Вы находитесь на странице: 1из 21

Journal of Earthquake Engineering

ISSN: 1363-2469 (Print) 1559-808X (Online) Journal homepage: http://www.tandfonline.com/loi/ueqe20

The Estimation of Peak Ground-motion


Parameters from Spectral Ordinates

Edmund Booth

To cite this article: Edmund Booth (2007) The Estimation of Peak Ground-motion
Parameters from Spectral Ordinates, Journal of Earthquake Engineering, 11:1, 13-32, DOI:
10.1080/13632460601123156

To link to this article: http://dx.doi.org/10.1080/13632460601123156

Published online: 25 May 2007.

Submit your article to this journal

Article views: 3482

View related articles

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=ueqe20

Download by: [27.54.121.152] Date: 01 August 2017, At: 09:20


Journal of Earthquake Engineering, 11:1332, 2007
Copyright A.S. Elnashai & N.N. Ambraseys
ISSN: 1363-2469 print / 1559-808X online
DOI: 10.1080/13632460601123156

The Estimation of Peak Ground-motion


1559-808X
1363-2469
Journal
UEQE of Earthquake Engineering
Engineering, Vol. 11, No. 1, December 2006: pp. 125

Parameters from Spectral Ordinates

EDMUND BOOTH
Estimating
E. Booth Peak Motions from Spectral Ordinates

Edmund Booth Consulting Engineer, Walthamstow, London, UK

Relationships are derived between peak ground-motion parameters and the peaks in the corre-
sponding 5% damped response spectrum. The relationships are based on a study of the smoothed
spectra from a database of over 1,000 horizontal ground-motion records. A more consistent, but
more complex, relationship is also derived between peak ground velocity (PGV) and an integration
Downloaded by [27.54.121.152] at 09:20 01 August 2017

of the 5% damped spectrum, using random vibration theory. A recent proposal to estimate PGV
from the 0.5 sec spectral ordinate is found to give less accurate results than the methods derived in
this study, when compared with values derived from the strong ground-motion database; in particu-
lar, it consistently underestimates PGV for records with unusually long or short predominant peri-
ods. However, when compared with predictive equations for PGV, this method gives results
comparable to the other two and is also much the simplest. Relationships between spectral ordinates
and PGV are useful for the purpose of estimating PGV from a smoothed design spectrum or from a
spectrum derived from ground-motion equations, for uses including estimation of structural damage
and geotechnical effects. The relationships between response spectrum peaks and peak ground-
motion parameters may also be useful in the selection of real time histories compatible with a
smoothed design spectrum.

Keywords PGV; PGD; PGA; Spectral Peak; Smoothed Spectrum

1. Introduction
In a recent paper, Bommer and Alarcon [2006] discussed the importance of peak ground
velocity (PGV) in earthquake engineering. They observed that there is a shortage of
ground-motion equations predicting PGV, compared to those for peak ground acceleration
(PGA) and spectral ordinates; therefore the need may arise to derive PGV from spectral
values. Bommer and Alarcon showed that the relationship between 5% damped spectral
response at 0.5 sec and PGV is more stable with respect to magnitude and has less scatter
than is provided by spectral response at other periods, including 1.0 sec, which is often
used to derive PGV. However, the relationship is subject to significant uncertainty, and
Bommer and Alarcon do not advise its use when near-fault directivity effects may be
present. More generally, they recommend direct use of a PGV predictive equation wher-
ever possible, and urge that more such equations should be produced in future.
The present article reports a study which attempted to improve the reliability of pre-
dicting PGV from 5% damped spectral ordinates and to remove some of the shortcomings
of basing the relationship on spectral response at a single, fixed period. Two methods are
derived; a simple relationship between PGV and peak pseudo spectral velocity and a
more complex method based on random vibration theory, which is nevertheless easily

Received 21 February 2005; accepted 15 June 2006.


Address correspondence to Edmund Booth, Edmund Booth Consulting Engineer, 67 Orford Road,
Walthamstow, London, E17 9NJ, UK. E-mail: Edmund@Booth-seismic.co.uk

13
14 E. Booth

implementable in a spreadsheet. The study also investigated the relationship between peak
spectral values and peak ground displacement (PGD) or acceleration (PGA), using similar
methods. In the rest of the paper, spectral accelerations and velocities are taken as pseudo-
spectral values, without further clarification.

2. Relationship between Peak Spectral Values and Peak Ground-motion


Parameters

2.1 Theoretical Basis


There is a clear and simple link between PGA and spectral values; PGA is of course
exactly equal to the spectral acceleration (independently of damping level) at zero period.
No such rigorous theoretical link exists for PGV or PGD; although for low damping, PGD
approximates to the spectral value at very long periods, the periods involved are longer
than are normally available. However, there are some grounds for supposing that PGV and
Downloaded by [27.54.121.152] at 09:20 01 August 2017

PGD might be related to the corresponding peak spectral values. For a single frequency
ground-motion of infinite duration, such a theoretical link does exist: peak spectral values
(acceleration, velocity or displacement) are related to peak ground parameters by the well-
known ratio 1/(2), where is the damping ratio. Thus for 5% damping, an infinitely long
single frequency ground-motion would produce 5% damped spectral peaks of 10 times the
ground-motion peaks, and this applies equally to acceleration, velocity, and displacement.
Real earthquakes, of course, have finite duration and a range of frequencies, and therefore
as expected the ratios of spectral to ground-motion peaks are much less than 10. Neverthe-
less, it seems intrinsically more probable that the ground-motion peaks should be related
to the spectral peaks than to spectral values at a fixed period. Indeed, 25 years ago, New-
mark and Hall [1982] proposed that design spectra should be constructed on the basis of
constant ratios to peak ground-motion parameters and Eurocode 8 [EN1998-1, 2004] also
implies that these ratios can be taken as constant, as shown in Table 1.

