Engineering Geology 226 (2017) 184–191

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

New statistical and graphical assessment of CPT-based empirical MARK

correlations for the shear wave velocity of soils
Mingfei Zhanga,b, Liyuan Tonga,c,⁎
a
Institute of Geotechnical Engineering, Southeast University, 210096 Nanjing, China
b
Civil Engineering and Architecture Institute, Zhengzhou University of Aeronautics, No.15, Wenyuan West Road, Zhengzhou New District, 450046 Zhengzhou, China
c
Jiangsu Key Laboratory of Urban Underground Engineering and Environmental Safety, Southeast University, Nanjing 210096, China

A R T I C L E I N F O A B S T R A C T

Keywords: The object of this paper is to gather, analyze and assess CPT–Vs correlation equations. Consequently, in this
Shear wave velocity paper the validity of some of these correlations between Vs and CPT parameters for Jiangsu clays was assessed
CPTU using piezocone data from Yangtze Delta deposits through six existing and two new indexes and the cumulative
Statistical and graphical assessment frequency curve graph. Common existing correlations are briefly reviewed and discussed. New empirical for-
The perfect distance
mulae are suggested to correlate CPT-based parameters and Vs, using regression analysis. The comparison is
conducted between the existing and proposed equation based on a dataset collected at seven sites in Jiangsu,
China. This paper proposes a new statistical assessment method taking the location of the points into account,
including the perfect distance and the slope of the specific line that links the point D10 and the point D90. The
method is more obvious and accurate. The results of these indexes and the cumulative frequency curve graph
show that a new improved expression gave better performance. The new indexes and the cumulative frequency
curve graph can surely be used to analyze and assess correlations including the new expression, which is very
helpful to evaluate correlations.

1. Introduction
Shear wave velocity (Vs) has been recognized as one of the im-
portant soil properties in earthquake and geotechnical engineering
(Ohta and Goto, 1978). It has been widely used to define site categories,
to assess site conditions, to estimate earthquake ground motion, to
evaluate the liquefaction resistance of soils and to implement remedial
measures (Borcherdt, 1994; Dobry et al., 2000; Andrus and Stokoe,
2000; Kayen et al., 2013). Correlations and empirical relationships
extensively abound in geotechnical engineering and are always used for
local regions. There seem to be several reasons. The first and funda-
mental reason, obviously, is that there is no/little theoretical relation-
ship between the wave velocity parameters (e.g. Vs) and general Cone
penetration test-based (CPT-based) parameters (e.g. fs). The second one
is that the use of empirical relationships gives a fast, effective way of
predicting the value of a parameter based on the values of some other,
possibly more easily determined, parameters. The third is that, in the
preliminary stages including feasibility studies, when there are limited
funds available for soil exploration, empirical correlations become very
valuable (Onyejekwe et al., 2015). The fourth is the space constraints
associated with these tests, especially in urban areas. The last but not
least reason is that, at the concept design stage of a project, most
practitioners are left with the use of correlations and typical values as
site investigation data may not be available or limited. Such correla-
tions may be the sole weapon available to the designer when no or poor
soil data is available and/or only limited testing has been carried out. In
some situations, correlations are even required as a test of reason-
ableness on the derived design parameters (Ameratunga et al., 2016).
Hence, the easiest way to estimate Vs is to use empirical relationships
with the common CPT-based parameters (e.g. fs).
In geotechnical engineering, many design parameters of soil are
associated with the piezocone penetration test (CPTU) because not only
is it an effective method but also it can provide Vs and other common
values, such as cone penetration tip resistance qc, and sleeve friction fs.
Common methods for obtaining Vs mainly consist of cross-hole,
down-hole, uphole and PS suspension logging (Kitsunezaki, 1980), and
the surface wave inversion (SWI) method is also applied as low cost yet
providing a knowledge of the elastic soil properties (Mokhtar et al.,
1988; Stokoe et al., 1988; Jongmans et al., 1990; Pitilakis et al., 1999;
Maheswari et al., 2010). Another increasing approach is termed as
seismic CPTU (SCPTU) (Campanella et al., 1986; Lunne et al., 1997; Cai
et al., 2010, 2011), through adding sensors to the CPTU.

⁎
Corresponding author at: Institute of Geotechnical Engineering, Southeast University, Nanjing 210096, China, Jiangsu Key Laboratory of Urban Underground Engineering and
Environmental Safety, Southeast University, 210096 Nanjing, China
E-mail address: atmu@seu.edu.cn (L. Tong).

http://dx.doi.org/10.1016/j.enggeo.2017.06.007
Received 29 December 2016; Received in revised form 3 May 2017; Accepted 11 June 2017
Available online 19 June 2017
0013-7952/ © 2017 Published by Elsevier B.V.
M. Zhang, L. Tong Engineering Geology 226 (2017) 184–191

Table 1
Correlations between CPT parameters (qc for CPT, qt for CPTU) and shear wave velocity.

Applicability # Correlation(Vs or Vs1) References

All soils A 2.62qt0.395IC0.912D0.124SFa Andrus et al. (2007)

