Ocean Engineering 111 (2016) 116–127

Contents lists available at ScienceDirect

Ocean Engineering
journal homepage: www.elsevier.com/locate/oceaneng

Wave-induced oscillatory response in a randomly heterogeneous

porous seabed
L.L. Zhang a, Y. Cheng a,b, J.H. Li c, X.L. Zhou a,n, D.S. Jeng a,d, X.Y. Peng a
a
State Key Laboratory of Ocean Engineering, Collaborative Innovation Center for Advanced Ship and Deep-Sea Exploration, Department of Civil Engineering,
Shanghai Jiaotong University, Shanghai, China
b
Shanghai Underground Space Architectural Design & Research Institute Co., Ltd, Shanghai, China
c
Department of Civil and Environmental Engineering, Harbin Institute of Technology Shenzhen Graduate School, Shenzhen, China
d
Griffith School of Engineering, Griffith University Gold Coast Campus, Queensland 4222, Australia

art ic l e i nf o a b s t r a c t

Article history: The seabed response under wave loading is important for the stability of foundations of offshore
Received 7 May 2015 structures. Unlike previous studies, the wave-induced seabed response in a spatially random porous
Accepted 9 October 2015 seabed is investigated in this study. A stochastic finite element model which integrates random field
simulation of spatially varied soil properties and finite element modeling of wave-induced seabed
Keywords: response is established. Spatial variability of soil shear modulus, soil permeability, and degree of
Porous seabed saturation, are simulated using the covariance matrix decomposition method. The results indicate that
Wave the pore water pressure and stress distribution in the seabed are significantly affected by the spatial
Spatial variability variability of the shear modulus. The mean of maximum oscillating pore pressure in a randomly het-
Random field
erogeneous seabed is greater than that in a homogenous seabed. The uncertainty of the maximum
Stochastic
oscillating pore pressure first increases and then reduces with the correlation length. The effects of
Monte Carlo simulation
spatial variability of permeability and degree of saturation are less significant than that of shear modulus.
The spatial variability of soil permeability increases the uncertainty of oscillating pore pressure in the
seabed at shallow depths. The spatial variability of shear modulus and degree of saturation affects the
uncertainty of oscillating pore pressure of the whole seabed.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction permeability, shear modulus, and degree of saturation (Jeng,

Evaluation of the wave-induced soil response is vital for the
design of foundations around marine structures. When a wave
propagates over the ocean, it generates dynamic fluctuations of
excess pore water pressure and effective stress in the seabed. This
wave-induced soil response may become significant enough to
cause instability of the seabed in front of marine structures and
may induce liquefaction. Three types of theories have been used to
analyze the wave-induced seabed response with different
assumptions about the compressibility of the pore fluid and the
soil skeleton (Jeng, 2013): (i) Laplace's equation (Sleath, 1970); (ii)
Terzaghi's theory of consolidation (Moshagen and Torum, 1975)
and (iii) Biot's consolidation theory for poroelastic media (Madsen,
1978; Yamamoto et al., 1978; Mei and Foda, 1981; Okusa, 1985;
Gatmiri, 1990; Thomas, 1989, 1995; Hsu et al., 1993; Jeng and Hsu,
1996; among others). The important seabed soil properties that
influence the wave-induced response have been identified as soil
n
Corresponding author.
E-mail address: zhouxl@sjtu.edu.cn (X.L. Zhou).
2013).
Generally, the physical properties of soils vary in marine sedi-
ments due to complicated natural geological processes. Res-
earchers have attempted to consider heterogeneous soil char-
acteristics (Jeng and Seymour, 1997; Jeng and Lin, 1996), aniso-
tropic soil behavior (Hsu and Jeng, 1994; Jeng, 1997; Kitano and
Mase, 1999), and multi-layered seabed (Hsu et al., 1995; Gao et al.,
2003; Zhou et al., 2011). In their studies, soil properties of a seabed
are usually considered as a specific form (e.g., linear or exponen-
tial) of function of depth because marine sediments may undergo
consolidation due to both the overburden soil pressure and the
water pressure, which may result in a general decrease in soil
permeability and an increase of soil rigidity with depth.
Inherent spatial variability exists in a supposedly homogeneous
soil layer due to small-scale variations in mineral composition,
environmental conditions during deposition, past stress history,
variations in moisture content, and so on (Ang and Tang, 2007;
Dasaka and Zhang, 2012; Lloret-Cabot et al., 2014). The inherent
spatial variability cannot feasibly be characterized by deterministic
approaches. Some researchers considered the spatial variability
of soil properties using probabilistic approaches for various

http://dx.doi.org/10.1016/j.oceaneng.2015.10.016
0029-8018/& 2015 Elsevier Ltd. All rights reserved.
 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 117

geotechnical engineering problems, such as consolidation and z

Wave length L
settlement (Fenton and Griffiths, 2002; Badaoui et al., 2007; Wave crest
Huang et al., 2010; Le et al., 2013; Manjari et al., 2014), water Amplitude
flow and contaminant transport (Tartakovsky et al., 2003; Li et al. Wave
Free surface
2009, 2012; Mousavi Nezhad et al., 2013; Kanning and Calle, 2013), height Hs
slope stability (Cho, 2007; Srivastava et al., 2009; Santoso et al.,
d Wave trough
2011; Zhu et al., 2013; Li et al. 2014, 2015a; Jiang et al. 2014a,b;
Zhang et al., 2008; Zhu et al., 2015; Jamshidi Chenari and Alaie,
Sea surface O
2015), foundations (Fan et al. 2014; Fan and Liang 2015; Li et al.,
A B x
2015b; Zhang and Chen, 2012; Zhang et al., 2014; Bari and Shahin, h Porous seabed
2015), and so on. However, most of the previous studies of seabed
or marine soils focused generally on the characterization of field
A ' Rigid, impermeable bottom B'
spatial variability of marine deposits (Keaveny et al., 1990; Valdez-
Llamas et al., 2003; Goff et al., 2008; Cheon, 2011; Gilbert et al., Fig. 1. Sketch of wave propagation above a porous seabed.
2014). The effects of random waves on seabed have been con-
sidered by many researchers (Suh et al. 1997; Liu and Jeng, 2007; where Kx and Kz are the hydraulic permeability of soil in the x and
Myrhaug et al. 1998 among others). Suh et al. (1997) used the z directions, respectively; p is the wave-induced oscillating pore
Green's second identity and Lagrangian formula to develop two pressure; γw is the unit weight of the pore water; n is the soil
ultimately equivalent hyperbolic equations for random waves. Liu porosity; β is the compressibility of the pore-fluid; t is time; ε is
and Jeng (2007) developed a semi-analytical solution for the the volume strain for the two-dimensional problem, which is
random wave induced soil response. The influence of random defined by
wave loading on the soil response was investigated by comparing
∂u ∂w
with the two different wave spectra. Myrhaug et al. (1998) ε¼ þ ð2Þ
∂x ∂z
investigated the seabed shear stresses under random waves and
compared the model predictions with field measurements. The where u and w are the soil displacements in the x and z directions,
sediment transport and scour of seabed by random waves were respectively. The compressibility of the pore fluid β can be related
investigated assuming the random waves to be a stationary to the bulk modulus of the pore-water and the degree of satura-
Gaussian narrow-band random process (Myrhaug and Ong 2010; tion Sr (Verruijt, 1969),
Myrhaug et al. 2014). Xu and Dong (2011) conducted a numerical 1 1
Sr
β¼ þ ð3Þ
study to investigate the effect of random waves on excess pore Kw P w0
pressure build-up and liquefaction processes. However, the effects where Kw is the true bulk modulus of water (which can be taken as
of randomly heterogeneous soil properties on wave-induced sea- 1.95  109 N/m2, Yamamoto et al., 1978), and Pw0 is the absolute
bed response have not been investigated. pore-water pressure.
In this paper, a stochastic finite element model which inte- The governing equations of porous seabed based on the force
grates random field modeling of spatially varied soil properties equilibrium of the soil skeleton and the effective stress concept
and finite element modeling of wave-induced seabed response is can be written as (Biot, 1941):
established to investigate the effect of a randomly heterogeneous 8 0
seabed on wave-induced response. The spatial variability of soil < ∂σ x þ ∂τzx ¼ ∂p
∂x ∂z ∂x
ð4Þ
: ∂τxz þ ∂σ z ¼ ∂p
0
properties, including shear modulus, soil permeability, and degree
∂x ∂z ∂z
of saturation, is simulated using the covariance matrix decom-
position method based on the random field theory. The simulated where σ 0x is the effective normal stress in the x direction, σ 0z is the
spatially random soil properties are imported into an interactive effective normal stress in the z direction, τxz is the shear stress in
finite element software environment, COMSOL, to solve the gov- the z direction on the plane perpendicular to the x axis, and τzx is
erning partial differential equations (PDEs) for wave-induced the shear stress in the x direction on the plane perpendicular to
seabed response. The effects of randomly heterogeneous soil the z axis.
properties on the uncertainties of the oscillating excess pore water Assume the soil skeleton is an elastic material, which follows
pressure and wave-induced incremental change in effective the generalized Hooke's law; then the equations of force equili-
stresses are discussed. brium within the soil matrix can be rewritten as:
8
< G∇2 u þ ð1
G2μÞ ∂∂xε ¼ ∂p
∂x
ð5Þ
2. Theory of wave-induced seabed response : G∇2 w þ ð1
G2μÞ ∂∂zε ¼ ∂p
∂z

