For Volume 1:

Geotechnical and Geophysical Site Characterisation 5 – Lehane, Acosta-Martínez & Kelly (Eds)
© 2016 Australian Geomechanics Society, Sydney, Australia, ISBN 978-0-9946261-1-0

Liquefaction
Liquefaction
Forassessment
assessment
Volume 2: based
basedon
oncombined
combineduse
useofofCPT
CPTand
andshear
shearwave
wave
velocity
velocitymeasurement
measurement
Geotechnical and Geophysical Site Characterisation 5 – Lehane, Acosta-Martínez & Kelly (Eds)
© 2016 Australian Geomechanics Society, Sydney, Australia, ISBN 978-0-9946261-2-7
Z.Z.Bán
Bán&&A.A.Mahler
Mahler
Budapest
BudapestUniversity
UniversityofofTechnology
Technologyand
andEconomics,
Economics,Budapest,
Budapest,Hungary
Hungary
T.J.
T.J.Katona
Katona
MVM
MVMNuclear
NuclearPower
PowerPlant
PlantPaks
PaksLtd.,
Ltd.,Paks,
Paks,Hungary
Hungary
E.E.Győri
Győri
MTA
MTACSFK
CSFKGGI
GGISeismological
SeismologicalObservatory,
Observatory,Budapest,
Budapest,Hungary
Hungary

ABSTRACT:
ABSTRACT:EmpiricalEmpiricalliquefaction
liquefactionpotential
potentialassessment
assessmentmethods
methodsare aregenerally
generallybased
basedononthetheresult
resultofofCPT,
CPT,
SPT
SPTororshear
shearwave
wavevelocity
velocitymeasurement.
measurement.InInmoremorecomplex
complexororhighhighrisk
riskprojects
projectsCPT
CPTand andVV S Smeasurement
measurement
areoften
are oftenperformed
performedatatthe thesame
samelocation
locationcommonly
commonlyininthe theform
formofofseismic
seismicCPT.
CPT.However,
However,combined
combineduse useofof
bothin-situ
both in-situindices
indicesininone
onesingle
singleempirical
empiricalmethod
methodisislimited.
limited.ForForthis
thisreason
reasonthe
thegoal
goalofofthis
thisresearch
researchwas wastoto
developsuch
develop suchananempirical
empiricalmethodmethodwhere
wherethe
theresult
resultofofCPT
CPTand andVVS Smeasurement
measurementare areused
usedininparallel
paralleland
andcan
can
supplement
supplementeach eachother.
other.After
Afterthe
theaccumulation
accumulationofofa acase
casehistory
historydatabase
databasewhere
whereboth
bothmeasurements
measurementsare areavail-
avail-
able
ableand
andperforming
performingthe thenecessary
necessarycorrections
correctionsofofthe
thevariables,
variables,logistic
logisticregression
regressionwaswasperformed
performedtotoobtain
obtainthethe
probability
probabilitycontours
contoursofofliquefaction
liquefactionoccurrence.
occurrence.InInthisthiscase
casethethegraphical
graphicalrepresentation
representationofofthe thecyclic
cyclicre-re-
sistance
sistanceratio
ratiocurve
curvefor
fora agiven
givenprobability
probabilitycan
canbebereplaced
replacedbybya asurface.
surface.

1 1 INTRODUCTION
INTRODUCTION ties
tiescomplement
complementeach
eachother
othersince
sincethey
theycharacterize
characterizethe
the
behavior
behaviorofofgranular
granularsystems
systemsatatdifferent
differentlevels
levelsofof
Soil
Soilliquefaction
liquefactionisisone oneofofthe themost
mostdevastating
devastatingsec- sec- strain.
strain.
ondary
ondaryeffects
effectsofofearthquakes
earthquakesand andcancancause
causesignifi-
signifi-
cant
cantdamage
damageininthe thebuilt
builtinfrastructure.
infrastructure.For Forthis
thisreason
reason 2 2 LIQUEFACTION
LIQUEFACTIONPOTENTIAL
POTENTIALASSESSMENT
ASSESSMENT
liquefaction
liquefactionhazard hazardshall
shallbebeconsidered
consideredininallallregions
regions BASED
BASEDON
ONCPT
CPTAND
ANDVV
SS
where
wheremoderate-to-high
moderate-to-highseismic seismicactivity
activityencounters
encounters
2.1
2.1 The
TheCPT
CPTand S based
andVV S based
empirical
empiricalmethods
methods
with
withsaturated,
saturated,loose,
loose,granular
granularsoilsoildeposits.
