Nonlinear Site Response Analysis with Pore-Water

Pressure Generation for Liquefaction

Triggering Evaluation
Scott M. Olson, Ph.D., P.E., M.ASCE 1; Xuan Mei, S.M.ASCE 2; and
Youssef M. A. Hashash, Ph.D., P.E., F.ASCE 3
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Abstract: The cyclic-stress approach is widely used to evaluate level-ground liquefaction triggering. Although easy to use, several lim-
itations introduce significant uncertainty in the analysis, including: (1) several correction factors are required, including the depth reduction,
magnitude scaling, and overburden correction factors; (2) seismic demand is quantified using a total-stress framework to capture an effective
stress phenomenon [pore-water pressure (PWP) generation and liquefaction]; and (3) because it is based on surface manifestations, its appli-
cability outside of database parameters (e.g., depths > 10 m) is unknown. In this study, the authors performed a broad parametric study to
assess the viability of using nonlinear site response analysis with validated constitutive and PWP generation models to evaluate level-ground
liquefaction. For a wide range of conditions, the parametric results agreed with published empirical liquefaction-triggering relations. The
nonlinear site response analysis with PWP generation also correctly predicted liquefaction for dynamic centrifuge tests and field cases,
demonstrating that this approach can assess level-ground liquefaction while avoiding highly uncertain correction factors required in the
cyclic stress method. DOI: 10.1061/(ASCE)GT.1943-5606.0002191. © 2019 American Society of Civil Engineers.

Introduction 0 = initial effective vertical

stress at the depth being evaluated; σvo
stress at the depth being evaluated; and rd = depth reduction
In the past 40 years, methods to evaluate liquefaction triggering of coefficient.
level to mildly sloping ground have not changed greatly. The most A factor of safety against liquefaction triggering then can be
widely used method is the cyclic stress method pioneered by Seed computed as
and Idriss (1971) and Whitman (1971), which compares the cyclic
resistance ratio (CRR) of a soil layer (i.e., the cyclic shear stress
CRR7.5
ratio required to trigger liquefaction) to the earthquake-induced FSliq ¼ · MSF · K σ ð2Þ
cyclic (shear) stress ratio (CSR). The CRR for level ground is a CSR
function of the soil relative density (Dr ) and effective vertical stress
(σv0 ) as represented by a field index measurement [e.g., standard where CRR7.5 = cyclic resistance ratio corresponding to a moment
penetration test (SPT) blow count (N 60 ), cone penetration test magnitude (M) 7.5 earthquake; MSF = magnitude scaling factor
(CPT) tip resistance (qc ), or shear wave velocity (V s )]. Seismic de- that is used to empirically account for the number of cycles (i.e., du-
mand is estimated using the simplified equation proposed by Seed ration of shaking) corresponding to a M7.5 earthquake; and K σ =
and Idriss (1971), as adjustment factor to incorporate the effect of effective vertical stress
(overburden stress) on liquefaction resistance. Numerous liquefac-
τ avg amax σvo tion triggering curves using this cyclic stress approach have been
CSR ¼ 0 ¼ 0.65 0 rd ð1Þ
σvo g σvo proposed (e.g., Youd et al. 2001; Cetin et al. 2004; Idriss and
Boulanger 2010, among others for the SPT; Stark and Olson 1995;
where amax = ground surface peak ground acceleration (PGA; in Robertson and Wride 1998; Moss et al. 2006; Boulanger and Idriss
units of g); g = acceleration of gravity; σvo = initial total vertical 2012, among others for the CPT; Andrus and Stokoe 2000; Zhou
and Chen 2007; Kayen et al. 2013, among others for shear wave
1
Professor, Dept. of Civil and Environmental Engineering, Univ. of velocity) and are used widely in practice.
Illinois at Urbana–Champaign, Urbana, IL 61801. Email: olsons@illinois Although simple, there are numerous limitations related to
.edu evaluating liquefaction using the cyclic stress method. First, several
2 corrections such as the depth reduction factor (rd ), magnitude scal-
Postdoctoral Scholar, Dept. of Civil Engineering, Tongji Univ.,
Shanghai 200092, China; formerly, Graduate Student, Dept. of Civil ing factor (MSF), and overburden stress correction factor (K σ ) are
and Environmental Engineering, Univ. of Illinois at Urbana–Champaign, required, greatly increasing uncertainty in the results. These fac-
Urbana, IL 61801 (corresponding author). ORCID: https://orcid.org tors depend not only on soil properties but also on ground motion
/0000-0002-8816-3714. Email: xuanmei2@gmail.com characteristics such as magnitude, duration, and frequency content.
3
Professor, Dept. of Civil and Environmental Engineering, Univ. of Second, published liquefaction resistance curves were developed
Illinois at Urbana–Champaign, Urbana, IL 61801. Email: hashash@illinois
using PGA values interpolated from sites that did not liquefy or
.edu
Note. This manuscript was submitted on January 7, 2019; approved on derived from total stress site response analysis—inconsistent with
August 9, 2019; published online on November 27, 2019. Discussion per- the fact that liquefaction is an effective stress phenomenon, thereby
iod open until April 27, 2020; separate discussions must be submitted for leading to additional uncertainties. Last, because the cyclic stress
individual papers. This paper is part of the Journal of Geotechnical and method is empirical, extrapolating beyond the available database
Geoenvironmental Engineering, © ASCE, ISSN 1090-0241. (e.g., to depths greater than 10–15 m) involves great uncertainty.

© ASCE 04019128-1 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 At the same time, various investigators (e.g., Matasovic 1993; generation model proposed by Vucetic and Dobry (1986), termed
Carlton 2014; Kramer et al. 2015; Gingery et al. 2015) have per- GQ=H þ u (Mei et al. forthcoming). The GQ=H þ u model incor-
formed parametric site response analysis using nonlinear site re- porates an implied strength correction (Groholski et al. 2016) to
sponse analysis with pore-water pressure (PWP) generation (termed capture both small-strain soil behavior as well as large-strain shear
nonlinear effective stress site response) to better understand the strength. Mei et al. (forthcoming) showed that this coupled
influence of particular variables such as shear modulus, Poisson’s GQ=H þ u model can provide realistic response for a liquefiable
ratio, soil profile, and groundwater table depth on ground motion soil where either of the following criteria are met: (1) the computed
excess PWP ratio ðru ¼ Δu=σvo 0 Þ < 0.8; or (2) the computed r >
propagation and dynamic soil response. In particular, these studies u
have focused on the effect of PWP generation on acceleration 0.8 and the computed maximum cyclic shear strain (γ max ) < limit
time histories and response spectra. Capabilities of effective stress shear strain (γ limit ). The limit shear strain represents the shear strain
analysis constitutive models for predicting PWP generation have at which significant dilation behavior is first observed in laboratory
cyclic shear stress - shear strain response. For σvo 0 ∼ 35–180 kPa,
been evaluated by several investigators, [e.g., Arulmoli et al.
(1992), Elgamal et al. (2005), Ziotopoulou et al. (2012), and Karimi Mei et al. (forthcoming) defined γ limit ¼ 2%, 1.5%, and
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

and Dashti (2015)]. However, these validation studies generally in- 1.2% for relatively loose (Dr ∼ 30%–50%), medium dense
volve a limited number of case histories and centrifuge tests with a (Dr ∼ 50%–70%), and dense (Dr > ∼70%) clean sands. As detailed
limited variety of soil profiles and ground motions. Gingery (2014) by Mei et al. (forthcoming), when these ru and γ max criteria are not
performed one-dimensional (1-D) effective stress site response met, spectral accelerations may be underestimated at both short and
analyses for a variety of soil profiles and input motions after cali- long periods because dilative soil response is not captured by the
brating a soil constitutive model to a semiempirical liquefaction- simple GQ=H þ u constitutive model.
triggering relationship. The resulting simulations agree well with Hysteretic (unload-reload) behavior for GQ=H þ u is mod-
the liquefaction-triggering relationship used for calibration. How- eled using extended Masing rules, and small-strain damping is
ever, the analysis results computed using this approach depend represented using a frequency-independent viscous damping for-
strictly on the liquefaction-triggering relationship selected for mulation (Phillips and Hashash 2009). For hysteretic damping, a
calibration. To date, the feasibility of using 1-D nonlinear effective reduction factor that modifies the extended Masing unload-reload
stress site response to evaluate liquefaction triggering for a variety of rules to provide better agreement with empirical damping curves is
soil profiles and ground motion characteristics independent of an used (Phillips and Hashash 2009).
existing liquefaction triggering curve has not yet been illustrated. Nonlinear soil properties for sand layers were estimated using
To address these uncertainties and knowledge gaps, the authors the Darendeli (2001) shear modulus reduction and damping curves.
performed a broad parametric study of liquefaction triggering for Input parameters for the Darendeli (2001) curves include coefficient
level ground using 1-D nonlinear effective stress site response of earth pressure at rest (K o ), plasticity index (PI), overconsolidation
analysis. Because PWP generation is tracked during nonlinear ef- ratio (OCR), number of cycles (Nc ), and loading frequency. Where
fective stress site response analysis, excess PWP (Δu) can be used measured data were available for published soil profiles, these mea-
directly to evaluate liquefaction triggering. The parametric study sured data were used in the site response analyses. For the synthetic
described here used a suite of synthetic and published soil profiles profiles or where data were not available at the published profiles,
excited by a total of 145 representative western US (Ancheta et al. OCR was set as 1, and K o was calculated as (Jaky 1944)
2014) and 176 representative central and eastern US (McGuire
et al. 2001) strong ground motions. Results from the parametric K o ¼ 1 − sinðϕ 0 Þ ð3Þ
study were then compared with field observations of liquefaction
case histories (i.e., published liquefaction resistance curves) and
dynamic centrifuge tests to validate the use of nonlinear effective where ϕ 0 = effective-stress friction angle, estimated using the
stress site response analysis to evaluate liquefaction triggering. In Andrus et al. (2004) V s1 -ðN 1 Þ60 correlation and the Terzaghi et al.
addition, the authors used the site response analysis results to evalu- (1996) ðN 1 Þ60 -ϕ 0 correlation. The number of cycles and loading fre-
ate the effect of effective vertical stress on CRR, and compared the quency were assumed to be 10 and 1 Hz, respectively, as recom-
resulting back-calculated K σ values with published correlations. mended by Darendeli (2001). When not measured in sand layers, V s
was estimated from Dr using the Shahien (1998) Dr -ðN 1 Þ60 corre-
lation and the Andrus et al. (2004) V s1 -ðN 1 Þ60 correlation. For the
Parametric Study Using Nonlinear, Effective Stress clay layers, measured dynamic properties were used directly where
Site Response Analysis available. Where not available, the Vucetic and Dobry (1991) shear
modulus reduction and damping curves were used (based on the
soil PI).
One-Dimensional Site Response Analysis Soil
Mei et al. (2018) evaluated the Vucetic and Dobry (1986) PWP
Profile Parameters
generation model for use in DEEPSOIL and concluded that the
The authors performed the parametric study using the 1-D site re- model reasonably captured PWP generation in clean sands over
sponse analysis software, DEEPSOIL v6.1 (Hashash et al. 2016). In wide ranges of relative density and effective stress. Parameters re-
DEEPSOIL, nonlinear site response analysis is performed by dis- quired for the model were selected using correlations described in
cretizing the soil column as a multi-degree-of-freedom lumped mass Mei et al. (2018). For clay layers, PWP generation was estimated
model, where each layer is represented by a corresponding mass, using the Matasovic and Vucetic (1995) model with input param-
nonlinear spring, and a dashpot representing viscous damping. eters based on correlations developed by Carlton (2014).
Lumping half of the mass of each of two consecutive layers at their Dissipation and redistribution of Δu during shaking was incor-
common boundary forms the mass matrix, and the stiffness matrix is porated in the site response analyses via Terzaghi 1-D consolidation
updated at each time increment to incorporate soil nonlinearity. theory, as formulated in DEEPSOIL. Coefficients of consolida-
The soil backbone and cyclic stress-strain constitutive response tion (cv ) were assumed to be 0.02–0.1 m2 =s for loose to dense
is approximated using a generalized quadratic/hyperbolic (GQ=H; sands using correlations from Pestana et al. (1997) and Domenico
Groholski et al. 2016) constitutive model coupled with the PWP and Mifflin (1965) to estimate permeability and coefficient of

