A model to predict the water retention curve from

basic geotechnical properties

M. Aubertin, M. Mbonimpa, B. Bussire, and R.P. Chapuis
Abstract: The water retention curve (WRC) has become a key material function to define the unsaturated behavior
of soils and other particulate media. In many instances, it can be useful to have an estimate of the WRC early in a
project, when little or no test results are available. Predictive models, based on easy to obtain geotechnical properties,
can also be employed to evaluate how changing parameters (e.g., porosity or grain size) affect the WRC. In this paper,
the authors present a general set of equations developed for predicting the relationship between volumetric water con-
tent, , (or the corresponding degree of saturation, S
r
) and suction, . The proposed model assumes that water retention
results from the combined effect of capillary and adhesion forces. The complete set of equations is given together with
complementary relationships developed for specific applications on granular materials and on fine-grained soils. It is
shown that the model provides a simple and practical means to estimate the water retention curve from basic geo-
technical properties. A discussion follows on the capabilities and limitations of the model, and on additional tools de-
veloped to complement its use.
Key words: water retention curve, unsaturated soils, prediction, porosity, grain size, liquid limit.
Rsum : La courbe de rtention deau (CRE) est devenue une fonction cl pour dfinir le comportement non satur
des sols et autres matriaux meubles. Dans beaucoup de cas, il peut tre utile davoir une valuation de la CRE durant
les premires phases dun projet, lorsque peu ou pas de rsultats dessais sont disponibles. Des modles prdictifs,
bass sur les proprits gotechniques de base, peuvent aussi tre utiliss pour valuer comment le changement des
paramtres (en termes de porosit ou de granulomtrie) affecte la CRE. Dans cet article, les auteurs prsentent un en-
semble dquations dveloppes pour prdire la relation entre la teneur en eau volumique (ou le degr de saturation
S
r
correspondant) et la succion . Le modle propos est bas sur lhypothse que la rtention deau rsulte de laction
combine de forces capillaires et dadhsion. Le systme dquations est donn avec des relations complmentaires
dveloppes pour des applications spcifiques sur des matriaux granulaires et sur des sols fins. On montre ainsi que
le modle constitue un moyen simple et pratique pour estimer la courbe de rtention deau partir des proprits go-
techniques de base. Une discussion suit sur les capacits et les limitations du modle, ainsi que sur des outils addition-
nels dvelopps pour complter son usage.
Mots cls : courbe de rtention deau, sols non saturs, prdiction, porosit, granulomtrie, limite de liquidit.
Aubertin et al. 1122
Introduction
In geotechnical engineering, subsurface water is often di-
vided into free water in the saturated zone and moisture re-
tained in the unsaturated (vadose) zone (e.g., Bowles 1984;
Smith 1990). More attention is now being paid to better de-
fine the response of water in the vadose zone. Geotechnique
for unsaturated media has become a rapidly expanding field,
which is related to a wide variety of applications including:
estimation of field capacity (Meyer and Gee 1999); effi-
ciency of covers with capillary barrier effects (Bussire et al.
2003); bearing capacity of foundation materials (Rassam and
Williams 1999); seepage through dams (Chapuis and
Aubertin 2001); compressibility and swelling soil response
(Rampino et al. 2000); shear strength and constitutive be-
havior (e.g., Alonso et al. 1990; Vanapalli et al. 1996); con-
taminant transport (Yong 2001); land subsidence (Thu and
Fredlund 2000); and freezethaw of road structures (Konrad
and Roy 2000). Fredlund (2000) and Looney and Falta
(2000) present recent overviews of various applications of
unsaturated soil mechanics and physics.
The distribution and motion of water in the unsaturated
zone are closely related to forces of molecular attraction
which are responsible for water adhering to solid surfaces
Can. Geotech. J. 40: 11041122 (2003) doi: 10.1139/T03-054 2003 NRC Canada
1104
Received 7 January 2002. Accepted July 4 2003. Published on
the NRC Research Press Web site at http://cgj.nrc.ca on
15 October 2003.
M. Aubertin,
1,2
M. Mbonimpa,
2
and R.P. Chapuis.
Department of Civil, Geological and Mining Engineering,
cole Polytechnique de Montral, P.O. Box 6079, Stn
Centre-Ville, Montreal, QC H3C 3A7, Canada.
B. Buissire.
2
Department of Applied Sciences, Universit du
Qubec en Abitibi-Tmiscamingue (UQAT), 445 boul.
de lUniversit, Rouyn-Noranda, QC J9X 5E4, Canada.
1
Corresponding author (e-mail: michel.aubertin@polymtl.ca).
2
Industrial NSERC PolytechniqueUQAT Chair on
Environment and Mine Wastes Management.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:39 AM
Color profile: Disabled
Composite Default screen
(i.e., hygroscopic or adsorbed water on soil particles) and for
surface tension at the interface with air causing capillary
retention (Bear 1972; Marshall et al. 1996). In saturated me-
dia, the adsorption forces tend to reduce the pore space
available for water flow, hence reducing the area for free
water flow and hydraulic conductivity, while capillary forces
disappear in the absence of a waterair interface. In unsatu-
rated conditions, these two distinct but complementary types
of force affect the behavior of soil. The corresponding prop-
erties, including hydraulic conductivity and shear strength,
are no longer material constants but rather depend on the
relative amount of water and air in the pore space.
In a porous system, increasing the value of suction, ,
(defined by u
a
u
w
, where u
a
is the air pressure and u
w
is the
water pressure) tends to reduce the volumetric water content,
. The quantity of water retained in a soil by suction depends
on many factors, namely: shape, size, and distribution of
pore space; mineralogy and surface activity of solid grain
particles; and the chemical composition of interstitial water.
The desaturation is typically more pronounced in coarse-
grained materials (such as sand and gravel) than in fine-
grained materials (such as silt and clay). The value of at a
given also depends on the path, whether it occurs during a
wetting or drying phase. Different paths may induce some-
what different curves (e.g., Maqsoud et al. 2002), but such
hysteresis phenomena will not be addressed directly herein,
as only the drainage path is considered here (to simplify the
presentation).
For a given porous media, parameters and are related
to each other, and form a fundamental material property
known as either the water retention curve (WRC) (Marshall
et al. 1996; Aubertin et al. 1998; Delleur 1999), the soil
water characteristic curve (Fredlund and Rahardjo 1993;
Barbour 1998), the soil suction curve (Yong 2001), the soil
moisture-retention curve (Kovcs 1981), and various other
names (e.g., Klute 1986; Carter 1993; Hillel 1998; Looney
and Falta 2000). The WRC is used in many applications to
represent the relationship between and . The former pa-
rameter is related to porosity, n, and degree of saturation, S
r
( = nS
r
); can also be expressed as a function of the more
common gravimetric water content, w, (i.e., = w(1 n)D
r
,
where D
r
is the relative density of the solid particles). The
value, on the other hand, is typically expressed with pressure
units (e.g., kPa). Suction can also be expressed as a pressure
head, as is done here for the proposed model because it is
based on the use of an equivalent capillary rise, h
co
(see
explanations in the section below). The total is the sum of
the matric suction,
m
, and of the osmotic suction, ; how-
ever, the latter is not considered explicitly in this presenta-
tion. For engineering applications, typically takes a po-
sitive value under negative pore-water pressure, u
w
, above
the phreatic water surface, where u
w
= u
a
or = u
a
u
w
= 0
(with u
a
taken as the reference air pressure).
A large number of experimental investigations have been
devoted to developing and applying techniques to obtain the
relationship of soils. Accordingly, various direct and in-
direct measurement methods have been proposed, and many
of these have been reviewed in textbooks and monographs
(e.g., Klute 1986; Carter 1993; Fredlund and Rahardjo 1993;
Marshall et al. 1996; Looney and Falta 2000). The evolution
and understanding of the WRC and of the corresponding
measurement techniques have been recently presented by
Barbour (1998), who provided a historical perspective.
When the volumetric water content diminishes following
a suction increase, the flow of water becomes more difficult
because of the reduced area and increased tortuosity of the
water phase. The unsaturated hydraulic conductivity, k
u
,
thus, depends on (or ). At a sufficiently low water con-
tent, the water phase becomes discontinuous and k
u
is re-
duced to a very small (near zero) value.
To perform the groundwater flow analyses required in a
number of applications in geotechnique and soil physics, en-
gineers and scientists use Richards (1931) equation, which
is a combination of Darcys law and the mass conservation
equation. In this equation, k
u
is a function of the state vari-
ables, which are typically expressed in terms of or of
(e.g., Leong and Rahardjo 1997a; Leij et al. 1997). As direct
measurement of k
u
can be difficult, time consuming, and
costly, it is customary to use, as a starting point, some rela-
tionship between and to estimate the k
u
function (e.g.,
Mualem 1976, 1986; Fredlund et al. 1994), because the
WRC can be evaluated more easily than k
u
in the laboratory
or in the field. The resulting versus data, obtained from
direct measurements, are plotted and used to derive specific
mathematical functions with curves that run through the data
points. Such equations have been proposed by a large num-
ber of authors including Brooks and Corey (1964), van
Genuchten (1980), Bumb et al. (1992), Fredlund and Xing
(1994), to name only a few. Many of these equations have
been compared to each other over the years (e.g., Khire et al.
1995; Leij et al. 1997; Leong and Rahardjo 1997b; Burger
and Shackelford 2001), showing that each has some advan-
tages and limitations, depending on the technique and data
bank used for comparison.
Measurement of the WRC in the laboratory and in the
field, although less demanding than that of k
u
, can also be
relatively time consuming and expensive. In some situations,
it may be useful to have estimates of the WRC beforehand,
especially during the preliminary stages of a project when
the available information is limited. Predictive models for
the WRC, which are typically based on simple geotechnical
(pedologic) properties such as grain size and porosity, can
also be useful when analyzing and validating test results,
and also for evaluating the variations that can be expected in
the field within a nonhomogeneous deposit.
Quite a few predictive models, sometimes referred to as
pedotransfer functions (e.g., Bouma 1989; Vereecken et al.
1992; Schaap and Leij 1998), have been proposed over the
last two decades or so (see Elsenbeer 2001, and the corre-
sponding Monograph Issue on the topic). These include a
number of functional regression methods using empirical
WRC equations, where the fitting parameters are related to
particular textural parameters (e.g., Cosby et al. 1984;
Schaap et al. 1998). Other discrete regression methods make
no a priori assumption regarding the shape of the WRC, and
construct empirical models from the regression equations
linking the values and basic properties (e.g., Gupta and
Larson 1979; Rawls and Brackensiek 1982; Rawls et al.
1991; Vereecken et al. 1992; Tietje and Tapkenhinrichs
1993). It appears preferable, however, to use predictive mod-
2003 NRC Canada
Aubertin et al. 1105
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:39 AM
Color profile: Disabled
Composite Default screen
els for which the WRC equation is based on physical charac-
teristics of the media. This latter type of model includes
those of Arya and Paris (1981), Haverkamp and Parlange
(1986), Tyler and Wheatcraft (1989, 1990), Haverkamp et
al. (1999), and Arya et al. (1999).
A model of this type was also proposed by Kovcs (1981).
Still relatively unknown and seldom used in the geotechnical
field, the Kovcs (1981) model assumes that a distinction ex-
ists between capillary and adhesive forces, which are both
acting simultaneously to induce suction. This approach po-
tentially provides a conceptual view that can be related to
the actual processes involved (e.g., Celia et al. 1995; Nitao
and Bear 1996). However, the Kovcs (1981) model, in its
original version, did not easily lend itself to practical engi-
neering applications because some of the key parameters
needed for its use were not completely defined. A few years
ago, the authors applied this model after some modifications
(hence, the name Modified Kovcs or MK model) to the
WRC of tailings and silts (Aubertin et al. 1998). In this pa-
per, the MK model is extended for general applications to
various types of materials, from coarse sand to some fine-
grained soils. The basic assumptions and theoretical consid-
erations behind these extensions are briefly presented. The
descriptive and predictive capabilities of the MK model are
shown using test results taken from the literature and ob-
tained in the authors facilities. It should be mentioned at the
onset that the model has been developed only for isotropic
and homogeneous materials, under a drainage path; influ-
ence factors such as internal microstructures, anisotropy, and
volume changes are neglected in this presentation.
The MK model
The proposed MK model presented in this paper retains
the same physical concepts from which the original Kovcs
(1981) model was constructed. The modifications introduced
serve to generalize the statistical function used to describe
the pore-size distribution of the media appearing in the capi-
llary component. Some constitutive parameters are also
expressed more specifically as a function of basic soil prop-
erties, as will be described in the following sections.
The equivalent capillary rise
The MK model makes use of a parameter defined as the
equivalent capillary rise, h
co
[L], in the porous medium. Due
to the central role played by h
co
, which is expressed as an el-
evation, the MK model equations presented here rely on
pressure head units to define (given in cm of water). The
role of the parameter h
co
is the same as the average capillary
rise in the original Kovcs (1981) model. This parameter is
derived from the well-known expression used for the rise h
c
[L] of water in a capillary tube having a diameter d [L]. The
value of h
c
is given by (Smith 1990; Chin 2000):
[1] h
d
c
w w
w
=
4