deposits.Several
Several
approaches
approachesexist existtototake
takeinto
intoaccount
accountthisthishazard,
hazard,fromfrom Since
Sincethetheintroduction
introductionofofcyclic
cyclicshear
shearstress
stressapproach
approach
which
whichthe thein-situ
in-situtest
testbased
basedempirical
empiricalmethods
methodsare arethe
the (Seed
(Seed&&Idriss
Idriss1971)
1971)several
severalempirical
empiricalmethods
methodshave have
most
mostcommonly
commonlyused usedininpractice.
practice.Traditional
Traditionalmeans meansofof been
beenpublished
publishedbybydifferent
differentauthors
authorsthat
thatcan
cangive
givea arel-
rel-
these
thesetests
testsarearethe
theStandard
StandardPenetration
PenetrationTest Test(SPT),
(SPT), atively
ativelyreliable
reliablequantification
quantificationofoffactor
factorofofsafety
safetyoror
Cone
ConePenetration
PenetrationTest Test(CPT)
(CPT)and andshear
shearwavewavevelocity
velocity probability
probability ofof liquefaction.
liquefaction. InIn current
current engineering
engineering
(V(VS)S)measurement.
measurement.InInroutine routineororlow-budget
low-budgetprojectsprojects practice
practicethethemost
mostcommonly
commonlyused usedCPTCPTbased
basedmethods
methods
often
oftenonlyonlyCPT CPTtesting
testingisisperformed
performedthat thatwillwillthen
then are
arethe
theprocedures
proceduresproposed
proposedbybyRobertson
Robertson&&Wride Wride
serve
serveasasthe thebasis
basisofofliquefaction
liquefactionpotential
potentialassessment.
assessment. (1998),
(1998),Moss
Mossetetal.al.(2006),
(2006),Idriss
Idriss&&Boulanger
Boulanger(2008)(2008)
InInmore
morecomplex
complexororhigh-risk
high-riskprojects
projectssuch
suchasasnuclear
nuclear and
andBoulanger
Boulanger&&Idriss Idriss(2014).
(2014).As Asthe
theuse
useofofCPT
CPTfor for
power
powerplants,
plants,CPTCPTand andVV S Smeasurement
measurementare areoften
oftenper-
per- ground
groundprofile
profilecharacterization
characterizationisisvery verypopular
popularitsitsap-
ap-
formed
formedatatthe thesame
samelocation,
location,commonly
commonlyininthe theform
formofof plication
plicationfor forliquefaction
liquefactionpotential
potentialevaluation
evaluationisisalsoalso
Seismic
Seismic Cone Cone Penetration
Penetration Test Test (sCPT).
(sCPT). However,
However, prevalent.
prevalent.
even
evenififthetheresults
resultsofofthethetwotwotests
testsare
areavailable
availablefor forthe
the Compared
Comparedwith withCPTCPTand andSPTSPTbased
basedmethods,
methods,the the
same
samespot,spot,empirical
empiricalliquefaction
liquefactionpotential
potentialevaluation
evaluation methods
methodsbasedbasedononVV S Stests
testsare
areless
lesswidely
widelyusedusedinin
can
canbebeperformed
performedusing usingeither
eitherofofthem,
them,but butcombined
combined practice
practiceforforliquefaction
liquefactionsusceptibility
susceptibilityevaluation.
evaluation.For For
use
useofofthethedata
dataininoneonesingle
singlemethod
methodhas hasbeen
beenlimited.
limited. very
verylong
longtime
timethethemethod
methodofofAndrus
Andrus&&Stokoe
Stokoe(2000)
(2000)
InInorder
ordertotosurmount
surmountthis thisissue,
issue,ananattempt
attempthas hasbeen
been was
wasused
usedalmost
almostexclusively.
exclusively.Very Veryrecently
recentlythethework
work
made
madetotodevelop
developananempirical
empiricalmethod,
method,whichwhichexploits
exploits ofofKayen
Kayenetetal.al.(2013)
(2013)made madea ahuge
hugestep
stepininthethead-
ad-
both
boththetheresults
resultsofofCPTCPTand andVV S Smeasurements.
measurements.The Theef-ef- vancement
vancementofofVV S Sbased
basedmethods.
methods.Besides
Besidesthe thead-ad-
fort
fortisisbased
basedononthe theassumption
assumptionthat thatthese
thesesoil
soilproper-
proper- vanced
vancedstatistical
statisticalframework
frameworkadopted
adoptedbybythe theauthors,
authors,

597
the most remarkable accomplishment was the compi-
lation of a global catalog of 422 case histories.