B [10(0.55Ic + 1.68)(qt − σv0) / pa]0.5 Robertson (2009)
[10.1 log(qc) − 11.4]1.67 (100 fs / qc) 0.3 Hegazy and Mayne (1995)
118.8log(fs) + 18.5 Mayne (2006)
32.3 qc0.089 fs0.1219D 0.215 Piratheepan (2002)
C Vs1 = 102 qc10.23 Robertson et al. (1992)
D Vs1 = 135 qc10.23 Fear and Robertson (1995)
E Vs1 = 149 qc10.205 Karray et al. (2011)
F Vs1 = 0.0831 qc1N0.103 (σ′v0/pa) 0.25e 1.786Ic Hegazy and Mayne (2006)
Clay G 14.13 qc0.359 e0–0.473 Hegazy and Mayne (1995)
3.18 qc0.549 fs0.025 Hegazy and Mayne (1995)
H 9.44qc0.435 e0–0.532 Mayne and Rix (1995)
11.9 qc0.269 fs109D 0.127 Piratheepan (2002)
I 1.75 qc0.627 Mayne and Rix (1995)
J 0.1 qc Jaime and Romo (1988)
K 2.944qc0.613 Long and Donohue (2010)
L 65.00qt0.150 e0–0.714 Long and Donohue (2010)
M 1.96(1 + Bq)1.202 qt0.579 Long and Donohue (2010)
Sand N 134.11 + 0.0052 qc Sykora and Stokoe (1983)
O 17.48 qc0.33(σ′v0) 0.27 Baldi et al. (1989)
P 13.18 qc0.192(σ′v0) 0.179 Hegazy and Mayne (1995)
12.02 qc0.319 fs− 0.0466 Hegazy and Mayne (1995)
25.3 qc0.103 fs0.029D 0.155 Piratheepan (2002)

Notes: Pa = 100 kPa; SFa = 0.92 for Holocene and 1.12 for Pleistocene, Stress unit in kPa and depth (D) in meters (m).

CPT-based correlations for evaluating Vs have advanced owning to
the contributions of numerous researchers and progressed with the
improvement of people's knowledge and experimental devices. There
are dozens of correlation equations, and, common representative
equations are listed in Table 1. Early equations mainly adopted a linear
relationship and single independent variable (e.g., Sykora and Stokoe,
1983; Jaime and Romo, 1988). Jaime and Romo (1988) suggested an-
other correlation, which was a linear relationship as opposed to the
nonlinear relationships proposed by other researchers (Sykora, 1987).
The exact reason is not yet clear, but may be related to a lack of ac-
curate tools and profound comprehension. Since Baldi et al. (1989)
gave an initial correlation in power exponent form for sand, countless
correlations have always been proposed (e.g., Hegazy and Mayne,
1995; Long and Donohue, 2010). Primary correlations involved in two
variables: qt (qc) and fs (σv0′, e0orBq), consequent occurred three
(Hegazy and Mayne, 1995; Piratheepan, 2002; Andrus et al., 2007) or
four variables (Hegazy and Mayne, 2006; Robertson, 2009) and in the
log form (Hegazy and Mayne, 1995; Hegazy and Mayne, 2006). Several
studies concluded that, however, the use of stressnormalized qc values
in Vs correlations proved to be considerably less accurate than corre-
lations based on non-normalized values (Sykora and Stokoe, 1983;
Lodge, 1994; Hasancebi and Ulusay, 2007; Piratheepan, 2002). And
numerical investigators (Robertson et al., 1992; Fear and Robertson,
1995; Hegazy and Mayne, 2006; Karray et al., 2011) presented the non-
normalized correlation between Vs1and qc1orqc1N values (see Table 1),
in whichVs1is the shear wave velocity normalized for the vertical ef-
fective stress, expressed by:
Vs1 = Vs (pa σv′0 )0.25
whereσv0′is the vertical effective stress (kPa). Cone penetration re-
sistance is often corrected for overburden stress (resulting in the term
ofqc1), and the truly normalized cone penetration resistance normalized
for overburden stress (qc1N, dimensionless) is given by (Robertson and
Wride, 1998; Hegazy and Mayne, 2006):
qc1N = (qc pa )(pa σv′0 )n
were generally developed for specific soil types (i.e., “Sand” or “Clay”)
or for more general “All Soils”. Some equations shown in Table 1 are
not named for lack of some of the related data. Two methods were
evaluated for selecting which correlation equations to use for design.
The first method involved using “All Soils” equations for the entire soil
profile. The second method involved selection of soil-type dependent
correlation equations based on the CPT and SBT. For sandy soils
(IC < 2.05), Vs was estimated using the value from Correlation N
(Sykora and Stokoe, 1983) through Correlation P (Piratheepan, 2002)
shown in.
Table 1. For intermediate soils (2.05 < IC < 2.60), Vs was esti-
mated using the average result from Correlation A (Andrus et al., 2007)
to Correlation F (Hegazy and Mayne, 2006). For clayey soils
(IC > 2.60), Vs was estimated using the average value through Cor-
relation G (Hegazy and Mayne, 1995) to Correlation M (Long and
Donohue, 2010). The soil type-specific method developed, however,
used an average of existing correlations for each soil type, which pro-
duced spikes (high and low) in the predicted Vs profile as material
transitions where differential equations were used for adjacent CPT sub-
layers. For this reason and ease of implementation, the “All Soils”
method was considered to be more desirable (Wair et al., 2012).
Obviously, primary correlations involved two variables: qt (qc) and
fs (σv0′, e0orBq), then gradually there occurred three (Hegazy and
Mayne, 1995; Piratheepan, 2002; Andrus et al., 2007) or four variables
(Hegazy and Mayne, 2006; Robertson, 2009). Epoch is significant along
with other geotechnical parameters such as Vs and only Correlation A
(Andrus et al., 2007) involves it.
There are so many CPT–Vs correlation equations that how to assess
(1)
them, how to distinguish them, option variables becomes important.
Conventional assessment is the simple application of a correlation
coefficient (e.g. Cai et al., 2014) that appears to be general and single.
The literature (Onyejekwe et al., 2015; Dikmen, 2009; Maheswari et al.,
2010) will give a clue to analyzing and assessing equations.
2. Data
(2)

wherepais a reference pressure in the same units as qc (i.e., equal to The study area, mainly overlying sensitive clay deposits, is located
0.1 MPa for qc or qc1 in MPa). If the value of IC is < 2.6, n = 0.5, else in Jiangsu Province of eastern China (see Fig. 1). The dataset used in
n = 0.75. this study consists of seven main sources: Nanjing, Lianyungang,
The published CPT–Vs correlation equations presented in Table 1 Changzhou, Yancheng, Suzhou, Taizhou and Yangzhou, respectively. A

185
M. Zhang, L. Tong Engineering Geology 226 (2017) 184–191

Fig. 1. Location map for different types of Jiangsu clays CPTU sites.