2.2 Methodology Adopted in this Study


When plotting the 5% damped spectrum of a typical ground-motion record, it is usual to
find that peaks in spectral acceleration occur at a number of different periods, and the
same applies to spectral velocity and displacement. For each spectral parameter, one peak
will be greater than the rest, and so at first sight the problem is that of finding the value of
this maximum peak, and hence its ratio to the corresponding peak ground-motion parame-
ter. When dealing with the spectra of individual records, this approach works well, and
yields stable relationships of relatively low variability. However, when dealing with
smoothed code design spectra or those produced by ground-motion equations, this
approach is not correct. Code spectra invariably have a single maximum value for spectral
acceleration and velocity; in the writers experience, the same is true for spectra generated
from ground-motion equations. These maxima represent an average of the spectral values
of the individual records used to generate the smoothed spectrum, but in general the true
peaks of the spectra of individual records will occur at a range of different periods. There-
fore, by relating PGV to the peak spectral velocity of each individual record, a true ensem-
ble average value is not obtained, because the spectral values are always at their peak,
rather than some being above the ensemble average and some below. The same applies to
acceleration and displacement values. The corresponding ratios of peak spectral value to
peak ground-motion value are higher than those relating to a smoothed spectrum. It would
therefore be expected that finding the ratio of individual peak spectral values to peak
Downloaded by [27.54.121.152] at 09:20 01 August 2017

TABLE 1 Ratios of peak 5% damped spectral values to peak motion parameters, from various sources
This study
Newmark & Peaks based on
Hall [1981] Malhotra [2006] Individual peaks smoothed trend
Mean + 1
Mean Mean + 1 Eurocode 8 Mean (Note 1) Mean Mean + 1 Mean Mean + 1
Peak SA PGA 2.12 2.71 2.5 2.14 3.1 3.35 4.1 2.65 3.65
Peak SV PGV 1.65 2.30 (not defined) 1.63 2.7 2.85 3.65 2.3 3.3
Peak SD PGD 1.39 2.01 2.53 1.84 2.8 2.8 3.7 2.3 3.2
Note 1: 1 confidence limits are not stated by Malhotra, but are drawn on his Figure 8, from which the values given in this table have been derived.

15
16 E. Booth

ground-motion parameter, and applying this parameter to a smoothed spectrum to predict


PGA, PGV, or PGD, would result in an underprediction of the ground-motion parameter.
This study found that the ratios based on individual peaks gave results that were about
40% higher than those based on the smoothed trend of individual records, as shown in
Table 1. The method adopted for finding the smoothed, trend spectra of individual records
is described in the following paragraphs.
The first attempt at finding the trend spectrum for an individual record was based on
finding the period corresponding to the peak spectral value at high damping levels. The
5% damped spectral value was then found at this period, and the ratio to peak ground-
motion parameter was calculated. It was hoped that by finding the peak period for increas-
ing values of damping, an underlying trend would be established. This procedure did
indeed result in a lowering of the spectral to ground-motion peak ratios as the damping
level increased. However, it was hoped that the ratio would show convergence to a stable
value at very high levels of damping. No such convergence was found for damping levels
up to 100%.
Downloaded by [27.54.121.152] at 09:20 01 August 2017

Therefore, another approach was used. Moving averages of the 5% damped spectra of
individual records were calculated. The period ranges for the averaging, to either side of
the period of interest, are shown in Table 2. Two different ranges were used for spectral
velocity and displacement and three for acceleration, in order to see if the results were
dependent on the period range adopted. For most records, the averaged spectra thus pro-
duced were still found to contain a number of peaks, often over a considerable period
range. The period adopted as that corresponding to the spectral peak of the underlying
trend was taken as the weighted average of the periods of spectral peaks within 75% of the
overall maximum peak value. The weights for calculating the average period were the val-
ues of the spectral peaks (Eq. 2.1). Here, SVPeak,1 is the first peak spectral velocity of at
least 75% of the overall peak and TPeak,1 is the corresponding period. Similar expressions
were used for spectral acceleration and displacement.

(TPeak ,1 .SVPeak ,1 + TPeak ,2 .SVPeak ,2 + TPeak ,3 .SVPeak ,3 ......)


TPeak ,Trend = (2.1)
(SVPeak ,1 + SVPeak ,2 + SVPeak ,3 .....)

The ratio of spectral peak to ground-motion peak for each record was then based on
the 5% damped spectral value at TPeak,Trend. The average ratio was calculated in this way
for the averaging period ranges shown in Table 2; the average ratio was found vary within
a range of less than 5% for the averaging periods investigated for velocity and accelera-
tion, and by less than 1% for the averaging periods investigated for displacement.
The methodology described above was then applied to PGV ratios for 1161 horizontal
records from 55 mainly North American earthquakes, with distances from the causative
fault reported as ranging from under 1 km to 223 km. The records were chosen from a
larger data set of 1292 horizontal records originally assembled by Abrahamson and Silva

TABLE 2 Period ranges for calculating moving average values of 5% damped spectra
Period ranges (seconds)
SA, spectral acceleration 0.025 0.05 0.075
SV, spectral velocity 0.0625 0.125
SD, spectral displacement 0.25 0.5
Estimating Peak Motions from Spectral Ordinates 17

[1997]; 131 records were excluded, either because the PGV reported in the database did
not match that calculated by single integration of the acceleration time histories to within
5%, or because the calculated period at peak spectral velocity exceeded 2.4 sec; the latter
value was chosen as 80% of 3 sec, the maximum period to which spectral values were cal-
culated. The same dataset was used for calculating peak spectral displacement to PGD
ratios; the results were based on 1,058 of the records, excluding those with a period at
peak spectral displacement exceeding 12 sec, since spectral values were only calculated to
a maximum period of 15 sec. Also, records were rejected where the reported PGD in the
database did not match that calculated by double integration of the time histories to within
15%. Integration of acceleration time histories to produce velocities and (especially) dis-
placements is sensitive to the processing applied to the accelerograms; the only check car-
ried out for this study was, as reported above, that peak velocity and displacement agreed
within 5% (for PGV) and 15% (for PGD) of the values reported by Abrahamson and Silva
[1997].
The results are shown in Figures 1 and 2 and summarized in Eqs. (2.2) and (2.3).
Downloaded by [27.54.121.152] at 09:20 01 August 2017

PGV = (peak 5% damped spectral velocity from smoothed spectruum) / 2.3


(coefficient of variation) = 0.16 (2.2)

PGD = (peak 5% damped spectral displacement from smoothed speectrum ) / 2.3


(coefficient of variation) = 0.15 (2.3)

The same methodology was then applied to finding the ratio between peak spectral
acceleration and peak ground acceleration, PGA. Of course, PGA is known with certainty
from the response spectrum, so the ratio is not useful for predictive purposes but is still of
intrinsic interest. Using exactly the same database of 1,161 records used to derive Eq.
(2.2), the following relationship was found.