© ASCE 04019128-2 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 Japan, Turkey). Peak ground accelerations range from approximately
0.016–1.03 g with a log mean value of 0.13 g. Figs. 1(a and c)
present the acceleration response and Fourier amplitude spectra of
the 145 representative WUS ground motions.
The representative CEUS ground motion records were selected
from the motions published in NUREG-6728 (McGuire et al.
2001). These 176 motions include weak to moderate ground mo-
tions recorded in the CEUS as well as representative motions
developed by spectrally matching WUS records to CEUS target
spectra. The PGA values range from approximately 0.018 to 3.64 g
with a log mean value of 0.25 g. Figs. 1(b and d) present the re-
sponse and Fourier amplitude spectra of the 176 representative
CEUS ground motions.
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Soil Profiles
To perform the parametric study, the authors created several syn-
thetic sand profiles that include several key variables: (1) the pres-
ence of a looser sand layer (as well as variations in its thickness),
where the V s1 in the looser layer was varied from 125 to 250 m=s
(in 25 m=s increments); (2) variations of soil compressibility
(more/less compressible); (3) variations of water table depth (0, 3,
and 4.5 m); and (4) variations in bedrock depth (20, 25, and 30 m).
Table 1 summarizes the synthetic soil profiles. Values of bedrock
V s were assigned as 760 m/[B/C boundary condition; Building
Fig. 1. Acceleration response spectra and smoothed Fourier amplitude Seismic Safety Council (2015)] for WUS sites and 3,000 m=s for
spectra for input ground motions used in this study: (a) response spectra CEUS sites (Hashash et al. 2014).
for WUS motions; (b) response spectra for CEUS motions; (c) Fourier In addition to the synthetic profiles, six profiles were simpli-
amplitude spectra for WUS motions; and (d) Fourier amplitude spectra fied from published soil profiles where accelerometers were (or
for CEUS motions. are) installed, as shown in Figs. 2 and 3. These profiles include
three National Earthquake Hazards Reduction Program (NEHRP)
site class D sites in the Christchurch region that were affected by
the 2010–2011 Canterbury earthquake sequence (Markham 2015),
compressibility (mv ), respectively. For clay layers, cv was taken as which consist chiefly of sands and silty sands to a depth of about
1 × 10−5 m2 =s. 30 m overlying bedrock (Fig. 2). The three remaining profiles
(Fig. 3) are NEHRP site class E sites that were used in a site re-
Strong Ground Motions sponse calibration study by Baturay and Stewart (2003). These pro-
files consist of silts and clays with layers of sands at different
A total of 145 representative western US (WUS) and 176 repre- depths.
sentative central and eastern US (CEUS) strong ground motions
were used as input motions in this study. The representative
WUS ground motion records were selected from the NGA-West2 Parametric Study Results
database (Ancheta et al. 2014) for sites with V s30 ≥ 700 m=s,
where V s30 is the average shear wave velocity in the upper Using the input soil parameters, input ground motions, and syn-
30 m of a profile. Among these 145 motions, 101 are WUS records thetic and published soil profiles described previously, the authors
and 44 were measured in other active crustal regions (e.g., Iran, conducted a total of 16,612 nonlinear total stress and nonlinear

Table 1. Synthetic sand deposit profiles used for parametric study

V s1 (m=s)
Depth loose Loose sand Water table Bedrock Input
Case Dense sand Loose sand sand (m) thickness (m) depth (m) depth (m) motions Sand type
1a 250 125–250 9 3 0 30 WUS Compressible
1b 250 125–250 9 3 0 30 WUS Less compressible
1c 250 125–250 9 3 0 30 CEUS Compressible
1d 250 125–200 22 3 0 30 WUS Compressible
1e 250 125–200 3 3 0 30 WUS Compressible
2a 250 125–200 9 1.5 0 30 WUS Compressible
2b 250 125–250 9 3 0 30 WUS Compressible
2c 250 125–200 9 4.5 0 30 WUS Compressible
3a 250 125–250 9 3 0 30 WUS Compressible
3b 250 125–200 9 3 3 30 WUS Compressible
3c 250 125–200 9 3 4.5 30 WUS Compressible
4a 250 125–200 9 3 0 20 WUS Compressible
4b 250 125–200 9 3 0 25 WUS Compressible
4c 250 125–250 9 3 0 30 WUS Compressible

© ASCE 04019128-3 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 strength and stiffness” is too broad to be applied systematically.
Thus, a more objective criterion is needed. The most widely used
criteria to define liquefaction triggering in the laboratory are PWP
and strain criteria.
The PWP criterion typically is quantified using the excess PWP
ratio, ru . Under level-ground conditions, liquefaction is commonly
associated with ru ¼ 1. The first occurrence of this condition in
laboratory tests was termed “initial liquefaction” by Seed and Lee
(1966). However, residual excess PWP is computed using the
Vucetic and Dobry (1986) PWP model. In contrast to the maximum
excess PWP commonly used in practice, residual excess PWP is
defined as the excess PWP occurring during each cycle when shear
stress equals zero. Hereafter, the excess PWP ratio using the
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

residual excess PWP is termed ru;residual .

In addition, there are laboratory and field situations where
Fig. 2. Shear wave velocity and stratigraphy for site class D profiles: ru ¼ 1 is not required to trigger significant softening. For example,
(a) station SHLC; (b) station NNSH; and (c) station HPSC. (Adapted Ishihara and Li (1972) stated that although specimens of sand with
from Markham 2015.) a small amount of fines can liquefy (i.e., develop large shear strain
and low shear stiffness during cycling), ru may peak around 0.9 to
0.95, and ru ¼ 1 is not achieved. Similarly, cyclic direct simple
shear tests performed by Wu et al. (2003) and Kenan (2005) for
clean sands covering wide range of Dr and consolidation stress
showed that ru ¼ 1 is not reached in some tests, although the spec-
imens exhibited large shear strains and liquefaction-like behavior.
However, laboratory equipment compliance also may result in an ru
value slightly smaller than unity when liquefaction-like response
develops. In the field, Kramer et al. (2015) suggested that a thick,
shallow layer of loose sandy soil can develop relatively high excess
PWP ratio (ru ∼ 0.8) and produce sand boils without the significant
strain softening commonly associated with initial liquefaction
(ru ¼ 1). Previously, Poulos et al. (1985) also had concluded that
defining liquefaction based on a momentary ru ¼ 1 condition is
misleading. Thus, a ru ¼ 1 criterion may be inconsistent with some
field observations.
In the laboratory, shear strain criteria also are commonly used to
define liquefaction. However, there is no general agreement on the
Fig. 3. Shear wave velocity and stratigraphy for site class E profiles: shear strain corresponding to liquefaction triggering. For example,
(a) Appel #2 in Redwood City, California; (b) Meloland in El Centro, Ishihara (1993) proposed a single amplitude (SA) shear strain of 3%
California; and (c) Pacific Park Plaza in Emeryville, California. and double amplitude (DA) shear strain of 5% as the criterion,
(Adapted from Baturay and Stewart 2003.) whereas other criteria such as SA strain of 1.4% to 2% (Boulanger
and Idriss 2004) and DA of 6% (Wu et al. 2003) have been recom-
mended. Many shear strain criteria were defined by correlating mea-
sured laboratory shear strains to initial liquefaction (ru ¼ 1), and
effective stress site response analyses. The results of the parametric therefore these shear strain criteria have limitations for use in non-
study are described in terms of computed cyclic stress ratios, PWP, linear effective stress site response analysis (as discussed in the
and shear strains. However, to gage whether the parametric study following section). Furthermore, several investigators (e.g., Darve
results are consistent with published CRR curves (i.e., case history- 1996; Andrade et al. 2013) have proposed mechanics-based criteria
based liquefaction triggering curves, termed field CRR curves here), to define liquefaction; however, these criteria were found to be dif-
the computed results must be: (1) related to specific liquefaction ficult to apply to the results of 1-D site response analysis using sim-
triggering criteria and (2) must be interpreted in a manner consistent plified hyperbolic stress-strain constitutive models.
with published CRR curves. These issues are addressed in the fol- In this study, liquefaction is defined based on both ru;residual and
lowing two sections. shear strain criteria in concert with observed cyclic shear behavior.
Fig. 4 presents results from a stress-controlled, cyclic direct simple
shear (DSS) test on Monterey sand with Dr ¼ 47% and σvo 0 ¼
Defining Liquefaction Triggering from Site Response
85 kPa. The results illustrate the increase in shear strain and ru with
Analysis
increasing number of cycles, with the solid circles indicating
Liquefaction is described as the loss of soil strength and stiffness ru;residual during each cycle. Here, an obvious increase in the rate
resulting from PWP increase and concurrent decrease in effective of shear strain accumulation occurs at ru;residual ≈ 0.8, which cor-
stress. When triggered by seismic shaking, it can manifest as sand responds to the intersection of the slopes of maximum shear strain
blows, surface settlement, lateral spreads, flow failures, flotation of accumulation rate [Fig. 4(b)]. At ru;residual > 0.8, maximum shear
buried structures, or bearing capacity failure. Therefore, these ob- strain amplitude increases greatly and the rate of increase is almost
servations commonly are used as evidence of liquefaction trigger- constant with increasing number of cycles, suggesting that the
ing in the field. These liquefaction-related phenomena, however, shear modulus softened significantly and that ru;residual ¼ 0.8 can
seldom occur in laboratory tests, and the definition of “loss of soil be treated as a boundary of marginal liquefaction triggering for

© ASCE 04019128-4 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Cyclic response of Monterey sand with Drc ¼ 47% and σvc 0 ¼ 85 kPa (test MS29j; data from Wu et al. 2003) illustrating proposed definitions

of thresholds of marginal liquefaction and liquefaction: (a) excess PWP and number of cycles; and (b) shear strain and number of cycles and summary
of undrained cyclic behavior exhibited in laboratory tests collected by Mei (2018); histograms of ru values corresponding to (c) slope change of shear
strain accumulation rate used in defining marginal liquefaction; and (d) 6% double amplitude shear strain used in defining liquefaction.

