Академический Документы
Профессиональный Документы
Культура Документы
1785/0120170028
Short Note
Never Fear Velocity Reversals
by Gony Yagoda-Biran, Bibi Kerpel, and Ronnie Kamai
Abstract Velocity reversals are layers within the soil/rock profile that have a veloc-
ity lower than that of the overlying strata. When performing site-response analysis,
these reversals are often avoided because of the complexity they introduce. In this
study, we explore the effect of a velocity reversal on the fundamental mode of the
1D linear elastic response for the case of two layers over a half-space. We define
a criterion for tolerable velocity reversals, in which the amplitude change imposed
by the reversal is relatively insignificant. We do so by modeling three-layer velocity
profiles with different combinations of thickness and velocities. We observe that the
amplitude is affected by both the impedance ratio and the thickness ratio between
the upper two layers. We find that the amplitude of the linear transfer function at
the fundamental frequency is, on a first order, a function of the effective impedance
ratiothe impedance ratio between the half-space and a layer having the combined
thickness of the upper two and their time-averaged shear-wave velocity. We suggest a
model to compute the minimum allowable velocity in the reversal layer while main-
taining an amplitude change smaller than 10% with respect to a nonreversal velocity
profile. The reversal velocity in our model is a function of the overlaying velocity and
the thickness ratio.
Introduction
Shear-wave velocity profiles are an inherent component considered atypical, such reversals are usually avoided
of site-response analysis for seismic-hazard purposes, be- when performing site-response simulations on real or syn-
cause the site response is typically considered to be con- thetic velocity profiles, especially when such synthetic
trolled by upward-propagating shear waves. It is common profiles are randomly generated (e.g., Rathje and Navidi,
to assume that the shear-wave velocity, density, and associ- 2013; Yagoda-Biran and Anderson, 2015; Jahanandish et al.,
ated stiffness will increase with depth due to the effects of 2017).
geologic age, cementation, and overburden stress. Analysis Pehlivan et al. (2015) studied the effects of velocity
of large databases of measured velocity profiles (e.g., Elec- reversals on the 1D surface response of spatially variable
tric Power Research Institute [EPRI], 1993; Boore, 2016; V S profiles, centered around a common base profile. They
Kamai et al., 2016) shows that on average, velocity increases compared the 1D equivalent linear response of two sets of
with depth, and that such an increase could be different in 60 randomly generated profiles, with and without reversals.
different regions. In addition, the use of V S30 as a proxy They found that the presence of reversals reduces the median
for site characterization is based on the assumption that be- surface response spectra by almost 10% at short periods and
cause the velocity is expected to gradually increase, the top that the reduction depends on the intensity of the incoming
30 m of the profile are an indicator for the velocities at depth; motion. They suggest that this reduction is due to higher
hence those profiles with the same V S30 will generally show damping caused by the low-velocity layers within the
the same amplification patterns (Kamai et al., 2016). profile.
Velocity reversals are defined as a layer within the In some countries, such as Italy, the site response is para-
profile that has a lower shear-wave velocity than the overly- meterized in terms of the expected horizontal-to-vertical
ing strata. This can occur in alluvial fans, where interlayering spectral ratio (HVSR). The effect of a velocity reversal on
of different materials is common (e.g., Bordoni, Milana, the HVSR has been studied by some Italian researchers
et al., 2011). It is also observed in volcanic settings, in which (e.g., Castellaro and Mulargia, 2009; Bordoni, Haines, et al.,
a stiff volcanic formation may cover softer formations that 2011; Bordoni, Milana, et al., 2011; Di Giulio et al., 2016).