2.1. Governing equations where G is the shear modulus, and μ is the Poisson's ratio. G is
related to Young's modulus E and the Poisson's ratio μ in the form
The problem considered in this paper is two-dimensional. of E/2(1 þ μ). In these formulations of the wave–seabed interaction
Assume a porous seabed of finite thickness h (Fig. 1). The x-axis is problem, the soil shear modulus G, soil permeability, and degree of
taken on the seabed surface and the z-axis is taken vertically saturation are expressed as constant parameters, but they will be
downward from the seabed surface. The waves travel from left to simulated as spatially random soil properties by random field
right along the positive direction of the x-axis. Based on Biot's simulation in this study.
consolidation theory (Biot, 1941), together with the storage
equation (Verruijt, 1969), the governing equation for the wave–soil 2.2. Boundary conditions
interaction problem within a compressible pore fluid in a com-
The governing equations, Eqs. (1) and (5), describing the
pressible porous seabed is given as:


 water–soil interaction problem, can be solved by incorporating the
∂ ∂p ∂ ∂p ∂p ∂ε boundary conditions specified at the seabed surface and the

Kx þ
Kz þ γ w nβ þ γ w ¼ 0 ð1Þ
∂x ∂x ∂z ∂z ∂t ∂t impermeable bottom, respectively. It is commonly accepted that
118 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127

the vertical effective normal stresses and shear stresses vanish at
the seabed surface, and the pore pressure is equal to the wave
pressure at the seabed surface (Jeng, 2013):
σ 0z ¼ τxz ¼ 0 at z ¼ 0
p ¼ p0 cos ðkx
ωtÞ at z¼0
where p0 is the amplitude of the dynamic wave pressure based on
the short-crested wave theory to the first order, which can be
expressed as
γw Hs
p0 ¼
2 cosh kd
in which Hs is the wave height; k is the wave number (k¼ 2π/L, in
which L is the wave length); ω is the angular frequency of the
wave (ω ¼ 2π/T, T is the wave period); and d is the water depth
above the seabed surface. In this study, the linear wave theory is
adopted to determine the wavelength with given wave period and
water depth based on the wave dispersion equation (Phillips,
1977). A matlab code is compiled to determine the wavelength
with given wave periods and water depth using the Newton's
numerical method.
For the soil resting on an impermeable rigid bottom below a
finite thickness, as depicted in Fig. 1, zero displacements and no
vertical flow occur:
∂p
u¼w¼0 and ¼0 at z ¼
h
∂z
When modeling an infinitely long seabed in a numerical model,
the length of the seabed in the numerical model can be assumed
to be equal to multiple wavelengths. Since the boundary condition
of the lateral boundaries is periodic in both time and space, it is
reasonable to assume that u, w, and p are the same for the two
lateral boundaries using the principle of repeatability.
3. Stochastic finite-element model of a heterogeneous seabed
3.1. Spatial variability of marine soils
Inherent spatial variability exists in seafloor sediments. Fig. 2
presents the field-measured cone tip resistances up to 25 m depth
at one artificial island, which was later used as a drilling platform
for oil exploration on the Canadian Beaufort sea shelf (Popescu et
al., 1997). The soil layer is a homogenous hydraulically placed sand.
These six in situ cone penetration tests were performed at 9 m
intervals in a straight line. As shown in the figure, the mean values
ð6Þ of soil properties (systematic trends) vary in space. There are
always some fluctuations about mean values even within the
ð7Þ homogenous layer. Some degrees of coherence between the fluc-
tuations can be observed, which become stronger as the measur-
ing points become closer. For even deeper depths of seabed, nat-
ural spatial variability of the soil properties is also observed. Fig. 3
shows the soil property profiles of the Pleistocene clay deposit in
Osaka Bay from the site investigation of Kansai International Air-
ð8Þ
port (Watabe et al., 2002). The Pleistocene clay deposit includes
several sandy soil layers between the clay layers. Profiles of the
liquid limit (wL), plastic limit (wP), natural water content (wn), void
ratio e, consolidation yield stress (σ 0y ), and over-consolidation ratio
(OCR) for the Pleistocene clay deposit from 25 to 230 m below the
seabed are presented on the graph. As shown on the graph, natural
spatial variability can be observed for different soil properties.
The spatial variation of soil properties can be described using
the geotechnical random field theory (Vanmarcke 1977; Fenton
and Griffiths, 2008). The soil property in the field is modeled as
the sum of a trend component and a residual term. The residuals
of the trend to exhibit spatial correlation, which can be expressed
through the autocovariance function and the autocorrelation
function. The autocovariance function and autocorrelation func-
ð9Þ tion are usually estimated from soil samples from a site (Lloret-
Cabot et al., 2014). Analytical models can be used to fit the sample
autocorrelation functions using regression analysis. A commonly
used autocorrelation function is the exponential model (Van-
marke, 1977; Fenton and Griffiths, 2008):


2τ
ρðτÞ ¼ exp
ð10Þ
δ
where τ is the separation distance between two points; δ is the
correlation length, which describes the distance within which the
spatially random values will tend to be significantly correlated (i.e.,
by more than about 10%). Within a range of δ, soil property values
are highly correlated, but in a range over δ, soil property values
can be considered as statistically independent.
Research studies about the characterization of spatial varia-
bility of the geotechnical properties of marine soils are relatively
limited. Hoeg and Tang (1977) and Tang (1979) reported that the
horizontal correlation length of CPT cone tip resistance in a marine

0
MAC31 MAC04 MAC05 MAC06 MAC07 MAC08

-5
Elevation (m)

-10

-15

-20

0 10 20 0 10 20 0 10 20 0 10 20 0 10 20 0 10 20
Cone tip resistance (MPa)
Fig. 2. In situ cone tip resistances of a homogenous sand layer in the Canadian Beaufort Sea (modified from Popescu et al. 1997). The grey lines represent systematic trends.
 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 119

Fig. 3. Profiles of physical soil properties of Pleistocene clay in Osaka Bay (modified from Watabe et al., 2002).

Table 1
Correlation length of geotechnical soil properties for marine soils.

