You are on page 1of 31

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/329628329

In situ assessment of soil dynamic parameters for characterizing nonlinear


seismic site response using KiK-net vertical array data

Article  in  Bulletin of Earthquake Engineering · December 2018


DOI: 10.1007/s10518-018-00539-3

CITATIONS READS

4 312

4 authors, including:

Wang Suyang Yu Miao


Huazhong University of Science and Technology International Flavors & Fragrances Inc.
16 PUBLICATIONS   61 CITATIONS    90 PUBLICATIONS   634 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Study on site effects using strong-motion data from geotechnical arrays View project

All content following this page was uploaded by Wang Suyang on 15 December 2018.

The user has requested enhancement of the downloaded file.


Bulletin of Earthquake Engineering
https://doi.org/10.1007/s10518-018-00539-3

ORIGINAL RESEARCH

In situ assessment of soil dynamic parameters


for characterizing nonlinear seismic site response using
KiK‑net vertical array data

Hai‑Yun Wang1   · Wei‑Ping Jiang1   · Su‑Yang Wang1,2   · Yu Miao2 

Received: 1 June 2018 / Accepted: 10 December 2018


© Springer Nature B.V. 2018

Abstract
In this study, we utilize acceleration time histories data at eight KiK-net vertical arrays
in Japan, recorded between January 2009 to June 2014, to analyze nonlinear seismic site
response. Ground motion records provided by seismic vertical arrays allow estimation of
in situ soil dynamic parameters. First, the shear modulus is computed from the shear wave
velocity using seismic interferometry by deconvolution. Considering each record as an
in situ cyclic test under seismic loading, the shear strain is computed as the velocity ratio
between two sensors and the shear stress is assumed as a spatial function of accelerations.
Using classic hyperbolic regression models, we quantitatively assess the relationships
between the in situ soil dynamic parameters and the level of ground motion. We found non-
linearity starts at a quite low level of shear strain of 5 × 10−6–5 × 10−5, and becomes much
more significant beyond a relatively high level of strain of 1 × 10−4. The threshold strains
from ground motion observations are well consistent with those from previous laboratory
tests. Finally, a parametric analysis is presented to discuss the influence of plasticity index
on nonlinear soil behavior evaluations. It is concluded that the threshold acceleration of
soil nonlinear behavior (generally ranging from 20 to 100 cm/s2) is mainly controlled by
the stiffness and plasticity index, and the degree of soil nonlinear behavior is mainly con-
trolled by the stiffness, plasticity index and the level of ground motion.

Keywords  Nonlinear soil behavior · Seismic interferometry · Seismic vertical arrays ·


Shear modulus reduction curve · Plasticity index

* Su‑Yang Wang
suyangwang@whu.edu.cn
1
Institute of Engineering Mechanics, China Earthquake Administration, Xuefu Road 29,
Harbin 150080, People’s Republic of China
2
School of Civil Engineering and Mechanics, Huazhong University of Science and Technology,
Luoyu Road 1037, Wuhan 430074, People’s Republic of China

13
Vol.:(0123456789)
Bulletin of Earthquake Engineering

1 Introduction

It has been long recognized that soft shallow sediments can dramatically amplify the
ground motion because of the decrease of the wave impedance and the effect of resonance.
For weak ground motions, the site amplification can be predicted in terms of linear elastic-
ity (Hooke’s law). However, when the level of ground motion exceeds a certain thresh-
old, the relationships between stress and strain of soils for large deformation are hysteretic
and nonlinear, which has been confirmed by both laboratory tests (Hardin and Drnevich
1972; Dobry and Vucetic 1987; Vucetic 1994) and strong motion observations (Zeghal
and Elgamal 1994; Pavlenko and Irikura 2003; Assimaki et al. 2008). As a result, the site
amplification can be significantly affected by nonlinear soil behavior. It is well documented
that during strong ground motions nonlinear soil behavior tends to reduce site amplifica-
tion at high frequencies due to the increase of anelastic damping and to shift resonance
frequencies toward lower frequencies due to the degradation of shear modulus (Field et al.
1997; Frankel et  al. 2002; Bonilla et  al. 2005; Wu et  al. 2009; Régnier et  al. 2013; Ren
et al. 2017).
Geotechnical engineers have confirmed that the nonlinear soil behavior is significant
for a long time based on numbers of laboratory tests. In terms of various laboratory tests,
Vucetic (1994) reported two types of shear strain threshold of soil nonlinearity: the lin-
ear threshold shear strain (γtl) and volumetric threshold shear strain (γtv). For γ < γtl, soil
behaves as a linearly elastic material; for γtl< γ < γtv, soil becomes markedly nonlinear but
remains largely elastic; for γ < γtv, soil exhibits significant nonlinear and strong inelastic
behavior. It was found that the order of magnitude of γtv for soils is around ­10−4 (Hardin
and Black 1969; Drnevich and Richart 1970; Dobry and Ladd 1980; Vucetic 1994), and
that of γtl is around ­10−5 (Vucetic 1994; Beresnev and Wen 1996; Chandra et  al. 2015;
Guéguen 2016). There had been a long-standing debate regarding the amplification asso-
ciated with strong motions between geotechnical engineers and seismologists due to the
lack of direct evidence from strong motion observations. However, with the increase in the
number of seismic stations and the improvement in data quality, nonlinear site response has
been reported after almost all large earthquakes during the last 30 years. The seismologists
have acknowledged that nonlinear soil behavior is more common than previously thought
(Aki 1993; Iwasak and Tai 1996; Field et al. 1998).
Although nonlinear soil behavior has been confirmed through various laboratory tests
and strong motion observations, its quantification and modeling remain a challenge. At
present, two common methods (i.e., equivalent-linear and nonlinear analysis methods) have
been developed to predict nonlinear site response. Compared with the nonlinear analysis
method, the equivalent-linear analysis method, which uses an iterative approach to obtain
strain-compatible soil properties for a linear-elastic analysis, is more widely used in prac-
tice because of its simplicity and high computational efficiency. Various studies have been
carried out to compare the results of equivalent-linear and nonlinear analyses. It has been
found that equivalent-linear analyses provide reasonable results for small-to-moderate
shear strains (i.e., smaller than about 0.1–0.3%). However, when the shear strain exceeds
a certain threshold (i.e., about 0.2–0.5%), which usually occurs in soft soils during strong
motion, a more rigorous nonlinear analysis with an appropriate constitutive law is required
(Kaklamanos et al. 2013, 2015; Kim and Hashash 2013; Carlton and Tokimatsu 2015; Kim
et al. 2016; Régnier et al. 2016).
In early studies, comparisons of the site response during strong and weak motions (e.g.
mainshock and aftershocks) provided a direct way to identify nonlinear soil behavior. Many

13
Bulletin of Earthquake Engineering

researchers found that significant nonlinear effects can develop beyond a threshold acceler-
ation of 100–200 cm/s2 (Chin and Aki 1991; Wen et al. 1994; Beresnev and Wen 1996). In
recent studies, more and more evidences of nonlinear soil behavior in medium-size earth-
quakes have appeared. Researchers found that the threshold is lower than the previously
recognized, and nonlinear models should be considered for smaller earthquakes. Rubin-
stein (2011) found that very modest levels of shaking (~ 35 cm/s2) can produce observable
changes in site response at the Turkey Flat site. Using seismic records from KiK-net in
Japan, Wu et al. (2010) estimated the peak ground acceleration (PGA) thresholds for 6 sites
ranging from 20 to 80 cm/s2; Ghofrani et al. (2013) estimated the PGA thresholds for 49
sites ranging from 4 to 150 cm/s2; Régnier et al. (2013) estimated the PGA thresholds for
36 sites ranging from 20 to 500 cm/s2. At present, the precise evaluation of nonlinear soil
behavior plays an important role in recent ground-motion prediction equations (GMPEs,
Abrahamson et al. 2014; Boore et al. 2014; Bozorgnia et al. 2014; Campbell and Bozorgnia
2014; Chiou and Youngs 2014; Idriss 2014), building codes (CEN 2013; BSSC 2015), site
response analysis (Kim and Hashash 2013; Zalachoris and Rathje 2015; Kaklamanos et al.
2015; Kim et al. 2016; Li et al. 2018; Kishida et al. 2018) and earthquake-hazard assess-
ment (Stewart et al. 2014).
Nonlinear soil behavior is usually represented by the stress–strain relation or modulus
reduction curve (G/G0–γ, and G0 is the elastic shear modulus). Such nonlinear soil prop-
erties are generally obtained from cyclic and dynamic laboratory tests on soil specimens
collected in the field. However, there is usually a certain difference between laboratory
measurements and in  situ observations because of the effects of sample disturbance, the
complexity of reproducing in  situ confining pressure in laboratory and the difficulty in
separating nonlinear effects from other site-amplification effects, including the basin and
topography effects (Chandra et al. 2016; Guéguen 2016; Bonilla et al. 2017). Besides the
laboratory tests, an effective in  situ analysis method using ground motion records from
seismic vertical arrays has been developed to estimate stress–strain relationship and modu-
lus reduction curve in the past decades (Zeghal et  al. 1995; Pavlenko and Irikura 2002,
2003; Chandra et al. 2015, 2016; Guéguen 2016; Bonilla et al. 2017). It was found that the
shear modulus degradation can be assessed from the variation of shear wave velocity (Vs)
using seismic interferometry. Furthermore, considering each record as an in situ cyclic test
under seismic loading, the shear strain can be computed as the velocity ratio or the rela-
tively horizontal displacement between two sensors, and the shear stress can be assumed as
a spatial function of accelerations.
In this study, we use seismic data from KiK-net vertical arrays in Japan to analyze non-
linear seismic site response. For that, we quantitatively assess the relationships between the
in situ soil dynamic parameters derived from strong motion observations and the level of
ground motion. Our goal is to evaluate the threshold and degree of soil nonlinear behav-
ior, and to discuss the influence of the plasticity index (PI) on nonlinear soil behavior
evaluations.