cos
where
w
[MT
2
] is the surface tension of water (
w
=
0.073 N/m at 20 C),
w
[] is the contact angle between wa-
ter and the tube surface (
w
= 0 for quartz and glass), and

w
[ML
2
T
2
] is the unit weight of water (
w
= 9.8 kN/m
3
at
20 C).
Equation [1] indicates that capillary rise is inversely pro-
portional to the diameter of the tube. When applied to pore
space in soils above the phreatic surface, this equation helps
to understand why there can be a much greater water rise in
fine-grained soils where the voids (somewhat similar to cap-
illary tubes) are small, than in coarse-grained soils where the
voids are typically larger. In soils, however, the pore size is
not uniform so h
c
is not easily defined with eq. [1]. This
pore system can be substituted by a system of regular chan-
nels with a diameter expressed as the equivalent hydraulic
pore diameter, d
eq
[L], defined as (Bear 1972; Kovcs 1981):
[2] d
V
A
eq
v
v
= 4
where V
v
[L
3
] and A
v
[L
2
] are the volume and surface of the
voids, respectively. In practice, A
v
approximately corre-
sponds to the surface area, A
G
[L
2
], of the solid grains. By
relating A
G
to the massic specific surface area S
m
[L
2
M
1
],
eq. [2] can be transformed as (Scheidegger 1974):
[3] d
e
S
eq
s m
= 4

In this equation, e [] is the void ratio and

s
[ML
3
] is the
solid grain density.
The h
co
in a soil is obtained by replacing d (in eq. [1]) by
d
eq
, and can, therefore, be expressed as:
[4] h
S
e
co
w w
w
s m
=

cos
This is one of the fundamental equations from which the
MK model is built. As it will be indicated below, h
co
is
somewhat equivalent (at least for granular soils) to the
height of the capillary fringe above the still water table in a
homogeneous deposit, as defined in many geotechnique text-
books (e.g., Lambe and Whitman 1979; Bowles 1984).
Although S
m
can be directly measured by various tech-
niques (e.g., Lowell and Shields 1984; Igwe 1991), in most
practical cases, the value of S
m
is not readily available to ap-
ply in eq. [4]. For coarse-grained soils, the specific surface
area can nevertheless be estimated from the grain-size distri-
bution using the following expression (Kovcs 1981):
[5] S
D
m
s H
=

where [] is a shape factor (6 18; = 10 is used here

as in the Kovcs original model), and D
H
[L] is an equiva-
lent particle diameter for a heterogeneous mixture. The D
H
for a heterogeneous mix of particles theoretically represents
the diameter of a homogeneous mix (with a single size) that
has the same specific surface area as the heterogeneous one.
Equation [4] can thus be rewritten to calculate the equiva-
lent capillary rise in a soil above the water table:
[6] h
eD
co,G
w w
w H
=

cos
where subscript G stands for granular (low-plasticity, low-
cohesion) materials, as opposed to fine-grained materials
(which will be discussed below). In this equation,
w
will be
taken as zero (e.g., Marshall et al. 1996).
2003 NRC Canada
1106 Can. Geotech. J. Vol. 40, 2003
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:40 AM
Color profile: Disabled
Composite Default screen
In granular soils, S
m
(and D
H
) can be evaluated by subdi-
viding the grain-size curve based on standard mesh sizes
(Chapuis and Lgar 1992). For practical geotechnical appli-
cations, the value of D
H
can also be approximated using the
following function (Aubertin et al. 1998; Mbonimpa et al.
2000, 2002):
[7] D
H
= [1 + 1.17 log (C
U
)]D
10
where D
10
[L] is the diameter corresponding to 10% passing
on the cumulative grain-size distribution curve, and C
U
[] is
the coefficient of uniformity (C
U
= D
60
/D
10
).
For the equivalent capillary rise in granular soils, eq. [6]
is then expressed as follows:
[8] h
b
eD
co,G
=
10
with
[9a] b
C
=
+

w w
U w
[1.17 log (
cos
) ] 1
For the values of ,
w
,
w
and
w
given above, with h
co
and D
10
expressed in cm, eq. [9a] becomes:
[9b] b
C
=
+
0 75
1
.
) 1.17 log (
U
According to eq. [9], for a sand with D
10
= 0.01 cm, C
U
=
5, and e = 0.5, then b = 0.412 cm
2
and the equivalent capil-
lary rise in a granular material, h
co,G
, is about 83 cm; for a
silt with D
10
= 0.0005 cm, C
U
= 20 and e = 0.5, and b =
0.297 cm
2
, h
co,G
would be approximately 1190 cm. These
values of h
co,G
are in the same range of magnitude as the
height of the capillary fringe obtained from the simplified
expression proposed by Bowles (1984), which only consid-
ers the influence of D
10
.
For some plasticcohesive soils, the above equations do
not provide reliable estimates of S
m
and h
co
, particularly
when the liquid limit, w
L
(%), is above about 30% to 40%.
For such fine-grained soils, other factors such as surface ac-
tivity influence their water retention capacity (lets recall
here that the effect of internal microstructure due to loading
history is not explicitly taken into account in this presenta-
tion). For these soils, S
m
(in eq. [4]) is better estimated using
the relationship that exists between the specific surface area
and w
L
. The following empirical expression, recently pro-
posed by Mbonimpa et al. (2002), is used here:
[10] S w
m L
=

where [L
2
M
1
] and (unitless) are material parameters.
Using a relatively large number (78) of test results from var-
ious sources, it has been established that 0.2 m
2
/g and
1.45 for materials with 22 m
2
/g S
m
433 m
2
/g and
18% w
L
127% (see details in Mbonimpa et al. 2002).
Combining eqs. [4] and [10] gives:
[11] h
e
w
co,P L
1. 45
=

where subscript P stands for plasticcohesive materials.
From the previous developments, parameter [L] can be
expressed as:
[12a]