Soil property Correlation length δ Soil type Source

Cone tip resistance, qc δh ¼14–38 m Offshore soils Keaveny et al. (1990)

Undrained shear strength, Su
Soil layer thickness
Natural water content
Natural water content
Cone tip resistance, qc
Cone tip resistance, qc
Cone tip resistance, qc
δv ¼ 0.3–0.6 m Offshore soils Keaveny et al. (1990)
δh ¼ 22.2 m Superficial soft clay Valdez-Llamas et al. (2003)
δv ¼ 0.8–2.0 m Superficial soft clay Valdez-Llamas et al. (2003)
δv ¼ 21 m, δh ¼ 1000 m Deep deposits with alternating clayey and sandy soils Valdez-Llamas et al. (2003)
δh ¼35–60 m (different levels) Marine clay Hoeg and Tang (1977); Tang (1979)
δh ¼55 m (0–3 m below sea bottom)
δv ¼ 0.40–0.44 m Filled sand in artificial island Lloret-Cabot et al. (2014)
δh ¼1.69–15.86 m

clay was 55 m for the shallow soil layer within 3 m depth of the Table 2
seabed. For greater depths of soils, the correlation length in the List of deterministic input parameters.
horizontal direction varied from 35 to 60 m. Keaveny et al. (1990)
Parameter Unit Value
investigated the spatial variability of offshore soils and found that
the horizontal correlation length for the CPT cone tip resistance Seabed thickness h m 15
varies from 14 to 38 m. Valdez-Llamas et al. (2003) estimated the Model length l m 38.076
spatial variability of the marine soils in the Sonda de Campeche Water depth d m 5
Wave height Hs m 4
area of the Gulf of Mexico. The thickness of the soft cohesive Wave length L m 38.076
superficial soil layer and the soil water content of marine deposits Wave period T s 6
up to 100 m depth were investigated based on data from 89 Density of water ρw kg/m3 1000
boreholes. The correlation length for the thickness of the super- Density of seabed ρs kg/m3 2000
Modulus of volume Kw N/m2 2  109
ficial soft soil layer is estimated to be 22 m. The correlation length
Porosity n – 0.3
of the natural water content for the superficial soft soil layer is Poisson ratio ν – 0.333
0.8–2.0 m along the vertical direction. For deeper marine deposits,
the vertical correlation length of the natural water content is 21 m.
Table 1 summarizes the correlation lengths of geotechnical soil (Kleiber and Hien, 1992), each response quantity is represented
properties for marine soils reported in the literature. As shown in using a series of random Hermite polynomials. Due to its robust-
the table, the correlation length along the vertical direction varies ness and simplicity, Monte Carlo simulation (Papadrakakis and
from tens of centimeters to a few meters. There is no significant Papadopoulos, 1996; Fenton and Griffiths, 2008) is often used in
difference in spatial variability in the vertical direction between the literature as a reference method in order to check the accuracy
land and shallow marine sediments (Phoon and Kulhawy, 1999). of other approaches and is sometimes combined with other SFEM
However, for deeper marine soils, the vertical correlation length is methods. In the MCS, a deterministic problem is solved a number
usually larger than it is for land soils. The horizontal correlation of times and the response variability is calculated using simple
length of basic soil properties, for example water content and relationships of statistics. In this study, the deterministic model is
density, can be a few kilometers in the seafloor (Valdez-Llamas et a complex dynamic problem and hence the MCS method is
al., 2003; Goff et al., 2008; Verfaillie et al., 2009). adopted for simplicity and a feasible computation load.
Different algorithms are available for discretization of random
3.2. The stochastic finite element model field. The most popular methods include the spectral methods
based on Fourier transformation (Cooley and Tukey, 1965; Mejia
There are two main variants of Stochastic Finite Element and Rodriguez-Iturbe, 1974; Dietrich and Newsam, 1996), the
method (SFEM), i.e. the perturbation approach and the spectral moving average method (Matérn, 1986, Yaglom, 1987, Oliver, 1995;
approach (Stefanou, 2009). The perturbation approach (Ghanem Cressie and Pavlicová, 2002; Chiles and Delfiner, 1999), the local
and Spanos, 1991) is based on a Taylor series expansion of the average subdivision method (Fenton and Vanmarcke, 1990; Fenton
response vector. In the spectral stochastic finite element method and Griffths, 2008), the covariance matrix decomposition method
120 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127

Table 3
List of stochastic model parameters.

Parameter Distribution Autocorrelation function Mean COV (%) Correlation length (m)

Shear modulus G LN Exponential 1.0  107 Pa 250 δlnG ¼ 0.2, 2, 10

Permeability Kz LN Exponential 5  10
4 m/s 250 δln(kz) ¼0.2, 2
Degree of saturation Sr LN Exponential 0.975 50 δln(Sr) ¼ 0.2, 2

Note: the soil layer is hydraulically isotropic, i.e., Kx ¼ Kz.

0 0.0
Present study
-0.1 Jeng (2013) T = 1.5 s
This study -0.2
d = 0.488 m
-0.2 L = 2.8 m
H = 0.056 m
-0.3 -0.4 h = 0.33 m
ν = 0.333

z/h
-0.4 n=1
|σz '|/p0 |p|/p0 -0.6 G = 5.0 10 Pa
z/h

-0.5 S = 0.985 (dense)

0.965 (loose)
-0.6 -0.8 K = 5.0 10 m/s (dense)
T = 15 s G = 1.0 107 Pa Dense sand
5.0 10 m/s (loose)
-0.7 d = 70 m Kz = 10-4 m/s Loose sand
L = 311.59 m Sr = 1 -1.0
-0.8 0.0 0.2 0.4 0.6 0.8 1.0
Hs = 3 m n = 0.3
-0.9 |p|/p0
h = 25 m ν = 0.333
Fig. 5. Vertical distributions of the maximum pore pressure |p|/p0 compared with
-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 experimental data of Tsui and Helfrich (1983).

Fig. 4. Vertical distributions of the maximum pore pressure |p|/p0 and effective
stresses |σz0 |/p0 versus z/h for a homogeneous porous seabed.
(Davis, 1987; Deutsch and Journel, 1997; Fenton and Griffths,
2008; Jamshidi Chenari and Kamyab Farahbakhsh, 2015), the
turning bands method (Matheron, 1973; Mantoglou and Wilson,
1982), the sequential simulation methods (Bellin and Rubin, 1996;
Deutsch and Journel, 1997).
In this study, the covariance matrix decomposition method
(Davis, 1987; Deutsch and Journel, 1997; Fenton and Griffths,
2008) is adopted to generate realizations of a random field. The
basic idea of the covariance matrix decomposition method is to
assume that the parameters at different locations in the field are
correlated random variables with a prescribed covariance matrix
and to produce a homogeneous random field which is the product
of the vector of independent standard normal random variables
and the decomposed low triangular matrix of the covariance
matrix.
The following is the brief description of the covariance matrix
decomposition method. Assume that the prescribed covariance
matrix for these correlated random variables is C. If C is positive
definite, then a correlated standard normal random field Z can be
produced using independent standard normal random variables
according to the following equation:
Z ¼ LU
where L is a lower triangular matrix satisfying LL ¼ C; U is a vectorT
of n-independent standard normal random variables; Z is a vector
of n-correlated standard normal random variables. The lower tri-
angular matrix L is typically obtained using the Cholesky decom-
position method. If a soil property X of the random field is nor-
mally distributed with a known mean and variance, then a reali-
zation for the soil property can be obtained using the realization of
the correlated standard normal variate Z as follows:
X ¼ μI þ σ Z
where I is the unit matrix; μ and σ are the mean value and
standard variation of the normally distributed X. Non-normal
random fields can be obtained through a suitable transformation
of a normally distributed random field. For example, a log-
normally distributed random field can be obtained from:
X ¼ expðμlnX I þ σ lnX ZÞ ð13Þ
where μInX and σ ln X are the mean and standard variation of ln X,
respectively.
The advantage of the covariance matrix decomposition method
is that it is applicable for any type of covariance model and is not
restricted to the geometric configuration of the simulated loca-
tions. Compared with the turning bands method, the spectral fast
Fourier method, and the sequential Gaussian simulation method,
the covariance matrix decomposition method shows the least
artificial bias (Harter, 1994). For random fields with many reali-
zations, this is indeed a very effective means of random field
generation, since the covariance matrix must be decomposed only
once for an entire simulation.
The stochastic finite element model for a porous seabed with
the spatially random soil property is established by integration of
the finite element model and the random field simulation. First,
the coupled governing equations [Eqs. (1) and (5)] are incorpo-
rated in a multiphysics (PDEsolver based on finite element
method, COMSOL (COMSOL, 2012). The Galerkin method is used to
discretize the PDEs. A nonlinear differential algebraic equation
solver, IDA (Hindmarsh et al., 2005), is used as a time integrator for
the differential algebraic system, in which the backward differ-
ð11Þ entiation formulas (BDFs) are used to discretize the time
derivative terms.
The thickness of the seabed is 15 m and the length of the two-
dimensional finite element model is set to be equal to the wave-
length of 38.076 m. The soil within the seabed is assumed to be
fine sand with the soil parameters presented in Table 2. The
densities of the seabed and water, the porosity, and the Poisson's
ratio are assumed to be constant. The shear modulus G, the per-
meability Kz, and the degree of saturation Sr are assumed to be
spatially random parameters, as listed in Table 3. The mean value
ð12Þ
of G is 1  107 Pa. The mean of permeability in the z direction is
assumed to be 5  10
4 m/s. The seabed is assumed to be
hydraulically isotropic. The seabed is unsaturated with a mean
 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 121