were present before the eruption (e.g., Rahpeyma et al., They found that a significant velocity reversal has a promi-
2016). Because profiles containing velocity reversals are nent effect on the dispersion curve, leading to mode jumps
The Model
We create a three-layer velocity model, in which the
third layer (L3) represents the half-space and has a constant
shear-wave velocity of 900 m=s (V SL3 ). The velocities and
thicknesses of the two upper layers (L1 and L2) vary as fol- VSL1
lows. The velocity of L1 (V SL1 ) ranges from 400 to 800 m=s VSL1
with 100 m=s intervals. The velocity of L2 (V SL2 ) is always VSL1
equal to or lower than V SL1 , ranging from 300 to 800 m=s VSL1
A
with those calculated for the correspond-
ing three-layer models. If our hypothesis
is correct, we would expect these ampli-
T T T
tudes to be identical. The results for the
Figure 3. Three sections from Figure 2, along the three highest values of impedance velocity model with V SL2 300 m=s are
ratio. Figure notation is consistent with that in Figure 2. The color version of this figure presented in Figure 4a. The comparison
is available only in the electronic edition. shows that although the amplitudes are
not identical, the difference between the
on as follows: circles, 10/40; triangles, 20/30; squares, 30/20; two-layer and three-layer model is less than 10% (calculated
and diamonds, 40/10. The different shades stand for different by equation 1), and increases as a function of the upper-layer
V values. velocity (Fig. 4b):
SL1
Let us look at the group of circles at the higher end of the
IR axis. This is a group of models that share the same V SL2 and EQ-TARGET;temp:intralink-;df1;313;541 change %
V SL3 , and therefore the same IR and TR: 10/40. What changes
amplitudethree-layer model amplitudeeffective
between the models is V SL1 . As can be observed, there is very 100 :
amplitudethree-layer model
little change in the amplitude as a result of the V SL1 change.
However, when looking at the group of diamonds at the high 1
end of the IR axis, these models all share a TR of 40/10, and
Next, we calculate an IRE for all different velocity mod-
here a change in V SL1 does have an effect on the amplitude.
els shown in Figure 2. Figure 5a shows the peak amplitude as
This is further emphasized in Figure 3, showing an increased computed for the full three-layer models versus their respec-
effect of changing V SL1 as TR increases (hence increasing the tive IRE. Collapsing the top two layers to a single layer using
thickness of L1 with respect to L2). the thicknesses and velocities allows us to collapse the 3D
It seems that the amplitude is more sensitive to the plot to a 2D plot, because the thickness dimension is ac-
thicker layer: when L1 is thinner than L2, a change in V SL1 counted for in the averaged velocity and therefore in the
has barely any impact on the amplitude of the peak, and the IRE calculations. Figure notation is consistent with Figure 2.
amplitude is affected mainly by the IRthat is, by the As seen in Figure 5a, the calculated amplitude and the IRE
changes in V SL2 . However, when the thickness of L1 in- are very strongly linearly correlated, yielding an R2 value of
creases and becomes thicker than L2, a change in V SL1 af- 0.98. Still, some trends are observable within the cloud, both
fects the amplitude more than a change in the V SL2 . within different TRs and within different V SL1 values. To re-
This observation leads us to hypothesize that the ampli- move these trends, we normalize the IRE with the IR of the
tude might be affected by the effective impedance ratio base model, a two-layer model with no reversals, so that L1,
(IRE). The IRE is the IR between L3 and a 50-m thick layer with its velocity, is 50 m thick. Then we calculate the change
in the amplitude compared to the amplitude of the base
(a) (b) model with no reversal, and plot it in Fig-
ure 5b, so that Figure 5b is Figure 5a with
both axes normalized by the base model.
Next, we characterize the profiles that
result in an up to 10% change in amplitude
compared to the base model with no rever-
sal. We identify the profiles that give am-
plitude just above and just below the 10%
boundary, and interpolate to find the exact
V SL2 that gives a 10% change given a
V SL1 . These boundaries are plotted as dif-
ferent lines for the different TRs in Fig-
ure 6a. Any value combination that falls
below a boundary line will result in a
Figure 4. (a) The amplitude of the peak as a function of V SL1 . Figure notation is con-
sistent with that in Figure 2. The filled symbols are for three-layer models, and the open change higher than 10%, and a combina-
symbols are for their corresponding two-layer models. (b) Amplitude change as a function of tion that falls above the boundary will re-
V SL1 . The color version of this figure is available only in the electronic edition. sult in a change lower than 10%.
BSSA Early Edition
4 Short Note
TR = 10/70
TR = 30/50
Discussion
TR = 50/30
VSL2
VSL2
TR = 70/10
Here we demonstrate that in the case
of velocity reversals within a three-layer
V V V
SL1 SL2 SL2
model, the peak amplitude of the linear
Figure 6. (a) Combinations of V SL1 and V SL2 that result in a 10% change in am- transfer function is directly correlated with
plitude compared to the base model. (b) V SL2 as calculated by the suggested model in the IR between the half-space and the
equation (2), versus V SL2 interpolated for the 10% boundary for (a). Figure notation is overlying (reversal) layer. Then we show
consistent with that in Figure 2. (c) Same as (b), for a combined thickness of 80 m for L1 that on a first order, the peak amplitude is
and L2. Shade notation is similar to that in Figure 2. The color version of this figure is also a function of the IRE and hence the
available only in the electronic edition.
two layers above the half-space can be
treated as an equivalent single layer with
the same time-averaged velocity. This sim-
ple exercise yields two inevitable ques-
tions: (1) do the trends we see in the
reversal containing models hold true for
nonreversal models as well, and (2) if
the amplitude is a function of an IR be-
tween the half-space and the averaged
velocity, is there any significance to the or-
der of the layers?