2 Site condition and seismic data

The Kiban-Kyoshin Network (KiK-net) is composed of about 700 vertical arrays with
an uphole/downhole pair of strong-motion seismometers. These arrays were built by the
National Research Institute for Earth Science and Disaster Resilience (NIED) after the
great Hanshin Awaji earthquake disaster (Okada et  al. 2004). The velocity profiles and

13
Bulletin of Earthquake Engineering

geotechnical cross sections for most KiK-net stations are available on the website (see Data
and Resources). These velocity profiles are obtained from downhole P–S logging measure-
ments. Because of the local geologic conditions, a majority of KiK-net stations in Japan are
located on rock or thin sedimentary, while more than a half of KiK-net stations have time-
averaged velocities to 30 m (Vs30) ranging between 360 and 760 m/s (Boore et al. 2011;
Wang and Wang, 2016; Wang et al. 2017).
In this study, horizontal PGA at the surface is selected to represent the level of ground
motion, and it will be abbreviated as PGA for short in the rest of this article. At first, we
collected about a million accelerograms recorded between 2009 January and 2014 June
during 5007 earthquakes. The PGAs of most accelerograms are smaller than 20  cm/s2,
and those of only 391 accelerograms are higher than 200 cm/s2. And then, to ensure there
exists enough data for evaluating nonlinear soil behavior at large deformations, eight verti-
cal arrays were selected based on the number of strong motions. As shown in Table 1, for
each selected station, there are at least one accelerogram with PGA higher than 800 cm/
s2, at least two accelerograms with PGA higher than 400 cm/s2, at least four accelerograms
with PGA higher than 200 cm/s2, and at least eight accelerograms with PGA higher than
100 cm/s2.
From Fig. 1a, it is obvious that all the eight stations are located on Northeast Honshu
(the main island of Japan) because this area is prone to earthquakes. FKSH10, FKSH19,
IBRH11, IBRH13, IWTH05, MYGH10, TCGH07 and TCGH16 stations recorded 1574,
1450, 1501, 1446, 941, 1712, 485 and 1646 events during January 2009 to June 2014,
respectively. The cumulative number of events for each station is shown in Fig. 1b. For all
the eight stations, only about 12% of events were recorded during the 800 days before the
2011 ­Mw9.0 Tohoku earthquake, about 50% of events were recorded during the 400 days
right after the 2011 ­Mw9.0 Tohoku earthquake, and about 38% of events were recorded
during the next 800 days.
In total, 21,510 sets of accelerograms (each set includes two accelerograms, i.e., surface
and borehole components) from 2816 events are used in this study. As shown in Fig.  2,
these ground motion data encompass a wide range of moment magnitudes (from 2.0 to 9.0,
with an averaged magnitude of 4.48), of focal depths (from 0 to 650 km, with an averaged
depth of 40.7 km), of PGAs (from 0.1 to 1200 cm/s2, with an averaged PGA of 11.7 cm/s2),
of hypocentral distances (from 1 to 2400 km, with an averaged distance of 138 km).
The velocity profiles and geotechnical cross sections for the eight KiK-net stations are
shown in Fig.  3. Among these eight KiK-net stations, TCGH16 is located on the north

Table 1  Distribution of the Station PGA (cm/s2) Total


horizontal PGA at the surface for
the eight KiK-net stations < 100 > 100 > 200 > 400 > 800

FKSH10 3128 20 6 2 1 3148


FKSH19 2864 36 5 2 1 2900
IBRH11 2967 35 8 2 2 3002
IBRH13 2802 90 35 11 1 2892
IWTH05 1864 18 4 4 1 1882
MYGH10 3390 34 10 4 2 3424
TCGH07 941 29 11 5 2 970
TCGH16 3273 19 7 2 1 3292

13
Bulletin of Earthquake Engineering

Fig. 1  a Locations of the eight KiK-net stations; b cumulative numbers of events with respect to time for
the eight KiK-net stations

Fig. 2  a Magnitude and focal depth of the selected events; b PGA and hypocentral distance of the selected
recordings from the eight KiK-net stations

of Kanto Plain with deep sediments, while the other seven stations are located in hilly or
valley terrain with shallow sediments. According to the geotechnical cross sections, the
soil depth of TCGH16 station is 88.4  m, while those of other stations range between 0
and 27 m. The station information including location, depth to engineering bedrock with
a shear wave velocity no less than 760 m/s (Z760), time-averaged near-surface velocity to
30 m (Vs30) and site class (BSSC 2015) for these eight stations is listed in Table 2.

13
Bulletin of Earthquake Engineering

Fig. 3  Geotechnical cross sections and shear wave velocity profiles of the eight KiK-net stations

13
Bulletin of Earthquake Engineering

Table 2  Station information for the eight KiK-net stations


Station Lat. Lon. Alt. (m) Dep. (m) Z760 (m) Vs30 (m/s) Site class

FKSH10 37.16 140.09 565 200 16 487 C


FKSH19 37.47 140.72 510 100 20 338 D
IBRH11 36.37 140.14 67 103 30 242 D
IBRH13 36.80 140.58 505 100 34 335 D
IWTH05 38.86 141.35 120 100 26 429 C
MYGH10 37.94 140.89 18 205 114 348 D
TCGH07 36.88 139.45 1085 100 22 419 C
TCGH16 36.55 140.08 105 112 > 112 213 D

3 Methods

3.1 Shear modulus assessment

An effective solution for observing nonlinear site response consists of measuring the shear
wave velocity variations under different levels of excitation (Haskell 1953; Dobry 2013; Chan-
dra et  al. 2015). For seismic vertical arrays, seismic interferometry is a powerful noninva-
sive tool to image the structure of shallow crust from earthquake records or ambient noise.
It is a technique based on the variations of signals recorded at different sensors, which can
estimate the Green’s function between them and extract the travel time of wave propagate.
Both cross correlation and deconvolution can be used to retrieve the Green’s function rep-
resentations between different receiver locations in seismic interferometry. These two types
of Green’s function representation have been verified as effective methods for vertical array
to extract shear wave velocity in many previous studies (Miyazawa et al. 2008; Nakata and
Snieder 2012; Takagi and Okada 2012; Chandra et al. 2015). In this study, we prefer decon-
volution-based interferometry because it can eliminate the effect of the incident wavefields to
get a more reasonable estimation of shear wave velocity (Nakata and Snieder 2012). In addi-
tion, it was found that the removal of low signal-to-noise ratio records has little influence on
the results, indicating that signal quality concerns are less of an issue for deconvolution-based
interferometry applied to vertical arrays.
Before computing deconvolution function, we applied a bandpass filter from 1 to 13 Hz to
the whole seismic records. The deconvolution function in frequency domain D(f) can be cal-
culated by Eq. 1 and it can be converted into time domain by inverse Fourier transform. Then,
the time-domain deconvolution function is resampled 100 times using cubic spline interpola-
tion to increase the accuracy of the travel time picking. As shown in Fig. 4, the travel time is
selected from the maximum amplitude of the pulse in the negative part. Finally, under the
one-dimensional wave propagation assumption (i.e., assumption of vertical incidence of plane
waves), the shear wave velocity can be estimated by dividing the distance between two receiv-
ers (borehole depth in this study) by the travel time. Besides, the shear modulus is directly
related to the shear wave velocity by Eq. 2. It should be noted that a wealth of studies have
demonstrated that the one-dimensional wave propagation assumption is generally insufficient
to capture the characteristics of complex site response due to 2D/3D local geological irregu-
larities or isotropic material properties (Chávez-García and Faccioli 2000; Thompson et  al.
2012; Régnier et al. 2013; Maufroy et al. 2015, 2017).