PG = (peak 5% damped spectral acceleration from smoothed speectrum) / 2.65


(coefficient of variation) = 0.13 (2.4)

3. Prediction of Peak Ground-motion Parameters from Random


Vibration Theory

3.1 Theoretical Basis for Prediction of PGV


The peak spectral to ground-motion ratios calculated as described above were derived
purely empirically. They were also to some extent influenced by the method used to find
the peaks in the spectral records. An attempt was therefore made to develop a method of
calculation based on a firmer theoretical basis. It used the fact that there is a reasonably
close relationship between the pseudo-spectral velocity spectrum and the Fourier spec-
trum of the stationary part of the underlying record. Together with knowledge of the
record duration and predominant period, this enables statistical estimates to be made of
PGV and PGD; the details are now explained.
The Fourier spectrum of the stationary part of a record can be shown to be related to
the 0% damped pseudo-velocity response spectrum, but usually only a 5% spectrum will
be available. Therefore, an adjustment to convert from 5% to 0% damping must be made.
18 E. Booth

10 10
R2 = 0.006

Predicted PGV
Observed PGV
Predicted PGV
Observed PGV

1 1

0.1 0.1
0.1 1 10 100 1000 0 0.5 1 1.5 2 2.5
PGV (cm/sec) Period at peak SV

a) Peak SV/2.3: eq.(2.2)

10 10
R2 = 0.015

Predicted PGV
Observed PGV
Predicted PGV
Observed PGV

1 1
Downloaded by [27.54.121.152] at 09:20 01 August 2017

0.1 0.1
0.1 1 10 100 1000 0 0.5 1 1.5 2 2.5
PGV (cm/sec) Period at peak SV

b) Random vibration theory: eq.(3.11)

10 10
R2 = 0.34
Predicted PGV
Observed PGV
Predicted PGV
Observed PGV

1 1

0.1 0.1
0.1 1 10 100 1000 0 0.5 1 1.5 2 2.5
PGV (cm/sec) Period at peak SV

c) SA(0.5)/20: Bommer and Alarcon [2006]

FIGURE 1 Predictions of PGV from spectral ordinates by three methods.

10 Peak SD/2.3: eq. (2.3) 10 Random vibration theory: eq. (3.13)


Observed PGD
Predicted PGD
Observed PGD
Predicted PGD

1 1

0.1 0.1
0.01 0.1 1 10 100 0.01 0.1 1 10 100
PGD (cm) PGD (cm)

FIGURE 2 Predictions of PGD from spectral ordinates by two methods.

Denoting the 5% damped acceleration spectrum as SA5%(T) and the 0% damped velocity
spectrum as SV0%(T) (where T is the structural period), one can write:

SV0% (T ) = (T / 2 p ). f (T ).SA5% (T ) (3.1)


Estimating Peak Motions from Spectral Ordinates 19

where f(T) is some function of structural period, which will depend on the ground-motion
record in question. At zero period, f(0) is exactly 1, and then rises to a peak at the predominant
periods of the ground-motion. It then tends to reduce at longer periods, but will never fall
below 1. Eurocode 8 Part 1 [EN1998-1, 2004] eq. (3.6) gives a period-independent value
for damping correction, understood to have been derived for the range 5% to 28%
damping (although the lower limit is not stated). Assuming the equation to be valid at 0%
damping gives a period independent value for f:

f = {10/(5 + 0)}1/2 = 1.41 (3.2)

Having obtained the 0% damped velocity spectrum from Eq. (3.1), the Fourier accel-
eration spectrum can be estimated. Hudson [1962] has shown that the former represents a
fairly close upper bound to the latter. Writing Fa as the Fourier acceleration spectrum, the
relationship may be expressed as:
Downloaded by [27.54.121.152] at 09:20 01 August 2017

Fa = g(T ). SV0% (T ) (3.3)

where g(T) is some dimensionless parameter related to period, and depending on the
record in question, but never greater than one.
The Fourier acceleration spectrum can then be converted to the Fourier velocity spec-
trum by dividing by the circular frequency (Eq. 3.4), and the rms (root mean square)
velocity, vrms, follows directly from the well known relationship (Parsifals equation),
given here as Eq. (3.5).

Fv = (T / 2p ).Fa (3.4)

T = T =
vrms = [1 / p {Fv }2 dT ]1 / 2 = [1 / p {T /(2 p).Fa}2 dT ]1 / 2 (3.5)
0 0

The rms velocity can then be used to obtain a statistical estimate of PGV. Boore
[2003] reports that the following relationship applies

PGV = vrms [{2 ln( N v )}1 / 2 + 0.58 /{2 ln( N v )}1 / 2 ] (3.6)

Here, Nv is the number of zero crossings during the stationary part of the velocity
record; in other words, the number of times the velocity changes sign. This will depend on
the predominant periods of the record, and the duration of the strong motion portion of the
record. In fact, the right hand side of Eq. (3.6) varies only slowly with Nv, and so deriving
PGV from vrms requires only a crude estimate of Nv.
Calculating PGV from a 5% damped response spectrum using Eqs. (3.1), (3.3), (3.4),
(3.5), and (3.6) therefore requires knowledge of the following:
20 E. Booth

1. f (T), the ratio of 0% to 5% damped spectral ordinates


2. g(T), the ratio of 0% damped velocity response spectrum to Fourier acceleration
spectrum
3. Nv, the number of zero crossings of the record, or equivalently its duration and pre-
dominant period.
Initially, f (T) and g (T) were taken as period independent constants of 1.4 and 0.8, respec-
tively. Reasonable assumptions were made on predominant period and duration, based on
the reported soil type at the location of the recording and the magnitude of the causative
earthquake. These values produced good predictions for PGV. However, it was realized
that the form of the equations can also be used to predict PGA, which is of course known
with certainty from the 5% damped response spectrum as the zero period acceleration. It
can easily be shown that Eq. (3.7) applies.