this test. Gingery (2014) also used this change in shear strain ac- increase the pore-water pressure ratio and ru flattens out and re-
cumulation rate to calibrate the parameter y1 in the PDMY2 model mains constant at 0.9. The induced shear stress decreases with time
(Elgamal et al. 2002; Yang et al. 2008; Gingery 2014). The y1 and becomes extremely low as ru;residual ¼ 0.9 is reached.” Based on
parameter controls the magnitude of postliquefaction plastic shear these observations, ru;residual ∼ 0.9 appears to be a typical upper
deformation in the highly yielded “neutral phase,” during which a limit and represents a condition where the soil is greatly softened
significant amount of permanent shear strain can accumulate prior and large shear strains can accumulate. Therefore, ru;residual ¼ 0.9
to dilation with minimal changes in shear stress and effective stress. could serve as a threshold for “full” liquefaction.
Other tests performed by Wu et al. (2003) and Polito (1999) However as noted previously, Mei et al. (forthcoming) showed
on Monterey and Yatesville sands, respectively, also exhibited that 1-D nonlinear, effective stress site response analysis performed
increases in the rate of shear strain accumulation at ru;residual ≈ using the GQ=H þ u model may provide unreliable results when
0.8–0.9. Fig. 4(c) summarizes the ru;residual values corresponding ru;residual > 0.8. Therefore, an excess PWP criterion alone may
to a rapid increase in the rate of shear strain accumulation [as illus- not reliably define liquefaction in site response analyses. Wu et al.
trated in Fig. 4(b)] for the undrained cyclic shear tests collected by (2003), Wijewickreme and Soysa (2016), and Mei et al.
Mei (2018). As illustrated in Fig. 4(c), it is clear that the number of (forthcoming) employed shear stiffness and shear strain criteria
occurrences greatly increases when ru;residual ≈ 0.8. Therefore, we to define liquefaction, where “full” liquefaction corresponds to
defined ru;residual ¼ 0.8 as a threshold for marginal liquefaction. In the condition when the shear stress—shear strain hysteresis loops
these tests, ru;residual begins to stabilize and increase very slowly transition from nearly hyperbolic (with a small decrease in secant
when ru;residual > ∼0.9 [e.g., Fig. 4(a)]. shear stiffness with increasing cycles) to “banana-shaped” (with
Fig. 4(d) summarizes the ru;residual values corresponding to a hysteresis loops that are relatively flat at small to intermediate shear
double amplitude shear strain ¼ 6%, which is defined as liquefac- strains and experience significant stiffening/dilation at larger shear
tion by Wu et al. (2003). Here, the number of occurrences increases strains). Mei et al. (forthcoming) defined the shear strain corre-
greatly when ru;residual > ∼0.9, indicating relatively small differen- sponding to this transition as the limit shear strain (γ limit ).
ces between ru -based and strain-based liquefaction criteria for tests γ limit ¼ 2%, 1.5%, and 1.2% for relatively loose (Dr ∼ 30%–50%),
with ru;residual > ∼0.9. Similarly, Kenan (2005) performed cyclic medium dense (Dr ∼ 50%–70%), and dense (Dr > ∼70%) clean
strain-controlled DSS tests on Monterey #0/30 and an aggregate sands when ru;residual > 0.8. Above these limit strain values, agree-
sand and observed that “when a maximum of 90% pore-water pres- ment between the hyperbolic GQ=H þ u model and the measured
sure ratio was reached, further loading of the specimen did not stress-strain response greatly decreases (Mei et al. forthcoming).

© ASCE 04019128-5 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 0
These observations and conclusions relate to ru defined as Δu=σvo overlying (and underlying) soil layers. Therefore, it may be unre-
and would need to be reconsidered if ru is defined in terms of ef- liable to derive data from a nonliquefied soil layer if other strata
fective mean stress. have softened. Last, when both shallow and deep sandy layers gen-
Based on these observations, following definitions were used in erated ru;residual > 0.8, the shallower layer was selected as critical
this study: (1) marginal liquefaction corresponds to 0.8 ≤ ru;residual ≤ to be consistent with current practice. If no sandy layers generated
0.9 and γ max < γ limit ; and (2) full liquefaction corresponds to either ru;residual > 0.8, the profile was defined as nonliquefied and the
0.8 ≤ ru;residual ≤ 0.9 and γ max ≥ γ limit or ru;residual ≥ 0.9, which- layer with the highest ru;residual value was selected as critical.
ever is reached first in the nonlinear, effective stress site response After evaluating the site response analyses, logistic regression
analysis; and (3) nonliquefaction corresponds to ru;residual < 0.8. was used to delineate “simulated CRR curves” to facilitate the com-
parison between the parametric study results and the field CRR
curves. A logistic regression framework for liquefaction analysis
Comparing Site Response-Based Liquefaction was first proposed by Liao et al. (1988) and later used by Youd
Triggering with CRR Curves and Noble (1997), Toprak et al. (1999), Juang et al. (2002), among
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

To compare the parametric study results to published CRR curves, others, to define probabilistic field CRR curves. However, the
the site response results must be interpreted in a manner consistent method of regression used here is not critical as the authors only
with these CRR curves. For example, liquefaction triggering in a used the simulated CRR curves to compare to the field CRR curves.
given nonlinear effective stress site response analysis was judged In the logistic framework, each case study is represented by
based on the ru;residual and γ max values computed for each sublayer a binary variable, Y, that indicates whether liquefaction occurs
and the liquefaction triggering criteria defined above. However, to (Y ¼ 1) or does not occur (Y ¼ 0) and a vector of explanatory var-
compare these “liquefied” or “nonliquefied” results to field CRR iables, X ¼ ½x1 ; x2 ; : : : :; xm T that represents ground motion and
curves, the total stress-based CSR is needed because the field cases soil characteristics. The probability of triggering liquefaction (PL )
are defined by CSR values computed from PGA measured at sites can be expressed as a function of X based on available n observa-
where excess PWP generation was minimal (in conjunction with tions of (X i ; Y i ) as (Cox and Snell 1970; Liao et al. 1988)
rd ) or from total stress-based site response analysis. Therefore, the
authors used the nonlinear total stress site response analyses to PL ðXÞ ¼ P½Y ¼ 1jX
define CSR profiles for the simulations, whereas the nonlinear ef- 1
fective stress site response analyses were used to define whether ¼ ð6Þ
1 þ exp½−ðβ 0 þ β 1 x1 þ β 2 x2 þ · · · þβ n xn Þ
liquefaction was triggered. Specifically, CSR was computed in two
different manners: (1) directly from computed shear stress time his- where β 0 ; β 1 ; : : : β n are regression coefficients determined from
tories in the nonlinear total stress site response analysis as shown in binary logistic regression analysis through maximization of the
Eqs. (4) and (2) using the peak ground acceleration computed from likelihood function, L, which is defined as
nonlinear total stress analysis with rd as shown in Eq. (1)
Y
n
τ L¼ ½PL ðX i ÞY i ½1 − PL ðX i Þ1−Y i ð7Þ
CSRNL-TS ¼ 0.65CSRNL-TS;max ¼ NL-TS;max
0 ð4Þ
σvo i¼1

where CSRNL-TS;max = maximum value of CSR corresponding to Several explanatory variables for modeling liquefaction prob-
τ NL-TS;max , which is the maximum shear stress computed by non- ability have been proposed by Liao et al. (1988), Youd and Noble
linear total stress analysis for a given layer. (1997), Toprak et al. (1999), and Juang et al. (2002). Among these,
Shear wave velocity is a required input for the site response the two most widely used variables are: (1) soil liquefaction resis-
analyses; therefore, the authors compared the results from the para- tance, often expressed by a field index such as ðN1 Þ60 , qc1N , or
metric study with the V s -based field CRR curves from Andrus and Vs1 ; and (2) seismic demand, often represented by CSR7.5, with a
Stokoe (2000) and Kayen et al. (2013). To be consistent with meth- functional form of lnðCSR7.5 Þ often used in the regression. There-
ods Andrus and Stokoe (2000) and Kayen et al. (2013) used to de- fore, liquefaction probability based on Vs measurements can be
velop the case histories, we used the rd and MSF correlations defined as
proposed by Andrus and Stokoe (2000) and Kayen et al. (2013),
1
respectively, when comparing with their curves and the specific PL ¼ ð8Þ
input motion earthquake magnitude to adjust the nonlinear total 1 þ exp½−ðβ 0 þ β 1 lnðCSR7.5 Þ þ β 2 VS1 Þ
stress-based CSR to CSR7.5 . Similar to Andrus and Stokoe (2000)
and Kayen et al. (2013) methods, we did not use the overburden
stress correction factor, K σ . Therefore, cyclic shear stress ratios Baseline Condition Results
from the parametric study were corrected as follows for comparison
to the V s -based field CRR curves For initial comparison to the Andrus and Stokoe (2000) and Kayen
et al. (2013) CRR curves, the authors used results for synthetic-
CSR7.5 ¼ CSRNL-TS =MSF ð5Þ profile cases with a looser, 3-m thick sand layer at an effective
vertical stress of about 100 kPa excited by WUS input motions
Note that although ru;residual and γ max were computed through- (case 1a, case 1b, Table 1). This case is termed “baseline” because
out the soil profile, only ru;residual and γ max corresponding to the the results do not require correction factors related to effective ver-
critical layer (i.e., the maximum ru;residual and γ max in a sandy tical stress, i.e., V s ¼ V s1 and K σ ¼ 1. (As noted previously, K σ
layer) were used for comparisons to published V s -based field CRR was not used for any cases in this study when comparing with the
curves. This approach is consistent with the current practice of Andrus and Stokoe (2000) and Kayen et al. (2013) CRR curves to
using one critical depth per boring/CPT sounding/V s profile when be consistent with the original procedures.)
developing field CRR curves. Furthermore, when high ru;residual Mei et al. (2018) concluded that in addition to Dr , sand
values (> ∼ 0.8) are generated in nonlinear effective stress site re- compressibility also affects PWP generation during cyclic load-
sponse analysis, the softened layer can influence the response of ing. They used the grain size distribution parameter, uniformity

© ASCE 04019128-6 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 0.6
coefficient (CU ), to reflect compressibility. Based on these obser-
vations, Mei et al. (2018) proposed correlations to define Vucetic Mw = 7.5
and Dobry (1986) PWP generation model parameters for clean σ v0 = 100 kPa
sands that consider both Dr and compressibility. However, as 0.4 Cu = 1.4
WUS

CSR 7.5
the parametric soil profiles could take any value of CU , and V s gen-
erally is controlled by relative density and effective stress (not com-
pressibility), the authors used two sets of Vucetic and Dobry (1986)
0.2
PWP generation model input parameters for each nonlinear, effec-
tive stress site response analysis—one set corresponding to more
compressible conditions (CU ¼ 3.8) and one set corresponding
(a)
to less compressible conditions (CU ¼ 1.4) identified by Mei et al. 0
(2018). 0.6
Using the methodology and input parameters described, two
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

sets of nonlinear site response analyses were performed for baseline Mw = 7.5
case 1a and 1b using the 145 WUS input ground motions. Figs. 5(a σ v0 = 100 kPa
and b) compare baseline case 1a and 1b liquefaction triggering re- 0.4 Cu = 3.8
WUS