2.2 Independence of CPT and VS measurement
Although, both VS and CPT tip resistance (qc) strongly
depend on relative density and effective stress state,
they can supply complementary information about the
soil, as the former is a small strain property, while the
latter is a large strain one. As discussed by Schneider
et al. (2004), VS in sands is controlled by the number
and area of grain-to-grain contacts, on the other hand,
penetration resistance in sands is controlled by the in-
teraction of particles being sheared by and rotating
around the penetrometer.
For the purpose of developing an empirical method
exploiting both parameters a database was compiled
where both measurements are available. To investigate
independence of the two parameters, the correlation
coefficient, R2 between VS and qc (measured in the
same critical layer for each case history) was deter-
mined for the dataset (Figure 1). Note that in the Fig-
ure the values of the two variables were normalized to
100 kPa effective overburden pressure to exclude the
influence of different effective overburden pressure on
them. The obtained correlation coefficient R2=0.156
indicates that in their untransformed state the two pa-
rameters have poor correlation.
were found where both measurements were available.
As locations where liquefaction occurred are more en-
ticing for post-earthquake field investigators than sites
where no apparent liquefaction occurred, the assem-
bled dataset over represents liquefied sites (68 sites),
relative to non-liquefied sites (30 sites).
The core of the database was assembled from the
CPT case history catalog of Moss (2003) and VS da-
taset of Kayen et al. (2013), from which 73 and 53 lo-
cations could be used, respectively. Additional case
histories were gathered from the publications of Bay
& Cox (2001), Ku et al. (2004), Moss et al. (2005),
Moss et al. (2009), Cox et al. (2013) Boulanger &
Idriss (2014), Batillas et al. (2014). The final database
consists case histories from 12 earthquakes (1975
Haicheng, 1976 Tangshan, 1979 Imperial Valley,
1981 Westmoreland, 1983 Borah Peak, 1987 Elmore
Ranch, 1987 Superstition Hills, 1989 Loma Prieta,
1999 Chi-Chi, 1999 Kocaeli, 2008 Achaia-Elia, 2011
Great Tohoku). For each case the following infor-
mation was available: the triggering ground motion’s
moment magnitude (Mw) and maximum horizontal ac-
celeration (amax), depth of the critical layer, groundwa-
ter level, total (σv) and effective overburden pressure
(σ’v), stress reduction factor of the soil profile, CPT
tip resistance and sleeve friction (fs) and shear wave
velocity in the critical layer. A summary of the dataset
is presented in Table 1.

Table 1. Summary of the compiled case history database

Liquefied cases Non-liquefied cases
Standard Standard
Mean Mean
deviation deviation
Mw 7.33 0.64 7.16 0.90
amax (m/s2) 0.30 0.15 0.23 0.17
σv (kPa) 92.73 45.43 98.13 48.80
σ'v (kPa) 60.83 27.35 66.65 27.64
qc1 (MPa) 5.18 2.68 10.14 6.95
RF (%) 0.92 0.57 1.05 0.56
VS1 (m/s) 152.72 29.89 181.17 67.41

4 OVERVIEW OF THE EMPIRICAL

Figure 1. Correlation between Vs1 and qc1N
3 FIELD CASE HISTORY DATABASE
The first and most time consuming step of recent
study was the collection of liquefaction/non-
liquefaction field case history catalog. Through careful
review of existing CPT and VS databases 98 cases
FRAMEWORK
4.1 General approach
In their study’s Moss et al. (2006) and Kayen et al.
(2013) applied Bayesian framework to quantify prob-
ability of liquefaction in which they treated the seis-
mic demand and soil related variables separately in
their logistic regression. On the other hand Idriss &
Boulanger (2008) applied a semi-empirical approach
in which correction and normalization of the variables
were based on experimental data and theoretical con-
siderations, and the final relationship was derived us-
ing maximum likelihood estimation treating only two
variables: the magnitude and effective overburden

598
stress corrected cyclic stress ratio (CSRM=7.5,σ’v=1atm)
and the equivalent clean sand value of normalized
overburden corrected cone tip resistance (qc1Ncs). Be-
cause the dataset for the present study was relatively
small compared to the aforementioned catalogs, a
wide enough range of magnitude and other variables
to accurately quantify their effect on liquefaction po-
tential was lacking. For this reason, the resistance var-
iables (qc and VS) and the seismic demand were cor-
rected and normalized before performing the logistic
regression.