13
Bulletin of Earthquake Engineering

Fig. 4  a Deconvolution functions of four events for the NS direction at TCGH16 station during the Mw 9.0
Tohoku-Oki earthquake sequence; b enlargement of a. The four vertical lines indicate the peak times of the
deconvolution function

A2 (f )A∗1 (f )
D(f ) =
|A1 (f )|2 + 𝜀 (1)
| |

in which f is frequency, ­A1(f) and ­A2(f) are the Fourier amplitude spectra of the accelero-
grams recorded at the surface and bottom, respectively. * denotes complex conjugate and
ε is a regularization parameter, set as 1% of the average power spectrum of the borehole
receiver (Nakata and Snieder 2012), to enhance the stability of deconvolution function.
G = 𝜌Vs2 (2)
in which ρ is the thickness-averaged density derived from the shear wave velocity profiles.
According to Cadet et  al. (2012), soil densities are assumed to increase with increasing
shear wave velocity and their values are set as follows:
3
⎧ 1.7 g/cm for Vs ≤ 180 m/s
⎪ 2.0 g/cm3 for 180 < Vs ≤ 360 m/s
𝜌=⎨ 3 (3)
⎪ 2.2 g/cm3 for 360 < Vs ≤ 1500 m/s
⎩ 2.5 g/cm for Vs > 1500 m/s

3.2 Shear strain assessment

There are two general approaches to assess shear strain from ground motion records: the
displacement-based approach and the wave propagation-based approach (Rathje et al. 2004).
These two approaches utilize the relative displacement between adjacent sensors and the ratio
of particle velocity to shear wave velocity to compute shear strain, respectively. Based on the
wave propagation approach, the shear strain (γ) can be calculated by Eq. 4.

13
Bulletin of Earthquake Engineering

du du∕dt PGV
𝛾= = = (4)
dz dz∕dt Vs

in which u, z, t are displacement, depth and time, respectively. PGV is the peak ground
velocity.
PGV/Vs30 is also considered as a strain proxy in several recent studies (Idriss 2011;
Chandra et al. 2015, 2016), where Vs30 is a site parameter widely used in building codes
and ground motion prediction equations. However, it should be noted that PGV/Vs and
PGV/Vs30 provide only a global estimate of near-surface shear strain for a given profile,
but they cannot predict the variation of shear strain with depth or the maximum shear strain
within the soil profile.
According to Chandra et al. (2015, 2016), the ratio between the average peak particle
velocity (v*) and the local average shear wave velocity (Vs) between two successive sensors
can be used to present a more accurate shear strain along with depth. The efficiency of the
wave-based v*/Vs strain has been verified by numerical simulations and centrifuge tests in
Chandra et  al. (2016). In this study, the average peak particle velocity can be calculated
by Eq. 5. It should be noted that, for KiK-net stations, there are only sensors at the top and
bottom of the borehole, so we can only get an average Vs throughout the entire profile.
[ ]
| v1 (t) + v2 (t) |
v ∗= max | | | (5)
|
| 2 |
in which v1(t) and v2(t) are the horizontal velocity time histories at the surface and bottom,
respectively.

3.3 Shear stress assessment

The soil column is usually considered as a shear deformation system in 1D response analy-
sis (Dobry, 2013). To be consistent with the definition of shear strain, the shear stress (τ) is
defined by Eq. 6, based on the shear deformation beam model (Chandra et al. 2015).
𝜏 = 𝜌 × z ∗ ×a ∗ (6)
in which a* is the average peak particle acceleration calculated by Eq.  7, and z* is the
equivalent depth calculated by Eq. 13.
[ ]
| a (t) + a2 (t) |
a ∗= max || 1 |
| (7)
| 2 |
in which a1(t) and a2(t) are the horizontal velocity time histories at the surface and bottom,
respectively.
Hardin and Drnevich (1972), Seed et  al. (1986) and Ishihara (1996) applied a classic
two-parameter nonlinear hyperbolic model to fit the stress–strain curve as listed in Eq. 8.
Meanwhile, a number of studies also recommended a three-parameter model that includes
an exponent on the γ/γ0 term (Askan et al. 2007; Vardanega and Bolton 2013; Yamada et al.
2008; Senetakis et al. 2012). Compared with the three-parameter model, Eq. 8 is simpler
and more convenient because it has fewer regression parameters.

G0 𝛾
𝜏= 𝛾 (8)
1+ 𝛾0

13
Bulletin of Earthquake Engineering

in which G0 is the elastic/maximum shear modulus and γ0 is the reference shear strain that
is the ratio between maximum shear stress τ0 and maximum shear modulus G0.
From Eq. 8, a hyperbolic model is proposed to fit the a*–γ relationship in this study, as
listed in Eq. 9.
b1 𝛾
a ∗= (9)
1 + k1 𝛾
in which k1 and b1 are fitting parameters. From Eqs. 8 and 9, b1 is equivalent to G0, and k1
is equivalent to 1/γ0, which is the ratio between maximum shear modulus G0 and maximum
peak particle acceleration a*max.
Based on the regression relationship between a* and γ, the value of z* can be deter-
mined as follows. For small deformation, as shown in Eq.  10, a*–γ can be regarded as
linear relationship, and tangent modulus approximately equals to secant modulus, as shown
in Eq. 11.
b1 𝛾
a ∗= ≈ b1 𝛾 (10)
1 + k1 𝛾

d𝜏 𝜏
G0 = ≈ (11)
d𝛾 𝛾
Substituting Eqs. 2, 6 and 10 into Eq. 11, we can get Eq. 12.
𝜏 𝜌 × z ∗ ×b1 × 𝛾
(12)
2
G0 = = = 𝜌 × z ∗ ×b1 = 𝜌Vs0
𝛾 𝛾
in which Vs0 is the elastic shear wave velocity corresponding to small deformation.
From Eq. 12, we can easily obtain Eq. 13. Substituting Eq. 13 into Eq. 6, the shear stress
can be calculated by Eq. 14.

(13)
2
z ∗= Vs0 ∕b1

(14)
2
𝜏 = 𝜌Vs0 a ∗ ∕b1

4 Estimating shear wave velocity using seismic interferometry

Utilizing the ground motion records from the eight KiK-net vertical arrays, the average
shear wave velocity between two sensors is computed by the seismic interferometry based
on deconvolution. As we mentioned above, the use of the average shear wave velocity (Vs)
provides an average estimation of the nonlinear behavior of the entire profile in this study,
without consideration of the vertical inhomogeneity of the shear wave velocity variation,
because each KiK-net vertical array consists of only a pair of seismometers located at the
surface and the bottom of borehole. Although there are many vertical seismic arrays with
additional borehole seismometers available in other regions, it is difficult to use for nonlin-
ear characterization, because strong motions are relatively scarce for most of these arrays.
On the other hand, as shown in Fig. 2, the near-surface low-velocity layers have a greater
influence on the average velocity than the deep high-velocity layers, and it is commonly
assumed that the nonlinearity appears essentially in shallow soft layers.

13
Bulletin of Earthquake Engineering

Hardin and Black (1969) expressed the shear modulus of soil as a function of eleven
factors, mainly including effective stress, void ratio, saturation, strain history, temperature,
grain properties and soil structure. Hence the near-surface shear wave velocity is subjected
to a variety of environment conditions, such as ground surface perturbation, crustal stress
and atmospheric forcing. And it is widely known that soil is a kind of heterogeneous ani-
sotropic material. To clarify the relevance between the shear wave velocity variation and
nonlinear soil behavior, the horizontal anisotropy and the long-term change of elastic shear
wave velocity are discussed in this study.

4.1 Horizontal anisotropy

According to Coutant (1996) and Chandra et al. (2015), the relative percentage difference
of shear wave velocities between east–west and north–south directions (η) is employed to
evaluate the horizontal anisotropy for the eight stations, as shown in Eq. 15.

V − VsNS
𝜂 = √sEW × 100% (15)
VsEW VsNS

in which VsEW and VsNS are the velocities in the east–west and north–south directions,
respectively.
As shown in Fig.  5, no strong anisotropy is observed at four stations (FKSH10,
FKSH19, IBRH11 and TCGH16) with mean velocity anisotropy of less than 2.5%. By con-
trast, obvious anisotropy occurs at the other four stations (IBRH13, IWTH05, MYGH05
and TCGH07) with mean velocity anisotropy of greater than 4%. The strongest values are
found at TCGH07 station, where they are generally greater than 15%. Therefore, east–west
and north–south components are considered separately for the latter four stations. Further-
more, Chandra et al. (2015) suggested that there is no relationship between anisotropy and
the level of deformation. In this study, no variation of anisotropy is observed with respect
to PGA, which indicates the same conclusion for the eight stations.