CSR 7.5
sults for the more and less compressible PWP generation model
parameters, respectively, with the field CRR curves. Fig. 5 includes
case history data compiled by Kayen et al. (2013), which comprises 0.2
301 case histories from North America, Asia, and Europe, includ-
ing 121 case histories reported previously by Robertson et al.
(1992), Kayen et al. (1992), Mitchell et al. (1994), Lodge (1994),
0
Andrus and Stokoe (2000), and Bay and Cox (2001). Although this 0 100 200 300
more recent field dataset is shown in most subsequent figures, Mei (b) Vs1 (m/s)
(2018) compares all of the simulation results described later to both
Kayen et al. (2013) Liq Simulation Liq
the Kayen et al. (2013) and Andrus and Stokoe (2000) datasets,
Kayen et al. (2013) NLiq Simulation Marg
using the method-specific MSF correlations. Kayen et al. (2013) P(L) = 15% Simulation NLiq
Each data point for baseline case 1a and 1b simulations in Fig. 5 Andrus and Stokoe (2000) Simulation P(L) = 15%
represent the combination of V s1 and CSR7.5 in the critical layer
0
(where V s1 ¼ V s for σvo ≈ 100 kPa in case 1a) and the CSR7.5 Fig. 5. Comparison of liquefaction (Liq), marginal liquefaction
computed using Eq. (5) for an individual ground motion. In this (Marg), and no liquefaction (NLiq) data from nonlinear effective stress
figure, CSRNL-TS were computed using Eq. (4). These data were site response simulations using WUS ground motions and MSF corre-
classified as liquefied, marginally liquefied, or not liquefied using lation proposed by Kayen et al. (2013) with case history data and field
the ru;residual and γ max criteria discussed previously. The Kayen CRR curves developed using data from Kayen et al. (2013) and Andrus
et al. (2013) MSF correlation was used in Eq. (5) to compare di- and Stokoe (2000): simulations using PWP generation model with
rectly to the Kayen et al. (2013) CRR curve. The logistic regression (a) less compressible sand parameters (CU ¼ 1.4); and (b) more com-
analysis described was performed to define simulated CRR curves pressible sand parameters (CU ¼ 3.8).
for comparison to the field CRR curves. Because Kayen et al.
(2013) recommended that their PL ¼ 15% curve be used as a deter-
ministic boundary, the simulated logistic PL ¼ 15% CRR curve in for V s1 < ∼200 m=s [Fig. 5(a) and 6(a)]. This result is sensible
this study also was selected as a deterministic boundary, so that the because a sandy soil should be less likely to liquefy when sub-
simulated CRR curves in this study can be compared witj the jected to a given seismic demand if PWP generation occurs
Kayen et al. (2013) CRR curve. more slowly during shaking (corresponding to the less compres-
Similarly, Figs. 6(a and b) compare baseline case 1a and 1b sible PWP generation model parameters). This observation also
liquefaction triggering results for the more and less compressible provides a reason why some nonliquefied data plot above field
PWP generation model parameters, respectively, with the field CRR CRR curves.
curves and clean sand case history data compiled by Andrus and 2. As illustrated in Figs. 5(b) and 6(b), the simulated CRR curves
Stokoe (2000). Again, CSRNL-TS were computed using Eq. (4) for more compressible PWP generation model parameters agree
and the Andrus and Stokoe (2000) MSF correlation was used in closely with field CRR curves for V s1 < ∼200 m=s. We note
Eq. (5) to compare directly to the Andrus and Stokoe (2000) CRR that for V s1 ¼ 175–200 m=s, the simulated CRR curve with
curve. To be consistent with Andrus and Stokoe (2000), the simu- more compressible sand parameters plots below the Andrus and
lated CRR curve was determined by regressing the curve-fitting Stokoe (2000) curve [Fig. 6(b)], largely because the Andrus
parameters a and b in the boundary curve formulation proposed and Stokoe (2000) curve approaches a vertical asymptote near
by Andrus and Stokoe (2000) V s1 ¼ 215 m=s. This agreement also is reasonable because
  2   deterministic field CRR curves generally correspond to field
V 1 1
CRR7.5 ¼ a s1 þ b − ð9Þ conditions that are most conducive to triggering liquefaction
100 215 − Vs1 215 (Ambraseys 1988) and the simulations using the more compres-
sible PWP generation model parameters correspond to sands
The regressed parameters a and b minimized the number of mis- that most easily liquefy during shaking.
classified liquefaction and nonliquefaction cases. 3. As expected, simulated marginal liquefaction cases plot near the
Some key observations can be made based on the comparisons deterministic simulated CRR curve and near the field CRR
in Figs. 5 and 6. curves in Figs. 5(b) and 6(b).
1. The simulated CRR curves for less compressible PWP genera- 4. For V s1 > 200 m=s, many simulated liquefied cases fall below
tion model parameters plot slightly above the field CRR curves the field CRR curves. Dobry and Abdoun (2011, 2015) suggested

© ASCE 04019128-7 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Comparison of liquefaction (Liq), marginal liquefaction

(Marg), and no liquefaction (NLiq) data from nonlinear effective stress Fig. 7. Comparison of liquefaction (Liq), marginal liquefaction
site response simulations using WUS ground motions and MSF corre- (Marg), and no liquefaction (NLiq) data from nonlinear effective stress
lation proposed by Andrus and Stokoe (2000) with case history data site response simulations using WUS ground motions, PWP generation
and field CRR curves developed using data from Kayen et al. (2013) model with more compressible sand parameters (CU ¼ 3.8) and CSR7.5
and Andrus and Stokoe (2000): simulations using PWP generation computed by simplified method and MSF, rd correlation proposed by
model with (a) less compressible sand parameters (CU ¼ 1.4); and Kayen et al. (2013) and Andrus and Stokoe (2000) with (a) case history
(b) more compressible sand parameters (CU ¼ 3.8). data and field CRR curves developed using data from Kayen et al.
(2013); and (b) Andrus and Stokoe (2000).

that sands with high shear wave velocities (V s1 > 200 m=s)
and high liquefaction resistance commonly are influenced by the MSF and rd correlations proposed by Kayen et al. (2013) and
aging, cementation, or preshearing. In contrast, they argued Andrus and Stokoe (2000), respectively. As shown in the figures,
laboratory-reconstituted specimens, recent fills, and geologically the simulated CRR curves match closely to the respective field
very young deposits of clean to silty sands rarely exhibit CRR curves from Kayen et al. (2013) and Andrus and Stokoe
V s1 > ∼190–200 m=s. The aging, cementation, or preshearing (2000). Although the simulated CRR curves using Eq. (1) result
effects required to achieve V s1 > 200 m=s greatly retard PWP in slightly more misclassified cases than the simulated curves using
generation. Because the PWP generation model parameters used Eq. (4), this comparison illustrates that the uncertainties in rd fac-
for nonlinear effective stress site response analysis were derived tors does not affect the ability of nonlinear effective stress site re-
from reconstituted specimens without these effects, it was ex- sponse analysis to reasonably evaluate liquefaction. As a result, all
pected that liquefaction resistance would be underestimated comparisons described subsequently are based on CSR7.5 com-
for sands with V s1 > 200 m=s. Furthermore, the database puted using Eqs. (4) and (5) (i.e., CSR determined directly from
collected by Mei et al. (2018) to define PWP generation model nonlinear total stress site response analysis).
parameters were limited to specimens with Dr < ∼80%. Coinci-
dentally, the Menq (2003) correlation predicts V s1 ≈ 200 m=s
for Dr ¼ 80% and σvo 0
¼ 100 kPa. For Dr > 80% Influence of Input Ground Motion Characteristics
(i.e., V s1 > 200 m=s), PWP model parameters are estimated It is widely known that site response analysis is affected strongly
by extrapolation, resulting in some uncertainty. For these two rea- by the input ground motion intensity and frequency content. Most
sons, the authors only used simulation results with V s1 ≤ level-ground liquefaction datasets are dominated by case records
200 m=s and did not plot site response analysis results for V s1 > and ground motions from active seismic regions considered similar
200 m=s in the figures. to the WUS. In contrast, very few case records originate from stable
For comparison, Figs. 7(a and b) present the simulated CRR tectonic regions similar to the CEUS. Nevertheless, engineers
curves for cases with more compressible sand parameters (corre- must rely on field CRR curves to evaluate liquefaction in stable
sponding to CU ¼ 3.8), CSR7.5 computed using Eqs. (1) and (5) for tectonic regions despite the understanding that these regions exhibit

© ASCE 04019128-8 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Comparison of simulated and field CRR curves for simulations

performed using CEUS input ground motions and PWP generation
model with more compressible sand parameters (corresponding to
CU ¼ 3.8).

differences in ground motions (e.g., motions that are richer in high Fig. 9. Comparison of ground motions recorded during the 1994
frequency energy; Hashash et al. 2014). Therefore, the authors ex- Northridge earthquake (M 6.69; reverse fault mechanism) with (motion
amined the effect of input motion characteristics on liquefaction RSN 1050) and without (motion RSN 989) near-fault effects. (a and
evaluation using nonlinear, effective stress site response analysis. b) CSR, (c and d) acceleration, (e and f) velocity, (g and h) displace-
Fig. 8 presents simulation results using the less compressible ment, and (i and j) ru time histories for (a, c, e, g, and i) RSN 1050 and
PWP model parameters, for baseline case 1 excited by 176 rep- (b, d, f, h, and j) RSN 989.
resentative CEUS ground motions. Again, the simulation results
are compared with field CRR curves using PL ¼ 15% where appro-
priate, as well as the simulated CRR curves developed for WUS As evident from the velocity and displacement time histories, mo-
ground motions (Fig. 5). As noted, the simulated CRR curves were tion RSN 1050 exhibited a near-fault effect (Hayden et al. 2014).
regressed using simulations with V s1 ≤ 200 m=s. The CSR time histories correspond to the layer with the highest
The baseline (case 1c) CEUS simulation results are generally
ru computed by nonlinear effective stress site response analysis.
similar to the baseline WUS simulation results and the simulated
As shown in Figs. 9(a and b), both motions exhibited a peak
CRR curve agrees closely with the field CRR curves. Interestingly,
CSR ≈ 0.4. Applying the Kayen et al. (2013) MSF correction,
the simulated CRR curves for CEUS and WUS input motions are
CSR7.5 ≈ 0.34 for both motions. The identical CSR7.5 values sug-
nearly identical, with the simulated CRR curve for WUS motions
gest that these two motions should impose similar seismic demand
only slightly higher than the CEUS. The difference likely results
on the soil (and generate similar values of excess PWP) in a given
from CEUS motions exhibiting a larger number of cycles for a
soil profile. However, as evident from the figure, RSN 1050 has far
given magnitude. However, this difference largely is offset by the
fewer cycles of shaking and generated less excess PWP in nonlin-
higher frequency of the CEUS motions, which generates smaller
ear, effective stress site response analysis compared to RSN 989.
shear strains.
Similar to the field cases, the simulations yielded some misclas- This example highlights the importance of input motion character-
sified data points (i.e., nonliquefied simulations falling above the istics on liquefaction potential, as well as the benefit of performing
simulated CRR curve; Figs. 5–8). Interestingly, most misclassified liquefaction analysis using a validated nonlinear, effective stress
data correspond to simulations with input ground motions that site response analysis.
exhibit near-fault effects, which often are characterized by a strong
acceleration pulse early in the motion (with a corresponding veloc- Influence of Effective Vertical Stress
ity pulse) with relatively weak shaking for the remainder of the time
history (Hayden et al. 2014; Shahi and Baker 2012). When the Seed A large number of relationships for the overburden stress correction
et al. (1975) method is used to count the equivalent number of cycles factor K σ have been proposed for use in cyclic stress-based lique-
(Nc;eq ), “near-fault” input motions commonly exhibit smaller Nc;eq faction analysis (Youd et al. 2001; Cetin et al. 2004; Idriss and
values than motions without a near-fault effect, even when M and Boulanger 2010; among others), however, there is still no consis-
peak CSR are similar. Thus Eq. (4) tends to overestimate CSR7.5 for tent agreement on its application in practice. This study used the
“near-fault” motions, and these simulations can result in nonlique- Vucetic and Dobry (1986) strain-based PWP generation model,
fied data plotting above the simulated CRR curve. which according to Dobry et al. (1982), is only modestly affected
Fig. 9 presents an example comparison of input motions with by effective vertical stress. And in the nonlinear effective stress site
and without near-fault effects. Both recorded during the 1994 response analyses described in this article, it is not necessary to
Northridge earthquake (M6.69; reverse fault mechanism), motion compute CSR7.5 , because liquefaction triggering is defined by
RSN 1050 had a rupture distance (Rrup ) of 7 km, whereas motion ru;residual and γ max , and PWP generation is computed directly
RSN 989 had Rrup ¼ 20.5 km. Fig. 9 includes time histories of from shear strain. Nevertheless, it is instructive to compare back-
CSR, acceleration, velocity, displacement, and ru for each motion. calculated values of K σ from the nonlinear effective stress site

© ASCE 04019128-9 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 0.6
Mw = 7.5
Watertable 3m
0.4 Cu = 3.8

CSR 7.5
WUS

0.2

(a) 0
0.6

Mw = 7.5
Watertable 4.5m
0.4
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Cu = 3.8