Table 2 summary of these sites is also presented in Table 2. At each site, samples
Summary of sites surveyed. were taken at different depths that corresponded to the depths where
the shear velocity measurement was made. The CPTU device used
Site location Soil type Geological Vs measured References
origin by
throughout the study comprised a versatile piezocone system equipped
with digital cone penetrometers fabricated with a 60° tapered, 10 cm2
Lianyungang Soft clay Marine SCPTU Liu et al. (2008), tip area cone, which provided measurements of qt, fs andu2with a 5-mm-
Liu et al. (2011) thick porous filter located just behind the cone tip. The rate of pene-
Nanjing Silty clay Backswamp SCPTU, Tong et al. (2013)
SASW
tration for all tests was 20 mm/s, enabling one set of readings to be
Yancheng Soft clay Lagoonal SCPTU, DH obtained for every 50 mm penetration. Shear wave velocities were
Suzhou Soft clay Alluvial and SCPTU, CH Tong et al. (2013) measured at intervals of 1.0 m. Tube samples of 76 mm diameter were
lacustrine collected from boreholes, using stainless steel fixed-piston tube sam-
Changzhou Firm clay Alluvial and SCPTU, DH
plers below ground level. Once the fixed-piston sampler was withdrawn
lacustrine
Taizhou Silty clay Alluvial and SCPTU, DH from the borehole, the soil sample at the end of the tube was excavated
lacustrine for wax sealing at both ends. The soil parameters for these sites are
Yangzhou Soft clay Lagoonal SCPTU, summarized in Table 2. Typical profiles of CPTU measurements, in-
SASW cluding qt, fs, and u2 versus depth recorded in Suzhou are presented in
Fig. 2. Data was collated and validated. Data validation entailed the
investigation of identified outliers so as to exclude only true outliers
qt(MPa) fs(kPa) u2(kPa) from the analysis. The in situ measurements could be influenced by a
0 2 4 6 8 101214 0 50 100 150 0 300 600 900
0 number of factors that are probably interrelated, such as compressi-
bility, particle distribution, mineralogy, and grain shape. Descriptive
statistics of the parameters in this paper and their frequency distribu-
5 tions are presented in Table 3.

3. Analysis
10
3.1. Formula analysis
Depth (m)

15
Existing correlations obtained from other sites may not be suitable
for local conditions. Uncritical use of these correlations could possibly
20
Table 3
Descriptive statistics of variables used in the study.
25
IC qc e0 σ′v0 σv0 D

30 Minimum 4.1 20,800 2.4 167.3 289.2 16.4

Maximum 1.79 460 0.6 12.34 33.3 1.79
Fig. 2. Typical CPTU soundings in Suzhou. Mean 2.75 3561.62 0.98 94.10 154.19 8.26
Standard deviation 0.534 3719.577 0.3797 35.896 69.598 3.707

186
M. Zhang, L. Tong Engineering Geology 226 (2017) 184–191

350 600 350

550
300 500 300
450
250 400 250
350
200 200
Vsc

Vsc

Vsc
300
150 250 150
200
100 150 100
100
50 50 50
50 100 150 200 250 300 350 50 100 150 200 250 300 350 400 450 500 550 600 50 100 150 200 250 300 350
Vsm Vsm Vsm
A I Q
350 900 350
800
300 300
700
250 600 250

200 500

Vsc

Vsm
Vsc

200
400
150 150
300
100 200 100
100
50 50
50 100 150 200 250 300 350 100 200 300 400 500 600 700 800 900 50 100 150 200 250 300 350
Vsm Vsm Vsm
B J R
350 350
1200
300 300
1000
250 250
800
Vsc1

Vsc

200 200

Vsc
600
150
150
400
100
200 100
50
50 100 150 200 250 300 350 50
200 400 600 800 1000 1200 50 100 150 200 250 300 350
Vsm1 Vsm Vsm
C K S
450 350 350
400
300 300
350
300 250 250
250 200
Vsc1

Vsc

200

Vsc
200
150 150
150
100 100 100
50 50
50 100 150 200 250 300 350 400 450 50
50 100 150 200 250 300 350 50 100 150 200 250 300 350
Vsm1 Vsm Vsm
D L T
500 650
450 600
550
400
500
350 450
400
300
Vsc1

350
Vsc

250 300
200 250
200
150
150
100 100
50
50 50 100150200250300350400450500550600650
50 100 150 200 250 300 350 400 450 500
VSm1 Vsm
E M
1600 350
1400
300
1200
250
1000
Vsc1

200
Vsc

800
600 150
400
100
200
50
200 400 600 800 1000 1200 1400 1600 50 100 150 200 250 300 350
Vsm1 Vsm
F N
600 1800
550 1600
500 1400
450
1200
400
1000
Vsc

350
Vsc

300 800
250 600
200 400
150 200
100
100 150 200 250 300 350 400 450 500 550 600 200 400 600 800 1000 1200 1400 1600 1800
Vsm Vsm
G O
600 350
550
500 300
450
400 250
350
200
Vsc