4.2 Normalization of seismic demand
According to the framework of simplified empirical
procedures seismic demand induced by an earthquake
can be represented by the cyclic stress ratio (CSR) at a
depth z below ground surface using the following ex-
pression (Seed & Idriss 1971):
𝑎𝑎𝑚𝑚𝑚𝑚𝑚𝑚 𝜎𝜎𝑣𝑣
𝐶𝐶𝐶𝐶𝐶𝐶 = 0.65 𝑔𝑔
𝑟𝑟
𝜎𝜎′𝑣𝑣 𝑑𝑑
(1)
where g = gravitational acceleration; and rd = stress
reduction factor.
In order to take into account duration – or number
of equivalent cycles – of different earthquakes a mag-
nitude scaling factor (MSF) is used to adjust CSR for
a duration typical of an average event of Mw=7.5. In
the present study the MSF equation of Idriss (1999)
was used:
𝑀𝑀𝑀𝑀𝑀𝑀 = 6.9 exp (
−𝑀𝑀𝑤𝑤
) − 0.058 (2)
4 classification cannot be made directly. Although for
Previous researchers noted that susceptibility of
soils to cyclic liquefaction strongly depends on effec-
tive overburden stress. This effect is often referred as
Kσ effect and is generally taken into account by a mul-
tiplying factor. In this study the expression of Boulan-
ger & Idriss (2004) developed using critical state
framework was adopted:
𝜎𝜎′
𝐾𝐾𝜎𝜎 = 1 − 𝐶𝐶𝜎𝜎 𝑙𝑙𝑙𝑙 ( 𝑃𝑃 𝑣𝑣 ) ≤ 1.1 (3)
𝑎𝑎
1
𝐶𝐶𝜎𝜎 = 37.3−8.27(𝑞𝑞 0.264 ≤ 0.3 (4)
𝑐𝑐1𝑁𝑁 )
where Pa = atmospheric pressure; and qc1N = normal-
ized overburden corrected cone tip resistance.
4.3 Normalization of CPT tip resistance and shear
wave velocity
Effective overburden stress can profoundly influence
CPT measurements. This effect is typically accounted
for by normalizing the tip resistance measured at a
given depth and vertical effective stress to a reference
effective stress of 100 kPa. Boulanger & Idriss (2004)
evaluated experimental and theoretical data for CPT
599
and they obtained the following iterative formula for
the effective stress normalization of qc:
𝐶𝐶𝑁𝑁 𝑞𝑞𝑐𝑐
𝑞𝑞𝑐𝑐1𝑁𝑁 = (5)
𝑃𝑃𝑎𝑎
𝛽𝛽
𝑃𝑃
𝐶𝐶𝑁𝑁 = (𝜎𝜎′𝑎𝑎 ) ≤ 1.7 (6)
𝑣𝑣
𝛽𝛽 = 1.338 − 0.249(𝑞𝑞𝑐𝑐1𝑁𝑁 )0.264 (7)
The role of fines on liquefaction susceptibility is a
somewhat contentious topic. The conclusions of recent
studies of the subject are rather controversial (Tan et
al. 2013). Nevertheless it is agreed that if the fines
content exceeds approximately 35-40% the coarser
grains will “float” in the matrix of fine-size particles
and the cyclic behavior of the soil will be governed by
the fines. Although many researchers reported an ini-
tial increase then a decrease in liquefaction potential
as fines content (FC) increases, all of the most widely
used empirical methods apply monotonically increas-
ing fines content correction. In the present study,
equivalent clean sand values of the tip resistance have
been determined using the updated equation of Bou-
langer & Idriss (2014):
𝑞𝑞𝑐𝑐1𝑁𝑁𝑁𝑁𝑁𝑁 = 𝑞𝑞𝑐𝑐1𝑁𝑁 + ∆𝑞𝑞𝑐𝑐1𝑁𝑁 (8)
𝑞𝑞𝑐𝑐1𝑁𝑁 9.7 15.7 2
∆𝑞𝑞𝑐𝑐1𝑁𝑁 = (11.9 + ) 𝑒𝑒𝑒𝑒𝑒𝑒 (1.63 − −( ) )(9)
14.6 𝐹𝐹𝐹𝐹+2 𝐹𝐹𝐹𝐹+2
A huge drawback of CPT compared to SPT that
soil samples are not obtained during the test thus soil
many of the case histories the actual fines content
were available in their source database, to keep con-
sistency between the input data and the intended fu-
ture use of the method, fines content of the critical
layers were determined empirically using the soil be-
havior index (Ic) recommended by Robertson & Wride
(1997). Boulanger & Idriss proposed the following
expression:
𝐹𝐹𝐹𝐹 = 80𝐼𝐼𝑐𝑐 − 137 (10)
0% ≤ 𝐹𝐹𝐹𝐹 ≤ 100% (11)
However, it should be noted that general correla-
tions between FC and Ic exhibit large scatter.