To answer the first question, we calcu-
Figure 7. (a) Reversal and nonreversal containing velocity profiles, (b) their corre- late the transfer function of seven velocity
sponding transfer functions, and (c) amplitude versus impedance ratio. Notice how the profiles that have the same V SL1 and
behavior of the nonreversal models is more complex, as the dominant reflector shifts the same TR but different V SL2 , with and
from L2/L3 to L1/L2. The color version of this figure is available only in the electronic without reversals (Fig. 7). From Fig-
edition.
ure 7b,c, we see that as V SL2 increases,
the amplitude of the fundamental frequency
We observe that the slopes of these boundaries are sim- decreases to a minimum and then starts to increase again. This
ilar, and the distances between them seem to correlate to the minimum occurs when V SL3 V SL2 V SL2 V SL1 , after
TR value, and so derive an equation that links V SL1 , V SL2 , which the fundamental frequency shifts from being controlled
and TR for profiles that result in a change of less than by the deeper reflector to the shallower one. Furthermore, as
10% in the amplitude that shift occurs, the frequency of the fundamental peak also
BSSA Early Edition
Short Note 5
Castellaro, S., and F. Mulargia (2009). The effect of velocity inversions on Rathje, E. M., and S. Navidi (2013). Identification of site parameters that
H/V, Pure Appl. Geophys. 166, 567592. improve predictions of site amplification, Report PEER 2013-19,
Di Giulio, G., R. de Nardis, P. Bonico, and G. Lavecchia (2016). Seismic PEER Berkeley.
response of a deep continental basin including velocity inversion: The Yagoda-Biran, G., and J. G. Anderson (2015). Investigation of the ground
Sulmona intramontane basin (Central Apennines, Italy), Geophys. J. motion variability associated with site response for sites with V S30 over
Int. 204, 418439. 500 m=s, Bull. Seismol. Soc. Am. 105, 10111028.
Electric Power Research Institute (EPRI) (1993). Guidelines for determining
design basis ground motions: Methods and guidelines for estimating
The Geological Survey of Israel
earthquake ground motion in Eastern North America, Report EPRI TR- 30 Malkhe Israel Street
102293, Vol. 1, Electric Power Research Institute. Jerusalem 95501, Israel
Jahanandish, M., H. Zafarani, and A. H. Shafiee (2017). Implementation of gonyb@gsi.gov.il
the square-root-impedance method to estimate site amplification in (G.Y.-B.)
Iran using random profile generation, Bull. Seismol. Soc. Am. 107,
doi: 10.1785/0120160119.
Kamai, R., N. A. Abrahamson, and W. J. Silva (2016). V S30 in the NGA Department of Geological and Environmental Sciences
GMPEs: Regional differences and suggested practice, Earthq. Spectra Ben-Gurion University of the Negev
32, 20832108. P.O. Box 653
NTC (2008). Norme Tecniche per le Costruzioni, DM 14/01/2008 in GU Beer Sheva 84990, Israel
n.29 del 04/02/2008, Suppl. Ord. 30 (in Italian). (B.K.)
Pehlivan, M., Y. Hashash, J. A. Harmon, E. M. Rathje, J. P. Stewart, W. J.
Silva, K. W. Campbell, and S. Nikolaou (2015). Influence of
shear wave velocity reversals on one-dimensional site response of Department of Structural Engineering
spatially varied profiles, 6th International Conference on Earthquake Ben-Gurion University of the Negev
Geotechnical Engineering, Christchurch, New Zealand, 14 Novem- P.O. Box 653
ber 2015. Beer Sheva 84990, Israel
Rahpeyma, S., B. Halldorsson, C. Olivera, R. A. Green, and S. Jnsson (R.K.)
(2016). Detailed site effect estimation in the presence of strong veloc-
ity reversals within a small-aperture strong-motion array in Iceland, Manuscript received 26 January 2017;
Soil Dynam. Earthq. Eng. 89, 136151. Published Online 11 July 2017