T =
[1 / p {Fv }2 dT ]1 / 2
. {2(ln N v ) + 0.58 /(ln N v ) }
1/ 2 1/ 2
Downloaded by [27.54.121.152] at 09:20 01 August 2017

PGV = PGA 0
T =
(3.7)
[1 / p {Fa } dT ] 2 1/ 2 {2(ln N a ) + 0.58 /(ln N a )1 / 2 }
1/ 2

where Na is the number of zero crossings of the acceleration record. In general, Na will be
greater than Nv. Writing

{2(ln N v )1 / 2 + 0.58 /(ln N v )1 / 2 }


X= (3.8)
{2(ln N a )1 / 2 + 0.58 /(ln N a )1 / 2 }

X is relatively insensitive to variations in Na and Nv, and would be expected to be in the


range 0.8 to 0.9 for most records.
Combining eqs. (3.7) and (3.8) then gives

T =
[1 / p { f (T ).g(T ).SA5% (T ).(T / 2 p ) } dT ]
2 2 1/ 2

PGV = X .PGA 0
T =
3 (3.9)
[1 / p { f (T ).g(T ).SA5% (T ).(T /2p )}2 dT ]1 / 2
0

Making the approximation that f (T) and g(T) are period independent greatly simpli-
fies Eq. (3.9) to yield Eq. (3.10)

T =
[ {SA5% (T ).(T / 2 p ) } dT ]
2 2 1/ 2

PGV = X .PGA 0
T =
(3.10)
[ {SA5% (T ).(T / 2 p )} dT ]
2 1/ 2
0

Given a 5% damped response spectrum, the only unknown on the right-hand side of
Eq. (3.10) is the parameter X, which as noted above might be expected to be around 0.85.
In fact, analysis of the same 1,161 records described in Sec. 2.2 gave a best fit value of X
as 0.79, assuming a log normal distribution of the ratio of predicted to actual PGV. X
allows not only for the differing values of Na and Nv but also the bias introduced by taking
f(T) and g(T) as frequency independent, and (more fundamentally) the differences
Estimating Peak Motions from Spectral Ordinates 21

between real records and the idealized records assumed by the theory. Also, the integra-
tion cannot in practice extend over an infinite period range. Nevertheless, the fact that X
falls within the range expected and the good agreement between predicted and actual PGV
shown in Figure 1 gives some confidence that Eq. (3.10) should hold for a wider range of
records than those investigated in this study.
For this study, integration was carried out numerically with a constant period incre-
ment T of 0.0125 sec. Integration was carried out to upper limits of both 3 sec and 5 sec
with little improvement in results for the larger period range, although it should be noted
that the few records with periods at peak spectral velocity exceeding 2.4 sec were
excluded from the dataset. Also, as reported later, integration to 5 sec did yield significant
improvement in results when applied to near-fault records. Rounding up the value of X
gives the final recommended form as Eq. (3.11). Note that it is assumed that consistent
units are used: e.g., PGV in cm/sec, PGA in cm/sec2 and T in seconds.

T =3
[ {SA5% (T ).(T / 2 p ) } dT ]
Downloaded by [27.54.121.152] at 09:20 01 August 2017

2 2 1/ 2

PGV = 0.8 PGA 0


T =3
(3.11)
[ {SA5% (T ).(T / 2 p)} dT ]
2 1/ 2
0

(coefficient of variation) = 0.10

3.1.1 Extension of random vibration theory to prediction of PGD. Eq. (3.10) can easily
be extended to PGD:

T =
[ {SA5% (T ).(T / 2p ) } dT ]
3 2 1/ 2

PGV = X .PGA 0
T =
(3.12)
[ {SA5% (T ).(T / 2 p)} dT ]
2 1/ 2
0

Here, X allows for the much smaller number of zero crossings in the displacement
trace, compared with acceleration trace. However, the theoretical basis assumes that the
displacement record over the strong motion part of the record has the form of a station-
ary random vibration, which is far from the case. Nevertheless, it was decided to see
how reliable Eq. (3.12) was in predicting PGD for the same 1,058 records used to inves-
tigate peak spectral displacement to PGD ratio. Somewhat to the authors surprise, using
an integration range up to 15 sec and a period increment T of 0.025 sec, Eq. (3.13) pro-
vided a slightly better fit to the results than that based on peak spectral displacement.
However, the very long spectral periods involved in the calculation mean that the results
are not very reliable, as indeed is the case for the underlying theory when applied to dis-
placements.

T =10
[ {SA5% (T ).(T / 2 p ) } dT ]
3 2 1/ 2

PGV = 0.7 PGA 0


T =10
(3.13)
[ {SA5% (T ).(T / 2 p )} dT ]
2 1/ 2
0

(coefficient of variation) = 0.12


Results are shown in Figure 2.
22 E. Booth

4. Prediction of Peak Ground-motion Parameters from 0.5 Second Spectral


Acceleration
Bommer and Alarcon [2006] have proposed that PGV implied by a smoothed spectrum
can be predicted from Eq. (4.1).

PGV = SA5% (T = 0.5) / 20 (4.1)

where SA5%(T = 0.5) is the 5% damped pseudo spectral acceleration at 0.5 sec. PGV and
SA5%(T = 0.5) must be in consistent units (eg cm/sec and cm/sec2). For the data set used in
this study, the coefficient of variation was found to be 0.19, and the average of pre-
dicted/actual PGV was 0.95. The results are plotted on Figure 1.
It is easy to show that Eq.(4.1) is equivalent to
Downloaded by [27.54.121.152] at 09:20 01 August 2017

PGV = SV5% (T = 0.5) / 1.59 (4.2)

Assuming that Eq. (2.2) is reliable, and comparing Eq. (4.2) with Eq. (2.2), would
lead to the expectation that Eq. (4.1) would tend to overpredict PGV when the period at
peak spectral velocity was around 0.5 sec, and underpredict it when the period was much
greater or much less. The trend lines plotted on Figure 1 support this expectation; the cor-
relation coefficient R between SA5%(T = 0.5)/2.3 and predominant period for the trendline
is shown on Figure 1 as 34%. For the other two methods, the trendlines are practically
constant at 1 and the correlation coefficients are less than 2%, suggesting a very low corre-
lation with period.
Equation (4.1) has the advantage that it is much easier to calculate than Eqs. (2.2) or
(3.11), because only one spectral period is involved. By contrast, to solve Eq. (2.2), the
peak spectral velocity must be found, involving calculation at a range of periods, and
Eq. (3.11) involves a complex integration over many periods.