As well as CPT tip resistance, VS is also routinely
normalized to an equivalent value measured at 100
kPa effective overburden stress (Robertson et al.
1992):
𝑃𝑃 0.25
𝑉𝑉𝑆𝑆1 = 𝑉𝑉𝑆𝑆 𝐶𝐶𝑉𝑉𝑆𝑆 = 𝑉𝑉𝑆𝑆 (𝜎𝜎′𝑎𝑎 ) 𝐶𝐶𝑉𝑉𝑆𝑆 ≤ 1.5 (12)
𝑣𝑣
Small strain shear modulus (Gmax) is closely related
to VS. As the small strain stiffness of sands, silts and
clays are of a similar range, VS measurement do not
allow detecting small differences in fines content, i.e.
VS is relatively insensitive to FC. Several studies
showed (Andrus & Stokoe 2000, Zhou & Chen 2007,
Kayen et a. 2013) that increase in fines content to 35%
has a maximum adjustment of 5 m/s. Compared to un-
certainties arising from other parts of the methodology
this uncertainty is fairly negligible. Thus, fines content
correction of shear wave velocity has been neglected.
4.4 Input variables for logistic regression
After performing all of the above discussed normaliza-
tion and corrections, three explanatory variables re-
mained to participate in the logistic regression: the
equivalent clean sand value of normalized overburden
corrected cone tip resistance (qc1Ncs), the overburden
corrected shear wave velocity (VS1), and the magni-
tude and effective stress corrected cyclic stress ratio
(CSRM=7.5,σ’v=1atm), which was obtained by the follow-
ing formula:
𝐶𝐶𝐶𝐶𝐶𝐶
𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀=7.5,𝜎𝜎′𝑣𝑣=1𝑎𝑎𝑎𝑎𝑎𝑎 = 𝑀𝑀𝑀𝑀𝑀𝑀∙𝐾𝐾 (13)
𝜎𝜎
There is no general agreement how the variables
should be incorporated in the logistic regression,
should they be used in their logarithmic, polynomial
or untransformed form. Baecher & Christian (2003)
recommended that variables should be transformed so
that their frequency distribution has to be approxi-
mately normal. Following this guideline
CSRM=7.5,σ’v=1atm has been included in the regression
by its natural logarithm, while qc1Ncs and VS1 remained
untransformed.
5 LOGISTIC REGRESSION
Logistic regression is often used to explore the rela-
tionship between a binary response and a set of ex-
planatory variables. The occurrence or absence of liq-
uefaction can be considered as binary outcome and the
previously summarized three parameters are the ex-
planatory variables. Moss et al. (2006) and Kayen et
al. (2013) adopted Bayesian updating technique (de-
veloped for probabilistic assessment of liquefaction
initiation by Cetin et al. (2002)), while Boulanger &
Idriss (2014) used maximum likelihood estimation to
develop their resulting correlation.
The key components of these methods are the for-
mulation of a limit state model that has a value of zero
at the limit state and is negative and positive for lique-
faction and non-liquefaction cases, respectively, and a
likelihood function that is proportional to the condi-
tional probability of observing a particular event as-
suming a given a set of parameters. Adopting the ap-
proach of Cetin et al. (2002) the following limit state
function has been formed:
600
𝑔𝑔 = 𝜃𝜃1 𝑉𝑉𝑆𝑆1 + 𝜃𝜃2 𝑞𝑞𝑐𝑐1𝑁𝑁𝑁𝑁𝑁𝑁 + 𝜃𝜃3 𝑙𝑙𝑙𝑙(𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀=7.5,𝜎𝜎′ 𝑣𝑣=1𝑎𝑎𝑎𝑎𝑎𝑎 ) +
𝜃𝜃4 + 𝜀𝜀 (14)
where 𝛳𝛳1…4 = the set of unknown parameters; and ε =
model error term. The latter is introduced to account
for the influences of the missing variables and the pos-
sible incorrect model form. If ε is assumed to be nor-
mally distributed with zero mean, the probability of
liquefaction can be expressed as:
𝑔𝑔−𝜀𝜀
𝑃𝑃𝐿𝐿 = 𝛷𝛷 [− 𝜎𝜎𝜖𝜖
] (15)
where Φ = standard normal cumulative probability
function; and σε = standard deviation of the model er-
ror term.