4.2 Long‑term change of elastic shear wave velocity

Many studies have reported reductions of the near-surface shear wave velocity after large
earthquakes due to long-term change of soil properties caused by strong motions (Sawa-
zaki et al. 2009; Nakata and Snieder 2011, 2012; Takagi and Okada 2012; Sawazaki and
Snieder 2013). To quantify the effect of large earthquakes on elastic shear wave velocity
(Vs0) and to clarify their influence on soil nonlinear evaluations, we plot the shear wave
velocities of weak motions with PGA less than 20 cm/s2 as a function of time for the eight
KiK-net stations as shown in Fig. 6.
From Fig. 6, it can be found that the shear wave velocity of FKSH10 hardly changed
after the 2011 M ­ w9.0 Tohoku-Oki main shock. The reason for this is probably that no
long-term damage was caused by the main shock for FKSH10 considering that FKSH10
is a relatively stiff soil site with the highest Vs30 and the smallest Z760 among the eight
stations. For the other seven stations, the mean elastic velocities dropped a few percent
(1–5%) within 200  days after the main shock than those before the main shock. As dis-
cussed in Nakata and Snieder (2012), the velocity reduction can be due either to the open-
ing and closing of existing fractures, to the creation of new fractures, or to the change in

13
Bulletin of Earthquake Engineering

Fig. 5  Horizontal velocity anisotropy for the eight KiK-net stations

the shear modulus caused by changes in the pore fluid pressure because of shaking-induced
compaction.
Meanwhile, there was a long-term recovery process of elastic shear wave velocity
after the main shock for five stations (i.e., FKSH19, IBRH11, TCGH16, IBRH13 and
MYGH10). Additionally, for FKSH19 and TCGH16, the mean elastic velocities in the
last 200  days (February 2014–July 2014) were still obviously lower than those before
the main shock. Logarithmic recovery of seismic velocity has been observed by previ-
ous studies (Schaff and Beroza 2004; Peng and Ben-Zion 2006). Besides, laboratory
studies also found logarithmic recovery under normal stress due to increasing contact
area of crack surfaces or grains for geomaterial and sedimentary rocks (Marone 1998;

13
Bulletin of Earthquake Engineering

Fig. 6  Long-term change of elastic shear wave velocity for the eight KiK-net stations. Day 0 represents
December 31, 2008, the start time of our study period

13
Bulletin of Earthquake Engineering

Vakhnenko et  al. 2005). However, irreversible changes (e.g., the creation of new frac-
tures) would lead to a permanent velocity change that will never return to the pre-shock
level. For IWTH05, no recovery process of elastic shear wave velocity was observed
after the main shock, probably due to the irreversible changes caused by the strong
motion. For TCGH07, the shear wave velocity decreased slightly after the Tohoku-Oki
main shock, and returned to the pre-shock level quickly during about 200  days. How-
ever, the elastic velocity was significantly reduced by a large earthquake on February
25th, 2013. After the large earthquake, we could observe a long-term recovery process
of shear wave velocity during the last 500 days of our analysis time period, and the elas-
tic velocity did not return to pre-shock level until July 2014.
So far, the mechanisms and the controlling parameters for the recovery processes are
still under investigation. As discussed in Wu et al. (2009), the fluid diffusion is suggested
to play an important role in the recovery process. Although there is no direct evidence, if
the assumption of fluid diffusion holds in our case, the recovery processes would be mainly
controlled by the permeability of the site, which could also be associated with amplitude of
ground motion.

4.3 Comparison between velocities computed from logging data and by seismic


interferometry

In order to verify the reliability of the seismic interferometry, we display the mean values
and standard deviations of the elastic velocities by seismic interferometry for those eight
KiK-net stations in Fig. 7, considering east–west and north–south directions separately for
four of them. The velocities computed by seismic interferometry are compared with those
from logging data. As shown in Fig. 7, the data are concentrated along the 1:1 line, which
indicates a good agreement between the shear wave velocities obtained from logging data
and those from seismic interferometry.
From Fig. 7, it can be clearly seen that the stations on soft sites generally have smaller
standard deviations than those on hard sites, which is consistent with the results of Nakata
and Snieder (2012). The reason for this is that long travel time between the surface and
the bottom of borehole could lead to small relative errors for identifying travel time from
deconvolution function. Besides, the small dispersion of Vs values also suggests the high
accuracy and robustness of the seismic interferometry by deconvolution for measuring
in situ shear wave velocity using ground motion data from seismic vertical arrays.

Fig. 7  Crossplot of velocities
computed from logging data
and by seismic interferometry.
The solid circles indicate mean
values of the elastic velocities by
seismic interferometry, and the
ends of the vertical line represent
mean value ± standard deviation.
VS0 represents the time-averaged
shear wave velocity throughout
the entire profile

13
Bulletin of Earthquake Engineering

5 Characterizing nonlinear seismic site response using soil dynamic


parameters

For the purpose of this study, we quantitatively assess the relationships between the in situ
soil-dynamic parameters and the level of ground motions using hyperbolic models. Spe-
cifically, (1) as above mentioned, we use a classic two-parameter hyperbolic model (Eq. 9)
to fit the relationship between a* and γ; (2) in order to relate the shear strain to the level
of ground motion, we use the same hyperbolic model with Eq.  9 to fit the relationship
between PGA and γ as follows.
b2 𝛾
PGA = (16)
1 + k2 𝛾
(3) based on a widely used shear modulus reduction model (Eq.  17, which can be also
derived from Eq. 8) proposed by Hardin and Drnevich (1972), a one-parameter hyperbolic
model is employed to fit the relationship between the G/G0 and γ as listed in Eq. 18, where
the shear modulus degradation can be obtained using V2s /V2s0to represent G/G0 from Eq. 2.

1
G∕G0 = (17)
1 + 𝛾𝛾
r

1
G∕G0 =
1 + k3 𝛾 (18)

in which, γr is the reference shear strain at which the shear modulus G has reduced to 0.5G0,
and k3 is equivalent to the reciprocal of the reference shear strain. As shown in Fig.  10,
the scatter of G/G0 at small deformations is considerable due to the variation of elastic
shear wave velocities. Hence, the G/G0–γ curves are fitted with data from strong motions
(PGA > 100  cm/s2) to avoid the influence of significant variability of data from weak
motions. The regression relationships at the eight KiK-net stations are listed in Table 3 and
shown in Figs. 8, 9 and 10.
From the correlation coefficients listed in Table  3, it can be found that a* and PGA
are strongly related to γ, while G/G0 is moderately-to-strongly related to γ. As shown in
Fig. 8, a*–γ curves are used to represent the stress–strain relationship for identifying non-
linear soil behavior. Visually, obvious nonlinearity is observed at shear-strain level of about
1 × 10−4. Besides, Fig. 9 shows that the relationship between PGA and γ has a similar pat-
tern with that between a* and γ, due to the high relevance between PGA and a*. From
Fig. 10, it can be found that the shear modulus degradation starts at a quite low level of
strain of 5 × 10−6–5 × 10−5 corresponding to the linear threshold shear strain (γtl), and
becomes much more significant beyond a relatively high level of strain of 1 × 10−4 cor-
responding to the volumetric threshold shear strain (γtv) for those eight KiK-net stations.
Hence, it can be concluded that the threshold strains from strong motion observations are
well consistent with those from previous laboratory tests.
To get a better understanding of nonlinear soil behavior, the regression relationships
for those eight KiK-net stations are compared in Fig. 11. It is obvious that the in situ G–γ
curves are comparable with the laboratory measurements previously obtained by Vucetic
and Dobry (1991) in Fig.  11a, and their results are widely used in nonlinear site effect
studies (Wu et al. 2009, Zhao et al. 2015). Vucetic and Dobry (1991) suggested the shear
modulus degradation depends strongly on the PI, thus PI is the main factor controlling

13
Bulletin of Earthquake Engineering

Table 3  Regression relationships between soil dynamic parameters and the level of ground motion
Station Relationships Fitting parameters Goodness-of-fit
k (× 103) b (× 106) r σRES

FKSH10 PGA–γ 1.24 2.59 0.99 2.73


a*–γ 1.08 4.98 0.99 5.34
G/G0–γ 1.50 – 0.64 0.0118
FKSH19 PGA–γ 3.28 2.36 0.97 3.21
a*–γ 3.42 4.56 0.98 6.37
G/G0–γ 1.34 – 0.52 0.0234
IBRH11 PGA–γ 1.86 1.53 0.98 3.36
a*–γ 2.22 3.04 0.98 6.45
G/G0–γ 1.33 – 0.88 0.0200
TCGH16 PGA–γ 1.01 0.749 0.98 2.82
a*–γ 0.682 1.35 0.98 6.36
G/G0–γ 0.740 – 0.91 0.0184
IBRH13 (EW/NS) PGA–γ 4.43/3.00 2.52/2.52 0.98/0.97 5.33/7.71
a*–γ 5.27/3.86 5.07/5.16 0.98/0.97 10.3/14.5
G/G0–γ 2.65/2.35 – 0.63/0.62 0.0411/0.0480
IWTH05 (EW/NS) PGA–γ 7.95/7.75 3.97/4.29 0.97/0.96 3.90/4.77
a*–γ 7.05/4.13 6.88/6.88 0.97/0.96 7.52/10.1
G/G0–γ 3.12/4.28 – 0.50/0.58 0.0212/0.0319
MYGH10 (EW/NS) PGA–γ 1.85/2.34 1.80/1.61 0.97/0.96 4.59/5.08
a*–γ 1.85/1.78 3.50/2.89 0.96/0.96 8.61/10.0
G/G0–γ 0.719/0.649 – 0.68/0.65 0.0147/0.0298
TCGH07 (EW/NS) PGA–γ 2.16/3.73 2.46/3.10 0.98/0.98 6.06/7.95
a*–γ 3.67/2.49 4.85/5.49 0.98/0.98 10.6/16.4
G/G0–γ 5.34/5.32 – 0.73/0.68 0.0302/0.0375