=
w w
w
s
cos
For values of
w
given in N/m,
w
[],
w
in kN/m
3
,
in m
2
/g, and
s
in kg/m
3
, eq. [12a] becomes:
[12b] 0.15
s
From eqs. [11] and [12], one can calculate that for a fine-
grained soil with w
L
= 40%,
s
= 2700 kg/m
3
, and e = 0.8,
then = 405 cm and h
co,P
= 106 500 cm; for w
L
= 80%,
s
=
2700 kg/m
3
, and e = 0.8, one obtains = 405 cm and h
co,P
=
290 968 cm.
The WRC equations
The MK model uses h
co
as a reference parameter to define
the relationship between the degree of saturation, S
r
(or vol-
umetric water content, ), and matric suction, (expressed
here as a pressure head). As mentioned above, both the orig-
inal Kovcs (1981) model and the MK model assumes that
water is held by capillary forces, responsible for capillary
saturation, S
c
, and by adhesive forces, causing saturation by
adhesion, S
a
. In these models, both components act simulta-
neously, and are thus, included in measurements made to
determine the relationship.
The S
c
component equation is obtained from a cumulative
pore-size distribution function, while the equation of S
a
is
given by an interaction law with van der Waals type attrac-
tion between grain surface and water dipoles. The S
c
compo-
nent is more important at relatively low suction values,
while the S
a
component becomes dominant at higher suction
when most capillary water has been withdrawn.
The proposed set of equations for the MK model is writ-
ten as follows for the degree of saturation:
[13] S
n
S S S
r c a c
= = +

*
( ) 1
In this equation, a truncated value of the adhesion compo-
nent, S
a
*
, is introduced in place of S
a
used in the original
model, to make sure that the adhesion component does not
exceed unity at low suction (0 S
a
*
1); it is expressed as:
[14] S S
a a
*
= 1 1
where represents the Macauley brackets ( y = 0.5(y +
*y*)); for S
a
1, S
a
*
= 1, and for S
a
< 1, S
a
*
= S
a
(defined be-
low).
The contributions of the capillary and adhesion compo-
nents to the total degree of saturation are defined as func-
tions of h
co
and using eqs. [15][17].
[15] S h m h
m
c co co
= + 1 1
2 2
[( / ) ] exp[ ( / ) ]
[16] S a C
h
e
a c
co n
n
=

/
( / )
( )
/
/ /
2 3
1 3 1 6
with
[17] C

=
+
+
1
1
1
0
ln( / )
ln( / )
r
r
Equation [15], providing the expression to evaluate S
c
(0
S
c
1) is a generalization of the one developed by Kovcs
(1981), in which the statistical exponential function has been
2003 NRC Canada
Aubertin et al. 1107
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:40 AM
Color profile: Disabled
Composite Default screen
expanded to better reflect the influence of pore-size distribu-
tion through the distribution parameter m. The statistical
distribution expression used for S
c
is one that can also be ap-
plied for grain-size curves, as there is a well-known similar-
ity between the latter and the WRC (e.g., Aubertin et al.
1998). On the WRC, parameter m influences the air entry
value
a
(or AEV), which theoretically corresponds to the
suction when the largest pores start to drain, and also the
rate of decline beyond
a
in the capillary range (Aubertin et
al. 2003). For practical applications, the value of m will be
expressed as a function of basic geotechnical properties.
Equation [16] is based on Kovcs (1981) developments
made from approximations used in theoretical physics, in
which a sixth-order hyperbola is used to relate the adhesion
saturation (due to the film of water adsorbed on grain sur-
faces) to suction. In this equation, a
c
is the adhesion coeffi-
cient (dimensionless) and
n
is a normalization parameter
introduced for unit consistencies (
n
= 1 cm when is given
in cm, corresponding to
n
10
3
atmosphere). Parameter
C

(eq. [17]), taken from Fredlund and Xing (1994), forces

the water content to zero when y reaches a limit imposed by
thermodynamic equilibrium ( = 0 at =
0
= 10
7
cm of wa-
ter, corresponding approximately to complete dryness). In
eq. [17],
r
represents the suction at residual water content;

r
is also equal to the water entry value, WEV, when no dis-
tinction is made between drainage and wetting path. As will
be shown below,
r
can be defined from basic soil properties
(as is the case with
a
and h
co
).
Figures 1a and 1b show typical curves drawn from the
MK model in a semi-log S
r
plane, illustrating the contri-
butions of S
c
and S
a
*
for granular (Fig. 1a) and plastic
cohesive (Fig 1b) materials. Hypothetical (but representa-
tive) values for D
10
, C
U
, e, w
L
,
s
, h
co
, m, a
c
, and
r
have
been used for these sample plots. The parameters
r
and
a
are also shown on Figs. 1a and 1b, together with
90
and
95
to define suctions for preset degrees of saturation (of 90%
and 95%, respectively). Suctions
90
and
95
are compared
to
a
in the discussion. The effect of varying parameters m,
a
c
, and
r
on the WRC is presented graphically in a compan-
ion technical report (Aubertin et al. 2003).
Parameter determination and model
applications
The MK model presented above includes a set of equa-
tions that provides an estimate of the WRC from full satura-
tion (S
r
= 1, = n) to complete dryness (S
r
= 0 = ). To apply
the model however, three parameters in the constitutive
equations have to be defined explicitly: parameter m in
eq. [15], a
c
in eq. [16], and
r
in eq. [17]. Based on investi-
gations carried out by the authors on a diversity of soils and
particulate media (identified in Tables 1 and 2), it has been
found that the values of m, a
c
and
r
can be predetermined
from basic geotechnical properties.
The experimental data used here for granular materials
have been taken from various investigations performed on
sands, low plasticity silts, and tailings. The authors results
have been obtained with either plate extractors or Tempe
cells according to procedures described in Aubertin et al.
(1995, 1998), while some others have been taken from the
literature (see Table 1). All the experimental results on fine-
grained materials have been taken from the literature (see
Table 2). However, to limit the possibility of significant
shrinkage during testing (see Discussion), results on fine-
grained soils cover a relatively small range of liquid limit
and porosity values.
Residual suction
The residual suction
r
introduced in the expression for
C

was evaluated first. For granular materials, values of

r
were determined using the tangent method (Fig. 1a), as de-
scribed by Fredlund and Xing (1994). The WRCs were ob-
tained by fitting the experimental data with a descriptive
equation (i.e., the van Genuchten (1980) model implemented
in the code RETC; van Genuchten et al. 1991); the tangent
method was then applied to the corresponding WRC best-fit.
The subsequent analysis indicated that the following rela-
tionship provides an adequate estimate of the residual suc-
tion (see Fig. 2):
[18]
r
H
=
0 42
1 26
.
( )
.
eD
A simple relationship was also established between
r
and
the equivalent capillary rise, h
co,G
(eq. [8]), for granular ma-
terials:
[19]
r co,G
= 0 86
1 2
.
.
h
This equation is illustrated in Fig. 3, with the correspond-
ing data. As can be seen, h
co,G
is typically somewhat lower
than the value of
r
obtained from the above described
method. In the following, the value of
r
for granular soils is
determined from either eqs. [18] or [19], which give almost
identical residual suction values (see Aubertin et al. 2003 for
more details).
Equation [18] is frequently impractical (or inapplicable)
for fine-grained soils because D
10
and C
U
are often un-
known. Furthermore, the WRC of plasticcohesive materials
is rarely bilinear around the residual water content in a semi-
log plot. Considering also that the experimental results are
not customarily available at high suctions, it can be argued
that the change in slope on the WRC corresponding to resid-
ual suction is generally not well defined (see also the discus-
sion by Corey (1994)). A direct determination of
r
, thus,
becomes difficult for these soils.
In the following applications of the MK model, it is as-
sumed that the
r
values for plasticcohesive soils can be re-
lated to h
co,P
(eq. [11]) using the same dependency as for
granular soils (eq. [19]). Hence, the expression used for fine-
grained soils becomes:
[20]