CSR 7.5
WUS

0.2

0
0 100 200 300
(b) Vs1 (m/s)

Kayen et al. (2013) Liq Simulation Liq (WT 3m/4.5m)

Kayen et al. (2013) NLiq Simulation Marg (WT 3m/4.5m)
Kayen et al. (2013) P(L) = 15% Simulation NLiq (WT 3m/4.5m)
Andrus and Stokoe (2000) Simulation P(L) = 15% (WT 3m/4.5m)
Simulation P(L) = 15% (100 kPa)

Fig. 11. Comparison of simulated CRR curves using WUS input mo-
Fig. 10. Comparison of simulated CRR curves using WUS input mo- tions and more compressible sand PWP model parameters (CU ¼ 3.8)
tions and more compressible sand PWP model parameters (CU ¼ 3.8) for (a) case 1a (water table at ground surface; critical layer with
0
for (a) case 1a (critical layer with σvo ≈ 100 kPa) and case 1d (critical 0
σvo ≈ 100 kPa) and case 3b (water table at 3 m; critical layer with
0
layer with σvo ≈ 200 kPa); and (b) case 1a and case 1e (critical layer 0
σvo ≈ 130 kPa); and (b) case 1a and case 3c (water table at 4.5 m;
0
with σvo ≈ 50 kPa). 0
critical layer with σvo ≈ 146 kPa).

response analyses to the empirical relationships noted previously, 1.2

which were developed in a total stress framework.
As shown in Table 1, the depth to the loose, liquefiable sand Dr < 40%
layer was varied in the synthetic profile as part of the parametric 0.8
Kσ

study. Fig. 10 presents the simulation data and simulated CRR

curve (cyan curve) with PL ¼ 15% for cases 1d and 1e (Table 1) 0.4 Dr = 60%
using WUS motions and more compressible sand PWP model Dr > 80%
0
parameters. The critical layer was located at a depth with σvo ≈
200 kPa and 50 kPa for cases 1d and 1e, respectively. For compari- (a) 0
son, Fig. 10 includes the simulated CRR curve for case 1a, where 1.2
0 ≈ 100 kPa [from Fig. 5(b)]. As illustrated
the critical layer had σvo (N1)60CS = 5
0
in Fig. 10, the simulated CRR curves for σvo ≈ 200 kPa and
50 kPa plot slightly below and above, respectively, the simulated 0.8
0
CRR curve for σvo ≈ 100 kPa.
Kσ

(N1)60CS = 15
The role of effective vertical stress also was examined by (N1)60CS = 25
0.4 Vs1 = 135 m/s
lowering the water table in the simulations. Fig. 11 presents Vs1 = 175 m/s
the simulation data and simulated CRR curve (green curve) with Vs1 = 200 m/s
PL ¼ 15% for cases 3b and 3c (Table 1) using WUS motions 0
0 200 400 600
and more compressible sand PWP model parameters. The critical
0 (b) Effective vertical stress, σ 'vo (kPa)
layer was located at a depth with σvo ≈ 130 kPa and 146 kPa for
cases 3b and 3c, respectively, and the corresponding simulated Fig. 12. Comparison of Kσ values back-calculated from nonlinear
boundary curves are slightly more liquefiable than simulated data effective stress site response analysis to empirical Kσ correlations
0
for σvo ≈ 100 kPa. developed using data from: (a) Youd et al. (2001); and (b) Idriss
Fig. 12 compares Kσ values back-calculated from the simu- and Boulanger (2010).
lated CRR curves in Figs. 10 and 11 with published correlations.
Because the simulations performed in this study were based on V s1 ,
Fig. 12 presents Kσ values corresponding to V s1 ¼ 135, 175, and
200 m=s. These V s1 values approximate Dr ¼ 40%, 60%, and comparison in Fig. 12 illustrates that the back-calculated Kσ values
80% and ðN1 Þ60 ¼ 5, 15, and 25 based on the Andrus et al. (2004) generally follow the same trend as the published correlations,
V s1 -ðN1 Þ60 and the Shahien (1998) ðN1 Þ60 -Dr -D50 correlations with back-calculated Kσ values decreasing slightly as effective ver-
(where D50 is median grain size, taken as 0.1 mm here). The tical stress increases. However, the published correlations show

© ASCE 04019128-10 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Fig. 13. Comparison of simulated CRR curve using WUS input mo- Fig. 14. Comparison of simulated CRR curve using WUS input mo-
tions and more compressible sand PWP model parameters (CU ¼ 3.8) tions and more compressible sand PWP model parameters (CU ¼ 3.8)
for the HPSC site with field CRR curves. for NEHRP site class D sites with field CRR curves.

0.6
significantly more variation with density [in terms of Dr or ðN1 Þ60 ] Mw = 7.5
than the simulation data using V s1 . Nevertheless, the general agree- σ v0 = 100 kPa
0.4

CSR 7.5
Cu = 3.8
ment demonstrates that the influence of effective vertical stress on
WUS
liquefaction resistance is implicitly incorporated in nonlinear effec-
0.2
tive stress site response analysis.
(a) 0
Influence of Soil Profile 0.6

The analysis results discussed previously involve simple synthetic

0.4
CSR 7.5

soil profiles. In these profiles, excluding the liquefiable layers, the

layers generally are stiff enough to preclude liquefaction. In real-
ity, soil profiles are more complex and the response of underlying 0.2
layers may influence potentially liquefiable layers. To examine the
influence of soil profile variations on the simulations, six published (b) 0
soil profiles (Fig. 2) were excited with WUS input motions. Sands 0.6
in the profiles were assigned the more compressible PWP gener-
ation model parameters. 0.4
CSR 7.5

Fig. 13 presents example nonlinear effective stress site response

analysis results for the HPSC site (Fig. 2) using the more compress- 0.2
ible sand PWP model parameters and WUS motions. In contrast
to the synthetic soil profiles, where the depth of the loose, liquefi- 0
able layer is predefined, the published profiles have different criti- 0 100 200 300
cal layers during different input motions, and thus each site can (c) Vs1 (m/s)
provide critical layers with different V s1 values. To be consistent Kayen et al. (2013) Liq Simulation Liq
with Kayen et al. (2013), the Kσ correction was not applied to CSR Kayen et al. (2013) NLiq Simulation Marg
values computed from nonlinear total stress site response analyses. Kayen et al. (2013) P(L) = 15% Simulation NLiq
This decision was further justified because critical layers in most Andrus and Stokoe (2000) Simulated P(L) = 15%

simulations had σvo 0 < 200 kPa. As shown in Fig. 13, the simulated

CRR curve for site HPSC agrees with the field CRR curves. Sim- Fig. 15. Comparison of simulation data and simulated CRR curve
ilarly, Fig. 14 illustrates that the simulation data and simulated CRR using WUS input motions and more compressible sand PWP model
curve for all NEHRP Site Class D sites (Fig. 2) are in excellent parameters (CU ¼ 3.8) for the Pacific Park Plaza site (NEHRP Site
agreement with the field CRR curves. Class E) with field CRR curves: (a) all cases; (b) clay layers with
Fig. 15 presents the simulation results for the sands in the Pacific ru;residual > 0.8 excluded; and (c) near-fault ground motions excluded.
Park Plaza profile [NEHRP site class E; Fig. 3(c)]. Unlike the
NEHRP site class D sites where the majority of the profiles consist
of clean sands, the site class E sites studied here generally have soft
clays underlying loose sands. These soft clays also generate large components), making the upper sand layers less likely to liquefy.
excess PWP during strong shaking, which measurably decreases As shown in Fig. 15(a), several nonliquefied sandy layer cases plot
the clay shear stiffness. In these cases, the softened clay layers can above the field and simulated CRR curves, likely as a result of the
dampen the ground motions (particularly higher frequency isolating effect of the soft clays underlying the looser sands.

© ASCE 04019128-11 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 importance and benefits of performing liquefaction evaluation in
an effective stress based framework.
The effect of varying loose sand layer thickness (1.5, 3, and
4.5 m) and bedrock depth (20, 25, and 30 m) were investigated
through case 2(a)-(c) and case 4(a)-(c). Observations from the si-
mulated results and CRR curves for these cases are nearly identical
to the simulated CRR curve of baseline case 1a, indicating the si-
mulated CRR curve is less sensitive to the changing of loose layer
thickness and bedrock depth within the range used in this study.
Mei (2018) details these simulation results.

Validation Using Dynamic Centrifuge Tests

Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

The parametric studies presented illustrated that liquefaction evalu-

ation using nonlinear effective stress site response analysis provides
results that are consistent with field observations (i.e., field CRR
curves) and provide potential explanations for some misclassified
Fig. 16. Comparison of simulation data (excluding cases with soft data. To further validate this framework, the following section
clays with ru;residual > 0.8) and simulated CRR curve using WUS input presents the simulation results for several dynamic centrifuge tests.
motions and more compressible sand PWP model parameters
(CU ¼ 3.8) for NEHRP site class E sites with field CRR curves.
Dynamic Centrifuge Tests
For comparison to 1-D site response analysis, the authors focused
on dynamic centrifuge tests conducted by Wilson et al. (1997),
Arulmoli et al. (1992), and Dashti (2009) that involved level-ground
Fig. 15(b) shows the results for Pacific Park Plaza site sand
conditions and saturated clean sand profiles where free-field records
layers after eliminating simulation cases where the soft clay layers of PWP and acceleration are available. Table 2 provides soil profile
generated ru;residual > 0.8. Although there are still some nonlique- details and shaking events for the dynamic shaking events studied
fied sand layers that plot above the field and simulated CRR curves, here (all dimensions are in prototype scale). As shown in Table 2,
these cases correspond to simulations involving motions with the selected centrifuge tests cover a wide range of PGA (0.038–
near-fault effects [as identified by Hayden et al. (2014) and Shahi 0.62 g), predominant period (0.21–0.8 s), liquefiable layer thickness
and Baker (2012)]. When the motions with near-fault effects are (3–10 m), and relative density (30%–90%) This section presents
removed, as shown in Fig. 15(c), the simulations for Pacific Park mainly the simulation of centrifuge test CSP02 (Wilson et al. 1997).
Plaza site sand layers are consistent with the field and simulated Mei (2018) details the simulations for the other centrifuge tests.
CRR curves, with nonliquefied sand layers plotting below the CRR In centrifuge test CSP02, Wilson et al. (1997) examined soil-
curves. These observations confirm that, as anticipated, cyclically pile-superstructure interaction in a liquefiable soil. The soil profile
softened clay layers affect the response of overlying soils and can consisted of 9 m of loose Nevada sand (Dr ¼ 35%) overlying
thereby reduce the likelihood of overlying looser sands liquefying 11.2 m of dense Nevada sand (Dr ¼ 75%). Detailed soil property
during shaking. In contrast, when liquefaction analysis is performed information is described in Table 3. The model was saturated using
in a total stress framework using the cyclic stress method, sand hydroxyl-propyl methylcellulose, with a viscosity about 10 times
layers in these cases often are predicted to liquefy during shaking. greater than water. The model was spun at a centrifugal acceleration
After eliminating cases with clay layers that generated of 30 g; therefore, model diffusion should occur about 3 times faster
ru;residual > 0.8, Fig. 16 presents the combined simulation results than the prototype (Wilson et al. 1997). Input motion CSP02E is a
for all NEHRP site class E sites. Again, except the cases that were strong shaking event with PGA ¼ 0.49 g. It generated significant
excited by WUS motions with near-fault effect, the simulation data excess PWP during shaking. Input motions were baseline-corrected
and simulated CRR curve for the site class E profiles are consistent and filtered using high (25 Hz) and low (0.1 Hz) band-pass Butter-
with the field CRR curves. This analysis again emphasizes the worth fourth-order filters for use in this study.