Realization no.2 Realization no.252 Realization no.109 Realization no.20

0 6 0 6 0 6 0 6

-5 4 -5 4 -5 4 -5 4
z (m)

z (m)

z (m)

z (m)
-10 2 -10 2 -10 2 -10 2

-15 0 -15 0 -15 0 -15 0

0 10 20 30 0 10 20 30 0 10 20 30 0 10 20 30
L (m) L (m) L (m) L (m)

0 0 0 0

-5 -5 -5 -5
z (m)

z (m)

z (m)

z (m)
-10 -10 -10 -10
δlnG = 2 δlnG = 2 δlnG = 2 δlnG = 2
deterministic case deterministic case deterministic case deterministic case
-15 -15 -15 -15
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
|p|/ p 0 |p|/ p 0 |p|/ p 0 |p|/ p 0

0 0 0 0
δlnG = 2 δlnG = 2 δlnG = 2
-5 deterministic case -5 deterministic case -5 -5 deterministic case
z (m)

z (m)

z (m)

z (m)
-10 -10 -10 -10
δlnG = 2
deterministic case
-15 -15 -15 -15
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
σz , / p 0 σz ,/p 0 σz ,/p 0 σz ,/p 0

Fig. 6. Effect of spatially random G (COVG ¼ 250%, δlnG ¼2 m) on seabed response: (a) realization of G (  107 Pa), (b) |p|/p0 and (c) |σz0 |/p0. The dashed lines represent a
homogenous seabed with the mean values in Table 3 and are denoted as the deterministic case.

degree of saturation of 0.975. As the experimental data for per- of degree of saturation is also rarely reported in the literature.
meability and shear modulus are usually found to follow a log- However, the variation of natural water content is often reported
normal distribution (Phoon and Kulhawy, 1999; Baecher and (Phoon and Kulhawy, 1999), with the COV varying from 5% to 50%.
Christian, 2003; Carsel and Parrish, 1988), all three parameters are The reported vertical correlation length of natural water content
assumed to have a lognormal distribution in this study. The log- varies from 1 to 10 m in the field. For ground formed by natural
normal distribution is chosen because it is suitable for strictly non- decomposition, the horizontal correlation length is one order lar-
negative random variables, and enables the large variability of soil ger than the vertical one (Phoon and Kulhawy, 1999; Popescu et
properties to be analyzed. The autocorrelation function of the al., 1997). Based on the previously reported spatial variability of
spatially random soil parameters is assumed to be an exponential in situ soils, the COV values of shear modulus and saturated per-
function. The other empirical autocorrelation functions which are meability are assumed to be 250% and the COV of degree of
also commonly adopted in literature include Gaussian, Spherical saturation is assumed to be 50%. The correlation lengths of the
function. When field data of soil properties are available, different normalized parameters are 0.2, 2, and 10 m in order to compre-
empirical autocorrelation functions can be fitted and a most sui-
hensively investigate the possible scenarios of seabed response
table function can be chosen based on the fitness. In this study, as
due to the effect of spatial variability (Table 3). In the following
no field data available and the illustrative example is hypothetical,
sections, we denote the COVs of the three parameters as COVG,
the exponential autocorrelation function is assumed.
COVKz, and COVSr and the correlation lengths as δlnG, δln(Kz), and δln
Since the effect of waves on the seabed is important for soil
(Sr), respectively. Along the horizontal direction, the correlation
layers near the seabed surface and there is no significant differ-
lengths for all three parameters are assumed to be 100 m as the
ence in spatial variability between land and shallow marine soils
horizontal correlation length is at least one order of magnitude
(Table 1), the variability of ground soils on land in the literature
can be used as reference values. The coefficient of variation (COV) larger than the vertical one (Phoon and Kulhawy, 1999; Popescu et
of permeability for in situ soils ranges from 60% to more than 200% al., 1997).
(Carsel and Parrish, 1988; Duncan, 2000; Zhang et al., 2005; Bae- A finite element mesh of 2400 identical four-node rectangular
cher and Christian, 2003). The correlation length of log(perme- elements, each with four integration points and an area of
ability) in the vertical direction ranges from 0.2 to around 10 m 0.25 m  0.95 m, is employed in numerical modeling. A random
(Rehfeldt et al., 1992; Rackwitz et al., 2002). The COV of shear field of identical rectangular cells is generated and superimposed
modulus has rarely been reported in the literature. However, there on the finite element mesh. Each FE element is then assigned the
are quite a lot studies about the variation of soil stiffness or value of the random field cell using interpolation. In this study,
resistance measured by cone penetration tests, standard penetra- 500 random realizations are generated for each soil property. The
tion tests, pressuremeter, and so on (Phoon and Kulhawy, 1999; finite element model is then solved using each realization of the
Rackwitz et al., 2002). The reported COV values of soil resistance random field, and a population of the random response quantities
are usually from 20% to 100%. The vertical correlation length of soil is obtained. This population can then be used to obtain statistics of
resistance or stiffness varies from 0.1 to 10 m in the field. The COV the seabed responses.
122 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127

0 4 0
-0.6 0.6 0.8 -0.-60.8
-0.
4
3 -0.4 0.4
-5 -5
0.4
z (m)

z (m)
2
-0.2 0 -0.2

0
-10 -10 0.2
1
-0.2 0.2

2
-0.
-15 0 -15
0 10 20 30 0 5 10 15 20 25 30 35
L (m) L (m)

0 4 0
-00..64
0.8 -0-0
.6.8
- .2 0.40.6 -0.4
-0 0.2 -0.2
3
-5 -5
0 0

z (m)
z (m)

2 -0.2 0.2 -0.2

-10 -10
1

-15 0 -15
0 10 20 30 0 5 10 15 20 25 30 35
L (m) L (m)

0 4 0
-0.6 0.60.8 -0-. 0.6

-0
4

.2
3 .4
-5 -5 -0
0.4
z (m)
z (m)

0.2

0.2
2
0

0
2
-0.