k refers to k1, k2 and k3; b refers to b1 and b2; “1” refers to the PGA–γ relationship, “2” refers to the a*–γ
relationship, and “3” refers to the G/G0–γ relationship. r and σRES represent the Pearson correlation coeffi-
cient and standard deviation of residuals between the fit and the data, respectively

the locations of the modulus reduction curve. In addition to PI, the mean effective confin-
ing pressure is also an important factor that influences the G/G0–γ curve (Darendeli 2001;
Zhang et  al. 2005). In our study, according to Eq.  14, the equivalent depths z* for these
eight stations range between 18.4 and 25.6 m, indicating that the mean effective confining
pressures would be similar across different stations. Hence, PI is one of the most important
index properties for nonlinear soil behavior evaluations in our case.
The PI for the eight KiK-net stations can be determined by comparing the in situ modu-
lus reduction curves with the laboratory modulus reduction curves, as shown in Fig. 11a.
It is obvious that (1) MYGH10 and TCGH16 have relatively high PI values of about 40;
(2) FKSH10, FKSH19 and IBRH11 have moderate PI values of about 25; (3) IBRH13
has a relatively low PI value of about 10; (4) TCGH07 and IWTH05 have the lowest PI
values of about 0. These results are basically consistent with their geotechnical cross sec-
tions. According to Fig. 1 and Table 2, it can be found that the PI at each of the KiK-net
sites is likely to be related with the site class and the depth to engineering bedrock as fol-
lows: (1) those two class D stations with high depths to engineering bedrock (MYGH10

13
Bulletin of Earthquake Engineering

Fig. 8  Relationship between the average peak particle acceleration and shear strain. The hollow circles
indicate the mean values of the measured data, as well as the standard deviation for different bands of
strain, that is, 1 × 10−7–5 × 10−7, 5 × 10−7–1 × 10−6, 1 × 10−6–5 × 10−6, 5 × 10−6–1 × 10−5, 1 × 10−5–5 × 10−5,
5 × 10−5–1 × 10−4, 1 × 10−4–5 × 10−4, 5 × 10−4–1 × 10−3, > 10−3

13
Bulletin of Earthquake Engineering

Fig. 9  Relationship between the average peak ground acceleration and shear strain. The hollow circles
indicate the mean values of the measured data, as well as the standard deviation for different bands of
strain, that is, 1 × 10−7–5 × 10−7, 5 × 10−7–1 × 10−6, 1 × 10−6–5 × 10−6, 5 × 10−6–1 × 10−5, 1 × 10−5–5 × 10−5,
5 × 10−5–1 × 10−4, 1 × 10−4–5 × 10−4, 5 × 10−4–1 × 10−3, > 10−3

13
Bulletin of Earthquake Engineering

Fig. 10  Relationship between the shear modulus ratio and shear strain. The hollow circles indicate
the mean values of the measured data, as well as the standard deviation for different bands of strain,
that is, 1 × 10−7–5 × 10−7, 5 × 10−7–1 × 10−6, 1 × 10−6–5 × 10−6, 5 × 10−6–1 × 10−5, 1 × 10−5–5 × 10−5,
5 × 10−5–1 × 10−4, 1 × 10−4–5 × 10−4, 5 × 10−4–1 × 10−3, > 10−3

13
Bulletin of Earthquake Engineering

Fig. 11  Regression relationships between shear strain and a shear modulus ratio; b average peak particle
acceleration; c peak ground acceleration. The in  situ G–γ relationships are compared with the laboratory
measurements previously obtained by Vucetic and Dobry (1991) in panel (a)

and TCGH16) have the highest PI values; (2) the other three class D stations with low
depths to engineering bedrock (IBRH11, FKSH19, IBRH13) have lower PI values; (3) with
an exception of FKSH10, the other two class C stations (IWTH05 and TCGH07) have the
lowest PI values. As shown in Fig. 11b, c, the representations of stress, a* and PGA, sat-
isfy Hooke’s law at small deformation. And it can be found that, for a given strain, the
stress generally increases with the increase of average shear wave velocity. Because the
locations of the a*–γ curves as well as the PGA–γ curves are mainly controlled by the aver-
age shear wave velocity, considering that the shear wave velocity has a direct relation to
shear modulus as shown in Eq. 2.
Furthermore, the shear modulus degradation for a given PGA can be determined by the
PGA–γ and G/G0–γ curves. Based on the regression relationships, the in  situ soil dynamic
parameters at four different levels of ground motion are calculated and listed in Table  4.
From Table 4, it can be found that the degree of nonlinear soil behavior increases as the PGA
increases beyond the nonlinear threshold: for the eight KiK-net stations, (1) when PGA comes
to 100 cm/s2, the soil exhibits slight nonlinearity with a shear modulus degradation ranging
from 2 to 10%; (2) when PGA comes to 200 cm/s2, the soil exhibits obvious nonlinearity with
a shear modulus degradation ranging from 4 to 18%; (3) when PGA comes to 400 cm/s2, the
soil exhibits strong nonlinearity with a shear modulus degradation ranging from 9 to 33%; (4)
when PGA comes to 800 cm/s2, the soil exhibits extreme nonlinearity with a shear modulus
degradation ranging from 22 to 69%.

13
Bulletin of Earthquake Engineering

Table 4  In-situ soil dynamic parameters for different levels of ground motion


Station PGA Soil dynamic parameters
(cm/s2) γ (× 10−4) τ (× 103 N/m2) G (× 108 N/m2) 100 (1 − G/G0)

FKSH10 100 0.205 29.2 14.2 2.76


200 0.420 58.2 13.8 5.27
400 0.880 116 13.1 10.1
800 1.94 229 11.8 19.1
FKSH19 100 0.237 27.9 12.4 2.22
200 0.516 56.1 12.1 4.51
400 1.25 113 11.5 9.83
800 4.39 229 9.37 26.2
IBRH11 100 0.355 24.6 7.08 4.20
200 0.770 49.8 6.78 8.35
400 1.86 102 6.11 17.3
800 6.33 215 4.44 39.9
TCGH16 100 0.780 23.5 3.08 4.89
200 1.65 45.8 2.94 9.42
400 3.71 79.6 2.41 18.3
800 9.95 146 1.88 36.2
IBRH13 (EW/NS) 100 0.220/0.209 22.6/20.1 10.8/9.89 4.10/2.91
200 0.498/0.456 45.9/40.9 10.3/9.59 8.45/5.87
400 1.35/1.11 95.0/84.6 9.10/8.91 19.1/12.6
800 9.37/3.86 205/182 4.56/6.94 59.5/31.9
IWTH05 (EW/NS) 100 0.162/0.155 28.0/30.1 18.9/20.9 3.00/4.17
200 0.366/0.330 55.3/57.3 18.3/20.0 6.25/8.18
400 0.985/0.765 108/105 16.7/18.2 14.5/16.4
800 6.45/2.24 205/178 9.76/14.1 50.0/35.1
MYGH10 (EW/NS) 100 0.302/0.369 22.8/24.0 7.84/6.96 1.71/1.58
200 0.640/0.790 45.6/47.1 7.70/6.85 3.42/3.18
400 1.45/1.84 91.2/90.8 7.41/6.59 7.12/6.80
800 3.96/5.46 182/170 6.66/5.87 16.5/16.9
TCGH07 (EW/NS) 100 0.223/0.191 21.6/26.5 9.25/13.6 9.00/8.41
200 0.486/0.401 44.7/51.9 8.43/12.6 17.1/15.6
400 1.18/0.890 95.8/99.5 6.89/10.7 32.3/28.2
800 4.18/2.29 223/184 3.96/7.61 61.1/48.9

6 The influence of the PI on soil nonlinear behavior evaluations

As mentioned above, Vucetic and Dobry (1991) demonstrated that PI is the main factor
controlling the locations of the G/G0–γ curve for a wide variety of soils ranging from
clays to sands. For a given γ, as the PI goes up, G/G0 rises. And soils with higher PI
tend to have a higher linear threshold strain. From a view of microstructure, soils with
high PI consist of small particles and have large interparticle contacts, so they have the
ability to take relatively large shear strain before the particles are permanently displaced

13
Bulletin of Earthquake Engineering

and stiffness degradation effects appear (Vucetic and Dobry 1991). Hence, PI is one of
the most important index properties for nonlinear soil behavior evaluations. Meanwhile,
considering that PI is a widely used soil index that can be obtained at low cost, discus-
sions on the influence of the PI on soil nonlinear behavior evaluations have high poten-
tial value for engineering application.
As discussed previously, the locations of the PGA–γ curves are mainly controlled by the
average shear wave velocity. For a given γ, as the Vs goes up, PGA increases. By dividing the
numerator and denominator by the common factor b2γ in the right side of Eq. 16, the relation-
ship between PGA and γ can be expressed by Eq. 19.