r L
=

0 86
1 2
1 74
.
.
.
e
w
Parameters a
c
and m
Once
r
was determined for each material in the database
(see Tables 1 and 2), the remaining parameters (a
c
and m)
were evaluated from a fitting procedure, so that calculated
WRCs match experimental data as closely as possible (with
the MK model). Figures 410 compare typical fitted curves
2003 NRC Canada
1108 Can. Geotech. J. Vol. 40, 2003
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:41 AM
Color profile: Disabled
Composite Default screen
2003 NRC Canada
Aubertin et al. 1109
Fig. 1. (a) Illustration of the capillary and adhesion saturation contributions to the total degree of saturation for a noncohesive
(low plasticity) soil, showing the WRC obtained with the MK model (for D
10
= 0.006 cm, C
U
= 10, e = 0.6;
r
= 190 cm, m = 0.1
and a
c
= 0.01);
a
is the pressure (suction) corresponding to the air entry value (AEV),
r
is the pressure corresponding to the resid-
ual water content (also called water entry value, WEV), and
90
and
95
are suctions corresponding to a degree of saturation, S
r
, of
90% and 95%, respectively. (b) Illustration of the capillary and adhesion saturation contributions to the total degree of saturation for a
cohesive (plastic) soil, showing the WRC obtained with the MK model (for w
L
= 30%, e = 0.6,
s
= 2700 kg/m
3
,
r
= 9.7 10
5
cm,
m = 3 10
5
, and a
c
= 7 10
4
);
a
is the pressure (suction) corresponding to the air entry value (AEV),
90
and
95
are suctions
corresponding to a degree of saturation, S
r,
of 90% and 95%, respectively.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:41 AM
Color profile: Disabled
Composite Default screen
(identified as MK model best-fit) using experimental
data on different granular and fine-grained materials; basic
properties (D
10
, C
U
, and e for granular materials; w
L
, e, and
r
s
for plasticcohesive soils) are also given in each figure.
The values of a
c
and m that lead to the best-fit curves pre-
sented in Figs. 410 are given in Table 3.
Further investigations have been conducted to relate the
parameter values m and a
c
to basic geotechnical properties.
For granular soils, where h
co,G
is given by eq. [8], the value
of the pore-size distribution parameter m can be closely ap-
proximated by the inverse of the uniformity coefficient: m =
1/C
U
. When C
U
= 1, m = 1 and Kovcs (1981) original
equation for S
c
is then recovered from the MK model. For
these same granular materials, analysis shows that a
c
can be
considered approximately constant, with a
c
= 0.01.
For fine-grained (plasticcohesive) soils, for which h
co,P
is
given by eq. [11], both m and a
c
values can be taken as con-
stants (with m = 3 10
5
and a
c
= 7 10
4
) in the predictive
applications. In this case, the influence of grain-size distri-
bution appears to be somewhat superseded by the dominant
effect of the surface activity (defined here through the w
L
dependency).
Sample results
The relationships and parameter values defined above are
used to evaluate the WRC of various materials. Tables A1
and A2, in the Appendix, give detailed calculation results to
show explicitly how the presented set of equations are used
to obtain the WRC with the MK model. Figures 410 com-
pare representative fitted curves (with best-fit a
c
and m val-
ues) and predicted curves (with preset a
c
and m values). As
can be seen on these figures, there is generally a good agree-
ment between the predicted and the measured WRC, despite
the differences sometimes observed between the best-fit and
predicted parameter values (especially with m for loose or
fine-grained materials). Such good agreement is obtained
with the majority of results identified in Tables 1 and 2.
The proposed set of equations has also been successfully
used by the authors on actual field projects (e.g., Aubertin et
al. 1999; Nastev and Aubertin 2000) where in situ data and
2003 NRC Canada
1110 Can. Geotech. J. Vol. 40, 2003
Source Material D
10
(cm) C
U
() e ()
Sydor (1992) Coarse sand 0.05800 1.3 0.750
Borden sand 0.00910 1.7 0.590
Modified Borden sand 0.00800 1.8 0.640
Bruch (1993) Beaver Creek sand consolidated at 5 kPa 0.00930 2.6 0.269
Beaver Creek sand consolidated at 10 kPa 0.00930 2.6 0.267
Ricard (1994) Tailings Bevcon 0.00035 11.4 0.674
0.00035 11.4 0.710
0.00035 11.4 0.795
Tailings Senator 0.00031 11.9 0.798
0.00031 11.9 0.929
Tailings Sigma 0.00033 14.6 0.695
0.00033 14.6 0.746
0.00033 14.6 0.802
Tailings Sigma + 10% bentonite 0.00010 35.0 0.698
0.00010 35.0 0.944
Kissiova (1996) Tailings Bevcon 0.00038 11.1 0.570
0.00038 11.1 0.680
0.00038 11.1 0.920
Tailings Sigma 0.00034 14.7 0.660
0.00034 14.7 0.720
Sacrete sand 0.01450 3.5 0.570
0.01450 3.5 0.630
MacKay (1997) London Silt 0.00060 5.5 0.634
Ottawa sand 0.00937 1.7 0.587
Lim et al. (1998) Beaver Creek sand consolidated at 5 kPa 0.00930 2.6 0.618
Rassam and Williams (1999) Tailings at 50 m 0.00600 5.0 0.637
Tailings at 150 m 0.00184 9.5 0.637
Authors results Tailings Sigma (fine) 0.00040 9.8 0.720
0.00040 9.8 0.740
Tailings Sigma (coarse) 0.00040 8.6 0.640
0.00040 8.6 0.710
0.00040 8.6 0.780
Till 0.00035 4.5 0.620
0.00035 4.5 0.700
0.00035 4.5 0.720
0.00035 4.5 0.790
Table 1. Nature, origin, and basic geotechnical properties of the granular materials.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:42 AM
Color profile: Disabled
Composite Default screen
laboratory independent measurements have confirmed that
the predicted WRCs correspond well to actual values. It has
also been validated recently during additional investigations
(including ongoing graduate student projects on other soils,
using column tests); these additional results will be the ob-
ject of future publications.
The MK model can equally be used in a more classical
(descriptive) manner to derive the WRC from a few relevant
testing results. In this instance, the model can then be
applied to evaluate the expected influence of changing prop-
erties on the WRC, such as varying grain size, porosity, or
liquid limit.
Despite the encouraging results obtained so far, there are
nevertheless some limitations and expected discrepancies be-
tween the predicted curves and actual data; some factors are
discussed below.
2003 NRC Canada
Aubertin et al. 1111
Fig. 2. Relationship between the residual suction,
r,exp
, determined from the experimental data, the void ratio, e, and the equivalent
grain-size diameter, D
H
, for granular materials identified in Table 1 (D
H
is defined by eq. [7]).
Fig. 3. Relationship between the residual suction,
r,exp
, determined from the measured data, and the equivalent capillary rise, h
co,G
(eq. [8]), for granular materials.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:42 AM
Color profile: Disabled
Composite Default screen
Discussion
Recent updates to the MK model
The proposed model presented above provides a simple
means to estimate the WRC for granular and fine-grained
soils, as well as for other particulate media. The MK model
equations have been developed as an extension of the
Kovcs (1981) model, which was selected initially because
of its physical bases regarding water retention phenomena
(Ricard 1994; Aubertin et al. 1995, 1998). The model makes
2003 NRC Canada
1112 Can. Geotech. J. Vol. 40, 2003
Fig. 4. Application of the MK model to a coarse, uniform, and relatively loose sand (data from Sydor (1992)).
Source Material w
L
(%) D
r
() e ()
Alimi-Ichola and Bentoumi (1995) Gault clay 40 2.650 0.942
Vanapalli et al. (1996) Sandy clay till (consolidated at 200 kPa) 35.5 2.730 0.430
Sandy clay till (consolidated at 25 kPa) 35.5 2.730 0.540
Sandy clay till (consolidated at 25 kPa) 35.5 2.730 0.545
Huang et al. (1998) Silty sand PPCT11 22.2 2.680 0.536
Silty sand PPCT21 22.2 2.680 0.502
Silty sand PPCT16 22.2 2.680 0.463
Silty sand PPCT26 22.2 2.680 0.425
OKane et al. (1998) Till cover 40 2.770 0.493
Vanapalli et al. (1998) Guadalix Red silty clay 33 2.660 0.480
Fredlund (1999) Record 3713 35.5 2.65 0.545
Record 3714 35.5 2.65 0.545
Record 3715 35.5 2.65 0.449
Record 3716 35.5 2.65 0.474
Record 3717 35.5 2.65 0.474
Record 3718 35.5 2.65 0.518
Record 3720 35.5 2.65 0.546
Record 3728 35.5 2.65 0.438
Record 55 35.5 2.65 0.475
Record 65 35.5 2.73 0.546
Record 66 35.5 2.73 0.438
Record 70 35.5 2.73 0.444
Record 71 35.5 2.73 0.518
Record 72 35.5 2.73 0.472
Record 73 35.5 2.73 0.545
Record 75 35.5 2.73 0.430
Record 76 35.5 2.73 0.372
Table 2. Nature, origin, and basic geotechnical properties of the plasticcohesive materials.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:42 AM
Color profile: Disabled
Composite Default screen
a distinction between capillary and adhesive forces responsi-
ble for moisture retention; this helps to understand the some-
what different approaches taken here to define the material
parameters for granular (S
c
dominated; see Fig. 1a) and
plasticcohesive (S
a
dominated; see Fig. 1b) materials. Re-
cent efforts have been aimed at developing the necessary re-
lationships so that the evaluation of these two components
for predictive purposes could be accomplished using only
basic geotechnical properties, such as the grain size (through
D
10
, C
U
), porosity, n (or void ratio e), solid grain density
(
s
), and liquid limit (w
L
).
The main modifications included in this generalized ver-
sion of the model, compared to the previous (Aubertin et al.
1998) version, can be summarized as follows:
(1) The equivalent capillary rise, h
co
, is defined explicitly
using eqs. [8] and [11], and its physical meaning is
better understood in terms of capillary fringe and resid-
ual suction.
2003 NRC Canada
Aubertin et al. 1113
Fig. 5. Application of the MK model to a fine, uniform, and dense sand (data from Bruch (1993)).
Fig. 6. Application of the MK model to tailings Sigma (silty material, well-graded, and loose) (data from authors).
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:43 AM
Color profile: Disabled
Composite Default screen
2003 NRC Canada
(2) The specific surface, S
m
, is evaluated with more explicit
equations for granular (eq. [5]) and fine-grained
(eq. [10]) soils. The former equation was already in-
cluded in the previous version, while the latter expres-
sion stems from the authors recent work on hydraulic
conductivity functions (Mbonimpa et al. 2002). Equa-
tion [10] extends the use of the MK model to some
fine-grained (plasticcohesive) soils for which the WRC
does not seem to be significantly influenced by com-
pressibility and internal structure.
(3) Instead of a fixed value for the residual suction,
r
(15 000 cm), the updated model uses a variable
r
in the
C

