own reasons.
biased estimates of response (Luco and Bazzurro For one, realism means incorporating all per-
2007). tinent sources of material and geometric
A standard choice used in the literature is the nonlinearity that are expected to arise. This
5 % damped first-mode (pseudo)spectral acceler- should include, for example, plastic-hinge forma-
ation Sa(T1) (Shome et al 1998; Shome and Cor- tion zones for moment-resisting frames, brace
nell 1999). This is generally adequate for first- buckling for braced frames, and P-Delta effects.
mode-dominated structures that do not displace Nonsimulated failure modes, such as the shear
far into the nonlinear region, as is the case of most failure of members or the brittle failure of
existing brittle or moderately ductile low/mid- beam–column joints, can be incorporated in the
rise buildings. For taller or asymmetric structures analysis a posteriori. Still, they essentially
where higher modes become important or mod- remove the model’s ability to track structural
ern buildings that exhibit significant ductility, behavior beyond their first occurrence. This
improved IM alternatives should be sought. One means that whenever nonsimulated failures are
particularly attractive option is the geometric found to have occurred, one cannot trust the
mean of spectral acceleration values at several model to provide estimates beyond that point.
periods (Cordova et al 2000; Vamvatsikos and The requirement of low-to-moderate com-
Cornell 2005; Bianchini et al 2009) or an inelastic plexity is necessary for ensuring that the compu-
spectral displacement-based IM (Luco and Cor- tationally intensive IDA will generally be
nell 2007). Both can largely alleviate the effect of confined to reasonable amounts of time. Complex
spectral shape, being able to capture the period nonlinear models can complicate the execution to
elongation characterizing ductile structures and no end, easily forcing each nonlinear analysis to
the effect of higher modes. last several hours or, worse, often forcing the
In terms of selecting ground motion records, analyst to consider reducing the number of
the general idea is to reduce the bias that a less records or dynamic runs per record employed.
than ideal IM may introduce. A general rule of This can be a tricky situation that may lead to
thumb is to utilize accelerograms that can signif- inaccurate or biased results. Thus, it is best to
icantly damage the investigated structure without strike a trade-off between model complexity and
much scaling. For modern ductile structures, this fidelity, striving to find a good balance that allows
generally means records having naturally high using a rich set of at least 20–30 ground motion
Sa(T1) values. Still, IDA is structured in such a records with six or more time history analyses
way that the same set of records is used through- for each.
out the entire range of structural behavior Finally, robustness becomes important in
(elastic–inelastic collapse). Whenever site condi- tracking structural behavior in the postelastic
tions are deemed to have different impact at region, especially beyond the structure’s
Incremental Dynamic Analysis 1167
maximum-strength point, where it starts deterio- computational cost. Due to the nearly perfect
rating and approaching collapse. To be able to computational independence of each dynamic
ensure such results, a highly reliable software analysis, IDA execution is an easily
platform should be used. IDA by nature will parallelizable problem (Vamvatsikos 2011).
drive both the model and the analysis software Hence, one can benefit from using a cluster of
to their limits, making robustness difficult to N computers to divide the total running time
achieve. Even then, typical complex finite ele- (almost) by N. Such savings make possible the
ment models that have been used successfully use of Monte Carlo–based algorithms
for elastic or mildly inelastic analysis tend to (Vamvatsikos 2014) to economically estimate
fare poorly when driven close to global dynamic any bias and additional variability introduced by
instability, being consistently less reliable than modeling uncertainty.
their simpler versions besides significantly rais-
ing the computational cost. For example, it is
often the case that distributed plasticity models Postprocessing
are numerically less stable and always more
expensive than lumped-plasticity ones. There- The results of IDA initially appear as distinct
fore, considerable attention should be paid to points, one per each dynamic analysis, in the
I
the formation and testing of the model before EDP versus IM plane, as observed in Fig. 1 for
IDA is performed, checking its behavior and sta- a nine-story steel moment-resisting frame. Linear
bility via nonlinear static pushover analyses. or spline interpolation (Vamvatsikos and Cornell
2004) is then employed to generate continuous
IDA curves, one for each individual ground
Execution motion record, shown in Fig. 2. The variability
offered by such curves is actually one of the
Performing IDA is conceptually simple. One only eye-opening features of IDA, visually
needs to take one record at a time, incrementally representing the probabilistic nature of seismic
scale it at constant or variable IM steps, and loading and the differences introduced even by
perform a nonlinear dynamic analysis each (seemingly) similar ground motions. At the
time. Start from a low IM value where the struc- highest IM level that each IDA curve can reach,
ture behaves elastically, and stop when global one can identify the characteristic flatline
collapse is encountered. The latter is defined as representing global collapse. In the example of
the occurrence of a nonsimulated failure mode or Fig. 2, this is the intensity where the maximum
the appearance of a global dynamic instability as (over all stories) peak interstory drift (the EDP of
a collapse mechanism showing “infinite” EDP choice) increases without bound for each record,
values at a given IM level. For a well-executed leading to global system collapse.
analysis and robust structural model, global The complex picture of structural response
dynamic instability manifests itself as numerical shown by the IDA curves can be significantly
nonconvergence. simplified by summarizing them into the 16/50/
Selecting constant IM steps is often the sim- 84 % fractile curves shown in Fig. 3. This is
plest but also the most wasteful approach to simply the determination of the 16/50/84 % per-
IDA. As global collapse is encountered at widely centile values of EDP at each level of the IM,
varying IM levels for each record, some ground taking horizontal cross sections of the IDA curves
motions will require twice or thrice the dynamic or, equivalently, the 84/50/16 % percentiles of
analyses than others. Using variable IM steps the IM given the EDP when vertical cross sec-
estimated via the hunt & fill procedure of tions are assumed instead. The three curves of
Vamvatsikos and Cornell (2002, 2004), offers Fig. 3 thus provide the full characterization of the
instead consistent accuracy at a predefined distribution of structural response via the central
1168 Incremental Dynamic Analysis
0.4
0.2
0
0 0.01 0.02 0.03 0.04 0.05 0.06
maximum interstory drift ratio, θmax
0.6