-10 -10
1

-15 0 -15
0 10 20 30 0 5 10 15 20 25 30 35
L (m)
L (m)
Fig. 8. Contours of p/p0 of random seabed (COVG ¼ 250%) (a) δlnG ¼ 0.2 m (b)
Fig. 7. Typical realizations of shear modulus G(  107 Pa) of random seabed
δlnG ¼2 m (c) δlnG ¼10 m.
(COVG ¼250%) (a) δlnG ¼ 0.2 m (b) δlnG ¼ 2 m (c) δlnG ¼ 10 m.

maximum |p|/p0 and effective stresses |σz0 |/p0 with respect to

4. Results and discussion
4.1. Verification of the present model for homogenous seabeds
To verify the finite element model established in COMSOL, the
wave-induced response in a homogenous seabed is first examined
by comparing the present study with the numerical results from
Jeng (2013) for a fine sand seabed as shown in Fig. 4. The model
parameters of the porous seabed are as follows: wave height
Hs ¼3 m, wave period T¼ 15 s, water depth d ¼70 m, wave length
L¼ 311.59 m, seabed thickness h¼25 m, density of water
ρw ¼ 1000 kg/m3, density of porous seabed ρs ¼ 1650 kg/m3, Pois-
son's ratio ν ¼0.333, porosity n¼ 1, degree of saturation Sr ¼1,
shear modulus G ¼1.0  107 Pa, permeability Kx ¼Kz ¼ 10-4 m/s, and
modulus of volume Kw ¼ 2  109 N/m2. Fig. 4 presents the profile of
depth. As shown on the graph, the results of the present numerical
model agree well with the numerical solutions in Jeng (2013).
A comparison with experimental data is presented in Fig. 5. The
experimental data are from Tsui and Helfrich (1983) for dense and
loose sands. The solid lines in the figure represent the simulated
results obtained by the present study. The present study is in good
agreement with the experimental data for both loose and dense
sands, with appropriate values assigned for the soil parameters as
shown on the graph.
4.2. Typical realizations of random heterogeneous seabed and
response
To investigate the effect of non-homogenous soil properties on
wave-induced seabed response using deterministic approaches,
 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127 123

Fig. 6 also illustrates that the random field simulation can

0 generate various scenarios to represent complicated situations in
the field. On the graphs, the dashed lines represent the simulated
5% results for a homogenous seabed with the mean value of shear
modulus in Table 3 and are denoted as the deterministic case. The
-5 pore water pressure |p|/p0 has much greater oscillations along the
mean
z (m)

95% depth of the seabed than a homogenous seabed. For some extreme
cases (e.g., realization no. 109), high pore pressure close to 0.8p0
can be observed within 5 m depth of seabed.
-10
4.3. Effect of spatial variability of soil properties
Deterministic case
Five hundred random field realizations of stochastic soil para-
-15 meters are generated using the covariance matrix decomposition
0 0.2 0.4 0.6 0.8 1
method. Fig. 7 illustrates three typical realizations of the shear
|p|/p0 modulus G with a correlation length of ln(G), δlnG, varying between
0.2, 2, and 10 m. As shown on the graph, the heterogeneity of G
0 becomes more significant with smaller δlnG. Fig. 8 presents the
corresponding pore water pressure profiles for the three realiza-
5% tions of a spatially random seabed. When δlnG is only 0.2 m, the
soft layers and the stiff layers are intersliced and hence the wave
-5 mean cannot penetrate smoothly down to the seabed bottom. Therefore,
the high value zones of pore pressure are scattered. As δlnG
z (m)

95% increases, the random seabed is more homogenous and the pore
water pressure distribution is closer to the response of a uniform
-10 seabed.
Fig. 9 presents the mean and 90% uncertainty bounds of max-
Deterministic case imum |p|/p0 based on the 500 realizations with spatially varied
δlnG. Compared with the pore pressure response in a deterministic
-15 homogenous seabed, the mean |p|/p0 for a spatially random seabed
0 0.2 0.4 0.6 0.8 1 is higher. This increase in the mean |p|/p0 is more pronounced for
|p|/p0 small correlation lengths and decreases with increasing correla-
tion length. The dependence of the expected value on spatial
variability is described as "counter-intuitive" because the common
0 perception is that the performance will have a constant mean if
5% the soil property has a fixed mean value. However, this variation of
the expected performance with spatially varied soil properties
mean given a fixed mean has been observed in many other geotechnical
-5 problems. Fenton and Griffiths (1996) found that the mean flow
95% rate through an earth dam with spatially random permeability
z (m)

decreases from the deterministic value with increases in the var-

-10
Deterministic case
-15
0 0.2 0.4 0.6 0.8
|p|/p0
Fig. 9. Mean and uncertainty bounds of maximum |p|/p0 distribution (COVG ¼ 250%)
(a) δlnG ¼ 0.2 m (b) δlnG ¼ 2 m (c) δlnG ¼ 10 m.
only a limited number of scenarios can be explored. However, the
response of a natural seabed is much more complicated and a
deterministic approach may lead to significant risk.
Fig. 6(a) shows four typical realizations of a spatially random
shear modulus in the 15 m thick seabed. The correlation length δlnG
is 2 m. The four realizations represent conditions with large shear
modulus layers at the top, middle, bottom, and both top and bottom
parts of the porous seabed, respectively. As shown in Fig. 6(b) and
(c), the maximum amplitude of pore pressure |p|/p0 increases and
the vertical effective normal stresses |σz0 |/p0 decrease in the zone
with a relatively small G. This observation agrees with the results in
the literature (Jeng and Hsu, 1996; Jeng, 2013).
iance of permeability. Le et al. (2012) found that if flow in an
unsaturated zone of the dam is considered, the mean flow rate
increases with increasing standard deviation of saturated perme-
ability. Le et al. (2012) explained this difference by a simple ana-
logy. The case of restricted flow through the saturated domain
could be assimilated into seepage perpendicular to a layered soil
1 system, whose overall permeability is dominated by the layer with
the smallest permeability. However, if flow in the unsaturated
region is considered, this spread flow through both saturated and
unsaturated domains could be assimilated into seepage parallel to
a layered soil system, whose overall permeability is dominated by
the layer with the highest permeability. Therefore, the expected
flow rate is larger for a higher variance of permeability. In this
study, the degree of saturation of the seabed does not affect the
permeability of the seabed but only the compressibility of pore
fluid. In addition, the whole domain of a seabed is not separated
into unsaturated and saturated zones. Hence, the propagation of
wave-induced pore pressure in the spatially random seabed is
similar to the case of water flow in a saturated dam. The only
difference is that the output variable is outflow for the problem of
flow through a dam, while for the wave-induced seabed response
problem, we focus on the wave-induced excess pore pressure. As
the direction of propagation is vertical, the low permeability zone
has a greater influence on the transition of pore pressure. The
interslicing of soft and stiff layers obstructs the propagation of
124 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127

0 0.98 0 -0.8
0.6
--0.4.2 0.4 0.60.8 -0-.40.6
-0 -0.2
0.97 0.2
-5 -5
0.96 0 0
z /m

z /m
0.95
-10 -10
0.94

-15 0.93 -15

0 10 20 30 0 5 10 15 20 25 30 35
L /m L /m

0 0.98 0
-0.4 0.6 0.8 -0 -0.6
-0.2 0.20.4 -0.2 .4
0.97 0 0
-5 0.2

-0.2
-5

-0
.2
0.96
z /m

z /m
0.95
-10 -10
0.94

-15 0.93 -15

0 10 20 30 0 5 10 15 20 25 30 35
L /m L /m
Fig. 10. Typical realizations of degree of saturation Sr of spatially random porous seabed and the corresponding |p|/p0 (COVSr ¼ 50%) (a) δln(Sr) ¼ 0.2 m (b) δln(Sr) ¼ 2 m.

0 0
5%
5% mean 95%
95%
-5 -5
z (m)
z /m

mean

-10 -10

Deterministic case
Deterministic case
-15 -15
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
|p|/p0 |p|/ p0