equation (eq. [17]), which depends on material prop-

erties.
1114 Can. Geotech. J. Vol. 40, 2003
Fig. 7. Application of the MK model to tailings Sigma mixed with 10% bentonite (data from authors).
Fig. 8. Application of the MK model to Guadalix red silty clay (data from Vanapalli et al. (1998)).
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:43 AM
Color profile: Disabled
Composite Default screen
(4) The pore-size distribution parameter m, appearing in the
capillary component, S
c
(eq. [15]), has been explicitly
related to the uniformity coefficient (for granular mate-
rials), hence, reflecting its grain-size dependency.
(5) The adhesion component is expressed with a truncated
term (S
a
*
) to eliminate the possibility of exceeding the
maximum degree of saturation. This term is particularly
important for plasticcohesive soils (when w
L
> 30%
40%); it may also be important for other relatively fine-
grained materials without cohesion such as hard rock
tailings. The value of coefficient a
c
is now defined more
precisely.
Also, the available results indicate that the predictions
made with the MK model appear to compare well with those
2003 NRC Canada
Aubertin et al. 1115
Fig. 9. Application of the MK model to a till (data from OKane et al. (1998)).
Fig. 10. Application of the MK model to Indian Head Till (data from Fredlund (1999)).
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:43 AM
Color profile: Disabled
Composite Default screen
2003 NRC Canada
obtained with many other predictive models, which often
include more input parameters (and thus, require more infor-
mation for their application). For instance, predicted volu-
metric water content values with MK, at a given suction,
generally approach the measured values within a range of
0.020.05, which is comparable to the best available pedo-
transfer functions (e.g., Wsten et al. 2001; Schaap et al.
2001; Rawls et al. 2001; Zhuang et al. 2001), even when
these rely on particular measured points on the WRC (e.g.,
water content at suctions of 10, 33, or 1500 kPa). It must be
said, however, that these are preliminary and qualitative re-
marks, as the authors have mainly concentrated their recent
efforts on the practical development of the model equations
for geotechnical applications. More work on this aspect will
thus, have to be performed in the future, to obtain a formal
comparison with other models.
The potential user should also be aware of some remain-
ing limitations for the MK model, as discussed below. Some
additional relationships for
a
and
r
, using the same mate-
rial properties as above, are proposed to complement the
tool kit made available to estimate WRC and related key
parameters.
Prediction of the residual suction and air entry value
The residual suction,
r
and the air entry value,
a
(or
AEV), are familiar points on the WRC, which are often used
for practical applications. The residual suction has already
been defined above with eqs. [18] and [19]. However, the ac-
tual value of
r
(and of the corresponding volumetric water
content) can often be difficult to determine precisely; even
its physical meaning has been questioned (e.g., Corey 1994).
For the applications shown here, more than one approach
can be used to define
r
. It can be measured, for example, on
the fitted curve adjusted to the available experimental data,
with either the MK model or with another model (such as
the van Genuchten (1980) equation);
r
then corresponds to
the intersection point between two tangents (see Fig. 1a). It
can also be estimated from the predicted WRC obtained with
the MK model (using eqs. [13][17]). Here, the calculated
r
value (from eq. [18] or [20]) may be somewhat different
than the one deduced with the tangent method with the com-
plete WRC obtained from the MK model, because the for-
mer gives a point-wise estimate while the latter represents a
predicted curve ranging over several orders of magnitude.
Alternatively,
r
could also be estimated with MK by using
the suction at which S
c
becomes negligible (S
c
0.010.05
for instance). Unfortunately, at this point, not enough is
known about the significance and measurement of residual
suction to be more specific about the best way to predict its
value (particularly for fine-grained soils).
More than one method also exists for the determination of
the AEV (e.g., Corey 1994; Aubertin et al. 1998). As the
MK model aims at obtaining the entire WRC, it is not par-
ticularly well suited for defining precisely specific points on
the curve, such as
a
(or
r
). As a general indication, it has
been observed that the MK model typically overestimates
the AEV by about 17%, on average, for granular materials
(Aubertin et al. 2003). A more precise predictive estimate of
the AEV can nevertheless be obtained directly from basic
geotechnical properties (for granular materials), using the
following expression (Aubertin et al. 2003):
[21]
a,est
H
=
b
eD
x
1
1
( )
where b
1
and x
1
are fitting parameters, and D
H
is given by
eq. [7]. As can be seen in Fig. 11, similar predictions can be
obtained with b
1
= 0.43, x
1
= 0.84, and with b
1
= 0.15, x
1
=
1.0, in eq. [21] (for the range of data considered here). Equa-
tion [21] can also be used to estimate the AEVdeduced from
the MK model WRC (i.e.,
a,MK
) by using b
1
= 0.6, x
1
= 0.8.
Additional methods to define and predict the AEV are
presented in a companion technical report (Aubertin et al.
2003), which also shows that the material AEV obtained
from experimental measurements on granular materials is
typically close to 75% of
95
and 60% of
90
, where
95
and

90
are, respectively, the suction for a degree of saturation of
95% and 90% (defined on the WRC obtained from the MK
model; see Fig. 1a).
For plasticcohesive materials, the authors have not yet
been successful at developing an acceptable correlation
between
a
and basic geotechnical properties. The approach
relating the AEV to
95
and
90
in the MK model appears to
be applicable in this case, but too few data have been ana-
lyzed at this stage to fully support this preliminary assump-
tion.
The equivalent capillary rise, h
co
The equivalent capillary rise, h
co
, constitutes a central pa-
rameter in the MK model. This parameter value is obtained
by eqs. [8] and [9] for granular materials and by eqs. [11]
and [12] for plasticcohesive soils. Here, this parameter the-
oretically represents the capillary head in a pore with a dia-
meter equal to the equivalent hydraulic diameter, d
eq
,
defined according to eq. [3]. As shown in Fig. 3, h
co
is rela-
tively close, but typically somewhat lower than
r
. Hence, it
1116 Can. Geotech. J. Vol. 40, 2003
Fitted WRC Predicted WRC
Material m a
c
m a
c
Coarse, uniform and relatively dense sand (data from Sydor (1992); see Fig. 4) 0.827 0.007 0.769 0.01
Fine and dense sand (data from Bruch (1993); see Fig. 5) 0.091 0.013 0.388 0.01
Tailings Sigma (silty material, coarse and loose) (Authors data, see Fig. 6) 0.161 0.010 0.102 0.01
Tailings Sigma mixed with 10% bentonite (Ricard (1994); see Fig. 7) 0.019 0.009 0.029 0.01
Guadalix Red silty clay (data from Vanapalli et al. (1998); see Fig. 8) 3.610
6
7.610
4
3.010
5
7.010
4
Till (data from OKane et al. (1998); see Fig. 9) 8.110
6
6.510
4
3.010
5
7.010
4
Indian Head Till (Record 728; data from Fredlund (1999); see Fig. 10) 1.010
9
7.010
4
3.010
5
7.010
4
Table 3. Parameters m and a
c
leading to the best-fit, for WRC shown in Figs. 4-10, and the selected values for predictive purposes.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:44 AM
Color profile: Disabled
Composite Default screen
can be seen as representative of the height of the capillary
fringe in homogeneous soils.
A recent investigation furthermore indicates that for gran-
ular materials, one can relate h
co,G
to the AEV using the fol-
lowing relationship (Aubertin et al. 2003):
[22] h b
x
co,G a
=
2
2