Assuming the statistical independence of the obser-
vations compiled from different sites, the likelihood
function can be written as the product of the probabili-
ties of the observations. As it was noted in section 3
the dataset contains significantly more liquefaction
cases than non-liquefaction cases; this bias is undesir-
able in logistic regression and can adversely affect the
result. A way to address this issue is to weight each
class of cases according to the proportion of the oth-
er’s class population in the total database (Cetin et al.
2002, Baecher & Christian 2003) while the sum of the
weights should remain 2.0 (Mayfield 2007). Accord-
ing to these guidelines the corresponding weight fac-
tors for liquefaction and non-liquefaction cases are
0.62 and 1.38, respectively. Then the likelihood func-
tion takes the following form:
𝑔𝑔−𝜀𝜀 0.62 𝑔𝑔−𝜀𝜀 1.38
𝐿𝐿 = ∏𝑙𝑙𝑙𝑙𝑙𝑙 [𝛷𝛷 (−
𝜎𝜎𝜖𝜖
)] × ∏𝑛𝑛𝑛𝑛𝑛𝑛−𝑙𝑙𝑙𝑙𝑙𝑙 [𝛷𝛷 (
𝜎𝜎𝜖𝜖
)] (16)
After taking the natural logarithm of the likelihood
function that is more convenient to work with, the un-
known parameters were determined using maximum
likelihood estimation.
6 PROBABILITY OF LIQUEFACTION
The logistic regression using the likelihood function in
Equation 16 yields the following result:
0.080𝑉𝑉𝑆𝑆1 +0.177𝑞𝑞𝑐𝑐1𝑁𝑁𝑁𝑁𝑁𝑁
−8.40𝑙𝑙𝑙𝑙(𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀=7.5,𝜎𝜎′ =1𝑎𝑎𝑎𝑎𝑎𝑎 )−46.04
𝑃𝑃𝐿𝐿 = 𝛷𝛷 [− 3.46
𝑣𝑣
] (17)
The denominator, that is the standard deviation of
the error term, is of particular interest since it de-
scribes the efficiency of the liquefaction relationship.
The regressed value is somewhat higher than that of
other commonly used methods, but the method is still
promising, since the method demonstrates certain re-
finement compared to the other methods in spite of
relative small size of the dataset used. It shall be noted
that the σE value only incorporates model uncertainty.
The measurements errors or parameter uncertainty
have not been taken into account in the analysis.
The cyclic resistance ratio for a given probability of
liquefaction can be expressed by rearranging Equation
17:
0.080𝑉𝑉𝑆𝑆1 +0.177𝑞𝑞𝑐𝑐1𝑁𝑁𝑁𝑁𝑁𝑁
−46.04+3.46𝛷𝛷−1 (𝑃𝑃𝐿𝐿 )
𝐶𝐶𝐶𝐶𝐶𝐶𝑀𝑀=7.5𝜎𝜎′ 𝑣𝑣=1𝑎𝑎𝑎𝑎𝑎𝑎 = 𝑒𝑒𝑒𝑒𝑒𝑒 ( 8.40
) (18)
This can be used in deterministic analysis by select-
ing a probability contour to separate liquefaction and
non-liquefaction states. Figure 2 shows the probability
surface corresponding to PL = 50%.
Figure 2. Cyclic resistance ratio surface corresponding to 50% of
liquefaction probability (solid squares – liquefaction cases, hol-
low circles – non-liquefaction cases)
7 CONCLUSION
For high-risk projects CPT and VS measurement are
often performed on the same location however the
possibility to characterize the soil’s resistance with
both indices in one single empirical liquefaction
method was limited. The goal of the research was to
develop such a method within the framework of sim-
plified empirical procedures that can reduce uncertain-
ty of empirical methods by combining two in-situ in-
dices; a small strain property measurement with a
large strain measurement. In the first step by careful
reviewing of the already existing liquefaction case his-
tory databases, sites were selected where the records
of both CPT and VS measurement are available. After
implementing the necessary corrections on the gath-
601
ered 98 case histories with respect to fines content,
overburden pressure and magnitude, a logistic regres-
sion was performed to obtain the probability contours
of liquefaction occurrence. The proposed formula is
an initial attempt to exploit the advantages offered by
the measurements of two soil parameters instead of
one.