Vsc

300
250 150
200
150 100
100
50 50
50 100 150 200 250 300 350 400 450 500 550 600 50 100 150 200 250 300 350
Vsm Vsm
H P
(caption on next page)
187
M. Zhang, L. Tong
Fig. 3. Measured (Vsm) versus calculated (Vsc) shear wave velocity.
yield a great deal of error, while subsequent uncertainty in the esti-
mated values of Vs would result in either inadequate, unsafe designs or
overly conservative, inefficient designs (Onyejekwe et al., 2015). The
measurement uncertainty from each test, cone (e.g., Kulhawy and
Mayne, 1990) and shear wave velocity (e.g., Moss, 2008), and then the
uncertainty in correlating these two together, model uncertainty. The
correlations between penetration resistance and Vs are based on re-
gression analysis of the datasets. These datasets typically contain a
significant amount of scatter in the measured data (as evident in Cor-
relations A through S in Fig. 3). Uncertainty in geotechnical properties
can be formally grouped into two classes: aleatory and epistemic un-
certainty (Lacasse and Nadim, 1996; Whitman, 1996; DNV, 2007). The
aim of analysis in this study is model-related epistemic uncertainty.
The published literature contains many equations describing the
correlation between CPT-based parameters and Vs. Some are material
dependent (sand or clay), while others depend on depth, qt (qc) or fs and
geological age.
Table 1 shows a gradual preponderance of power functional corre-
lations based on qt or qc in place of linear correlations. Correlations are
multifarious but the most common correlations used are those easy-to-
obtain index properties.
Table 1 also shows a number of correlations based on the cone tip
resistance (qc), soil behavior index (IC), initial void ratio (e0), vertical
(effective) stress σv0(σv0′) and depth (D). The cone tip resistance (qc) is
found to be most closely related to mean normal stress. And qc and fs are
the specific representatives of CPTU. The soil behavior index (IC) can be
related to the mean grain size (D50), which is related to the size of its
constituent particles and is essentially the radius of concentric circles
that define the boundaries of the soil type (Jefferies and Davies, 1993).
The initial void ratio (e0) is generally used to define the compressibility
of a soil that accounts for the in situ state of the soil and also describes
the structure of the soil through its geologic history. Hence, an ideal
correlation between Vs and index properties will include parameters
that account for both the specific representatives of CPT and the soils'
initial state, which have large variability, i.e., the ideal correlation is a
non-linear multi-parameter correlation. This is especially true for un-
disturbed samples. Correlations that are based on multiple regressions
are more accurate. In order to obtain reliable multiple parameters,
however, the variables considered must be absolutely independent of
each other and the physical significance of the mathematical operation
must be taken into consideration (Li and White, 1993; Cherubini and
Giasi, 2000; Giasi et al., 2003). Hence, correlations such as those that
contain qt and qc, σv0andσv0′, are unreliable as those parameters are
related and not independent. Note that most correlations are applicable
to all soils and clay. Considering that silts often contain clays and sands
and hence some of the sand correlations might apply to silts and clayey
silts, hence, some correlations applicable to sands were selected.
3.2. Regression analysis and a new modified expression
After preliminary analysis, Correlation A for the “All Soils” equa-
tions, Correlation L for “Clay” equations and Correlation N and P for
“Sand” equations give better performance. In addition, the soil type-
specific method mentioned above uses an average of published corre-
lations for each soil type, which could increase the workload and errors
once irrelevant equations are selected. For this reason, the piecewise
function equations (selecting one correlation in each section for sim-
plification), are also selected. The additional Correlations Q and R are
the combination of Correlations A-L-P and Correlations A-L-N, sepa-
rately, with a selected method detailed above. And Correlation S is
given by the average result of Correlations Q and R. Furthermore,
Correlation T is proposed in this paper.
The correlation coefficient (R2) has always occupied an important
188
Engineering Geology 226 (2017) 184–191
position in applications of linear regression analysis to assess the
availability of an equation. The measured (Vsm) versus calculated (Vsc)
shear wave velocity was plotted from Correlations A through S in Fig. 3,
and the dashed line in each figure was the perfect line (Vsm = Vsc).
In this case, a new modified expression (Correlation T) using mul-
tiple regression analysis can be developed that relates Vs directly with
qt, D and IC as variables as follows:
Vs = 10.915qt 0.317 IC 0.210D 0.057 SF a (3)
3.3. Statistical analysis using five indexes
The reliability of the adopted correlations was assessed by five in-
dexes: root mean square error (RMSE) (Grima and Babuška, 1999), the
first and second moment (standard deviation) statistics of the ratio of
the estimated to test-determined shear wave velocity (K) (Briaud and
Tucker, 1988), ranking index (RI) (Briaud and Tucker, 1988), ranking
distance (RD) (Cherubini and Orr, 2000), and relative error index (RE).
The first (mean μ) and second moment (SD standard deviation σ)
statistics of the ratio of the estimated shear wave velocity to the mea-
sured shear wave velocity are denoted by K. It is determined by using
the following equation:
K = Vsc Vsm (4)
The ranking index (RI) is determined by:
RI = μln(K ) + σln(K ) (5)
where: μ and σ represent the mean and standard deviation of the series
of analyzed data, respectively. Ranking distance (RD) is another method
that takes into consideration the mean value and the standard deviation
of all the K data.
Relative error (RE) is the ratio of the absolute difference between
the measured value and the estimated compression to the measured
shear wave velocity. It is mainly used to assess the pros and cons of the
close correlation and the formula is given as follows
RE = Vsc − Vsm Vsm = K − 1 (6)
The lower the RMSE, K, RE, RI and RD values are, the better the
correlation is.
3.4. New statistical analysis using variable coefficient
A variable coefficient (CVK) can usually be applied to assess equa-
tions. It has hardly ever been denoted by K, and is expressed by
CVK = μ K SDK (7)
3.5. New graphical analysis using cumulative frequency
Actually, the cumulative frequency curve is always applied to the
particle-size distribution curve. Yet, it may be used for the assessment
of equations in this paper (see Fig. 4) since not only can it reveals the
variation range of an equation, but it also presents the variation ten-
dency of an equation.
4. The proposed statistical assessment method
The abovementioned statistical indexes were on the basis of isolated
points ignoring the location of these points in the figure of measured
versus calculated shear wave velocity (see Fig. 3), which may introduce
some errors, especially when compared with two close equations.
Hence, this paper proposed a new statistical assessment method taking
the location of the points into account. In the graph of measured (Vsm)
versus calculated (Vsc) shear wave velocity, the perfect line means the
M. Zhang, L. Tong Engineering Geology 226 (2017) 184–191