1
PGA = 1
+
k2 (19)
b2 𝛾 b2

Based on the regression parameters of the eight KiK-net stations listed in Table 3, we find
b2 is strongly correlated with the average velocity as shown in Fig. 12a. Using a linear regres-
sion between b2 and Vs0, we can get Eq. 20.
b2 = 9.18 × 103 × Vs0 − 1.99 × 106 (20)

Fig. 12  a Relationship between b2 and Vs0; b distribution of k2/b2 with respect to Vs0; c Relationship
between k3 and PI; d Relationship between γtl and PI. For panels a and b, those twelve points are derived
from the regression relationships for the eight stations, since east–west and north–south components are
considered separately for four of these eight stations. For panels c and d, those four points are derived from
the modulus reduction curves from Vucetic and Dobry (1991), which correspond to the curves with PI = 0,
15, 35, 50, respectively

13
Bulletin of Earthquake Engineering

Assuming k2/b2 to be a constant (as shown in Figs.  12b) and substituting Eq.  20 into
Eq. 19, the empirical velocity-dependent PGA–γ relationship is given by Eq. 21 (as shown in
Fig. 13a).

1
PGA = 1
+ 6.64 × 10−4 (21)
(9.18×103 ×Vs0 −1.99×106 )×𝛾

Based on the laboratory measurements from Vucetic and Dobry (1991), the regression-
based G/G0–γ relationships for PI = 0, 15, 35, 50 are established using the one-parameter
hyperbolic model of Eq. 18. Meanwhile, it can be found that k3 is strongly correlated with
PI as shown in Fig. 12c and listed in Eq. 22.

3.71 × 103
k3 = (22)
1.043PI
Substituting Eqs. 21 and 22 into Eq. 18, we can get the Vs0- and PI-based G/G0-PGA
relationship as shown in Eq. 23.
1
G∕G0 =
1+ 3.71
(
1
) (23)
1.043PI ×(9.18Vs0 −1.99×103 )× PGA −6.64×10−4

As shown in Fig. 13b, the shear modulus degradation is somewhat overestimated by the


empirical G/G0–γ curves at very small shear strains, which the shear modulus degradation
is negligible for γ < γtl. To gain further insight into the threshold of nonlinear soil behavior,
based on the laboratory measurements from Vucetic and Dobry (1991), an empirical rela-
tionship between the linear threshold shear strain (γtl) and PI is built as shown in Eq. 24 and
Fig. 12d, where γtl corresponds to G/G0 = 0.99 in this study as shown in Fig. 13b. Vucetic
(1994) also proposed a correlation between the linear threshold shear strain (γtl) and PI as
shown in Fig. 12d. It can be found that for a given PI, the γtl in this study is slightly higher
than that in Vucetic (1994).
𝛾tl = 5.74 × 10−6 × 1.041PI (24)

Fig. 13  a Regression-based PGA–γ relationship; b regression-based G/G0–γ relationships

13
Bulletin of Earthquake Engineering

Table 5  Nonlinear threshold Vs0 (m/s) PI


PGA for different PI and Vs0
(cm/s2) 0 10 20 30 40 50

400 10 14 21 32 47 69
600 20 30 44 65 94 137
800 30 45 66 96 139 199
1000 40 59 87 126 181 256

Fig. 14  Nonlinear threshold PGA


for different PI and Vs0

Substituting Eq. 24 into Eq. 21, we can get the Vs0- and PI-based linear threshold PGA
­(PGAtl, also called nonlinear threshold PGA) model as shown in Eq. 25.

5.74 × 1.041PI
PGAtl =
103
+ 6.64 × 10−4 × 5.74 × 1.041PI (25)
9.18Vs0 −1.99×103

According to Eq. 25, the nonlinear threshold PGA is mainly controlled by the stiff-
ness (represented by Vs0) and PI. For a given PI, nonlinear threshold PGA increases
with the increase of Vs0; for a given Vs0, nonlinear threshold PGA increases with the
increase of PI. We calculate the P ­ GAtl for different PI and Vs0 as listed in Table  5
and shown in Fig. 14. From Table 6, it can be concluded that the nonlinear threshold
PGA generally ranges from 20 to 100 cm/s2. Besides, if the stiffness increases and PI
decreases, the nonlinear threshold PGA may increase or decrease. Hence, the nonlinear
threshold PGA of a stiff soil site may or may not be greater than that of a soft soil site,
and this may explain the poor correlation between nonlinear threshold PGA and site
classes in previous studies (Ghofrani et al. 2013; Régnier et al. 2013).
According to Eq. 23, the degree of soil nonlinear behavior is mainly controlled by
the stiffness, PI, and the level of ground motion. For a given Vs0 and a given PGA,
shear modulus degradation decreases with the increase of PI; for a given PI and a given
PGA, shear modulus degradation decreases with the increase of Vs0; for a given PI and
a given Vs0, shear modulus degradation increases with the increase of PGA. The shear

13
Bulletin of Earthquake Engineering

Table 6  Shear modulus Vs0 (m/s) PGA (cm/s2) PI


degradation for different PI, Vs0
and PGA (%) 0 10 20 30 40 50

400 200 34 25 18 13 9 6
400 55 44 34 25 18 13
600 69 59 49 38 29 21
800 79 71 62 52 41 31
600 200 20 14 9 6 4 3
400 36 27 20 14 10 7
600 51 41 31 23 16 11
800 64 54 44 34 25 18
800 200 14 9 6 4 3 2
400 27 20 14 10 7 4
600 41 31 23 16 11 8
800 54 44 34 25 18 13
1000 200 11 7 5 3 2 1
400 22 16 11 7 5 3
600 34 25 18 13 9 6
800 47 37 28 20 14 10

Fig. 15  Shear modulus degradation for different PI, Vs0 and PGA

13
Bulletin of Earthquake Engineering

modulus degradation for different PI, Vs0 and PGA is calculated as listed in Table  6
and shown in Fig. 15. Assume there are four typical sites, and their average velocities
are 400, 600, 800, 1000 m/s, respectively; their PI are 40, 20, 10, 0, respectively. From
Table 6, it can be found that their shear modulus degradations are quite close to each
other. When PGA comes to 200, 400, 600 and 800 cm/s2, the shear modulus degrada-
tion is 9–11%, 18–22%, 29–34% and 41–47%, respectively. Hence, for a given PGA, if
the stiffness increases and PI decreases, the shear modulus degradation may increase
or decrease; that is to say, the degree of nonlinear soil behavior of a soft soil site may
or may not be higher than that of a stiff soil site for a given level of ground motion.

7 Conclusions

In this study, we utilize ground motion records at eight KiK-net vertical arrays in Japan
to analyze nonlinear seismic site response. First, the in situ shear modulus degradation
is assessed from the variation of shear wave velocity (Vs) using seismic interferome-
try. Considering each record as an in situ cyclic test under seismic loading, the in situ
stress–strain relations of soils are extracted from seismic data. Then we quantitatively
assess the relationships between the soil-dynamic parameters and the level of ground
motion. Finally, a parametric analysis is presented to discuss the influence of the plastic-
ity index on nonlinear soil behavior evaluations. The main conclusions of this study are
summarized as follows:

1. The velocities computed by seismic interferometry are well consistent with those from
logging data. Besides, the small dispersion of Vs values computed by seismic interfer-
ometry also suggests the high accuracy and robustness of the seismic interferometry by
deconvolution for measuring in situ shear wave velocity using ground motion data from
seismic vertical arrays.
2. The shear stress proxies (a* and PGA) are strongly related to the shear strain proxy
(v*/Vs), while the shear modulus ratio (G/G0) is moderately-to-strongly related to the
shear strain proxy. This confirms the good prediction of the nonlinear hyperbolic mod-
els using the strain proxy derived from in situ data, which is also inverted with present
GMPE model that fit the data with moderate scattering using Vs30.
3. The threshold strains from ground motion observations are well consistent with those
from previous laboratory tests. For those eight KiK-net stations, nonlinearity starts at a
quite low level of strain of 5 × 10−6–5 × 10−5 corresponding to the linear threshold shear
strain, and becomes much more significant beyond a relatively high level of strain of
1 × 10−4 corresponding to the volumetric threshold shear strain.
4. In our case, the nonlinear threshold PGA generally ranges from 20 to 100 cm/s2, and
it is mainly controlled by the stiffness and PI. For a given PI, nonlinear threshold PGA
increases with the increase of Vs0; for a given Vs0, nonlinear threshold PGA increases
with the increase of PI. Considering that a stiff soil site probably has a relatively low PI
and a soft soil site probably has a relatively high PI, the nonlinear threshold PGA of a
stiff soil site may or may not be greater than that of a soft soil site.
5. For those eight KiK-net stations, when PGA comes to 100, 200, 400 and 800 cm/s2, the
shear modulus degradation is 2–10%, 4–18%, 9–33% and 22–69%, respectively. In our
case, the degree of soil nonlinear behavior is mainly controlled by the stiffness, PI, and
the level of ground motion. For a given Vs0 and a given PGA, shear modulus degrada-

13
Bulletin of Earthquake Engineering

tion decreases with the increase of PI; for a given PI and a given PGA, shear modulus
degradation decreases with the increase of Vs0; for a given PI and a given Vs0, shear
modulus degradation increases with the increase of PGA.