Figure 12 illustrates this equation (data identified in Ta-

ble 1) with b
2
5.3 and x
2
1 when
a
is determined from
the experimental curve (
a,exp
); eq. [22] is also shown on the
same figure with b
2
1.7 and x
2
1.2, applicable when
a
is
given by the predictive MK model (
a,MK
; see Fig. 1a).
For plasticcohesive soils, the physical meaning of h
co,P
remains unclear, as indicated by the high values calculated
with eq. [12] given above. Again, the fact that the residual
suction is not well defined in fine-grained soils does not fa-
vor a clear understanding of the conditions represented by
such high suction values (for
r
and h
co,P
). Hence, the equiv-
alent capillary rise, h
co,P
, should only be viewed, at this
point, as a required input for the MK model application.
Parameters with preset values
As shown above, the MK model can be applied in a de-
scriptive manner by adjusting directly parameters a
c
and m
to the experimental results (see Figs. 410). Its main use at
this point, however, is aimed at making predictions based on
easy to obtain basic properties.
In the predictive application of the MK model, values are
preassigned to the adhesion coefficient. A value of a
c
equal
to 0.01 is suggested in the case of granular materials (for
w
L
below 30% 40%), and of 7 10
4
in the case of
plasticcohesive materials.
A fixed value is also assigned to the pore-size distribution
parameter m (3 10
5
) in the case of plasticcohesive mate-
rials, while m has been related to the uniformity coefficient
(m = 1/C
U
) for granular materials.
The user could possibly improve predictions by defining
more precisely the underlying properties on which those pa-
rameters depend, including the specific surface area, S
m
, and
the particle shape factor, a (Aubertin et al. 2003); this aspect
needs to be studied further to evaluate the real influence of
such additional information.
When considering the quality of predictions made with
the MK model and the discrepancies sometimes observed,
the variety of the WRC measurement techniques and associ-
ated uncertainties have to be kept in mind. These may have a
significant effect on the curves and on the estimates obtained
for parameters
r
, m, and a
c
in the MK model, and it can be
expected that the quality of the predictions is influenced by
the data themselves. Nevertheless, the available data appear
to be representative of what can be expected under usual lab-
oratory conditions, and it is encouraging to see that the
agreement between predicted and measured WRC is gener-
ally good.
A final note of caution is required as many influence fac-
tors have not been taken into account in the present state of
the MK model. For instance, the influence of compaction
conditions that may affect the microstructure (fabric) and
pore-size distribution, is neglected; this should be kept in
mind by users who may wish to apply the model to com-
pacted clay liners, for example. Also neglected by the model
are phenomena such as hysteresis of the WRC (e.g., Mualem
1984; Maqsoud et al. 2002), volume change of the sample
during testing (e.g., Al-Mukhtar et al. 1999; Ng and Pang
2000), stressstrain history (Vanapalli et al. 1999), existence
of a heterogeneous or multimodal pore-size distribution
(Burger and Shackelford 2001), and presence of very coarse
particles (Yazdani et al. 2000). Further work is under way to
introduce some of these factors into the MK model.
Additional work being conducted also aims at using the
MK model to evaluate the relative hydraulic conductivity
function k
r
() by integrating the equations presented above.
2003 NRC Canada
Aubertin et al. 1117
Fig. 11. Proposed relationships to estimate the air entry value,
a,exp
(obtained from the experimental data), for granular materials.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:44 AM
Color profile: Disabled
Composite Default screen
This function, k
r
(), is combined with a predictive model for
the saturated hydraulic conductivity, k
s
(Mbonimpa et al.
2000, 2002), to obtain the complete unsaturated k
u
function.
Conclusion
This paper presents a set of equations developed to predict
the water retention curve (WRC). These proposed expres-
sions are derived from the Kovcs (1981) model, which has
been modified to define explicitly all input parameters, and
generalized to extend its application to a variety of porous
media, including granular and fine-grained soils.
In the proposed MK model, the degree of saturation, S
r
,
includes two components acting jointly: one created by cap-
illary forces (S
c
) whose contribution is more important at
relatively low suction, and one associated with adhesive
forces (S
a
) which mainly contributes at higher suction. Both
components can be evaluated from basic (and generally
available) material properties, including the effective diame-
ter, D
10
, uniformity coefficient, C
U
, liquid limit, w
L
, void
ratio, e, and solid grain density,
s
. These properties are first
used to define the equivalent capillary rise, h
co
, which con-
stitutes the central parameter in the MK model.
The set of equations used to predict the WRC contains
three parameters required for model application: the residual
suction
r
, the pore-size distribution parameter, m, and the
adhesion coefficient, a
c
. In the case of granular materials, a
relationship has been developed between
r
and basic geo-
technical properties; a simple relationship between
r
and
h
co
is also presented. The function,
r
(h
co
), established for
granular materials is also used for fine-grained soils where
the residual suction is difficult to define directly. The two
other parameters, m and a
c
, have been obtained first by a
curve-fit procedure so that calculated WRC match experi-
mental data as closely as possible with the MK model.
These values are then expressed from basic geotechnical
properties.
In all the cases considered here, the MK model allowed a
good representation of the experimental WRC, for granular
and fine-grained materials. It can, therefore, be presented as
a useful tool for predicting the WRC during the preliminary
phases of a project, and for estimating how such WRC may
vary with changing material properties.
The model is not meant to replace the required tests to ob-
tain representative data for particular conditions, but rather
provides potential users with a simple and practical means
of foreseeing the possible characteristics of the WRC. The
same can be said about the complementary relationships
proposed to estimate specific points on the curve, including
the air entry value, from material basic properties. When
experimental data become available, these can be used to
improve upon the predictive capabilities of the proposed
equations, by making the necessary adjustments for specific
materials; the MK model can then be used in a more classi-
cal descriptive manner. The discussion presented at the end
of the paper finally draws attention to the inherent limita-
tions of the proposed model.
Acknowledgments
The postdoctoral grant provided to M. Mbonimpa by the
Institut de Recherche en Sant et Scurit du Travail du
Qubec (IRSST) is thankfully acknowledged. Special ac-
knowledgements are also given to Antonio Gatien and to the
graduate students who performed the tests over the years,
and to Dr. John Molson for helping improve the quality of
the manuscript. The authors also received financial support
from the Natural Sciences and Engineering Research Coun-
cil of Canada (NSERC) and from a number of industrial par-
ticipants of the PolytechniqueUQAT Chair in Environment
and Mine Wastes Management.
2003 NRC Canada
1118 Can. Geotech. J. Vol. 40, 2003
Fig. 12. Relationships between the equivalent capillary rise, h
co,G
, and the air entry value for granular materials;
a,exp
and
a,MK
are
obtained on the experimental and on the MK-model-predicted WRC, respectively.
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:45 AM
Color profile: Disabled
Composite Default screen
References
Alimi-Ichola, I., and Bentoumi, O. 1995. Hydraulic conductivity
and diffusivity in vertical and horizontal inflow. In Unsaturated
soils / Sols non saturs. Vol. 1. Edited by E.E. Alonso and P.
Delage. A.A. Balkema, Rotterdam. pp. 335341.
Al-Mukhtar, M., Qi, Y., Alcover, J.-F., and Bergaya, F. 1999.
Oedometric and water retention behavior of highly compacted
unsaturated smectites. Canadian Geotechnical Journal, 36(4):
675684.
Alonso, E., Gens, A., and Josa, A. 1990. A constitutive model for
partially saturated soils. Gotechnique, 40(3): 405430.
Arya, L.M., and Paris, J.F. 1981. A physico-empirical model to
predict the soil moisture characteristic from particle size distri-
bution and bulk density data. Soil Science Society of America
Journal, 45: 10231030.
Arya, L.M., Leij, F.J., van Genuchten, M.T., and Shouse, P.J. 1999.
Scaling parameter to predict the soil water characteristic from
particle-size distribution data. Soil Science Society of America
Journal, 63: 510519.
Aubertin, M., Ricard, J.F., and Chapuis, R.P. 1995. A study of cap-
illary properties of mine tailings: Measurements and modeling.
In Proceedings of the 48th Canadian Geotechnical Conference,
Vancouver, B.C., 2527 September. pp. 1724.
Aubertin, M., Ricard, J.-F., and Chapuis, R.P. 1998. A predictive
model for the water retention curve: application to tailings from
hard-rock mines. Canadian Geotechnical Journal, 35: 5569.
[Erratum, 36: 401 (1999)]
Aubertin, M., Bussire, B. Monzon, M., Joanes, A.-M., Gagnon,
D., Barbera, J.-M., Aachib, M.,Bdard, C. Chapuis, R.P., and
Bernier, L. 1999. tude sur les barrires sches construites
partir des rsidus miniers. Phase II : Essais en place. Rapport
de recherche Projet CDT P1899. Rapport NEDEM/MEND
2.22.2c, 331 pp.
Aubertin, M., Mbonimpa, M., Bussire, M., and Chapuis, R.P.
2003. Development of a model to predict the water retention
curve using basic geotechnical properties. Technical Report
EPM-RT-2003-01. cole Polytechnique de Montral, Montral.
51 pp.
Barbour, S.L. 1998. Nineteenth Canadian Geotechnical Collo-
qium : The soilwater characteristic curve: a historical perspec-
tive. Canadian Geotechnical Journal, 35(5): 873894.
Bear, J. 1972. Dynamics of fluids in porous media. Dover Publica-
tions Inc., New York.
Bouma, J. 1989. Using soil survey data for quantitative land evalu-
ation. Advances in Soil Science, 9: 177213.
Bowles, J.E. 1984. Physical and geotechnical properties of soils.
2nd ed. McGraw-Hill Book Company, New York.
Brooks, R.H., and Corey, A.T. 1964. Hydraulic properties of
porous media. Colorado State University, Fort Collins, Colo.
Hydrology Paper No. 3.
Bruch, P.G. 1993. A laboratory study of evaporative fluxes in
homogeneous and layered soils. M.Sc. thesis, University of Sas-
katchewan, Saskatoon, Saskatchewan.
Bumb, A.C., Murphy, C.L., and Everett, L.G. 1992. A comparison
of three functional forms to representing soil moisture character-
istics. Ground Water, 3: 177185.
Burger, C.A., and Shackelford, C.D. 2001. Evaluating dual poros-
ity of pelletized diatomaceous earth using bimodal soil water
characteristic curve functions. Canadian Geotechnical Journal,
38(1): 5366.
Bussire, B. Aubertin, M., and Chapuis, R.P. 2003. The behavior
of inclined covers used as oxygen barrier. Canadian Geo-
technical Journal, 40: 512535.
Carter, M.R. 1993. Soil sampling and methods of analysis. Lewis,
Boca Raton, Fla.
Celia, M.A., Reeves, P.C., and Ferrand, L.A. 1995. Pore scale
models for multiphase flow in porous media. U.S. National
Report to the International Union of Geodesy and Geophysics
19911994. Reviews of Geophysics, 33: 10491057.
Chapuis, R.P., and Aubertin, M. 2001. A simplified method to esti-
mate saturated and unsaturated seepage through dikes under
steady state conditions. Canadian Geotechnical Journal, 38(6):
13211328.
Chapuis, R.P., and Lgar, P.P. 1992. A simple method for deter-
mining the surface area of fine aggregates and fillers in bitumi-
nous mixtures. In Effects of aggregates and mineral filler on
asphalt mixture performance. American Society for Testing and
Materials (ASTM), Special Technical Publication (STP), 1147:
177186.
Chin, D.A. 2000. Water-resources engineering. Prentice Hill, Up-
per Saddle River, N.J.