100 This study 1000

RMSE
Correlation S
RI
Correlation R
RD
Correlation Q
Cumulative frequency (%)

80 CVK
P-Hegazy & Mayne (1995) 100
M-Long & Donohue (2010)
L-Long & Donohue (2010)

RMSE/RI/RD
60 H-Mayne & Rix (1995)
E-Karray et al. (2011) 10
D-Fear & Robertson (1995)
40 B-Robertson (2009)
A-Andrus et al. (2007)
1

20

0.1
0 A B C D E F G H I J K L M N O P Q R S T
-50 0 50 100 150 Correlation
Scaled relative errors (%)
Fig. 6. Results of the RMSE, RI RD and CVK.
Fig. 4. Scaled relative errors of Vs predicted.

worst performance, while Correlation T, newly presented in this paper,

diagonal (Vsm = Vsc). On this occasion, the distance from one point to gave the best performance.
the perfect line with the analytic geometry formula (Ds) can be given In terms of RMSE, for the “All Soils” correlations, Correlation A gave
by. the best prediction with RMSE of 13.110. For the “Clay” correlations,
Vsmi − Vsci correlation L gave the best prediction with RMSE of 69.206. As to the
Dsi = “Sand” correlations, Correlation N unexpectedly gave the best predic-
1+1 (8)
tion with RMSE of 39.366. And for the sectional correlations,
where i means the ith equation. Correlation R gave the best prediction of Vs with RMSE of 38.724. The
Moreover, based on the graph of the distance from the perfect line, overall best performing correlation was correlation A
in this paper a new assessment index is presented to signify the slope of (RMSE = 13.110).
the specific line that links the point D10 and the point D90 shown in For the general overestimation (K > 1 or RE > 0) of Vs, five out of
Fig. 5: six “All Soils” correlations, all of the “Clay” correlations, one of the
D90i − D10i three “Sand” correlations, and none of the sectional correlations,
K2i = had > 50% of the K values > 1. These illustrated the overestimation of
90 − 10 (9)
most of the equations. The sectional combined Correlations Q through S
The reason why points D10 and D90 are used is that these two points were closer to 0.5 than the pre-combined because the combined gave
are not only closer to the top and bottom of the y-axis, but also in an more even values than the pre-combined.
almost straight line. Hence, these two points can present the trend of In terms of accuracy (closeness of the mean of K to 1 or RE to 0), for
the curve to a large extent. the “All Soils” correlations, Correlation A gave the best performance
with a K mean of 0.978. As to the “Clay” correlations, Correlation M
5. Results and discussion gave the most accurate evaluation of Vs with a mean K of 1.375. It was
unexpected that linear Correlation N gave the most accurate evaluation
A summary of the results of the RMSE, K, RE, RI and RD for the of Vs with a mean K of 1.002, which indicated that the mean value of K
Jiangsu clays is presented in Table 4. The results of the RMSE, RI, RD can barely serve as the assessment index of CPT-based empirical cor-
and CVK are illustrated in Fig. 6 in the log form. relations.
By comparing the difference between the scatter points and the The RI curve seems like an exaggeration of the RMSE curve
dashed line shown in Fig. 3, we can obtain that Correlation O gave the (Onyejekwe et al., 2015); however, in this study, the RMSE curve in-
versely seems like an exaggeration of the RI curve where the former
100 This study generally follows the profile of the latter but plots above it from Fig. 6.
D90 Correlation S
RD, which gives equal weight to accuracy and precision, is a better
Correlation R
Correlation Q
parameter for comparing the suitability of different correlation equa-
80
Cumulative frequency (%)

P-Hegazy & Mayne (1995) tions (Onyejekwe et al., 2015). In terms of RD, the overall best corre-
M-Long & Donohue (2010) lation for Jiangsu soft clays was the one proposed in this paper
L-Long & Donohue (2010) (RI = 0.158, RD = 0.127), followed by Correlation A (RI = 0.187,
60 H-Mayne & Rix (1995)
K2 RD = 0.155). This indicated that the new non-linear multi-parameter
E-Karray et al. (2011)
D-Fear & Robertson (1995) “All Soils” correlation gave better reliability.
40 B-Robertson (2009) As the common allowable limit of relative error (ALE) was 5%, the
A-Andrus et al. (2007) percentage of relative error less than the ALE (PRELA) is shown gra-
phically in Fig. 7. The higher PRELA was, the better the correlation
20 performance is. For higher ALE (5%), Correlation A (PRELA = 59.38%)
D10
gave the best performance, and the proposed correlation gave the next
best performance with the PRELA of 54.17%. Generally speaking, in
0
terms of PRELA, the most efficient correlations were Correlations A and
0 50 100 150
the proposed correlation.
The perfect distance
As to the variable coefficient CVK, the overall best correlation was
Fig. 5. Distance from the perfect line. the one proposed in this paper (12.84), followed by Correlation A

189
M. Zhang, L. Tong Engineering Geology 226 (2017) 184–191

60 deviation of Ds is denoted by SD2. Obviously, as seen in Table 4 and

55
ALE=1%
Fig. 4, the proposed correlation gives the best estimation (K2 = 3.94,
50 ALE=5% SD2 = 12.20), and Correlation A gives the next best estimation
(K2 = 2.97, SD2 = 13.48), while Correlations F and O perform second
45
worst and worst respectively. The results agree with the conclusions of
40
the above-mentioned statistical and graphical analysis.
35 This proposed correlation and Correlation A give better perfor-
PRELA (%)