Table 2. Information of centrifuge tests used for validation of GQ=H þ u model

Predominant Relative Liquefiable layer Total
Shaking event PGA (g) period (s) density (%) thickness (m) thickness (m) Reference
CSP02D 0.038 0.29 35, 75 9 20 Wilson et al. (1997)
CSP02E 0.49 0.29 35, 75 9 20
CSP03H 0.027 0.21 55, 75 9 20
CSP03I 0.49 0.29 55, 75 9 20
VELACS 0.23 0.5 40 10 10 Arulmoli et al. (1992)
SHD02M PI 0.049 0.88 30, 90 3 25 Dashti (2009)
SHD02L PI 0.62 0.37 65, 90 3 25
SHD04M PI 0.15 0.77 50, 90 3 25
SHD04 TCU 0.12 0.68 58, 92 3 25
SHD04L PI 0.38 0.37 66, 92 3 25
Note: All dimensions in prototype scale.

© ASCE 04019128-12 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 1
Table 3. Index properties of Nevada sand
0.9
Property Value
0.8 +0.1 1:1
Assumed specific gravity, Gs 2.67 0.7

Computed Peak ru
Maximum dry density, ρmax (g=cm3 ) 1.76 CSP02D
CSP02E
Minimum dry density, ρmin (g=cm3 ) 1.41 0.6 -0.1
CSP03H
Maximum void ratio, emax 0.89 0.5
CSP03I
Minimum void ratio, emin 0.52 0.4 VELACS
Coefficient of uniformity, CU 1.5 0.3
SHD02S PI
D50 (mm) 0.15 SHD02L PI
0.2 SHD04M PI
Source: Data from Wilson et al. (1997). 0.1 SHD04 TCU
SHD04L PI
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

Measured Peak ru

Fig. 18. Comparison of maximum measured ru after fluctuations of

PWP have stabilized and maximum computed ru;residual in loose
and dense layers during centrifuge tests.

0.6

Mw = 7.5
σ v0 = 100 kPa
0.4

Fig. 17. Profiles of computed response for dynamic centrifuge test CSR 7.5
CSP02E: (a) measured and computed maximum ru;residual ; (b) com- 0.2
puted γ max ; and (c) computed maximum CSR.

0
0 100 200 300
Model Input Parameters Vs1 (m/s)
Because only relative density is provided in the centrifuge tests, the Kayen et al. (2013) Liq Simulation Liq
shear wave velocity profile was estimated using correlations by Kayen et al. (2013) NLiq Simulation NLiq
Menq (2003). Modulus reduction curves and damping curves pro- Kayen et al. (2013) P(L) = 15% Simulated P(L) = 15%
posed by Darendeli (2001) were used to define dynamic soil prop- Andrus and Stokoe (2000)
erties, consistent with the parametric study. Effective stress friction
angles, ϕ 0 ¼ 31° and 34° (for Dr ¼ 35% and 75%, respectively) Fig. 19. Comparison of simulation results for dynamic centrifuge tests
were estimated based on cyclic DSS tests by Arulmoli et al. (1992) with field CRR curves and simulated CRR curve developed using WUS
and Kammerer et al. (2000) on Nevada sand. The GQ=H model was input motions and more compressible sand PWP model parameters
used to fit the Darendeli modulus reduction and damping curves (CU ¼ 3.8).
while remaining consistent with the implied shear strength. Param-
eters for the Vucetic and Dobry (1986) PWP model (p, F, and s)
were assigned to saturated sand layers using the Mei et al. (2018)
correlations for the reported Dr and CU values. Fig. 18 compares the computed maximum ru;residual and mea-
sured maximum ru values after PWP fluctuations had stabilized for
all sand layers in all centrifuge shaking events in Table 2. As shown
Centrifuge Test Simulation Results
in the figure, nearly all of the maximum computed ru;residual values
The authors performed nonlinear total stress and effective stress site fall within 0.1 of the maximum measured ru values. The com-
response analysis for centrifuge shaking events listed in Table 2. parison illustrates the ability of the GQ=H þ u model to reasonably
Fig. 17 presents profiles of computed maximum ru;residual , γ max , estimate PWP generation during shaking.
and maximum cyclic stress ratio (CSRmax ) for centrifuge test shak- Again, to compare results from the nonlinear effective stress site
ing event CSP02E. As described earlier, ru;residual and shear strain response analyses with the field CRR curves, CSRmax in the critical
were computed by nonlinear effective stress site response analysis, layer (computed using nonlinear total stress analysis) was corrected
whereas CSR was computed by nonlinear total stress analysis. In using the Kayen et al. (2013) MSF and reduced by a factor of 0.65 to
general, computed maximum ru;residual values agree closely with compute average CSR7.5 . Also, because the input motions were ap-
measured peak ru values. The maximum computed ru;residual and plied only in one direction in the centrifuge tests, the value of CSR7.5
measured ru in the loose sand layer both exceeded 0.9, indicating was reduced further by a factor of 0.9 to account for multidirectional
that the GQ=H þ u model correctly simulates liquefaction of the shaking effects on PWP generation (e.g., Idriss and Boulanger 2008),
upper loose layer. The simulated critical layer occurs at a depth yielding CSR7.5;2D . The V s value in the critical layer was normalized
∼2 m, which has ru;residual > 0.9 and γ max ≈ 1.5%. to V s1 per Robertson et al. (1992). This procedure produced the

© ASCE 04019128-13 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 Table 4. Field case history for validation of proposed correlations Table 5. Sandy soil PWP model parameters used for analysis of field sites
Measured γ tv
Hypocentral PGA (g) Site Correlation reference v f p F s (%)
distance Wildlife Vucetic and Dobry (1986) 1 2 1 2.6 1.7 0.02
Site Earthquake Magnitude (km) NS EW Port Island Mei et al. (2018) 1 2 1 3.5 1 0.05
Treasure Island Vucetic and Dobry (1986) 1 2 1 3.2 1.4 0.022
Wildlife 1987 Superstition Hills 6.6 28 0.21 0.18
1987 Elmore Ranch 6.2 23 0.12 0.12
Port Island 1995 Kobe 6.9 5 0.34 0.3
Treasure 1989 Loma Prieta 6.9 2.4 N/A 0.16
Island
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

liquefied data point with V s1 ¼ 167 m=s and CSR7.5;2D ¼ 0.188

shown in Fig. 19.
Following this procedure, simulations were performed for all
shaking events in Table 2. Fig. 19 presents the simulation results
and illustrates that the nonlinear effective stress site response analy-
sis yields liquefaction classifications consistent with the field and
simulated CRR curves.

Validation Using Field Case Histories

Similar to the parametric study, the centrifuge tests described in- Fig. 20. Comparison of simulation results for field sites with field CRR
volved simple profiles with uniform relative densities within indi- curves and simulated CRR curve developed using WUS input motions
vidual layers. In the field, soil profiles can be more complex with and more compressible sand PWP model parameters (CU ¼ 3.8).
interlayered sands and clays. To validate the nonlinear effective
stress site response analysis for liquefaction evaluation, the authors
also simulated the response of available field sites where acceler- Island site consists of fine to medium silty sand hydraulic fill
ations were recorded and liquefaction could be assessed. (fine content ∼ 15%). Mei (2018) provides more detailed informa-
tion for these sites, including soil profiles and in situ test data (V s
General Description of Field Conditions and SPT).
Table 4 lists the field case histories evaluated in this study. These
three sites have been used widely for validating site response analy- Model Input Parameters
sis (e.g., Matasovic 1993; Youd and Carter 2005; Ziotopoulou et al.
2012) and have several important characteristics in common, in- Similar to the centrifuge tests, the site response analyses required
cluding: (1) earthquakes that generated excess PWP in the sandy modulus reduction and damping curves, implied shear strength
layers; (2) records of accelerations available from the site or neigh- parameters, and PWP generation and dissipation parameters. For all
boring locations; and (3) reasonable documentation, including the sand, silty sand and sandy silt layers, Darendeli (2001) modulus
information on layer stratification and in situ testing. The North- reduction and damping curves were used. For clay layers, measured
South (N-S) input motion at Treasure Island is not included in this dynamic soil properties were used where available, and where not
study as the phase of the N-S input motion (obtained from Yuerba available, Vucetic and Dobry (1991) modulus reduction and damp-
Buena Island) does not correlate well with N-S motions recorded ing curves were used. Again, the GQ=H model was used to fit all
on Treasure Island (Youd and Carter 2005). modulus reduction and damping curves considering the implied
Among these sites, only the Wildlife site has PWP record- shear strength.
ings, which recorded small ru values during the Elmore Ranch Generation of PWP in clay layers was estimated using the
earthquake, whereas liquefaction was triggered during the Super- Matasovic and Vucetic (1995) model with model parameters de-
stition Hills earthquake. During the Superstition Hills earthquake, fined using the Carlton (2014) correlation. For the sandy layers,
Holzer and Youd (2007) normalized the ru time histories to a maxi- the Vucetic and Dobry (1986) PWP model was used. At the Wild-
mum value of unity by assuming ru ¼ 1 at the end of shaking life site, Vucetic and Dobry (1986) measured PWP model param-
(∼97 s). For the other two sites, although PWP were not recorded, eters (p, F, s) using cyclic DSS and cyclic triaxial tests on relatively
recorded ground motions at the Treasure Island site exhibited char- undisturbed samples. These measured parameters were used di-
acteristics of significant soil softening near the end of shaking, sug- rectly in this study. For the Masado fill at the Port Island site, the
gesting that liquefaction occurred (Idriss and Boulanger 2010). The authors used the Mei et al. (2018) model parameters for more com-
Port Island site experienced sand boils, ground fissures, and signifi- pressible sand. The sands at the Treasure Island site have a grada-
cant tilting of the quay walls, suggesting that liquefaction occurred tion similar to the Heber Road channel fill, therefore the authors
(Cubrinovski et al. 1996). used PWP model parameters (p, F, s) measured by Vucetic and
At the Wildlife site, the most liquefiable layer is the saturated Dobry (1986) for the Heber Road channel fill for the Treasure
Holocene floodplain sediments that consists of sandy silt (ML; Island sands. In each case, the multidirectional shaking parameter
fine content ∼ 78%Þ) and silty sand (SM; fine content ∼ 36%). was activated in the PWP model; therefore, no correction for multi-
At the Port Island site, the liquefiable layer consists of “Masado directional shaking was applied to CSR. Table 5 summarizes the
fill,” which is a well-graded soil with large fractions of gravel sand layer PWP generation model parameters used in this study,
and sand (fine content < 5%). The liquefiable layer at the Treasure and Mei (2018) provides additional details on the input parameters.