0 0
5%
5%
95% 95%
mean
-5
-5
z /m

z (m)

mean

-10
-10
Deterministic case
Deterministic case
-15
0 0.2 0.4 0.6 0.8 1
-15
|p|/p0 0 0.2 0.4 0.6 0.8 1
|p|/ p0
Fig. 11. Mean and confidence intervals of maximum |p|/p0 distribution
(COVSr ¼50%) (a) δln(Sr) ¼ 0.2 m (b) δln(Sr) ¼ 2 m.
Fig. 12. Mean and confidence intervals of maximum |p|/p0 distribution
(COVkz ¼250%) (a) δln(Kz) ¼ 0.2 m (b) δln(Kz) ¼2 m.
oscillating excess pore water pressure, as shown in Fig. 8(a). With
a smaller correlation length, it is more difficult for the wave- reduced. Therefore, the mean value of |p|/p0 decreases with the
induced excess pore pressure to propagate to a greater depth. As increase of correlation length.
the correlation length increases, this interference or obstruction is
 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
As shown in Fig. 9, the uncertainty of pore water pressure first
increases and then decreases with the increase of correlation
length. It is generally believed that if the correlation length is less
than the domain size or scale of interest, the variance of the output
is reduced with the decrease of correlation length due to the space
averaging effect (Fenton and Griffiths, 2008). However, this is not
necessarily true for a flow-related problem. In order to obtain a
statistically meaningful realization of the hydraulic field, the
domain size of flow should be at least 20–25 correlation lengths to
reach asymptotic behavior (Ababou et al., 1989). Elfeki et al. (2012)
illustrated that for transport in aquifers with correlation lengths,
which are relatively large compared with the travel distance (i.e.,
the distance traveled is less than 20 correlation lengths), the
ergodic condition is not fulfilled and hence no trend is obtained in
the results of spatial moments when the correlation length
increases. In this study, the propagation of wave-induced pore
pressure is confined within the 15 m thick seabed. Therefore, the
ergodic condition is not fulfilled for the cases with a correlation
length larger than 0.75 m.
Assume that the scale of fluctuations for ln(Sr) along the ver-
tical direction, δln(Sr), is 0.2 and 2 m. Fig. 10 illustrates the reali-
zations of Sr of a spatially random porous seabed and the corre-
sponding oscillating pore pressures. Compared with Fig. 8, the
oscillating pore pressure is less affected by the spatially random Sr.
Fig. 11 illustrates that the mean of maximum oscillating pore
pressure versus seabed depth is slightly greater than the deter-
ministic value. The uncertainty bounds of pore pressure for the
case with a correlation length of 2 m are wider than those of δln
(Sr) ¼ 0.2 m.
Fig. 12 illustrates the mean of maximum oscillating pore pres-
sure versus seabed depth for seabed with spatially varied per-
meability. For fine sand seabed, the wave-induced pore pressure
attenuates rapidly within the top 3 m of the seabed thickness and
finally decreases slightly with depth. The variance of the pore
pressure is also concentrated in shallow depths. As shown in
previous studies (Jeng and Seymour, 1997; Ulker et al., 2009; Jeng,
2013), the pore pressure is generally greater when the coefficient
of permeability is larger because the wave pressure can be easier
to transmit to deeper seabed. However, in a random hetero-
geneous seabed, the zones with different permeability are ran-
domly intersliced and hence the pore pressure near the surface of
the seabed cannot be smoothly transmitted to deeper depth of a
seabed due to the blockage of a possible low permeability zone.
Therefore, the uncertainty of pore pressure is mainly concentrated
in near the surface of a seabed as shown in Fig. 12.
Comparing Figs. 9, 11, and 12, the effect of spatial variability of
shear modulus is most significant. The spatial variability of shear
modulus and degree of saturation affects the uncertainty of oscil-
lating pore pressure of the whole seabed. The effect of variable per-
meability is least significant and mostly in shallow seabed depths.
5. Conclusions
In this study, a stochastic finite element model is developed to
investigate the wave-induced seabed response in a spatially ran-
dom porous seabed and the effects of spatial variability of soil
properties on seabed responses are investigated by Monte Carlo
simulation. The major conclusions are as follows:
 The pore water pressure and stress distribution in a spatially
random heterogeneous seabed are significantly affected by
spatial variability of the shear modulus. The effects of spatial
variability of permeability and degree of saturation are less
significant than the effect of spatial variability of shear modulus.
125
 The mean of oscillating pore pressure in a heterogeneous sea-
bed with spatially random shear modulus is greater than that in
a homogenous seabed. The increase of mean pore pressure is
more pronounced for small correlation lengths and decreases
with increasing correlation length. The uncertainty of the
maximum pore pressure first increases and then decreases with
the correlation length.
 The spatial variability of shear modulus and degree of saturation
affects the uncertainty of oscillating pore pressure of the whole
seabed while the random heterogeneity of permeability influ-
ences the seabed response at shallow depths.
Acknowledgments
The work in this paper was substantially supported by the
National Basic Research Program of China (973 Program, Project
no. 2014CB049100), the National Natural Science Foundation of
China (Project nos. 41172252, 41572243 and 51422905) and State
Key Laboratory of Ocean Engineering (GKZD010065). The authors
are grateful for the supports from the Top Grade Young
Researchers (Ba Jian) Talent Program by the Organization
Department of the Central Committee of the CPC.
References
Ababou, R., McLaughlin, D., Gelhar, L.W., Tompson, A.F.B., 1989. Numerical simu-
lation of three-dimensional saturated flow. Transp. Porous Media 4 (6),
549–565.
Ang, A.H.S., Tang, W.H., 2007. Probability Concepts in Engineering: Emphasis on
Applications to Civil and Environmental Engineering, second ed. Wiley, New
York.
Badaoui, M., Nour, A., Slimani, A., Berrah, M.K., 2007. Consolidation statistics
investigation via thin layer method analysis. Transp. Porous Media 67 (1),
69–91.
Baecher G.B. and Christian J.T., Reliability and Statistics in Geotechnical Engineer-
ing, 2003, John Wiley and Sons, London and New York, pp 605.
Bari, M.W., Shahin, M.A., 2015. Three-dimensional finite element analysis of spa-
tially variable PVD improved ground. Georisk: Assess. Manag. Risk Eng. Syst.
Geohazards 9 (1), 37–48.
Bellin, A., Rubin, Y., 1996. HYDRO GEN: a spatially distributed random field gen-
erator for correlated properties. Stoch. Hydrol. Hydraul. 10 (4), 253–278.
Biot, M.A., 1941. General theory of three-dimensional consolidation. J. Appl. Phys.
12 (2), 155–164.
Carsel, R.F., Parrish, R.S., 1988. Developing joint probability distributions of soil
water retention characteristics. Water Resour. Res. 24 (5), 755–769.
Cheon, J.Y., 2011. Analysis of spatial variability in geotechnical data for offshore
foundations (Ph.D. dissertation). University of Texas at Austin, Austin, Texas.
Chiles, J.P., Delfiner, P., 1999. Geostatistics: Modeling Spatial Uncertainty. Wiley,
New York.
Cho, S.E., 2007. Effects of spatial variability of soil properties on slope stability. Eng.
Geol. 92 (3–4), 97–109.
Cooley, J.W., Tukey, J.W., 1965. An algorithm for the machine calculation of complex