30 mance than other correlations for the Jiangsu soft clays. One can con-
25
firm the sensitivity parameters through the analysis, where qt provided
a better coefficient of correlation R value of 0.909 than IC
20
(R = − 0.719) and D (R = 0.433) with Vs. The reason why the sleeve
15 friction (fs) and natural void ratio (e0) are excluded is that these para-
10 meters may be difficult to estimate as they are highly susceptible to
5 sampling disturbance.
Evaluation of some of these published relationships revealed that
0
A B C D E F G H I J K L M N O P Q R S T most did not match the local data in the present study well. These re-
Correlation sults show that the proposed relationships give a better estimation than
those from previous existing correlations. However, all of the correla-
Fig. 7. Results of the PRELA. tion equations obtained in this study are close to most of the other
previously published results. The reason for the differences may be due
(15.67). to the specific geotechnical conditions of the study area, geological age,
To gain an insight into the capabilities of the proposed correlations, the differences observed or water table fluctuations affecting correla-
the scaled relative error (Maheswari et al., 2010), RE2, in per cent, i.e. tions considerably. The results support most of the conclusions of ear-
RE = | RE2 | is given by. lier works in the literature. These practical CPT-based Vs correlations
should be used with caution in geotechnical engineering and should be
RE2 = 100(Vsc − Vsm) Vsm (10)
checked against measured Vs.
Twelve of the best equations were revealed in Fig. 4. As seen in
Fig. 4, using Correlations A, B, D and E, for all soils, about 90%, 80%, 6. Conclusions
4% and 1% of the Vs values were predicted within a 20% error margin.
Using Correlations H, L and M, only about 10%, 20% and 40% of the Vs Twenty common empirical correlations between the shear wave
values were predicted within 20% error for clay soil. Using Correlation velocity and CPT-based parameters were evaluated in this study. The
P, only 1% of the Vs values were predicted within 20% error for sand study area located in Jiangsu in the eastern part of China. Statistical and
soil. graphical analysis was executed in terms of six existing and two new
The perfect distance (the distance from one point to the perfect line, indexes and the cumulative frequency curve graph, allowing the fol-
Ds) could describe the relative position relations more distinctly from lowing conclusions to be obtained:
the point of view of the graphics, which can be applied to compare
different corrections, especially two approximate ones. And the cumu- i) The old, linear and single-parameter correlation gave place to the
lative frequency curve of Ds for twelve equations is indicated in Fig. 5. non-linear multi-parameter correlation.
The results show that Correlation D, E and H give the worst estimation, ii) The proposed statistical and graphical assessment method is more
while the proposed one performs best, which is more obvious than in accurate than the previous methods.
Fig. 4 and Table 4. Yet, this graphic method shows a little complexity, iii) The overall best correlation was the proposed expression,
which should be indexed, and it is K2 (see Table 4). The standard (Vs = 10.915qt0.317IC0.210D0.057SFa), which yielded an R2 value of

Table 4
Results of the correlations.

# RMSE K RE RI RD R2 CVK SD2 K2

> 1(%) | μ| σ > 0(%) | μ| σ

A 13.110 0.260 0.978 0.153 0.260 − 0.022 0.115 0.187 0.155 0.779 15.67 13.48 2.97
B 16.646 0.542 1.079 0.320 0.542 0.079 0.271 0.294 0.329 0.500 29.67 23.18 2.09
C 46.971 0.893 1.275 0.286 0.893 0.275 0.247 0.437 0.397 0.421 22.45 17.02 1.88
D 113.850 0.977 1.687 0.379 0.977 0.687 0.366 0.718 0.785 0.421 22.45 26.22 1.17
E 141.862 0.985 1.862 0.418 0.985 0.862 0.410 0.816 0.958 0.421 22.45 29.95 1.04
F 95.622 0.756 1.792 1.676 0.756 0.792 1.618 0.951 1.854 0.055 93.54 169.51 0.37
G 119.860 0.969 1.568 0.332 0.969 0.568 0.317 0.641 0.658 0.671 21.16 40.71 0.75
H 110.564 0.962 1.586 0.510 0.962 0.586 0.498 0.700 0.776 0.537 32.14 43.69 0.73
I 186.658 0.832 1.622 0.700 0.832 0.622 0.657 0.641 0.936 0.542 43.14 87.05 0.38
J 386.444 0.702 1.973 1.400 0.702 0.973 1.277 1.134 1.705 0.545 70.99 227.79 0.17
K 277.588 0.969 2.043 0.724 0.969 1.043 0.713 0.988 1.270 0.732 35.41 138.55 0.24
L 69.206 0.908 1.385 0.276 0.908 0.385 0.240 0.514 0.474 0.581 19.97 21.11 1.48
M 94.914 0.824 1.375 0.460 0.824 0.375 0.430 0.568 0.594 0.584 33.48 48.04 0.74
N 39.366 0.344 1.002 0.300 0.344 0.002 0.213 0.287 0.300 0.545 29.95 16.55 2.07
O 694.917 1.000 5.041 1.391 1.000 4.041 1.391 1.842 4.273 0.571 27.59 184.75 0.17
P 45.865 0.168 0.859 0.225 0.168 − 0.14 0.143 0.410 0.266 0.560 26.18 18.39 1.92
Q 59.702 0.618 1.218 0.354 0.618 0.218 0.258 0.453 0.416 0.133 29.05 23.98 1.25
R 38.724 0.328 1.004 0.296 0.328 0.004 0.224 0.410 0.296 0.455 29.45 18.64 1.85
S 40.283 0.626 1.111 0.268 0.626 0.111 0.188 0.320 0.290 0.363 24.09 17.06 1.96
T 22.925 0.458 0.986 0.127 0.458 − 0.014 0.086 0.158 0.127 0.798 12.84 12.20 3.94