This equation only takes into account model uncer-
tainty but variation of the parameters has been ne-
glected. In spite of this the proposed equation can be a
useful tool in its current form (as being conditional on
known values for the input parameters). To assess the
probability of the phenomenon in a liquefaction haz-
ard evaluation, the conditional probability of liquefac-
tion provided by this equation needs to be combined
with the probabilities of the qc1Ncs, VS1 and
CSRM=7.5,σ’v=1atm; i.e., the parameter uncertainties.
The use of the recommended relationship is likely
to be limited to high-risk projects where probabilistic
seismic hazard analysis (PSHA) and detailed site
characterization is required. In the sequence of hazard
analysis liquefaction hazard analysis is preceded by
both the PSHA (which would account for the majority
of the uncertainty in the seismic demand) and the site
exploration (which would account for the majority of
the uncertainty in the soil resistance parameters). In
that scenario, it may be reasonable to only include
model uncertainty in the liquefaction triggering analy-
sis because the parameter uncertainties were already
accounted for in the previous stages of the analysis
(Boulanger & Idriss 2014).
8 FUTURE WORK
This research is far from being considered as complet-
ed, there is still considerable more work to be done.
As the dataset is relatively small, in the first round
more emphasis was put on the quantity rather than the
quality of the data. However, it is believed that if the
magnitude of the database will allow proper quality
assessment, omission of low quality data can improve
the reliability of the methodology. This problem refers
back to the most important flaw of the method that is
the relatively minor number of cases included in the
catalog. With the inclusion of cases from the 2010-11
Canterbury and other future earthquakes hopefully a
more robust dataset will be available that will allow
the separate treatment of all the independent variables
and further refinement of this approach.
As it was mentioned this equation only takes into
account model uncertainty but variation of the pa-
rameters were neglected. There is an intention to in-
clude measurement errors.
 Although this work is only the beginning of a road,
it is hoped that after more refinement and expanded
dataset this approach will be able to provide answers
with less degree of uncertainty that is essential for de-
sign of high-risk facilities.
9 REFERENCES
Andrus, R.D. & Stokoe, K.H., II. 2000. Liquefaction resistance
of soils from shear-wave velocity. Journal of Geotechnical
and Geoenvironmental Engineering 126(11): 1015–1025.
Batilas, A., Pelekis, P., Vlachakis, V., & Athanasopoulos, G.
2014. Soil Liquefaction/Nonliquefaction in the Achaia-Ilia
(Greece) 2008 Earthquake: Field Evidence, Site Characteriza-
tion and Ground Motion Assessment. International Journal of
Geoengineering Case Histories 2(4): 270-287.
Bay, J.A., & Cox, B.R. 2001. Shear wave velocity profiling and
liquefaction assessment of sites shaken by the 1999 Kocaeli,
Turkey earthquake. PEER Project Report SA3017-18336, Pa-
cific Earthquake Engineering Research, Berkeley.
Baecher, G.B. & Christian, J.T. 2003. Reliability and statistics in
geotechnical engineering. Chichester: John Wiley & Sons
Inc.
Boulanger, R.W. & Idriss, I.M. 2004. State normalization of pen-
etration resistance and the effect of overburden stress on liq-
uefaction resistance. 11th International Conf. on Soil Dynam-
ics and Earthquake Engineering and 3rd International
Conference on Earthquake Geotechnical Engineering; Pro-
ceedings, University of California, Berkeley, 7-9 January
2004.
Boulanger, R.W. & Idriss, I.M. 2014. CPT and SPT based Lique-
faction Triggering Procedures. Report No. UCD/CGM-14/01,
University of California, Davis.
Cetin, K.O., Der Kiureghian, A. & Seed, R.B. 2002. Probabilistic
models for the initiation of seismic soil liquefaction. Struc-
tural Safety 24(1): 67–82.
Cox, B.R., Boulanger, R.W., Tokimatsu, K., Wood, C.M., Abe,
A., Ashford, S., Donahue, J., Ishihara, K., Kayen, R., Katsu-
mata, K., Kishida, T., Kokusho, T., Mason, H.B., Moss,
R.E.S., Stewart, J.P., Tohyama, K., & Zekkos, D. 2013. Liq-
uefaction at strong motion stations and in Urayasu City dur-
ing the 2011 Tohoku-Oki earthquake. Earthquake Spectra
29(S1): S55-S80.