Corey, A.T. 1994. Mechanics of immiscible fluids in porous media.
Water Resources Publications, Highlands Ranch, Colo.
Cosby, B.J., Hornberger,G.M., Clapp, R.B., and Ginn, T.R. 1984.
A statistical exploration of the relationships of soil moisture
characteristics to the physical properties of soils. Water Re-
sources Research, 20: 682690.
Delleur, J.W. 1999. The handbook of groundwater engineering.
CRC Press, New York.
Elsenbeer, H. 2001. Editorial: Pedotransfer fuctions in hydrology.
Journal of Hydrology (Special Issue), 251(34): 121122.
Fredlund, M.D. 1999. Soilvision 2.0. Version 2.0 [computer pro-
gram]. A knowledge-based database system for unsaturated-
saturated soil properties. Soilvision Systems Ltd., Saskatoon,
Sask.
Fredlund, D.G. 2000. The 1999 R.M. Hardy Lecture: The imple-
mentation of unsaturated soil mechanics into geotechnical engi-
neering. Canadian Geotechnical Journal, 37: 963986.
Fredlund, D.G., and Rahardjo, H. 1993. Soil mechanics for unsatu-
rated soils. John Wiley and Sons, Inc., New York.
Fredlund, D.G., and Xing, A. 1994. Equations for the soilwater
characteristic curve. Canadian Geotechnical Journal, 31(4):
521532.
Fredlund, D.G., Xing, A., and Huang, S. 1994. Predicting the
permeability function for unsaturated soils using the soilwater
characteristic curve. Canadian Geotechnical Journal, 31: 533
546.
Gupta, S.C., and Larson, W.E. 1979. Estimating soil-water reten-
tion characteristics from particle size distribution, organic matter
percent, and bulk density. Water Resources Research, 15(6):
16331635.
Haverkamp, R., and Parlange, J.-Y. 1986. Predicting the water
retention curve from particle size distribution: 1. Sandy soils
without organic matter. Soil Science, 142: 325339.
Haverkamp, R., Bouraoui, F., Zammit, C., and Angulo-Jaramillo,
R. 1999. Soil properties and moisture movement in the unsatu-
rated zone. In The handbook of groundwater engineering. Edited
by J.W. Delleur, CRC Press, New York. pp. 5.15.47.
Hillel, D. 1998. Environmental soil physics. Academic Press, New
York.
Huang, S., Barbour, S.L., and Fredlund, D.G. 1998. Development
and verification of a coefficient of permeability function for
a deformable unsaturated soil. Canadian Geotechnical Journal,
35(3): 411425.
Igwe, G.J.I. 1991. Powder technology and multiphase systems: Gas
permeametry and surface area measurement. Ellis Horwood,
New York.
2003 NRC Canada
Aubertin et al. 1119
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:45 AM
Color profile: Disabled
Composite Default screen
2003 NRC Canada
1120 Can. Geotech. J. Vol. 40, 2003
Khire, M., Meerdink, J., Benson, C., and Bosscher, P. 1995. Unsat-
urated hydraulic conductivity and water balance predictions for
earthen landfill final covers. In Soil suction applications in
geotechnical engineering practice. Edited by W. Wray and S.
Houston. ASCE, New York. pp. 3857.
Kissiova, M. 1996. tude des modles de prdiction de la conduc-
tivit hydraulique des matriaux meubles non-saturs. M. Ing.,
cole Polytechnique de Montral, Montral, Qc.
Klute, A. 1986. Water retention : laboratory methods. In Methods
of soil analysis, part I. Physical and mineralogical methods. 2nd
ed. Edited by A. Klute. Agronomy Monograph No. 9, American
Society of Agronomy. Soil Science Society of America, Madi-
son Wis. pp. 635662.
Konrad, J.-M., and Roy, M. 2000. Flexible pavements in cold re-
gions: a geotechnical perspective. Canadian Geotechnical Jour-
nal, 37(3): 689699.
Kovcs, G. 1981. Seepage hydraulics. Elsevier Science Publishers,
Amsterdam.
Lambe, T.W., and Whitman, R.V. 1979. Soil mechanics, SI version.
John Wiley and Sons, New York.
Leij, F.J., Russel, W.B., and Lesch, S.M. 1997. Closed-form
expressions for water retention and conductivity data. Ground
Water, 35(5): 848858.
Leong, E.C., and Rahardjo, H. 1997a. Permeability functions for
unsaturated soils. Journal of Geotechnical and Geoenviron-
mental Engineering, ASCE 123(12): 11181126.
Leong, E.C., and Rahardjo, H. 1997b. Review of soil-water charac-
teristic curve functions, Journal of Geotechnical and Geoenvi-
ronmental Engineering, ASCE 123(12): 11061117.
Lim, P.C., Barbour, S.L., and Fredlund, D.G. 1998. The influence
of degree of saturation on the coefficient of aqueous diffusion.
Canadian Geotechnical Journal, 35(5): 811827.
Looney, B.B., and Falta, R.W. 2000. Vadose zone science and tech-
nology solutions. Vols. I and II. Battelle Press, Columbus, Ohio.
Lowell, S., and Shields, J.E. 1984. Powder surface area and poros-
ity. 2nd ed. Chapman and Hall, London.
MacKay, P.C. 1997. Evaluation of oxygen diffusion in unsaturated
soils. M.Sc. Eng., University of Western Ontario, London, Ont.
Maqsoud, A., Bussire, B., and Aubertin, M. 2002. Lhystrsis
des sols non saturs utilizs dans les couvertures avec effets de
barrire capillaire. In Proceedings of the 55th Canadian Geo-
technical Conference and 3rd Joint IAH-CNC and CGS Ground-
water Specialty Conference, Niagara Falls, Ontario, Canada.
Edited by D. Stolle, A. Piggot, and J.J. Crowd. pp. 181188.
Marshall, T.J., Holmes, J.W., and Rose, C.W. 1996. Soil physics.
3rd ed. Cambridge University Press, Cambridge, U.K.
Mbonimpa, M., Aubertin, M., Chapuis, R.P., and Bussire, B.
2000. Dveloppement de fonctions hydriques utilizant les pro-
prits gotechniques de base. In Proceedings of the 1st Joint
IAH-CNC and CGS Groundwater Specialty Conference and the
53rd Canadian Geotechnical Conference, Montreal, Quebec, Ca-
nada, 1518 October. pp. 343350.
Mbonimpa, M., Aubertin, M., Chapuis, R.P., and Bussire, B. 2002
Practical pedotransfer functions for estimating the saturated hy-
draulic conductivity. Geotechnical and Geological Engineering,
20: 235259.
Meyer, P.D., and Gee, G.W. 1999. Advective-diffusion contaminant
migration in unsaturated sand and gravel. Journal of Geotechni-
cal and Geoenvironmental Engineering, ASCE 122(12): 965
975.
Mualem, Y. 1976. A new model for predicting the hydraulic con-
ductivity of unsaturated porous media. Water Resources
Research, 12: 513522.
Mualem, Y. 1984. A modified dependent-domain theory of hyster-
esis. Soil Science, 137: 283291.
Mualem, Y. 1986. Hydraulic conductivity of unsaturated soil: Pre-
diction and formulas. In Methods of soil analysis, Part I: Physi-
cal and mineralogical methods. Edited by A Klute. American
Society of Agronomy (ASA) and Soil Science Society of Amer-
ica (SSSA), Madison, Wis. Agronomy Monograph No. 9.
pp. 799823.
Nastev, M., and Aubertin M. 2000. Hydrogeological modelling for
the reclamation work at the Lorraine Mine Site, Qubec. In Pro-
ceedings of the 1st Joint IAH-CNC and CGS Groundwater Spe-
cialty Conference, Montreal, Quebec, Canada, 1518 October.
pp. 311318.
Ng, C.W.W., and Pang, Y.W. 2000. Experimental investigations
of the soilwater characteristics of a volcanic soil. Canadian
Geotechnical Journal, 37: 12521264.
Nitao, J.J., and Bear, J. 1996. Potentials and their role in transport
in porous media. Water Resources Research, 32(2): 225250.
OKane, M., Wilson, G.W., and Barbour S.L. 1998. Instrumenta-
tion and monitoring of an engineered soil cover system for mine
waste rock. Canadian Geotechnical Journal, 35(5): 828846.
Rampino, C., Mancuso, C., and Vinale, F. 2000. Experimental be-
haviour and modeling of an unsaturated compacted soil. Cana-
dian Geotechnical Journal, 37(4): 748763.
Rassam, D.W., and Williams, D.J. 1999. A relationship describing
the shear strength of unsaturated soils. Canadian Geotechnical
Journal, 36(2): 363368
Rawls, W.J., and Brackensiek, D.L. 1982. Estimating soil water re-
tention from soil properties. Journal of Irrigation and Drainage
Division, ASCE 108: 166171.
Rawls, W.J., Gish, T.J., and Brakensiek, D.L. 1991. Estimating soil
water retention from soil physical properties and characteristics.
Advances in Soil Science, 16: 213234.
Rawls, W.J., Pachepsky Y.A., and Shen, M.H. 2001. Testing soil
water retention estimation with the MUUF pedotransfer model
using data from the southern United States. Journal of Hydrol-
ogy, 251: 177185.
Ricard, J.F. 1994. tude en laboratoire de la relation capillaire et
de la conductivit hydraulique des rsidus miniers. Mmoire de
Matrise, cole Polytechnique de Montral, Montral, Qc.
Richards, L.A. 1931. Capillary conduction of liquids in porous me-
diums. Physics, 1: 318333.
Schaap, M.G., and Leij, J.F. 1998. Database related accuracy and
uncertainty of pedotransfer functions. Soil Science, 163: 765
779.
Schaap, M.G., Leij, F.J., van Genuchten, M.Th. 1998. Neural net-
work analysis for hierarchical prediction of soil water retention
and saturated hydraulic conductivity. Soil Science Society of
America Journal, 64: 843855.
Schaap, M.G., Leij, F.J., van Genuchten, M.Th. 2001. ROSETTA:
a computer program for estimating soil hydraulic parameters
with hierarchical pedotransfer functions. Journal of Hydrology,
251: 163176.
Scheidegger, A.E. 1974. The physics of flow through porous me-
dia. 3rd ed. University of Toronto Press, Toronto.
Smith, G.N. 1990. Elements of soil mechanics. 6th ed. BSP Profes-
sional Books, Oxford.
Sydor, R.C. 1992. Engineered mine tailings cover: verification of
drainage behaviour and investigations of design. M.Sc. thesis,
University of Waterloo, Waterloo, Ont.
Thu, T.M., and Fredlund, D.G. 2000. Modelling subsidence in Ha-
noi City area, Vietnam. Canadian Geotechnical Journal, 37(7):
621637.
Tietje, O., and Tapkenhinrichs, M. 1993. Evaluation of pedo-
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:46 AM
Color profile: Disabled
Composite Default screen
transfer functions. Soil Science Society of America Journal, 57:
10881095.
Tyler, S.W., and Wheatcraft, S.W. 1989. Application of fractal
mathematics to soil water retention estimation. Soil Science So-
ciety of America Journal, 53(4): 987996.
Tyler, S.W., and Wheatcraft, S.W. 1990. Fractal processes in soil
water retention. Water Resources Research, 26(5): 10471054.
van Genuchten, M.Th. 1980. A closed-form equation for predicting
the hydraulic conductivity of unsaturated soils. Soil Science So-
ciety of America Journal, 44: 892898.
van Genuchten, M.TH., Leij, F.J., and Yates, S.R. 1991. The RETC
code for quantifying the hydraulic functions of unsaturated soils.
EPA/600/2-91/065.
Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., and Clifton, A.W.
1996. Model for the prediction of shear strength with respect to
soil suction. Canadian Geotechnical Journal, 33: 379392.
Vanapalli, S.K., Sillers, W.S., and Fredlund, M.D. 1998. The mean-
ing and relevance of residual state to unsaturated soils. In
Proceedings of the 51st Canadian Geotechnical Conference,
Edmonton, Alberta, Canada, 47 October. Preprint Vol. 1.
pp. 101108.
Vanapalli, S.K., Fredlund, D.G., and Pufahl, D.E. 1999. The influ-
ence of soil structure and stress history on the soil water charac-
teristics of a compacted till. Gotechnique, 49: 143159.
Vereecken, H., Diels, J., van Orshoven, J., Feyen, J., and Bouma, J.
1992. Functional evaluation of pedotransfer functions for the es-
timation of soil hydraulic properties. Soil Science Society of
America Journal, 56: 13711378.
Wsten, J.H.M., Pachepsky, Ya.A., and Rawls, W.J. 2001. Pedo-
transfer functions: bridging the gap between available basic soil
data and missing soil hydraulic characteristics. Journal of Hy-
drology, 251: 123150.
Yazdani, J., Barbour, L., and Wilson, W. 2000. Soil water charac-
teristic curve for mine waste rock containing coarse material. In
Proceedings of the 6th Environmental Speciality CSCE and 2nd
Spring Conference of the Geoenvironmental Divison of the
CGS, London, Ontario, 710 June. pp. 198202.
Yong, R.N. 2001. Geoenvironmental engineering Contaminated
soils, pollutant fate, and mitigation. CRC Press, Boca Raton,
Fla.
Zhuang, J., Jin, Y., and Miyazaki, T. 2001. Estimating water reten-
tion characteristic from soil particle-size distribution using a
non-similar media concept. Journal of Hydrology, 251: 308
321.
List of symbols
a
c
adhesion coefficient ()
A
v
surface of the voids (cm
2
)
A
G
surface area of the solid grains [L
2
]
b coarse-grained material parameter to calculate h
co,G
(cm
2
)
b
1
coarse-grained material parameter to estimate
a
(cm
x +1
1
)
b
2
coarse-grained material parameter to correlate h
co,G
and