190
M. Zhang, L. Tong
0.798, an RD value of 0.127 and a K2 value of 3.94.
iv) This method has been selected thanks to its greater accuracy and
truthfulness compared with other possible approaches.
Acknowledgement
Majority of the work presented in this paper was supported by the
National Natural Science Foundation of China (Grant No. 41572273)
and the “Twelfth Five-Year” National Science and Technology Support
Plan (Grant No.2012BAJ01B02). Dr. Guo-jun Cai is also appreciated
deeply for his sharing data and suggestion. The authors would like to
express their appreciations to the editor and anonymous reviewer for
their valuable comments and suggestions.
References
Ameratunga, J., Sivakugan, N., Das, B.M., 2016. Correlations of Soil and Rock Properties
in Geotechnical Engineering. Springerpp. 8–9.
Andrus, R.D., Stokoe II, K.H., 2000. Liquefaction resistance of soils from shear-wave
velocity. J. Geotech. Geoenviron. 126 (11), 1015–1025.
Andrus, R.D., Mohanan, N.P., Piratheepan, P., et al., 2007. Predicting shear-wave velocity
from cone penetration resistance. In: Proceedings, 4th International Conference on
Earthquake Geotechnical Engineering. Springer Netherlands, Thessaloniki, Greece.
Baldi, G., Bellotti, R., Ghionna, V.N., et al., 1989. Modulus of sands from CPTs and DMTs.
In: Proceedings of the 12th International Conference on Soil Mechanics and
Foundation Engineering. Rio de Janeiro. Vol 1. Balkema Pub., Rotterdam, pp.
165–170.
Borcherdt, R.D., 1994. Estimates of site dependent response spectra for design metho-
dology and justifications. Earthquake Spectra 10 (4), 617–654.
Briaud, J., Tucker, L.M., 1988. Measured and predicted axial response of 98 piles. J.
Geotech. Eng. 114 (9), 984–1001.
Cai, G.J., Liu, S.Y., Tong, L.Y., 2010. Field evaluation of deformation characteristics of a
lacustrine clay deposit using seismic piezocone tests. Eng. Geol. 116 (3–4), 251–260.
Cai, G.J., Liu, S.Y., Puppala, A.J., Tong, L.Y., 2011. Assessment of the coefficient of lateral
earth pressure at rest (Ko) from in situ seismic tests. Geotech. Test. J. 34 (4),
310–320.
Cai, G., Puppala, A.J., Liu, S., 2014. Characterization on the correlation between shear
wave velocity and piezocone tip resistance of Jiangsu clays. Eng. Geol. 171, 96–103.
Campanella, R.G., Robertson, P.K., Gillespie, D., 1986. Seismic cone penetration test. In:
Proceeding of In-Situ'86. GSP 6. ASCE, Reston, VA, pp. 116–130.
Cherubini, C., Orr, T.L.L., 2000. A Rational Procedure for Comparing Measured and
Calculated Values in Geotechnics. 1. Coastal Geotechnical Engineering in Practice,
Yokohama, pp. 261–265.
Cherubini, C., Giasi, C.I., 2000. Correlation equations for normal consolidated clays. In:
Coastal Geotechnical Engineering in Practice.
Dikmen, Ü., 2009. Statistical correlations of shear wave velocity and penetration re-
sistance for soils. J. Geophys. Eng. 6 (1), 61–72.
DNV, 2007. Recommended Practice: Statistical Representation of Soil Data (DNV-RP-
C207). Det Norske Veritas, Hovik.
Dobry, R., Borcherdt, R.D., Crouse, C.B., et al., 2000. New site coefficient and site clas-
sification system used in recent building code provisions. Earthquake Spectra 16 (1),
41–67.
Fear, C.E., Robertson, P.K., 1995. Estimating the undrained strength of sand: a theoretical
framework. Can. Geotech. J. 32 (5), 859–870.
Giasi, C.I., Cherubini, C., Paccapelo, F., 2003. Evaluation of compression index of re-
molded clays by means of Atterberg limits. Bull. Eng. Geol. Environ. 62 (4), 333–340.
Grima, M. Alvarez, Babuška, R., 1999. Fuzzy model for the prediction of unconfined
compressive strength of rock samples. Int. J. Rock Mech. Min. Sci. 36 (3), 339–349.
Hasancebi, N., Ulusay, R., 2007. Empirical correlations between shear wave velocity and
penetration resistance for ground shaking assessments. Bull. Eng. Geol. Environ. 66
(2), 203–213.
Hegazy, Y.A., Mayne, P.W., 1995. Statistical correlations between VS and cone penetra-
tion data for different soil types. In: Proceedings International Symposium on Cone
Penetration Testing, CPT'95. Vol 2. Spon Press, Linkoping, Sweden, pp. 173–178.
Hegazy, Y.A., Mayne, P.W., 2006. A global statistical correlation between shear wave
velocity and cone penetration data. In: Site and Geomaterial Characterization. ASCE,
pp. 243–248.
Jaime, A., Romo, M.P., 1988. The Mexico earthquake of September 19, 1985-correlations
between dynamic and static properties of Mexico City clay. Earthquake Spectra 4 (4),
787–804.
Jefferies, M.G., Davies, M.P., 1993. Use of CPTU to estimate equivalent SPT N60.
Geotech. Test. J. 16 (4), 458–468 ASTM.
Jongmans, D., Campillo, M., Demanet, D., 1990. The use of surface waves inversion and
seismic reflection methods for engineering applications. In: Balkema, A.A. (Ed.),
191
Engineering Geology 226 (2017) 184–191
Proceedings of the 6th Congress of International Association for Engineering Geology
and the Environment (IAEG), Amsterdam, pp. 979–985.
Karray, M., Lefebvre, G., Ethier, Y., et al., 2011. Influence of particle size on the corre-