5. Malhotras Smoothed Spectrum


Malhotra [2006] has proposed a method for constructing a smooth design spectrum from
the peak ground parameters PGA, PGV, and PGD. The method is an extension of that pro-
posed by Newmark and Hall [1981]. It involves normalising the period scale by dividing
all periods by Tcg, given by

pgd
Tcg = 2p . (5.1)
pga

The spectral ordinates are normalized by the factor pga. pgd . Malhotra used a data-
base of 63 horizontal records to derive a mean trend through the 5% damped spectra of all
63 records. The results of particular relevance to this study are the peak spectral to peak
ground-motion parameters, which are shown in Table 1. It can be seen that the ratios are
significantly lower than found in this study, and (for acceleration and velocity) very close
to the values given by Newmark and Hall [1981].
Figure 3 compares Malhotras central period Tcg with Tpeak,trend, the period corre-
sponding to peak spectral velocity from this study, taking the trend of each individual
Estimating Peak Motions from Spectral Ordinates 23

10

Period at peak SV
Malhotras Tcg

0.1
0 0.5 1 1.5 2 2.5 3
Downloaded by [27.54.121.152] at 09:20 01 August 2017

Period at peak SV (secs)

FIGURE 3 Comparison of Tcg: Eq. (5.1) from Malhotra [2006] with period at peak spec-
tral velocity: Eq. (2.1) from this study.

spectrum. The comparison is based on the 1,058 records from the Abrahamson and Silva
[1997] database where the reported and calculated PGD differed by less than 15%. It can
be seen that Tcg predicts the average value of the period at peak spectral velocity rather
well over a wide range of periods; Malhotra give a value of 1.1 Tcg compared with a loga-
rithmic average of 0.9Tcg for the data of Figure 3. However, there is a wide scatter. Since
Malhotras method involves smoothing over the combined results, while this study bases
the spectral ratio on the trend of each individual spectra, it would be expected that Mal-
hotras results would result in a much greater relative reduction in the peak spectral ordi-
nates. As seen from Table 1, Malhotras spectral ratios are indeed considerably lower than
those from this study. Which method of smoothing is more appropriate is discussed later
in this article.

6. Validation of Results
Two validation tests were carried out. The first involved comparison of the predictions of
peak ground-motion parameters from the methods discussed in this paper with those from
ground-motion predictive equations. It was intended as a general check on the methods
developed, and in particular to test whether the smoothing methods used to derive Eq.
(2.2) were appropriate. The second test involved using the methods of this paper on a set
of 108 ground-motion records taken near the source of large earthquakes. It was intended
to test the validity of the predictions in the presence of near-fault effects, namely directiv-
ity and velocity flings.

6.1 Comparison with the Values given by Ground-motion Predictive Equations


Six ground-motion predictive equations were chosen which gave both 5% damped spec-
tral ordinates and peak ground velocity, based on a consistent dataset. The sources used
were Tromans and Bommer [2002] together with Ambraseys et al. [1996] which used
24 E. Booth

essentially the same dataset, Campbell [1997], Sadigh [1997, 1998], Molas and Yamazaki
[1995, 1996], Joyner and Boore [1988], and Sabetta and Pugliese [1996]. A range of mag-
nitudes between 5 and 8, of distances between 10 km and 100 km and soil types were
taken, to give a representative spread of results. The magnitudes and distances were cho-
sen for each predictive equation to be within the ranged of applicability stated by the
authors. The results are shown in Figure 4 and the averages and dispersions are shown in
Table 3.
It can be seen that Eq. 2.2 (peak SV/2.3) tends to underestimate results by about 10%,
while Malhotras peak SV/1.65 overestimates by about 25%. Bommer and Alarcons
SA(0.5)/20 (Eq. 4.1) gives an average ratio close to 1 but has the largest scatter. Random
vibration theory overestimates PGV on average by 14%, but the results for Campbells
hard rock class show a consistently high ratio of between 1.7 and 2.3. Removing the hard
rock result reduces the overestimate to 7% and gives the lowest scatter of the four relation-
ships examined. Campbell [private communication] advises that hard rock refers to very
Downloaded by [27.54.121.152] at 09:20 01 August 2017

Tromans & Bommer Tromans & Bommer


10 Sadigh et al 10 Sadigh et al
Campbell - hard rock Campbell - hard rock
Campbell - soft rock & soil Campbell - soft rock & soil
Molas and Yamazaki Molas and Yamazaki
Joyner & Boore Joyner & Boore
Sabietta & Pugliese
Sabietta & Pugliese
pgv from grnd moti on eq

pgv from grnd moti on eq


predicted pgv

predi cted pgv

1 1

0.1 0.1
0.01 0.1 1 10 100 1000 0.01 0.1 1 10 100 1000
pgv (cm/sec) pgv (cm/sec)

a) Peak SV/2.3: eq. (2.2) b) Random vibration theory: eq. (3.11)

Tromans & Bommer Tromans & Bommer


10 Sadigh et al 10 Sadigh et al
Campbell - hard rock Campbell - hard rock
Campbell - soft rock & soil Campbell - soft rock & soil
Molas and Yamazaki Molas and Yamazaki
Joyner & Boore Joyner & Boore
Sabietta & Pugliese Sabietta & Pugliese
pgv from grnd m oti on eq
pgv from grnd m oti on eq

predi cted pgv


predicted pgv

1 1

0.1
0.1
0.01 0.1 1 10 100 1000
0.01 0.1 1 10 100 1000
pgv (cm/sec)
pgv (cm/sec)

c) SA(0.5)/20: Bommer and Alarcon [2006] d) Peak SV/1.63: Malhotra [2006]

FIGURE 4 Comparison of predictions of PGV from spectra generated by ground motion


equations with direct prediction from a ground motion equation.
Estimating Peak Motions from Spectral Ordinates 25

TABLE 3 Summary of results shown in Figure 4 averaged over six ground motion
predictive equations
Logarithmic average
of (predicted PGV)
(PGV from Coefficient of 84 percentile
predictive equation) variation confidence range
Note 1 Note 2 Note 1 Note 2 Note 1 Note 2
Peak SV/2.3: 0.91 0.90 0.11 0.11 0.71 to 1.17 0.70 to 1.16
eq (2.2)
Random vibration 1.14 1.07 0.12 0.09 0.86 to 1.50 0.87 to 1.32
theory: eq (3.11)
SA(0.5)/20: Bommer 1.00 0.98 0.14 0.14 0.72 to 1.38 0.71 to 1.35
& Alarcon [2006]
Peak SV/1.63: 1.28 1.26 0.11 0.11 0.99 to 1.65 0.98 to 1.62
Downloaded by [27.54.121.152] at 09:20 01 August 2017

Malhotra [2006]
Note 1: Figures including Campbell [1997] for hard rock.
Note 2: Figures excluding Campbell [1997] for hard rock.

hard basement rock material; the classification of rock in most ground-motion equations is
considerably softer. It is not clear why the discrepancy between random vibration theory
and Campbells hard rock results should occur; the fact that it does not apply to the other
methods suggests that it is associated with the long period spectral response included in
Eq. (3.11) but not the other methods. This is an issue which merits further study.