Fourier series. Math. Comput. 19 (90), 297–301.
COMSOL, AB, 2012. COMSOL Multiphysics User Guide (Version 4.3 a). COMSOL, AB
Stockholm, Sweden.
Cressie, N., Pavlicová, M., 2002. Calibrated spatial moving average simulations. Stat.
Model. 2, 267–279.
Dasaka, S.M., Zhang, L.M., 2012. Spatial variability of in-situ weathered soils. Geo-
technique 62 (5), 375–384.
Davis, M.W., 1987. Production of conditional simulations via the LU triangular
decomposition of the covariance matrix. Math. Geol. 19 (2), 91–98.
Deutsch, C.V., Journel, A.G., 1997. GSLIB: Geostatistical Software Library and User’s
Guide (Applied Geostatistics. Oxford University Press, Oxford, New York, p. 384.
Dietrich, C.R., Newsam, G.N., 1996. A fast and exact method for multidimensional
Gaussian stochastic simulations: extension to realizations conditioned on direct
and indirect measurements. Water Resour. Res. 32 (6), 1643–1652.
Duncan, J.M., 2000. Factors of safety and reliability in geotechnical engineering. J.
Geotech. Geoenviron. Eng. 126 (4), 307–316.
Elfeki, A.M., Uffink, G., Lebreton, S., 2012. Influence of temporal fluctuations and
spatial heterogeneity on pollution transport in porous media. Hydrogeol. J. 20
(2), 283–297.
Fan, H., Huang, Q., Liang, R., 2014. Reliability analysis of piles in spatially varying
soils considering multiple failure modes. Comput. Geotech. 57, 97–104.
Fan, H., Liang, R., 2015. Importance sampling based algorithm for efficient reliability
analysis of axially loaded piles. Comput. Geotech. 65, 278–284.
126 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
Fenton, G.A., Griffiths, D.V., 2008. Risk Assessment in Geotechnical Engineering.
Wiley, New York.
Fenton, G.A., Griffiths, D.V., 2002. Probabilistic foundation settlement on a spatially
random soil. J. Geotech. Geoenviron. Eng. 128 (5), 381–390.
Fenton, G.A., Griffiths, D.V., 1996. Statistics of free surface flow through stochastic
earth dam. J. Geotech. Eng 122 (6), 427–436.
Fenton, G.A., Vanmarcke, E.H., 1990. Simulation of random fields via local average
subdivision. J. Eng. Mech 116 (8), 1733–1949.
Gao, F.P., Jeng, D.S., Sekiguchi, H., 2003. Numerical study on the interaction between
non-linear wave, buried pipeline and non–homogeneous porous seabed.
Comput. Geotech. 30 (6), 535–547.
Gatmiri, B., 1990. A simplified finite element analysis of wave-induced effective
stresses and pore pressures in permeable seabed. Geotechnique 40 (1), 15–30.
Ghanem, R.G., Spanos, P.D., 1991. Stochastic Finite Element-A Spectral Approach.
Springer, New York.
Gilbert, R.B., Lacasse, S., Nadim, F., 2014. Advances in geotechnical risk and relia-
bility for offshore applications. In: Proceedings of the Fourth International
Symposium Geotechnical Safety and Risk. Geotechnical Safety and Risk IV.
Balkema, The Netherlands.
Goff, J.A., Jenkins, C.J., Williams, S.J., 2008. Seabed mapping and characterization of
sediment variability using the usSEABED data base. Cont. Shelf. Res. 28 (4),
614–633.
Harter, T., 1994. Unconditional and Conditional Simulation of Flow and Transport in
Heterogeneous, Variably Saturated Porous Media (PhD dissertation). University
of Arizona, Tucson, Arizona.
Hindmarsh, A.C., Brown, P.N., Grant, K.E., Lee, S.L., Serban, R., Shumaker, D.E.,
Woodward, C.S., 2005. SUNDIALS: suite of nonlinear and differential/algebraic
equation solvers. ACM Trans. Math. Softw. 31 (3), 363–396.
Hoeg, K, Tang, W.H., 1977. Probabilistic considerations in the foundation engi-
neering for offshore structures. In: Proceedings of the 2nd International Con-
ference on Structural Safety and Reliability. Munich, Germany, pp. 267–296.
Hsu, J.R.C., Jeng, D.S., 1994. Wave-induced soil response in an unsaturated aniso-
tropic seabed of finite thickness. Int. J. Numer. Anal. Methods Geomech. 18 (11),
785–807.
Hsu, J.R.C., Jeng, D.S., Lee, C.P., 1995. Oscillatory soil response and liquefaction in an
unsaturated layered seabed. Int. J. Numer. Anal. Methods Geomech. 19 (12),
825–849.
Hsu, J.R.C., Jeng, D.S., Tsai, C.P., 1993. Short crested wave induced soil response in a
porous seabed of infinite thickness. Int. J. Numer. Anal. Methods Geomech. 17
(4), 553–576.
Huang, J.S., Griffiths, D.V., Fenton, G.A., 2010. Probabilistic analysis of coupled soil
consolidation. J. Geotech. Geoenviron. Eng. 136 (3), 417–430.
Jamshidi Chenari, R., Alaie, R., 2015. Effects of anisotropy in correlation structure on
the stability of an undrained clay slope. Georisk: Assess. Manag. Risk Eng. Syst.
Geohazards 9 (2), 109–123.
Jiang, S.H., Li, D.Q., Cao, Z.J., Zhou, C.B., Phoon, K.K., 2014a. Efficient system relia-
bility analysis of slope stability in spatially variable soils using Monte Carlo
simulation. J. Geotech. Geoenviron. Eng. 141 (2), 04014096.
Jiang, S.H., Li, D.Q., Zhang, L.M., Zhou, C.B., 2014b. Slope reliability analysis con-
sidering spatially variable shear strength parameters using a non-intrusive
stochastic finite element method. Eng. Geol. 168, 120–128.
Jeng, D.S., 1997. Soil response in cross-anisotropic seabed due to standing waves. J.
Geotech. Geoenviron. Eng. 123 (1), 9–19.
Jeng, D.S., 2013. Porous Models for Wave-seabed Interactions. Shanghai Jiao Tong
University Press and Springer-Verlag, Berlin, Heidelberg.
Jeng, D.S., Hsu, J.R.C., 1996. Wave-induced soil response in a nearly saturated seabed
of finite thickness. Geotechnique 46 (3), 427–440.
Jeng, D.S., Lin, Y.S., 1996. Finite element modelling for water waves–seabed inter-
action. Soil Dyn. Earthq. Eng 15 (5), 283–300.
Jeng, D.S., Seymour, B.R., 1997. Response in seabed of finite depth with variable
permeability. J. Geotech. Geoenviron. Eng. 123 (10), 902–911.
Kanning, W., EOF, Calle, 2013. Derivation of a representative piping resistance
parameter based on random field modelling of erosion paths. Georisk: Assess.
Manag. Risk Eng. Syst. Geohazards 7 (2), 99–109.
Keaveny, J.M., Nadim, F., Lacasse, S., 1990. Autocorrelation function data for offshore
geotechnical data. In: Proceedings of the International Conference on Structural
Safety and Reliability. ICOSSAR 89’, San Francisco, California, pp. 263–270.
Kitano, T., Mase, H., 1999. Boundary layer theory for anisotropic seabed response to
sea waves. J. Waterw. Port Coast. Ocean Eng. 125 (4), 187–194.
Kleiber, M., Hien, T.D., 1992. The stochastic finite element method: basic pertur-
bation technique and computer implementation. Wiley, Chichester, U.K.
Le, T.M.H., Gallipoli, D., Sanchez, M., Wheeler, S.J., 2012. Stochastic analysis of
unsaturated seepage through randomly heterogeneous earth embankments.
Int. J. Numer. Anal. Methods Geomech 36 (8), 1056–1076.
Le, T.M.H., Gallipoli, D., Sanchez, M., Wheeler, S., 2013. Rainfall-induced differential
settlements of foundations on heterogeneous unsaturated soils. Geotechnique
63 (15), 1346–1355.
Li, D.Q., Jiang, S.H., Cao, Z.J., Zhou, W., Zhou, C.B., Zhang, L.M., 2015a. A multiple
response-surface method for slope reliability analysis considering spatial
variability of soil properties. Eng. Geol. 187, 60–72.
Li, D.Q., Qi, X.H., Phoon, K.K., Zhang, L.M., Zhou, C.B., 2014. Effect of spatially variable
shear strength parameters with linearly increasing mean trend on reliability of
infinite slopes. Struct. Saf. 49, 45–55.
Li, J.H., Tian, Y., Cassidy, M., 2015b. Failure mechanism and bearing capacity of
footings buried at various depths in spatially random soil. J. Geotech. Geoen-
viron. Eng. 141 (2), 401–409.
Li, W., Lu, Z., Zhang, D., 2009. Stochastic analysis of unsaturated flow with prob-