lation between shear wave velocity and cone tip resistance. Can. Geotech. J. 48 (4),
599–615.
Kayen, R., Moss, R.E.S., Thompson, E.M., Seed, R.B., Cetin, K.O., Kiureghian, A.D.,
Tokimatsu, K., 2013. Shear-wave velocity–based probabilistic and deterministic as-
sessment of seismic soil liquefaction potential. J. Geotech. Geoenviron. 139 (3),
407–419.
Kitsunezaki, C., 1980. A new method for shear-wave logging. Geophysics 45 (10),
1489–1506.
Kulhawy, F.H., Mayne, P.W., 1990. Manual on Estimating Soil Properties for Foundation
Design. Epri.palo Alto.report.el.
Lacasse, S., Nadim, F., 1996. Uncertainties in characterizing soil properties. In:
Shackleford, C.D., Nelson, P.P., Roth, M.J.S. (Eds.), Uncertainty in the Geological
Environment: From Theory to Practice, Geotechnical Special Publication No. 58.
ASCE, New York, pp. 49–75.
Li, K.S., White, W., 1993. Use and misuse of regression analysis and curve fitting in
geotechnical engineering. In: Probabilistic Methods in Geotechnical Engineering, pp.
145–152.
Liu, S.Y., Cai, G.J., Tong, L.Y., Du, G.Y., 2008. Approach on the engineering properties of
Lianyungang marine clay from piezocone penetration tests. Mar. Georesour.
Geotechnol. 26 (3), 189–210.
Liu, S.Y., Shao, G.H., Du, Y.J., Cai, G.J., 2011. Depositional and geotechnical properties of
marine clays in Lianyungang, China. Eng. Geol. 121 (1–2), 66–74.
Lodge, A.L., 1994. Shear Wave Velocity Measurements for Subsurface Characterization.
University of California, Berkeley.
Long, M., Donohue, S., 2010. Characterization of Norwegian marine clays with combined
shear wave velocity and piezocone cone penetration test (CPTU) data. Can. Geotech.
J. 47 (7), 709–718.
Lunne, T., Robertson, P.K., Powell, J.J.M., 1997. Cone Penetration Testing in
Geotechnical Practice. Spon Press, Chapman & Hall, London.
Maheswari, R.U., Boominathan, A., Dodagoudar, G.R., 2010. Use of surface waves in
statistical correlations of shear wave velocity and penetration resistance of Chennai
soils. Geotech. Geol. Eng. 28 (2), 119–137.
Mayne, P.W., 2006. Undisturbed sand strength from seismic cone tests. The 2nd James K.
Mitchell Lecture. Geomech. Geoeng. 1 (4), 239–247.
Mayne, P.W., Rix, G.J., 1995. Correlations between cone tip resistance and shear wave
velocity in natural clay. Soils Found. 35 (2), 107–110.
Mokhtar, T.A., Hermann, R.B., Russell, D.R., 1988. Seismic velocity and Q model for the
shallow structure of the Arabian shield from short-period Rayleigh waves. Geophysics
53 (11), 1379–1387.
Moss, R.E.S., 2008. Quantifying measurement uncertainty of thirty-meter shear-wave
velocity. Bull. Seismol. Soc. Am. 98 (3), 1399–1411.
Ohta, Y., Goto, N., 1978. Empirical shear wave velocity equations in terms of char-
acteristic soil indexes. Earthq. Eng. Struct. Dyn. 6 (2), 167–187.
Onyejekwe, S., Kang, X., Ge, L., 2015. Assessment of empirical equations for the com-
pression index of fine-grained soils in Missouri. Bull. Eng. Geol. Environ. 74 (3),
705–716.
Piratheepan, P., 2002. Estimating Shear-Wave Velocity from SPT and CPT Data. Clemson
University (Master of Science Thesis).
Pitilakis, K., Raptakis, D., Lontzetidis, K., et al., 1999. Geotechnical and geophysical de-
scription of EuroSeistests, using field and laboratory tests, and moderate strong
ground motions. J. Earthq. Eng. 3 (3), 381–409.
Robertson, P.K., 2009. Interpretation of cone penetration tests-a unified approach. Can.
Geotech. J. 46 (11), 1337–1355.
Robertson, P.K., Wride, C.E., 1998. Evaluating cyclic liquefaction potential using the cone
penetration test. Can. Geotech. J. 35 (3), 442–459.
Robertson, P.K., Woeller, D.J., Kokan, M., et al. (1992). Seismic techniques to evaluate
liquefaction potential. In Proceedings of the 45th Canadian Geotechnical Conference,
Toronto, Ont, 26–28, pp. 51–59.
Stokoe, Kenneth H., et al., 1988. In situ seismic testing of hard-to-sample soils by surface
wave method. In: Earthquake Engineering and Soil Dynamics II—Recent Advances in
Ground-motion Evaluation. ASCE.
Sykora, D.W., 1987. Examination of existing shear wave velocity and shear modulus
correlations in soils. In: Department of the Army, Waterways Experiment Station,
Corps of Engineers. Miscellaneous Paper GL-87-22.
Sykora, D.E., Stokoe, K.H.I.I., 1983. Correlations of in-situ measurements in sands of
shear wave velocity. Soil Dyn. Earthq. Eng. 20.1–4, 125–136.
Tong, L., Liu, L., Cai, G., Du, G., 2013. Assessing the coefficient of the earth pressure at
rest from shear wave velocity and electrical resistivity measurements. Eng. Geol. 163,
122–131.
Wair, B.R., DeJong, J.T., Shantz, T., 2012. Guidelines for Estimation of Shear Wave
Velocity Profiles, PEER Report 2012/08. University of California, Pacific Earthquake
Research Center, Berkeley.
Whitman, R.V., 1996. Organizing and evaluating uncertainty in geotechnical engineering.
In: Shackleford, C.D., Nelson, P.P., Roth, M.J.S. (Eds.), Uncertainty in the Geological
Environment: From Theory to Practice, Geotechnical Special Publication No. 58.
ASCE, New York, pp. 1–28.