© ASCE 04019128-14 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 Simulation Results and ultimately involves a more sophisticated and time-consuming
two-dimensional finite element or finite difference methods.
Using the procedures detailed previously, the authors performed
Although not within the scope of this paper, the proposed 1-D
nonlinear total stress and effective stress site response analyses
nonlinear site response analysis can be performed using three to
for the sites listed in Table 4. The analyses were used to identify
seven ground motions, which is the conventional standard for
critical layers, and values of Vs1 and CSR7.5 were computed for each
ground response analysis. Furthermore, this work represents the
critical layer. Fig. 20 compares the simulation results to the field
development of a deterministic framework for 1-D nonlinear site
and simulated CRR curves. For each field site, horizontal motions
response analysis. This deterministic framework is needed before
from both directions were analyzed and simulations for both direc-
developing a probabilistic framework that incorporates soil prop-
tions are plotted in Fig. 20. As illustrated in the figure, the nonlinear
erty (modulus reduction, damping, compressibility) and other un-
effective stress site response analyses yield results consistent with
certainties. This work currently is ongoing.
the field and simulated CRR curves, with liquefaction being com-
puted for the Superstition Hills event (Wildlife site), at Port Island,
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

and at Treasure Island for the East-West (E-W) direction motion,

whereas liquefaction was not predicted for the Elmore Ranch event Acknowledgments
(Wildlife site). The authors thank Drs. Wu and Polito for sharing their laboratory
test databases.
Conclusion

In this paper, the authors propose criteria to evaluate triggering of

References
level-ground liquefaction using 1-D nonlinear site response analy- Ambraseys, N. N. 1988. “Engineering seismology. I.” Earthquake Eng.
sis with pore-water pressure generation (termed nonlinear effec- Struct. Dyn. 17 (1): 1–50. https://doi.org/10.1002/eqe.4290170101.
tive stress site response). The effective stress-based procedure and Ancheta, T. D., et al. 2014. “NGA-West2 database.” Earthquake Spectra
the triggering criteria are demonstrated using a broad parametric 30 (3): 989–1005. https://doi.org/10.1193/070913EQS197M.
study using 145 WUS and 176 CEUS ground motions propagated Andrade, J. E., A. M. Ramos, and A. Lizcano. 2013. “Criterion for flow
through 44 synthetic sand profiles and six published soil profiles liquefaction instability.” Acta Geotech. 8 (5): 525–535. https://doi.org
(three NEHRP site class D and three site class E). The synthetic /10.1007/s11440-013-0223-x.
profiles involved variations in loose sand layer shear wave velocity, Andrus, R. D., P. Piratheepan, B. S. Ellis, J. Zhang, and C. H. Juang. 2004.
“Comparing liquefaction evaluation methods using penetration-Vs
loose sand layer thickness, depth to the loose sand layer, depth to
relationships.” Soil Dyn. Earthquake Eng. 24 (9): 713–721. https://doi
the water table, and depth to bedrock. The 16,612 parametric analy- .org/10.1016/j.soildyn.2004.06.001.
ses (using both synthetic and published soil profiles) yielded lique- Andrus, R. D., and K. H. Stokoe II. 2000. “Liquefaction resistance
faction resistance results that generally were in excellent agreement of soils from shear-wave velocity.” J. Geotech. Geoenviron. Eng.
with published liquefaction resistance curves (Andrus and Stokoe 126 (11): 1015–1025. https://doi.org/10.1061/(ASCE)1090-0241(2000)
2000; Kayen et al. 2013) for Vs1 < 200 m=s. Furthermore, the non- 126:11(1015).
linear effective stress site response analyses exhibited only minor Arulmoli, K., K. K. Muraleetharan, M. M. Hossain, and L. S. Fruth. 1992.
differences in liquefaction resistance for WUS and CEUS motions. VELACS: Verification of liquefaction analysis by centrifuge studies.
The agreement is not as good for sandy soils with V s1 > 200 m=s, Project No. 90-0562. Irvine, CA: Laboratory Testing Program, Soil
input motions that exhibit near-fault (velocity and displacement Data Report, The Earth Technology Corporation.
Baturay, M. B., and J. P. Stewart. 2003. “Uncertainty and bias in ground-
pulse) effects, and the presence of soft clay layers underlying loose
motion estimates from ground response analyses.” Bull. Seismol. Soc.
sand layers (site class E profiles). The reasons for the lack of agree- Am. 93 (5): 2025–2042. https://doi.org/10.1785/0120020216.
ment are explained and quantified, and the simulation results ac- Bay, J. A., and B. R. Cox. 2001. Shear wave velocity profiling and lique-
tually provide reasonable explanations for why some field sites faction assessment of sites shaken by the 1999 Kocaeli, Turkey earth-
are misclassified using empirical cyclic stress (simplified) liquefac- quake. PEER Project Rep. No. SA3017,18336. Berkeley, CA: Pacific
tion analyses. Earthquake Engineering Research.
The procedure then is validated using 10 shaking events from Boulanger, R. W., and I. M. Idriss. 2004. Evaluating the potential for
three dynamic centrifuge tests and 4 shaking events from three well- liquefaction or cyclic failure of silts and clays. Davis, CA: Center for
documented field sites. Simulations for these 14 shaking events also Geotechnical Modeling.
are in excellent agreement with the field CRR curves, predicting Boulanger, R. W., and I. M. Idriss. 2012. “Probabilistic standard penetra-
tion test–based liquefaction–triggering procedure.” J. Geotech. Geoen-
liquefaction in cases (i.e., specific layers in dynamic centrifuge tests
viron. Eng. 138 (10): 1185–1195. https://doi.org/10.1061/(ASCE)GT
or individual field sites) where liquefaction was observed and no .1943-5606.0000700.
liquefaction where liquefaction was not observed. Building Seismic Safety Council. 2015. NEHRP recommended seismic
The proposed nonlinear effective stress site response analy- provisions for new buildings and other structures. Part 1 Provisions,
sis approach for liquefaction triggering incorporates site-specific Part 2 Commentary, FEMA Rep. No. P-1050-1. Washington, DC:
ground motions, actual soil profiles, and soil mass characteristics Building Seismic Safety Council.
(e.g., density, gradation, compressibility) as direct inputs, thereby Carlton, B. 2014. “An improved description of the seismic response of sites
circumventing the uncertainties related to evaluating liquefaction with high plasticity soils, organic clays, and deep soft soil deposits.”
triggering using the cyclic stress method. Furthermore, the nonlin- Ph.D. dissertation, Dept. of Civil and Environment Engineering, Univ.
ear effective stress site response analysis approach may allow lique- of California.
Cetin, K. O., R. B. Seed, A. Der Kiureghian, K. Tokimatsu, L. F. Harder, Jr.,
faction to be assessed for parameters (e.g., depth) outside of the
R. E. Kayen, and R. E. Moss. 2004. “Standard penetration test-based
empirical liquefaction database, potentially broadening its applica- probabilistic and deterministic assessment of seismic soil liquefaction
tion. As such, the proposed approach for liquefaction can be used as potential.” J. Geotech. Geoenviron. Eng. 130 (12): 1314–1340. https://
part of a staged analysis that begins with the simplified cyclic stress doi.org/10.1061/(ASCE)1090-0241(2004)130:12(1314).
method, proceeds to the proposed 1-D nonlinear effective stress site Cox, D. R., and E. J. Snell. 1970. The analysis of binary data. London:
response analysis when ground motions for a project are defined, Methuen and Co.

© ASCE 04019128-15 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 Cubrinovski, M., K. Ishihara, and F. Tanizawa. 1996. “Numerical simula- Ishihara, K., and S. I. Li. 1972. “Liquefaction of saturated sand in triaxial
tion of the Kobe Port Island liquefaction.” In Proc., 11th World Conf. on torsion shear test.” Soils Found. 12 (2): 19–39. https://doi.org/10.3208
Earthquake Engineering. Acapulco, Mexico. /sandf1972.12.19.
Darendeli, M. B. 2001. “Development of a new family of normalized Jaky, J. 1944. “The coefficient of earth pressure at rest.” [In Hungarian.]
modulus reduction and material damping curves.” Ph.D. thesis, Dept. J. Soc. Hung. Architects Eng. 78 (22): 355–388.
of Civil and Environmental Engineering, Univ. of Texas at Austin. Juang, C. H., T. Jiang, and R. D. Andrus. 2002. “Assessing probability-
Darve, F. 1996. “Liquefaction phenomenon of granular materials and con- based methods for liquefaction potential evaluation.” J. Geotech. Geo-
stitutive stability.” Eng. Comput. 13 (7): 5–28. https://doi.org/10.1108 environ. Eng. 128 (7): 580–589. https://doi.org/10.1061/(ASCE)1090
/02644409610151539. -0241(2002)128:7(580).
Dashti, S. 2009. “Toward developing an engineering procedure for evalu- Kammerer, A., J. Wu, J. Pestana, M. Riemer, and R. Seed. 2000. Cyclic
ating building performance on softened ground.” Ph.D. dissertation, simple shear testing of Nevada sand for PEER Center. Project
Dept. of Civil and Environment Engineering, Univ. of California. 2051999. Rep. UCB/GT/00-01. Berkeley, CA: Univ. of California.
Dobry, R., and T. Abdoun. 2011. “An investigation into why liquefaction Karimi, Z., and S. Dashti. 2015. “Numerical simulation of earthquake
induced soil liquefaction: Validation against centrifuge experimental
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