6.2 Validity for Near-fault Records


PGV and PGD were estimated for 108 near-fault records shown in Table 4. They are taken
from Tables 1 and 2 given in Bray and Rodriguez-Marek [2004] which were chosen to
investigate one type of near-fault effect namely forward-directivity. Bray and
Rodriguez-Marek studied recordings from 54 stations which captured near-fault ground-
motions from 13 earthquakes of magnitude range 6.1 to 7.1. All but two of the sets of
recordings can be downloaded from the Pacific Earthquake Engineering Research (PEER)
strong motion database (http://peer.berkeley.edu/smcat). Two near-fault Chi-Chi records
were substituted for two Chi-Chi records attributed to CDMG and used by Bray and
Rodriguez-Marek, since they do not appear to be in the PEER database. The two horizon-
tal components of each record (shown in Table 4 as x and y components) are taken
separately, making a total 108 records. The database contains records which exhibit fling
as well as forward-directivity effects, but (unlike the work of Bray and Rodriguez-Marek
[2004], which was for a different purpose) no attempt was made to remove the former
effect.
Because of the long periods involved, the maximum spectral period investigated in
the analysis for peak SV/2.3, Eq. (2.2), was increased from 3 sec to 5 sec; similarly, for
random vibration theory, the upper integration limit in Eq. (3.11) was increased to 5 sec.
Results are shown in Table 5 and Figure 5. It can be seen that PGV is best predicted by
random vibration theory, with an average and coefficient of variation almost the same as
was found for the full Abrahamson & Silva dataset. Peak SV/2.3 predicts less well, with
an average value of predicted to observed PGV of 0.80. However, the value is
Downloaded by [27.54.121.152] at 09:20 01 August 2017

TABLE 4 Near-fault records used for Figure 5

26
PGA (g) PGV (cm/s) PGD (cm)
(Note 2) (Note 2) (Note 2)
Earthquake M (Note 1) Year Station R (km) x y x y x y
Parkfield 6.1 1966 Cholame #2 0.1 0.26 0.48 13.7 0.5 3.8 22.5
Temblor 9.9 0.36 0.27 21.5 0.3 3.9 3.4
San Fernando 6.6 1971 Pacoima dam 2.8 1.23 1.16 112.5 1.2 35.5 11.7
Imperial Valley 6.5 1979 Brawley airport 8.5 0.16 0.22 35.9 0.2 22.4 13.5
EC County center FF 7.6 0.21 0.24 37.5 0.2 16.0 39.4
EC Meloland overpass FF 0.5 0.31 0.30 71.7 0.3 25.5 31.7
El Centro array #10 8.6 0.17 0.22 47.5 0.2 31.1 19.4
El Centro array #3 9.3 0.27 0.22 46.8 0.2 18.9 23.3
El Centro array #4 4.2 0.49 0.36 37.4 0.4 20.2 59.0
El Centro array #5 1.0 0.52 0.38 46.9 0.4 35.4 63.0
El Centro array #6 1.0 0.41 0.44 64.9 0.4 27.7 65.9
El Centro array #7 0.6 0.34 0.46 47.6 0.5 24.7 44.7
El Centro array #8 3.8 0.60 0.45 54.3 0.5 32.3 35.6
El Centro Diff. Array 5.3 0.35 0.48 71.2 0.5 45.8 14.0
Holtville Post Office 7.5 0.25 0.22 48.8 0.2 31.5 32.0
Westmorland Fire Sta 15.1 0.07 0.11 21.2 0.1 16.6 10.0
Morgan Hill 6.2 1984 Coyote Lake Dam (SW Abut) 0.1 0.71 1.30 51.6 1.3 12.0 9.6
Gilroy Array #6 11.8 0.22 0.29 11.4 0.3 2.5 6.1
Superstition Hills(B) 6.6 1987 El Centro Imp. Co. Cent 13.9 0.36 0.26 46.4 0.3 17.5 20.2
Parachute Test Site 0.7 0.46 0.38 112.0 0.4 52.8 15.2
Loma Prieta 7.0 1989 Gilroy Gavilan Coll. 11.6 0.36 0.33 28.6 0.3 6.4 4.6
Gilroy Hist. Bldg 12.7 0.28 0.24 42.0 0.2 11.1 3.7
Gilroy Array #1 11.2 0.41 0.47 31.6 0.5 6.4 8.0
Gilroy Array #2 12.7 0.37 0.32 32.9 0.3 7.2 12.1
Gilroy Array #3 14.4 0.56 0.37 35.7 0.4 8.2 19.3
LGPC 6.1 0.56 0.61 94.8 0.6 41.2 11.5
Downloaded by [27.54.121.152] at 09:20 01 August 2017

Saratoga Aloha Ave 13.0 0.51 0.32 41.2 0.3 16.2 27.5
Saratoga W Valley Coll. 13.7 0.26 0.33 42.4 0.3 19.6 36.4
Erzincan 6.7 1992 Erzincan 2.0 0.52 0.50 83.9 0.5 27.4 22.8
Landers 7.3 1992 Lucerne 1.1 0.72 0.79 97.6 0.8 70.3 16.4
Northridge 6.7 1994 Jensen Filter Plant 6.2 0.42 0.59 106.2 0.6 43.1 24.0
LA Dam 2.6 0.51 0.35 63.7 0.3 21.2 15.1
Newhall Fire Sta 7.1 0.58 0.59 75.5 0.6 17.6 38.1
W. Pico Canyon Rd 7.1 0.46 0.33 92.8 0.3 56.6 16.1
Pacoima Dam (downstr) 8.0 0.42 0.43 45.6 0.4 5.1 4.8
Pacoima Dam (upper left) 8.0 1.59 1.29 55.7 1.3 6.1 23.8
Rinaldi Receiving Sta 7.1 0.84 0.47 166.1 0.5 28.8 19.8
Sylmar Converter Sta 6.2 0.61 0.90 117.4 0.9 53.5 47.0
Sylmar Converter Sta East 6.1 0.83 0.49 117.5 0.5 34.2 28.7
Olive View Med FF 6.4 0.60 0.84 78.2 0.8 16.1 32.7
Kobe 6.9 1995 KJMA 0.6 0.82 0.60 81.3 0.6 17.7 20.0
Takarazuka 1.2 0.69 0.69 68.3 0.7 26.7 16.8
OSAJ 8.5 0.08 0.06 18.3 0.1 9.3 8.0
Takatori 0.3 0.61 0.62 127.1 0.6 35.8 32.7
Kocaeli 7.4 1999 Arcelik 17.0 0.22 0.15 17.7 0.1 13.6 35.6
Duzce 12.7 0.31 0.36 58.8 0.4 44.1 17.6
Gebze 17.0 0.24 0.14 50.3 0.1 42.7 27.5
Chi-Chi 7.6 1999 TCU052 0.2 0.42 0.35 118.4 0.3 246.2 184.4
TCU068 1.1 0.46 0.57 263.1 0.6 430.0 324.1
TCU075 1.5 0.26 0.33 38.2 0.3 33.2 86.5
TCU101 2.9 0.25 0.20 49.4 0.2 35.1 75.4
TCU102 1.8 0.17 0.30 77.1 0.3 44.9 89.2
TCU103 4.0 0.16 0.13 26.8 0.1 16.0 87.5
Duzce 7.1 1999 Bolu 17.6 0.73 0.82 56.4 0.8 23.1 13.6