Acknowledgements  This study was financially supported by the National Natural Science Foundation of
China (Nos. 41372335, 51778260, 51378234). It is gratefully acknowledged that the velocity profiles and
earthquake records were provided by NIED in Japan.

Data and resources  The shear-wave velocity profiles and seismic data used in this study
were provided by National Research Institute for Earth Science and Disaster Prevention
(NIED) in Japan, and they are available online (http://www.kyosh​in.bosai​.go.jp/kyosh​in/
db/index​_en.html?all, last accessed April 2018). A detail introduction to KiK-net can be
found in Okada et al. (2004).

References
Abrahamson NA, Silva WJ, Kamai R (2014) Summary of the ASK14 ground motion relation for active
crustal regions. Earthq Spectra 30(3):1025–1055
Aki K (1993) Local site effects on weak and strong ground motion. Tectonophysics 218(1–3):93–111
Askan A, Akcelik V, Bielak J, Ghattas O (2007) Full waveform inversion for seismic velocity and anelastic
losses in heterogeneous structures. Bull Seismol Soc Am 97(7):1990–2008
Assimaki D, Li W, Steidl J, Schmedes J (2008) Quantifying nonlinearity susceptibility via site-response
modeling uncertainty at three sites in the Los Angeles Basin. Bull Seismol Soc Am 98(5):2364–2390
Beresnev IA, Wen KL (1996) Nonlinear soil response—A reality? Bull Seismol Soc Am 86(6):1964–1978
Bonilla LF, Archuleta RJ, Lavallée D (2005) Hysteretic and dilatant behavior of cohesionless soils and
their effects on nonlinear site response: field data observations and modeling. Bull Seismol Soc Am
95(6):2373–2395
Bonilla LF, Gueguen P, Lopez-Caballero F, Mercerat ED, Gélis C (2017) Prediction of non-linear site
response using downhole array data and numerical modeling: the Belleplaine (Guadeloupe) case study.
Phys Chem Earth Parts A/B/C 98:107–118
Boore DM, Thompson EM, Cadet H (2011) Regional correlations of VS 30 and velocities averaged over
depths less than and greater than 30 meters. Bull Seismol Soc Am 101(6):3046–3059
Boore DM, Stewart JP, Seyhan E, Atkinson GM (2014) NGA-West2 equations for predicting PGA, PGV,
and 5% damped PSA for shallow crustal earthquakes. Earthq Spectra 30(3):1057–1085
Bozorgnia Y, Abrahamson NA, Atik LA, Ancheta TD, Atkinson GM, Baker JW, Darragh R (2014) NGA-
West2 research project. Earthq Spectra 30(3):973–987
Building Seismic Safety Council (BSSC) (2015) NEHRP recommended seismic provisions for new build-
ings and other structures, Washington D.C., FEMA P-1050
Cadet H, Bard PY, Rodriguez-Marek A (2012) Site effect assessment using KiK-net data: Part 1. A simple
correction procedure for surface/downhole spectral ratios. Bull Earthq Eng 10(2):421–448
Campbell KW, Bozorgnia Y (2014) NGA-West2 ground motion model for the average horizontal com-
ponents of PGA, PGV, and 5% damped linear acceleration response spectra. Earthq Spectra
30(3):1087–1115
Carlton B, Tokimatsu K (2015) Comparison of equivalent linear and nonlinear site response analysis results
and model to estimate maximum shear strain. Earthq Spectra 32(3):1867–1887
Chandra J, Guéguen P, Steidl JH, Bonilla LF (2015) In situ assessment of the G–γ curve for characterizing
the nonlinear response of soil: application to the Garner Valley downhole array and the wildlife lique-
faction array. Bull Seismol Soc Am 105(2A):993–1010

13
Bulletin of Earthquake Engineering

Chandra J, Guéguen P, Bonilla LF (2016) PGA-PGV/Vs considered as a stress–strain proxy for predicting
nonlinear soil response. Soil Dyn Earthq Eng 85:146–160
Chávez-García FJ, Faccioli E (2000) Complex site effects and building codes: making the leap. J Seismol
4(1):23–40
Chin BH, Aki K (1991) Simultaneous study of the source, path, and site effects on strong ground motion
during the 1989 Loma Prieta earthquake: a preliminary result on pervasive nonlinear site effects. Bull
Seismol Soc Am 81(5):1859–1884
Chiou BSJ, Youngs RR (2014) Update of the Chiou and Youngs NGA model for the average horizontal
component of peak ground motion and response spectra. Earthq Spectra 30(3):1117–1153
Coutant O (1996) Observation of shallow anisotropy on local earthquake records at the Garner Valley,
Southern California, downhole array. Bull Seismol Soc Am 86(2):477–488
Darendeli MB (2001) Development of a new family of normalized modulus reduction and material damping
curves. Ph.D. thesis, University of Texas at Austin, Austin, Texas
Dobry R (2013) Radiation damping in the context of one-dimensional wave propagation: a teaching per-
spective. Soil Dyn Earthq Eng 47:51–61
Dobry R, Ladd RS (1980) Soil liquefaction and cyclic mobility evaluation for level ground during earth-
quakes and liquefaction potential: science versus practice. J Geotech Geoenviron Eng 106(6):720–724
Dobry R, Vucetic M (1987) Dynamic properties and seismic response of soft clay deposits. In: Proceedings
of international symposium on geotechnical engineering of soft soils, 13–14 August 1987, Mexico
City, Mexico
Drnevich VP, Richart FE (1970) Dynamic prestraining of dry sand. J Soil Mech Found Div 96(2):453–469
European Committee for Standardization (CEN) (2013) Eurocode 8: design of structures for earthquake
resistance, part 1: general rules, seismic actions and rules for buildings, EN 1998-1. European Com-
mittee for Standardization (CEN), Brussels
Field EH, Johnson PA, Beresnev IA, Zeng Y (1997) Nonlinear ground-motion amplification by sediments
during the 1994 Northridge earthquake. Nature 390(6660):599
Field EH, Kramer S, Elgamal AW, Bray JD, Matasovic N, Johnson PA, Dimitriu PP (1998) Nonlinear site
response: where we’re at (a report from a SCEC/PEER seminar and workshop). Seismol Res Lett
69(3):230–234
Frankel AD, Carver DL, Williams RA (2002) Nonlinear and linear site response and basin effects in Seattle
for the M 6.8 Nisqually, Washington, earthquake. Bull Seismol Soc Am 92(6):2090–2109
Ghofrani H, Atkinson GM, Goda K (2013) Implications of the 2011 M9.0 Tohoku Japan earthquake for the
treatment of site effects in large earthquakes. Bull Earthq Eng 11(1):171–203
Guéguen P (2016) Predicting nonlinear site response using spectral acceleration vs PGV/VS30: a case his-
tory using the Volvi-test site. Pure appl Geophys 173(6):2047–2063
Hardin BO, Black WL (1969) Closure to vibration modulus of normally consolidated clays. J Soil Mech
Found Div 94(2):1531–1537
Hardin BO, Drnevich VP (1972) Shear modulus and damping in soils: measurement and parameter effects. J
Soil Mech Found Div ASCE 98(SM6):603–624
Haskell NA (1953) The dispersion of surface waves on multilayered media. Bull Seismol Soc Am
43(1):17–34
Idriss IM (2011) Use of Vs30 to represent local site conditions. In: Proceedings of the 4th IASPEI/IAEE
international symposium. Effects of source geology on seismic motion, California, USA
Idriss IM (2014) An NGA-West2 empirical model for estimating the horizontal spectral values generated by
shallow crustal earthquakes. Earthq Spectra 30(3):1155–1177
Ishihara K (1996) Soil behaviour in earthquake geotechnics. Clarendon Press, Oxford
Iwasak Y, Tai M (1996) Strong motion records at Kobe Port Island. Soils Found 36(Special):29–40
Kaklamanos J, Bradley B, Thompson E, Baise L (2013) Critical parameters affecting bias and variability in
site-response analyses using KiK-net downhole array data. Bull Seismol Soc Am 103(3):1733–1749
Kaklamanos J, Baise LG, Thompson EM, Dorfmann L (2015) Comparison of 1D linear, equivalent-linear,
and nonlinear site response models at six KiK-net validation sites. Soil Dyn Earthq Eng 69:207–219
Kim B, Hashash YM (2013) Site response analysis using downhole array recordings during the March
2011 Tohoku-Oki earthquake and the effect of long-duration ground motions. Earthq Spectra
29(s1):S37–S54
Kim B, Hashash YM, Stewart JP, Rathje EM, Harmon JA, Musgrove MI, Silva WJ (2016) Relative dif-
ferences between nonlinear and equivalent-linear 1-D site response analyses. Earthq Spectra
32(3):1845–1865
Kishida T, Haddadi H, Darragh RB, Kayen RE, Silva WJ, Bozorgnia Y (2018) Apparent wave velocity and
site amplification at the California Strong Motion Instrumentation Program Carquinez Bridge geotech-
nical arrays during the 2014 M6.0 South Napa earthquake. Earthq Spectra 34(1):327–347