abilistic collocation method. Water Resour. Res. 45, 8.
Liu, H.J., Jeng, D.S., 2007. A semi-analytical solution for random wave-induced soil
response and seabed liquefaction in marine sediments. Ocean Eng. 34 (8–9),
1211–1224.
Lloret-Cabot, M., Fenton, G.A., Hicks, M.A., 2014. On the estimation of scale of
fluctuation in geostatistics. Georisk: Assess. Manag. Risk Eng. Syst. Geohazards
8 (2), 129–140.
Madsen, O.S., 1978. Wave-induced pore pressure and effective stresses in a porous
bed. Geotechnique 28 (4), 377–393.
Matérn, B., 1986. Spatial Variation, second ed. Springer-Verlag, Berlin.
Matheron, G., 1973. The intrinsic random functions and their applications. Adv.
Appl. Probab. 5 (3), 439–468.
Manjari, K.G., Rao, K.B., Babu, G.S., 2014. Stochastic model for settlement: footings
on cohesionless soil. Georisk: Assess. Manag. Risk Eng. Syst. Geohazards 8 (4),
269–283.
Mantoglou, A., Wilson, J.L., 1982. The turning bands method for simulation of
random fields using line generation by a spectral method. Water Resour. Res. 18
(5), 1379–1394.
Mejia, J., Rodriguez-Iturbe, I., 1974. On the synthesis of random field sampling from
the spectrum: an application to the generation of hydrologic spatial processes.
Water Resour. Res. 10 (4), 705–711.
Mei, C.C., Foda, M.A., 1981. Wave-induced response in a fluid-filled poro-elastic
solid with a free surface - a boundary layer theory. Geophys. J. Int. 66 (3),
597–631.
Moshagen, H., Torum, A., 1975. Wave-induced pressure in permeable seabeds. J.
Waterw. Harb. Coast. Eng. 101 (1), 49–57.
Mousavi Nezhad, M., Javadi, A.A., Al-Tabbaa, A., Abbasi, F., 2013. Numerical study of
soil heterogeneity effects on contaminant transport in unsaturated soil (model
development and validation). Int. J. Numer. Anal. Methods Geomech. 37 (3),
278–298.
Myrhaug, D., Holmedal, L.E., Ong, M.C., 2014. Seepage effects on bedload sediment
transport rate by random waves. Ocean Eng. 82, 123–127.
Myrhaug, D., Ong, M.C., 2010. Random wave-induced onshore scour characteristics
around submerged breakwaters using a stochastic method. Ocean Eng. 37 (13),
1233–1238.
Myrhaug, D., Slaattelid, O.H., Lambrakos, K.F., 1998. Seabed shear stresses under
random waves: predictions vs estimates from field measurements. Ocean Eng.
25 (10), 907–916.
Okusa, S., 1985. Wave-induced stress in unsaturated submarine sediments. Geo-
technique 35 (4), 517–532.
Oliver, D.S., 1995. Moving averages for Gaussian simulation in two and three
dimensions. Math. Geol. 27, 939–960.
Papadrakakis, M., Papadopoulos, V., 1996. Robust and efficient methods for sto-
chastic finite element analysis using Monte Carlo simulation. Comput. Method
Appl. Mech. Eng. 134 (3), 325–340.
Phillips, O.M., 1977. The Dynamics of the Upper Ocean. Cambridge University Press,
Cambridge.
Phoon, K.K., Kulhawy, F.H., 1999. Characterization of geotechnical variability. Can.
Geotech. J. 36 (4), 612–624.
Popescu, R., Prévost, J.H., Deodatis, G., 1997. Effects of spatial variability on soil
liquefaction: some design recommendations. Geotechnique 47 (5), 1019–1036.
Rackwitz, R., Denver, H., Calle, E., 2002. JCSS Probabilistic Model Code. Joint Com-
mittee on Structural Safety, Zurich, Switzerland 5th version.
Rehfeldt, K.R., Boggs, J.M., Gelhar, L.W., 1992. Field study of dispersion in a het-
erogeneous aquifer: 3. Geostatistical analysis of hydraulic conductivity. Water
Resour. Res. 28 (12), 3309–3324.
Santoso, A.M., Phoon, K.K., Quek, S.T., 2011. Effects of soil spatial variability on
rainfall-induced landslides. Comput. Struct. 89 (11–12), 893–900.
Sleath, J.F.A., 1970. Wave-induced pressures in beds of sand. J. Hydraul. Div. 96,
367–379.
Srivastava, A., Sivakumar, B.G.L., Haldar, S., 2009. Influence of spatial variability of
permeability property on steady state seepage flow and slope stability analysis.
Eng. Geol. 110 (3–4), 93–101.
Stefanou, G., 2009. The stochastic finite element method: past, present and future.
Comput. Method Appl. Mech. Eng. 198 (9), 1031–1051.
Suh, K.D., Lee, C., Park, W.S., 1997. Time-dependent equations for wave propagation
on rapidly varying topography. Coast. Eng. 32 (2–3), 91–117.
Tang, W.H., 1979. Probabilistic evaluation of penetration resistances. J. Geotech. Div
105, 1173–1191, GT10.
Tartakovsky, D.M., Lu, Z., Guadagnini, A., Tartakovsky, A.M., 2003. Unsaturated flow
heterogeneous soils with spatially distributed uncertain hydraulic parameters.
J. Hydrol. 275 (3–4), 182–193.
Thomas, S.D., 1989. A finite element model for the analysis of wave induced
stresses, displacements and pore pressures in anunsaturated seabed. I: theory.
Comput. Geotech. 8 (1), 1–38.
Thomas, S.D., 1995. A finite element model for the analysis of wave induced
stresses, displacements and pore pressures in an unsaturated seabed. II: model
verification. Comput. Geotech. 17, 107–132.
Tsui, Y., Helfrich, S.C., 1983. Wave-induced pore pressures in submerged sand layer.
J. Geotech. Eng. 109 (4), 603–618.
Ulker, M.B.C., Rahman, M.S., Jeng, D.S., 2009. Wave-induced response of seabed:
various formulations and their applicability. Appl. Ocean Res. 31 (1), 12–24.
Valdez-Llamas, Y.P., Auvinet, G., Nunez, J., 2003. Spatial variability of the marine soil
in the Gulf of Mexico. In: Proceedings of the Offshore Technology Conference,
Houston, Texas, OTC 15266.
 L.L. Zhang et al. / Ocean Engineering 111 (2016) 116–127
Vanmarcke, E.H., 1977. Probabilistic modeling of soil profiles. J. Geotech. Eng. Div,
103; , pp. 1227–1246.
Verfaillie, E., Du, Four, I., Van, Meirvenne, M., Van, Lancker, V., 2009. Geostatistical
modeling of sedimentological parameters using multi-scale terrain variables:
application along the Belgian Part of the North Sea. Int. J. Geogr. Inf. Sci. 23 (2),
135–150.
Verruijt, A., 1969. Elastic Storage of Aquifers, Flow through Porous Media. Academic
Press, New York, pp. 331–376.
Watabe, Y., Tsuchida, T., Adachi, K., 2002. Undrained shear strength of Pleistocene
clay in Osaka Bay. J. Geotech. Geoenviron. Eng. 128 (3), 216–226.
Xu, H., Dong, P., 2011. A probabilistic analysis of random wave-induced liquefaction.
Ocean Eng. 38 (7), 860–867.
Yaglom, A.M., 1987. Correlation Theory of Stationary and Related Random Func-
tions. Volume I: Basic Results. Springer, Berlin.
Yamamoto, T., Koning, H.L., Sellmeijer, H., 1978. On the response of a poro-elastic
bed to water waves. J. Fluid. Mech. 87 (01), 193–206.
127
Zhang, L., Chen, J.J., 2012. Effect of spatial correlation of standard penetration test
(SPT) data on bearing capacity of driven piles in sand. Can. Geotech. J. 49 (4),
394–402.
Zhang, J., Li, J.P., Zhang, L.M., Huang, H.W., 2014. Calibrating cross-site variability for
reliability-based design of pile foundations. Comput. Geotech. 62, 154–163.
Zhang, L.L., Zhang, L.M., Tang, W.H., 2005. Rainfall-induced slope failure considering
variability of soil properties. Geotechnique 55 (2), 183–188.
Zhang, L.L., Zhang, L.M., Tang, W.H., 2008. Similarity of soil variability in centrifuge
models. Can. Geotech. J. 45 (8), 1118–1129.
Zhou, X.L., Xu, B., Wang, J.H., 2011. An analytical solution for wave-induced seabed
response in a multi-layered poro-elastic seabed. Ocean Eng. 38 (1), 119–129.
Zhu, H., Griffiths, D.V., Fenton, G.A., Zhang, L.M., 2015. Undrained failure mechan-
isms of slopes in random soil. Eng. Geol. 191, 31–35.
Zhu, H., Zhang, L.M., Zhang, L.L., Zhou, C.B., 2013. Two-dimensional probabilistic
infiltration analysis with a spatially varying permeability function. Comput.
Geotech. 48, 249–259.