Note 1: Magnitudes given are moment magnitudes.


Note 2: x and y refer to the two horizontal components of motion given in the PEER database (http://peer.berkeley.edu/smcat.

27
28 E. Booth

TABLE 5 Summary of results shown in Figure 5 for near-fault records of Table 4


Logarithmic average
of predicted PGV Coefficient of 84 percentile
actual PGV variation confidence range
Peak SV/2.3: eq (2.2) 0.80 0.16 0.55 to 1.16
Random vibration 1.02 0.11 0.79 to 1.31
theory: eq (3.11)
SA(0.5)/20: Bommer 0.63 0.23 0.37 to 1.07
& Alarcon [2006]

unchanged at 0.16. For SA(0.5)/20, eq. (4.1), increases from 0.19 for the Abrahamson &
Silva dataset to 0.23 for the near-fault records, and the average drops from 0.95 to 0.63,
confirming Bommer and Alarcons [2006] observation that the relationship is not advis-
Downloaded by [27.54.121.152] at 09:20 01 August 2017

able for use where near-fault effects are present.


Trendlines have been fitted to Figure 5, and the correlation coefficients R2 are shown on
the graphs. It can be seen that peak SV/2.3 and random vibration theory show little or no
dependence on Tpeak,trend, the period at peak spectral velocity. However, SA(0.5)/20 shows a
significant dependence on Tpeak,trend, with same marked tendency shown in Figure 1c to
underpredict PGV for large values of Tpeak,trend and to overpredict at a period of 0.5 sec.

7. Discussion
The method used in this study to derive spectral ratios is based on the characteristics of
individual records. It would be expected to produce the most representative spectral
shapes corresponding to real ground-motions. A more drastic degree of smoothing is
likely to apply in the derivation of ground-motion equations, because of the wide range of
actual soil types and site periods lumped together under one soil type. Thus, the spectral
value given by a ground-motion equation for a particular period, magnitude, epicentral
distance, and soil type average will be based on the average of spectral values at different
positions relative to their spectral peaks, which will occur at periods that are significantly
affected by the site period and hence also by the soil conditions. This range of periods cor-
responding to spectral peak will tend to lower spectral values at peaks but increase them
elsewhere. Therefore, the values in this study would be expected to underestimate the
PGV corresponding to a given peak spectral velocity obtained from a ground-motion
equation. Table 3 suggests that this is indeed the case, the underestimate being around
10%. For similar reasons, the smoothing method used by Malhotra [2006] also tends to
lower the spectral peak, and Table 3 suggests that this effect may be as much as 35%
for PGV.
The random vibration theory considers values over the entire spectral range, and so
the effect of lower spectral peak from a ground-motion equation might be expected to be
offset (at any rate to some extent) by the higher values away from the peak. The results in
Table 3 excluding Campbells hard rock tends to support this expectation, but inclusion of
the Campbell hard rock results makes the argument rather less sustainable. These are
issues here which deserve further investigation.
Deriving PGA, PGV, and PGD from smoothed response spectrum may be useful for
finding real time histories which match a specified design spectrum, by choosing records
which have similar values of the three peak ground-motion parameters. This may be
attractive because it might ensure that matching the three parameters would help a similar
Estimating Peak Motions from Spectral Ordinates 29

10
R2 =0.09

Predicted PGV
Observed PGV
1

0.1
0 1 2 3 4 5
Period at peak SV (seconds)

a) Peak SV/2.3: eq (2.2)


Downloaded by [27.54.121.152] at 09:20 01 August 2017

10
R2 = 0.003
Predicted PGV
Observed PGV

0.1
0 1 2 3 4 5
Period at peak SV (seconds)

b) Random vibration theory: eq (3.11)

10
R2 = 0.51
Predicted PGV
Observed PGV

0.1
0 1 2 3 4 5
Period at peak SV (seconds)

c) SA(0.5)/20: Bommer & Alarcon [2006]

FIGURE 5 Predictions of PGV from spectral ordinates by three methods for near-fault
events.
30 E. Booth

matching of underlying frequency content over a broad range of periods. A trial use of this
application by the author proved promising, but it was not rigorously tested, and would
merit further investigation.
It is worth considering what degree and method of smoothing is appropriate for a
given circumstance. The method used in this paper in deriving Eqs. (2.2) to (2.4) might be
expected to produce the most representative spectral shapes for specific seismological and
geotechnical conditions. It is probably best when specifying the PGA, PGV, and PGD val-
ues of individual records required to be compatible with a smoothed design spectrum.
However, for structural design purposes, the uncertainties involved in all of the seismo-
logical and geotechnical parameters point to a broader envelope spectrum being required,
perhaps as provided by Malhotra at the mean plus a suitable number of standard devia-
tions. Once again, these are issues which deserve further investigation.

8. Conclusions
Downloaded by [27.54.121.152] at 09:20 01 August 2017

1. The following ratios of peak spectral values to peak ground-motion parameters


were found from the smoothed spectra of over 1,000 horizontal strong ground-
motion records.