13
Bulletin of Earthquake Engineering

Li G, Motamed R, Dickenson S (2018) Evaluation of one-dimensional multi-directional site response analy-


ses using geotechnical downhole array data in California and Japan. Earthq Spectra 34(1):349–376
Marone C (1998) laboratory-derived friction laws and their application to seismic faulting. Annu Rev Earth
Planet Sci 26(1):643–696
Maufroy E, Cruz-Atienza V, Cotton F, Gaffet S (2015) Frequency-scaled curvature as a proxy for topo-
graphic site-effect amplification and ground-motion variability. Bull Seismol Soc Am 105(1):354–367
Maufroy E, Chaljub E, Theodoulidis NP, Roumelioti Z, Hollender F, Bard PY et al (2017) Source-related
variability of site response in the Mygdonian Basin (Greece) from accelerometric recordings and 3D
numerical simulations. Bull Seismol Soc Am 107(2):787–808
Miyazawa M, Snieder R, Venkataraman A (2008) Application of seismic interferometry to extract P- and
S-wave propagation and observation of shear-wave splitting from noise data at Cold Lake, Alberta,
Canada. Geophysics 73(4):D35–D40
Nakata N, Snieder R (2011) Near-surface weakening in Japan after the 2011 Tohoku-Oki earthquake. Geo-
phys Res Lett 38(17):245–255
Nakata N, Snieder R (2012) Estimating near-surface shear wave velocities in Japan by applying seismic
interferometry to KiK-net data. J Geophys Res Solid Earth 117(B1):B01308
Okada Y, Kasahara K, Hori S, Obara K et al (2004) Recent progress of seismic observation networks in
Japan—Hi-net, F-net, K-net and KiK-net. Earth Planets Space 56(8):xv–xxviii
Pavlenko OV, Irikura K (2002) Changes in shear moduli of liquefied and nonliquefied soils during
the 1995 Kobe earthquake and its aftershocks at three vertical-array sites. Bull Seismol Soc Am
92(5):1952–1969
Pavlenko OV, Irikura K (2003) Estimation of nonlinear time-dependent soil behavior in strong ground
motion based on vertical array data. Pure Appl Geophys 160(12):2365–2379
Peng Z, Ben-Zion Y (2006) Temporal changes of shallow seismic velocity around the Karadere-Düzce
branch of the north Anatolian Fault and strong ground motion. Pure Appl Geophys 163(2):567–600
Rathje EM, Chang WJ, Stokoe KH, Cox BR (2004) Evaluation of ground strain from in  situ dynamic
response. In: Proceeding of the 13th world conference on earthquake engineering
Régnier J, Cadet H, Bonilla LF, Bertrand E, Semblat JF (2013) Assessing nonlinear behavior of soils
in seismic site response: statistical analysis on KiK-net strong-motion data. Bull Seismol Soc Am
103(3):1750–1770
Régnier J, Bonilla LF, Bard PY et al (2016) International benchmark on numerical simulations for 1D,
nonlinear site response (prenolin): verification phase based on canonical cases. Bull Seismol Soc
Am 106(5):2112–2135
Ren Y, Wen R, Yao X, Ji K (2017) Five parameters for the evaluation of the soil nonlinearity during the
Ms8.0 Wenchuan Earthquake using the HVSR method. Earth Planets Space 69(1):116
Rubinstein JL (2011) Nonlinear site response in medium magnitude earthquakes near Parkfield, Califor-
nia. Bull Seismol Soc Am 101(1):275–286
Sawazaki K, Snieder R (2013) Time-lapse changes of P- and S-wave velocities and shear wave split-
ting in the first year after the 2011 Tohoku earthquake, Japan: shallow subsurface. Geophys J Int
193(1):238–251
Sawazaki K, Sato H, Nakahara H, Nishimura T (2009) Time-lapse changes of seismic velocity in the
shallow ground caused by strong ground motion shock of the 2000 Western-Tottori earthquake,
Japan, as revealed from coda deconvolution analysis. Bull Seismol Soc Am 99(1):352–366
Schaff D, Beroza G (2004) Coseismic and postseismic velocity changes measured by repeating earth-
quakes. J Geophys Res Solid Earth 109(B10):B10302
Seed HB, Wong RT, Idriss IM, Tokimatsu K (1986) Moduli and damping factors for dynamic analyses
of cohesionless soils. J Geotech Eng 112(11):1016–1032
Senetakis K, Anastasiadis A, Pitilakis K (2012) Dynamic properties of dry sand/rubber (SRM) and
gravel/rubber (GRM) mixtures in a wide range of shearing strain amplitudes. Soil Dyn Earthq Eng
33(1):38–53
Stewart JP, Afshari K, Hashash YM (2014) Guidelines for performing hazard-consistent one-dimen-
sional ground response analysis for ground motion prediction. PEER Rep 2014:16
Takagi R, Okada T (2012) Temporal change in shear velocity and polarization anisotropy related to the
2011 M9.0 Tohoku-Oki earthquake examined using KiK-net vertical array data. Geophys Res Lett
39(9):9310
Thompson EM, Baise LG, Tanaka Y, Kayen RE (2012) A taxonomy of site response complexity. Soil
Dyn Earthq Eng 41:32–43
Vakhnenko O, Vakhnenko V, Shankland T (2005) Soft-ratchet modeling of end-point memory in the
nonlinear resonant response of sedimentary rocks. Phys Rev B 71(17):174103

13
Bulletin of Earthquake Engineering

Vardanega PJ, Bolton MD (2013) Stiffness of clays and silts: normalizing shear modulus and shear
strain. J Geotech Geoenviron Eng 139(9):1575–1589
Vucetic M (1994) Cyclic threshold shear strains in soils. J Geotech Eng 120(12):2208–2228
Vucetic M, Dobry R (1991) Effect of soil plasticity on cyclic response. J Geotech Eng 117(1):89–107
Wang SY, Wang HY (2016) Site-dependent shear-wave velocity equations versus depth in California and
Japan. Soil Dyn Earthq Eng 88:8–14
Wang SY, Wang HY, Li Q (2017) An alternative method for estimating Vs(30) from a shallow shear-
wave velocity profile (depth < 30 m). Soil Dyn Earthq Eng 99:68–73
Wen KL, Beresnev IA, Yeh YT (1994) Nonlinear soil amplification inferred from downhole strong seis-
mic motion data. Geophys Res Lett 21(24):2625–2628
Wu C, Peng Z, Assimaki D (2009) Temporal changes in site response associated with the strong ground
motion of the 2004 Mw6.6 mid-Niigata earthquake sequences in Japan. Bull Seismol Soc Am
99(6):3487–3495
Wu C, Peng Z, Ben-Zion Y (2010) Refined thresholds for non-linear ground motion and temporal changes
of site response associated with medium-size earthquakes. Geophys J Int 182(3):1567–1576
Yamada S, Hyodo M, Orense RP, Dinesh SV (2008) Initial shear modulus of remolded sand-clay mixtures. J
Geotech Geoenviron Eng 134(7):960–971
Zalachoris G, Rathje EM (2015) Evaluation of one-dimensional site response techniques using borehole
arrays. J Geotech Geoenviron Eng 141(12):04015053
Zeghal M, Elgamal AW (1994) Analysis of site liquefaction using earthquake records. J Geotech Eng
120(6):996–1017
Zeghal M, Elgamal AW, Tang HT, Stepp JC (1995) Lotung downhole array. II: evaluation of soil nonlinear
properties. J Geotech Eng 121(4):363–378
Zhang J, Andrus R, Juang C (2005) Normalized shear modulus and material damping ratio relationships. J
Geotech Geoenviron Eng 131(4):453–464
Zhao JX, Hu J, Jiang F, Zhou J, Zhang Y, An X, Rhoades DA (2015) Nonlinear site models derived from
1D analyses for ground-motion prediction equations using site class as the site parameter. Bull Seismol
Soc Am 105(4):2010–2022

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13
View publication stats

You might also like