a
()
C

correction factor ()
C
U
uniformity coefficient () (C
U
= D
60
/D
10
)
d diameter of a tube (cm)
D
10
diameter corresponding to 10% passing on the cumula-
tive grain-size distribution curve (cm)
D
60
diameter corresponding to 60% passing on the cumula-
tive grain-size distribution curve (cm)
d
eq
equivalent pore-size diameter (cm)
D
H
equivalent grain-size diameter for a heterogeneous mix-
ture (cm)
D
r
relative density of the solid particles ()
e void ratio ()
h
c
capillary rise in a tube (cm)
h
co
equivalent capillary rise in a porous material (cm)
h
co,G
equivalent capillary rise in a granular material (cm)
h
co,P
equivalent capillary rise in a plasticcohesive material
(cm)
k
r
relative hydraulic conductivity ()
k
s
saturated hydraulic conductivity [LT
1
]
k
u
unsaturated permeability function [LT
1
]
m pore-size distribution parameter in the MK model ()
n porosity ()
S
a
adhesion component of the degree of saturation ()
S
a
*
truncated adhesion component of the degree of satura-
tion ()
S
c
capillary component of the degree of saturation ()
S
m
specific surface area per unit mass of solid (m
2
/g)
S
r
degree of saturation ()
u
a
air pressure (cm)
u
w
water pressure (cm)
V
v
volume of the voids (cm
3
)
w gravimetric water content ()
w
L
liquid limit (%)
x
1
coarse-grained material parameter to estimate
a
()
x
2
coarse-grained material parameter to correlate h
co,G
and

a
()
y open variable
shape factor ()

w
contact angle ()

w
unit weight of water (kN/m
3
)
volumetric water content ()
material parameter used to estimate S
m
(m
2
/g)
plasticcohesive material parameter required to calcu-
late h
co,P
(cm)
osmotic suction

s
solid-grain density (kg/m
3
)

w
surface tension of water (N/m)
material parameter used to estimate S
m
()
suction (cm)

0
suction at complete dryness (S
r
= 0) (cm)

90
suction corresponding to a degree of saturation of 90%
(cm)

95
suction corresponding to a degree of saturation of 95%
(cm)

a
air entry value or AEV (cm)

a,exp
air entry value determined from the experimental data
(cm)

a,MK
air entry value determined from the WRC predicted
with the MK model (cm)

m
matric suction

n
normalization parameter

r
residual suction (corresponding to the residual water
content) (cm)

r,exp
residual suction determined from the experimental data
(cm)
2003 NRC Canada
Aubertin et al. 1121
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:46 AM
Color profile: Disabled
Composite Default screen
Appendix A: Sample calculations
2003 NRC Canada
1122 Can. Geotech. J. Vol. 40, 2003
Suction
(cm)
S
c
(eq. [15])
()
C

(eq. [17])
()
S
a
(eq. [16])
()
S
a
*
(eq. [14])
()
S
r
(eq. [13])
()
Predicted
(eq. [13]) ()
Measured
()
1 1.000 1.000 1.264 1.000 1.000 0.419 0.419
10 1.000 1.000 0.861 0.861 1.000 0.419 0.401
141 0.999 0.996 0.552 0.552 1.000 0.418 0.392
281 0.793 0.992 0.490 0.490 0.895 0.374 0.381
422 0.455 0.989 0.456 0.456 0.704 0.295 0.306
563 0.254 0.985 0.433 0.433 0.577 0.242 0.266
703 0.148 0.982 0.416 0.416 0.502 0.210 0.24
844 0.090 0.978 0.402 0.402 0.456 0.191 0.213
984 0.058 0.975 0.391 0.391 0.426 0.178 0.195
1125 0.038 0.972 0.381 0.381 0.405 0.169 0.183
1266 0.026 0.968 0.372 0.372 0.389 0.163 0.171
1406 0.019 0.965 0.365 0.365 0.376 0.158 0.157
1688 0.010 0.959 0.351 0.351 0.358 0.150 0.146
2110 0.004 0.951 0.336 0.336 0.339 0.142 0.138
1.010
4
0.000 0.850 0.231 0.231 0.231 0.097
1.010
7
0.000 0.000 0.000 0.000 0.000 0.000
Note: D
10
is 0.0004 cm; e is 0.72 (); C
U
is 9.8 (); b (eq. [9]) is 0.347 cm
2
; h
co,G
(eq. [8]) is 1206 cm;
r
(eq. [18]) is 4603 cm;
0
(fixed) is 1 10
7
cm;
n
(fixed) is 1 cm; a
c
(fixed) is 0.01 (); m = 1/C
U
= 0.102 ().
Table A1. Typical calculation results to predict the WRC of granular materials: Case of Sigma tailings (see Fig. 6).
Suction (cm)
S
c
(eq. [15])
()
C

(eq. [17])
()
S
a
(eq. [16])
()
S
a
* (eq.[14])
()
S
r
(eq. [13])
()
Predicted
(eq. [13]) ()
Measured
()
1 1.000 1.000 2.790 1.000 1.000 0.305 0.305
9.8 1.000 1.000 1.908 1.000 1.000 0.305 0.305
196 1.000 1.000 1.158 1.000 1.000 0.305 0.305
392 0.993 1.000 1.031 1.000 1.000 0.305 0.295
784 0.716 1.000 0.919 0.919 0.977 0.298 0.268
1 176 0.428 1.000 0.859 0.859 0.919 0.280 0.267
1 568 0.270 0.999 0.818 0.818 0.867 0.264 0.263
1 960 0.182 0.999 0.788 0.788 0.827 0.252 0.254
2 940 0.085 0.999 0.737 0.737 0.759 0.231 0.247
3 920 0.049 0.999 0.702 0.702 0.716 0.218 0.232
4 900 0.032 0.998 0.676 0.676 0.686 0.209 0.227
5 880 0.022 0.998 0.656 0.656 0.663 0.202 0.220
7 840 0.012 0.997 0.624 0.624 0.629 0.192 0.215
43 120 0.000 0.986 0.465 0.465 0.465 0.142 0.175
372 400 0.000 0.892 0.293 0.293 0.293 0.089 0.075
838 880 0.000 0.783 0.225 0.225 0.225 0.069 0.055
1 486 660 0.000 0.663 0.173 0.173 0.173 0.053 0.040
2 916 480 0.000 0.471 0.110 0.110 0.110 0.033 0.024
1 0000 000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Note: w
L
is 35.5%; e is 0.438;
s
is 2659 kg/m
3
; (eq. [12b] is 397.5 cm; h
co,P
(eq. [11]) is 160 581 cm;
r
(eq. [20]) is 1 518 208 cm;

0
is 1 10
7
cm;
n
(fixed) is 1 cm; a
c
(fixed) is 7 10
4
(); m (fixed) is 3 10
5
().
Table A2. Typical calculation results to predict the WRC of plasticcohesive soils: Case of Indian Head Till (see Fig. 10).
I:\cgj\CGJ40-06\T03-054.vp
October 8, 2003 9:38:47 AM
Color profile: Disabled
Composite Default screen