PGA = peak SA(smoothed) / 2.65


PGV = peak SV(smoothed) / 2.3
PGD = peak SD(smoothed) / 2.3

where peak SA(smoothed), peak SV(smoothed), and peak SD(smoothed) are mean values of
the 5% damped peak spectral acceleration, peak spectral velocity and peak spectral
displacement, taken from a smoothed response spectrum.
2. The coefficient of variation for these relationships is in the range 0.13 to 0.16.
3. The relationships appear to be valid over a wide range of seismological conditions,
although they become somewhat less reliable in the presence of near-fault effects.
They are not biased with respect to magnitude, distance or soil type.
4. These ratios of spectral ordinate to corresponding peak ground-motion parameter
may be about 10% too high when applied to the smoothed spectra derived from
ground-motion equations, at any rate for PGV.
5. The ratios are significantly higher than those originally reported by Newmark and
Hall [1981] and more recently by Malhotra [2006]. The Newmark and Hall and
Malhotra ratios appear to be around 25% too low, when used to estimate PGV
from smoothed spectra derived from ground-motion predictive equations.
6. The ratios of spectral peaks to PGA and PGD found in this study are close to those
adopted in Eurocode 8.
7. A more reliable estimate of PGV (Eq. 3.11), based on random vibration theory and
involving integration of the (smoothed or unsmoothed) 5% damped response spec-
trum, reduces the value to 0.09. Equation (3.11) works well in the presence of
near-fault effects, although it is advisable to increase the upper limit of integration
in Eq. (3.11) from 3 sec to 5 sec.
8. The same random vibration theory can be extended to provide an estimate of PGD
(Eq. 3.13), but it does not result in a significantly improved estimate, compared
with that based on peak spectral displacement (Eq. 2.3).
Estimating Peak Motions from Spectral Ordinates 31

9. The random vibration theory is therefore the most reliable of the methods studied
here, but it is also the most complex. However, it is still easily implementable in
a spreadsheet. One of its advantages with respect to use with smoothed spectra is
that it should be largely independent of the smoothing method used in the source
spectrum.
10. When applied to individual records, the estimate of PGV based on 5% spectral
ordinate at 0.5 sec period gives less accurate results than those based on either
peak SV(smoothed)/2.3 or random vibration theory, and appears consistently to
underestimate PGV for records with unusually long or short predominant peri-
ods. In particular, it is not reliable in the presence of near-fault effects. However,
when compared with results from predictive equations for PGV, this method
gives results broadly comparable to the other two. Moreover, the method is much
the simplest to implement.
11. Deriving PGV from smoothed response spectra may be useful where damage
estimates are required corresponding to spectra derived from ground-motion pre-
Downloaded by [27.54.121.152] at 09:20 01 August 2017

dictive equations providing spectral ordinates but not peak ground velocity.
PGV, as discussed by Bommer and Alarcon [2006], appears to be one of the best
single predictors of structural damage and geotechnical effects.
12. Deriving PGA, PGV, and PGD from smoothed response spectra may also be use-
ful for finding real time histories which match a specified design spectrum, by
choosing records which have similar values of the three peak ground-motion
parameters.

Acknowledgments
The strong motion database used in this study was kindly provided by Walt Silva of
Pacific Engineering and Analysis. Extensive and valuable discussions on the study and its
interpretation were held with Julian Bommer and John Alarcon of Imperial College,
Praveen Malhotra of FM Global and Kenneth Campbell of EQECAT. An anonymous
reviewer sent extensive and very useful comments and suggestions. The support and
encouragement of all these people is gratefully acknowledged.

References
Abrahamson, N. A. and Silva, W. J. [1997] Empirical Response Spectra Attenuation Relations for
Shallow Crustal Earthquakes, Seismological Research Letters, 68(1), 94127.
Ambraseys, N. N, Simpson, K. A. and Bommer, J. J. [1996] The prediction of horizontal response
spectra in Europe, Earthquake Engineering and Structural Dynamics, 25, 371400.
Bommer, J. J. and Alarcon, J. E. [2006] The prediction and use of peak ground velocity, Journal
of Earthquake Engineering, 10(1), 131.
Bray, J. D. and Rodriguez-Marek, A. [2004] Characterization of forward-directivity ground
motions in the near-fault region, Soil Dynamics and Earthquake Engineering, 24(2), 815828.
Boore, D. M. [2003] Simulation of ground motion using the stochastic method. Pure and Applied
Geophysics, 160, 636676.
Campbell, K. W. [1997] Empirical near-source attenuation relationships for horizontal and vertical
components of peak ground acceleration, peak ground velocity and pseudo-spectral acceleration
response spectra, Seismological Research Letters, 68(1), 154179.
EN1998-1 [2004] Eurocode 8: Design of structures for earthquake resistance. Part 1: General rules,
seismic actions and rules for buildings, (CEN, Brussels).
Hudson, D. E. [1962] Some problems in the application of spectrum techniques to strong motion
analysis. Bulletin of the Seismological Society of America, 53(2), 417430.
32 E. Booth

Joyner, W. B. and Boore, D. M. [1988] Measurement, characterization and prediction of strong


ground motion, Proceedings of Earthquake Engineering & Soil Dynamics II, Park City, Utah,
4397.
Malhotra, P. K. [2006] Smooth spectra of horizontal and vertical ground-motions, Bulletin of the
Seismological Society of America, 96(2), April.
Molas, G. L. and Yamazaki, F. [1995] Attenuation of earthquake ground-motion in Japan including
deep focus events, Bulletin of the Seismological Society of America, 85(5), 13431358.
Molas, G. L. and Yamazaki, F. [1996] Attenuation of response spectra in Japan using new JMA
records, Bull. Earthquake Resistant Struct., 29, 115128.
Newmark, N. N. and Hall, W. J. [1981] Earthquake Spectra and Design. Earthquake Engineering
Research Institute, Berkeley CA.
Sabetta, F. and Pugliese, A. [1996] Estimation of response spectra and simulation of nonstationary
earthquake ground motions, Bulletin of the Seismological Society of America, 86(2), 337352.
Sadigh, R. K., Chang C. Y., Egan, J. A., Makdisi, F., and Youngs, R. R. [1997] Attenuation rela-
tionships for shallow crustal earthquakes based on Californian strong motion data, Seismologi-
cal Research Letters, 68(1), 180189.
Sadigh, R. K. and Egan, J. A. [1998] Updated relationships for horizontal peak ground velocity and
Downloaded by [27.54.121.152] at 09:20 01 August 2017

peak ground displacements for shallow crustal earthquakes, Proc. of the Sixth U.S. National
Conference on Earthquake Engineering, Seattle, Paper 317.
Tromans, I. J. and Bommer, J. J. [2002] The attenuation of strong-motion peaks in Europe, Proc
12th European Conference on Earthquake Engineering, London.

Вам также может понравиться