charts work: A necessary step toward integrating the states of art and
practice.” In Proc., 5th Int. Conf. on Earthquake Geotechnical Engi- results.” In IFCEE 2015, 11–20. Reston, VA: ASCE.
neering, 13–44. Santiago, Chile: Chilean Geotechnical Society. Kayen, R., R. E. S. Moss, E. M. Thompson, R. B. Seed, K. O. Cetin, A. D.
Dobry, R., and T. Abdoun. 2015. “Cyclic shear strain needed for liquefac- Kiureghian, and K. Tokimatsu. 2013. “Shear-wave velocity–based
tion triggering and assessment of overburden pressure factor Kσ.” probabilistic and deterministic assessment of seismic soil liquefaction
J. Geotech. Geoenviron. Eng. 141 (11): 04015047. https://doi.org/10 potential.” J. Geotech. Geoenviron. Eng. 139 (3): 407–419. https://doi
.1061/(ASCE)GT.1943-5606.0001342. .org/10.1061/(ASCE)GT.1943-5606.0000743.
Dobry, R., R. S. Ladd, F. Y. Yokel, R. M. Chung, and D. Powell. 1982. Kayen, R. E., J. K. Mitchell, R. B. Seed, A. Lodge, S. Y. Nishio, and R.
“Prediction of pore water pressure buildup and liquefaction of sands Coutinho. 1992. “Evaluation of SPT-, CPT-, and shear wave-based
during earthquakes by the cyclic strain method.” In National Bureau methods for liquefaction potential assessment using Loma Prieta data.”
of Standards building science series. Gaithersburg, MD: National In Vol. 1 of Proc., 4th Japan–US Workshop on Earthquake Resistant
Bureau of Standards Publications. Design of Lifeline Facilities and Countermeasures for Soil Liquefac-
Domenico, P. A., and M. D. Mifflin. 1965. “Water from low-permeability tion, 177–192. New York: National Center for Earthquake Engineering
Research.
sediments and land subsidence.” Water Resour. Res. 1 (4): 563–576.
Kenan, H. 2005. “Pore pressure generation characteristics of sands and silty
https://doi.org/10.1029/WR001i004p00563.
sands: A strain approach.” Ph.D. thesis, Dept. of Civil and Environment
Elgamal, A., Z. Yang, T. Lai, B. L. Kutter, and D. W. Wilson. 2005.
Engineering, Univ. of Texas.
“Dynamic response of saturated dense sand in laminated centrifuge
Kramer, S. L., B. A. Asl, P. Ozener, and S. S. Sideras. 2015. “Effects of
container.” J. Geotech. Geoenviron. Eng. 131 (5): 598–609. https://doi
liquefaction on ground surface motions.” In Perspectives on earthquake
.org/10.1061/(ASCE)1090-0241(2005)131:5(598).
geotechnical engineering, 285–309. New York: Springer.
Elgamal, A., Z. Yang, and E. Parra. 2002. “Computational modeling of
Liao, S. S., D. Veneziano, and R. V. Whitman. 1988. “Regression models
cyclic mobility and post-liquefaction site response.” Soil Dyn. Earth-
for evaluating liquefaction probability.” J. Geotech. Eng. 114 (4): 389–411.
quake Eng. 22 (4): 259–271. https://doi.org/10.1016/S0267-7261(02)
https://doi.org/10.1061/(ASCE)0733-9410(1988)114:4(389).
00022-2.
Lodge, A. L. 1994. “Shear wave velocity measurements for subsurface
Gingery, J. R. 2014. “Effects of liquefaction on earthquake ground
characterization.” Ph.D. thesis, Dept. of Civil and Environment Engi-
motions.” Ph.D. thesis, Dept. of Structural Engineering, Univ. of
neering, Univ. of California.
California.
Markham, C. S. 2015. “Response of liquefiable sites in the central busi-
Gingery, J. R., A. Elgamal, and J. D. Bray. 2015. “Response spectra
ness district of Christchurch, New Zealand.” Ph.D. thesis, Dept. of Civil
at liquefaction sites during shallow crustal earthquakes.” Earthquake and Environment Engineering, Univ. of California.
Spectra 31 (4): 2325–2349. https://doi.org/10.1193/101813EQS272M. Matasovic, N. 1993. “Seismic response of composite horizontally-layered
Groholski, D. R., Y. M. A. Hashash, B. Kim, M. Musgrove, J. Harmon, and soil deposits.” Ph.D. thesis, Dept. of Civil and Environment Engineer-
J. P. Stewart. 2016. “Simplified model for small-strain nonlinearity and ing, Univ. of California.
strength in 1-D seismic site response analysis.” J. Geotech. Geoenviron. Matasovic, N., and M. Vucetic. 1995. “Generalized cyclic-degradation-
Eng. 142 (9): 04016042. https://doi.org/10.1061/(ASCE)GT.1943-5606 pore-pressure generation model for clays.” J. Geotech. Eng. 121 (1):
.0001496. 33–42. https://doi.org/10.1061/(ASCE)0733-9410(1995)121:1(33).
Hashash, Y. M. A., B. Kim, S. M. Olson, and S. Moon. 2014. “Geotech- McGuire, R. K., W. J. Silva, and C. J. Costantino. 2001. Technical basis for
nical issues and site response in the central US.” In Seismic hazard revision of regulatory guidance on design ground motions: Hazard-
design issues, 71–90. Reston, VA: ASCE. and risk-consistent ground motion spectra guidelines. Washington,
Hashash, Y. M. A., M. Musgrove, D. R. Groholski, C. A. Phillips, and DC: US Nuclear Regulatory Commission.
D. Park. 2016. DEEPSOIL 6.0, user manual. Urbana, IL: Univ. of Mei, X. 2018. “Porewater pressure generation and liquefaction analysis
Illinois at Urbana–Champaign. using nonlinear, effective stress-based site response analysis.” Ph.D.
Hayden, C. P., J. D. Bray, and N. A. Abrahamson. 2014. “Selection of near- thesis, Dept. of Civil and Environment Engineering, Univ. of Illinois,
fault pulse motions.” J. Geotech. Geoenviron. Eng. 140 (7): 04014030. Urbana–Champaign.
https://doi.org/10.1061/(ASCE)GT.1943-5606.0001129. Mei, X., S. M. Olson, and Y. M. Hashash. Forthcoming. “Evaluation of a
Holzer, T. L., and T. L. Youd. 2007. “Liquefaction, ground oscillation, and simplified soil constitutive model considering implied strength and
soil deformation at the Wildlife Array, California.” Bull. Seismol. Soc. porewater pressure generation for 1-D seismic site response.” Can. Geo-
Am. 97 (3): 961–976. https://doi.org/10.1785/0120060156. tech. J. https://doi.org/10.1139/cgj-2018-0893.
Idriss, I. M., and R. W. Boulanger. 2008. Soil liquefaction during earth- Mei, X., S. M. Olson, and Y. M. Hashash. 2018. “Empirical porewater pres-
quakes: Monograph MNO-12, 261. Oakland, CA: Earthquake Engi- sure generation model parameters in 1-D seismic site response analy-
neering Research Institute. sis.” Soil Dyn. Earthquake Eng. 114 (Nov): 563–567. https://doi.org/10
Idriss, I. M., and R. W. Boulanger. 2010. SPT-based liquefaction triggering .1016/j.soildyn.2018.07.011.
procedures. Rep. UCD/CGM-10. Davis, CA: Center for Geotechnical Menq, F. Y. 2003. “Dynamic properties of sandy and gravelly soil.” Ph.D.
Modeling, Univ. of California at Davis. dissertation, Dept. of Civil and Environment Engineering, Univ. of
Ishihara, K. 1993. “Liquefaction and flow failure during earthquakes.” Texas.
Geotechnique 43 (3): 351–451. https://doi.org/10.1680/geot.1993.43 Mitchell, J. K. 1994. Vol. 94 of In situ test results from four Loma Prieta
.3.351. earthquake liquefaction sites: SPT, CPT, DMT and shear wave velocity.

© ASCE 04019128-16 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128

 Berkeley, CA: Earthquake Engineering Research Center, College of Toprak, S., T. L. Holzer, M. J. Bennett, and J. C. Tinsley III. 1999. “CPT-
Engineering, Univ. of California at Berkeley. and SPT-based probabilistic assessment of liquefaction.” In Proc.,
Moss, R. E., R. B. Seed, R. E. Kayen, J. P. Stewart, A. Der Kiureghian, and 7th US–Japan Workshop on Earthquake Resistant Design of Lifeline
K. O. Cetin. 2006. “CPT-based probabilistic and deterministic assess- Facilities and Countermeasures against Liquefaction, 69–86. Buffalo,
ment of in situ seismic soil liquefaction potential.” J. Geotech. Geoen- NY: Multidisciplinary Center for Earthquake Engineering Research.
viron. Eng. 132 (8): 1032–1051. https://doi.org/10.1061/(ASCE)1090 Vucetic, M., and R. Dobry. 1986. Pore pressure build-up and liquefaction
-0241(2006)132:8(1032). at level sandy sites during earthquakes. Research Rep. No. CE-86-3.
Pestana, J. M., C. H. Hunt, and R. R. Goughnour. 1997. FEQDRAIN: A Troy, NY: Dept. of Civil Engineering, Rensselaer Polytechnic Institute.
finite element computer program for the analysis of the earthquake gen- Vucetic, M., and R. Dobry. 1991. “Effect of soil plasticity on cyclic
eration and dissipation of pore water pressure in layered sand deposits response.” J. Geotech. Eng. 117 (1): 89–107. https://doi.org/10.1061
with vertical drains. UCB/EERC-97/15. Berkeley, CA: Earthquake /(ASCE)0733-9410(1991)117:1(89).
Engineering Research Center, Univ. of California. Whitman, R. V. 1971. “Resistance of soil to liquefaction and settlement.”
Phillips, C., and Y. M. Hashash. 2009. “Damping formulation for nonlinear Soils Found. 11 (4): 59–68. https://doi.org/10.3208/sandf1960.11.4_59.
1D site response analyses.” Soil Dyn. Earthquake Eng. 29 (7): 1143– Wijewickreme, D., and A. Soysa. 2016. “Stress–strain pattern–based cri-
Downloaded from ascelibrary.org by University of South Carolina on 09/30/21. Copyright ASCE. For personal use only; all rights reserved.

1158. https://doi.org/10.1016/j.soildyn.2009.01.004. terion to assess cyclic shear resistance of soil from laboratory element
Polito, C. P. 1999. “The effects of non-plastic and plastic fines on the lique- tests.” Can. Geotech. J. 53 (9): 1460–1473. https://doi.org/10.1139/cgj
faction of sandy soils.” Ph.D. thesis, Dept. of Civil and Environment -2015-0499.
Engineering, Virginia Tech. Wilson, D. W., R. W. Boulanger, and B. L. Kutter. 1997. Soil-pile-
Poulos, S. J., G. Castro, and J. W. France. 1985. “Liquefaction evaluation superstructure interaction at soft or liquefiable soil sites—Centrifuge
procedure.” J. Geotech. Eng. 111 (6): 772–792. https://doi.org/10.1061 data report for CSP02. Rep. No. UCD/CGMDR-97/05. Davis, CA:
/(ASCE)0733-9410(1985)111:6(772). Center for Geotechnical Modeling, CEE Dept., Univ. of California.
Robertson, P. K., D. J. Woeller, and W. D. L. Finn. 1992. “Seismic cone Wu, J., R. B. Seed, and J. M. Pestana. 2003. Liquefaction triggering
penetration test for evaluating liquefaction potential under cyclic load- and post liquefaction deformations of Monterey 0/30 sand under uni-
ing.” Can. Geotech. J. 29 (4): 686–695. https://doi.org/10.1139/t92-075. directional cyclic simple shear loading. Geotechnical Engineering Re-
Robertson, P. K., and C. E. Wride. 1998. “Evaluating cyclic liquefaction search Rep. No. UCB/GE-2003/01. Berkeley, CA: Univ. of California.
potential using the cone penetration test.” Can. Geotech. J. 35 (3): 442–459. Yang, Z., J. Lu, and A. Elgamal. 2008. “OpenSees soil models and solid-
https://doi.org/10.1139/t98-017. fluid fully coupled elements.” In User’s manual: Version 1. La Jolla,
Seed, B., and K. L. Lee. 1966. “Liquefaction of saturated sands during CA: Univ. of California, San Diego.
cyclic loading.” J. Soil Mech. Found. Div. 92 (6): 105–134. Youd, T. L., et al. 2001. “Liquefaction resistance of soils: Summary report
Seed, H. B., and I. M. Idriss. 1971. “Simplified procedure for evalu- from the 1996 NCEER and 1998 NCEER/NSF workshops on evalu-
ating soil liquefaction potential.” J. Soil Mech. Found. Div. 97 (9): ation of liquefaction resistance of soils.” J. Geotech. Geoenviron.
1249–1273. Eng. 127 (10): 817–834. https://doi.org/10.1061/(ASCE)1090-0241
Seed, H. B., I. M. Idriss, F. Makdisi, and N. Banerjee. 1975. Representation (2001)127:4(297).
of irregular stress time histories by equivalent uniform stress series in Youd, T. L., and B. L. Carter. 2005. “Influence of soil softening and
liquefaction analyses. Rep. No. EERC 75-29. Berkeley, CA: Earth- liquefaction on spectral acceleration.” J. Geotech. Geoenviron. Eng.
quake Engineering Research Center, Univ. of California. 131 (7): 811–825. https://doi.org/10.1061/(ASCE)1090-0241(2005)
Shahi, S., and J. Baker. 2012. “Pulse classifications from NGA West2 data- 131:7(811).
base.” Accessed November 2, 2019. http://web.stanford.edu/∼bakerjw Youd, T. L., and S. K. Noble. 1997. Liquefaction criteria based on
/pulse_classification_v2/Pulse-like-records.html. statistical and probabilistic analyses. Technical Report NCEER-97.
Shahien, M. M. 1998. “Settlement of structures on granular soils subjected Buffalo, NY: National Center for Earthquake Engineering Research.
to static and earthquake loads.” Ph.D. dissertation, Dept. of Civil and Zhou, Y. G., and Y. M. Chen. 2007. “Laboratory investigation on assessing
Environment Engineering, Univ. of Illinois, Urbana–Champaign. liquefaction resistance of sandy soils by shear wave velocity.” J. Geotech.
Stark, T. D., and S. M. Olson. 1995. “Liquefaction resistance using CPT Geoenviron. Eng. 133 (8): 959–972. https://doi.org/10.1061/(ASCE)
and field case histories.” J. Geotech. Eng. 121 (12): 856–869. https://doi 1090-0241(2007)133:8(959).
.org/10.1061/(ASCE)0733-9410(1995)121:12(856). Ziotopoulou, K., R. W. Boulanger, and S. L. Kramer. 2012. “Site response
Terzaghi, K., R. B. Peck, and G. Mesri. 1996. Soil mechanics in engineer- analysis of liquefying sites.” In Proc., GeoCongress Conf. 2012.
ing practice. New York: Wiley. Reston, VA: ASCE.

© ASCE 04019128-17 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(2): 04019128