Idriss, I.M. 1999. An update to the Seed-Idriss simplified proce-
dure for evaluating liquefaction potential, Proceedings, TRB
Workshop on New Approaches to Liquefaction, Publication
No. FHWARD- 99-165, Federal Highway Administration.
Idriss, I.M. & Boulanger, R.W. 2008. Soil Liquefaction during
Earthquakes. Monograph MNO-12, Earthquake Engineering
Research Institute, Oakland.
Kayen, R., Moss, R.E.S., Thompson, E.M., Seed, R.B., Cetin,
K.O., Der Kiureghian, A., Tanaka, Y. & Tokimatsu, K. 2013.
Shear-Wave Velocity–Based Probabilistic and Deterministic
Assessment of Seismic Soil Liquefaction Potential. Journal of
Geotechnical and Geoenvironmental Engineering 139(3):
407-419.
Ku, C.-S., Lee, D.-H., & Wu, J.-H. 2004. Evaluation of Soil Liq-
uefaction in the Chi-Chi, Taiwan Earthquake Using CPT. Soil
Dynamics and Earthquake Engineering 24(9-10): 659-673.
Mayfield, R.T. 2007. The return period of soil liquefaction. Ph.D
Dissertation, University of Washington, Seattle.
602
Moss, R.E.S. 2003. CPT-based probabilistic assessment of seis-
mic soil liquefaction initiation. Doctor of Philosophy disserta-
tion, University of California, Berkeley.
Moss, R.E.S., Collins, B.D. & Whang, D.H. 2005. Retesting of
Liquefaction/Nonliquefaction Case Histories in the Imperial
Valley. Earthquake Spectra 21(1): 179-196.
Moss, R.E.S., Seed, R.B., Kayen, R.E., Stewart, J.P., Der Kiu-
reghian A. & Cetin, K.O. 2006. CPT-based probabilistic and
deterministic assessment of in situ seismic soil liquefaction
potential. Journal of Geotechnical and Geoenvironmental
Engineering 132(8): 1032–1051.
Moss, R.E.S., Kayen, R.E., Tong, L., Liu, S., Cai, G. & Wu, J.
2009. Reinvestigating liquefaction and nonliquefaction case
histories from the 1976 Tangshan earthquake. PEER Project
Report 2009/102, Pacific Earthquake Engineering Center,
Berkeley.
Robertson, P.K., Woeller, D.J., & Finn, W.D.L. 1992. Seismic
cone penetration test for evaluating liquefaction potential un-
der cyclic loading. Canadian Geotechnical Journal 29(4):
686–695.
Robertson, P.K. & Wride, C.E. 1997. Cyclic liquefaction and its
evaluation based on SPT and CPT. In Youd, T.L. & Idriss,
I.M. (eds.), NCEER Workshop on Evaluation of Liquefaction
Resistance of Soils, Proceedings, Salt Lake City, Technical
Report NCEER-97-0022: National Center for Earthquake En-
gineering Research.
Robertson, P.K. & Wride, C.E. 1998. Evaluating cyclic liquefac-
tion potential using the cone penetration test. Canadian Ge-
otechnical Journal 35(3): 442-459.
Schneider, J.A., McGillivray, A.V. & Mayne, P.W. 2004. Eval-
uation of SCPTU intra-correlations at sand sites in the Lower
Mississippi River valley, USA. In Viana Da Fonseca, A. &
Mayne, P.W. (eds), Geotechnical & Geophysical Site Char-
acterization; Proc. ISC-2, Porto, 19-22 September 2004. Rot-
terdam: Millpress.
Seed, H.B. & Idriss, I.M. 1971. Simplified procedure for evaluat-
ing soil liquefaction potential. Journal of the Soil Mechanics
and Foundations Division 97(SM9): 1249-1273.
Tan, C.S., Marto, A., Leong, T.K. & Teng, L.S. 2013. The Role
of Fines in Liquefaction Susceptibility of Sand Matrix Soils.
Electronic Journal of Geotechnical Engineering 18(L): 2355–
2368.
Zhou, T.-G. & Chen, T.-M. 2007. Laboratory investigation on as-
sessing liquefaction resistance of sandy soils by shear wave
velocity. Journal of Geotechnical and Geoenvironmental En-
gineering 133(8): 959–972.