The 1999 R.M.

Hardy Lecture: The implementation

of unsaturated soil mechanics into geotechnical
engineering
Delwyn G. Fredlund
Abstract: The implementation of unsaturated soil mechanics into geotechnical engineering practice requires that there
be a paradigm shift from classical soil mechanics methodology. The primary drawback to implementation has been the
excessive costs required to experimentally measure unsaturated soil properties. The use of the soil-water characteristic
curve has been shown to be the key to the implementation of unsaturated soil mechanics. Numerous techniques have
been proposed and studied for the assessment of the soil-water characteristic curves. These techniques range from
direct laboratory measurement to indirect estimation from grain-size curves and knowledge-based database systems. The
soil-water characteristic curve can then be used for the estimation of unsaturated soil property functions. Theoretically
based techniques have been proposed for the estimation of soil property functions such as (i) coefficient of permeabil-
ity, (ii) water storage modulus, and (iii) shear strength. Gradually these estimations are producing acceptable procedures
for geotechnical engineering practices for unsaturated soils. The moisture flux ground surface boundary condition is
likewise becoming a part of the solution of most problems involving unsaturated soils. The implementation process for
unsaturated soils will still require years of collaboration between researchers and practicing geotechnical engineers.
Key words: unsaturated soil mechanics, soil suction, unsaturated soil property functions, negative pore-water pressure,
matric suction, soil-water characteristic curve.
Rsum : Limplantation de la mcanique des sols non saturs dans la pratique de la gotechnique requiert quil y ait
un changement de paradigme par rapport la mthodologie de la mcanique des sols classique. Le principal
inconvnient cette implantation a t les cots excessifs requis pour mesurer exprimentalement les proprits des
sols non saturs. Il a t dmontr que lutilisation de la courbe caractristique sol-eau est la clef de limplantation de
la mcanique des sols saturs. De nombreuses techniques ont t proposes et tudies pour dterminer les courbes
caractristiques sol-eau. Ces techniques stendent de la mesure directe en laboratoire lestimation indirecte en partant
des courbes granulomtriques et base sur la connaissances des systmes de banques de donnes. La courbe
caractristique sol-eau peut alors tre utilise pour lestimation des fonctions de proprits des sols non saturs. Des
techniques fondes sur la thorie ont t proposes pour lestimation des fonctions de proprits des sol telles que:
(i) coefficient de permabilit, (ii) module dentreposage de leau, et (iii) rsistance au cisaillement. Ces estimations
produisent graduellement des procdures acceptables pour les pratiques de la gotechnique dans les sols non saturs.
La condition de frontire de lcoulement de lhumidit la surface du terrain devient galement une partie de la solu-
tion de la plupart des problmes impliquant les sols non saturs. Le processus dimplantation des sols non saturs va
ncessiter encore des annes de collaboration entre les ingnieurs gotechniciens en recherche et en pratique.
Mots cls : mcanique des sols non saturs, succion du sol, fonctions de proprits du sol non satur, pression
interstitielle ngative, succion matricielle, courbe caractristique sol-eau.
[Traduit par la Rdaction] Fredlund 986
Introduction
It is a long road from the discovery of the basic science
related to an engineering phenomenon to its implementation
into standard engineering practice. There can be a sound the-
oretical basis for material behavior, as well as mathematical
solutions, and still this may not bring about a change in en-
gineering practice. Implementation is a unique and important
step that brings theories and analytical solutions into engi-
neering practice. There are several stages in the development
of a science that must be brought together in an efficient and
appropriate manner in order for implementation to become a
reality. This paper focuses on the stages related to the imple-
mentation of unsaturated soil mechanics.
The engineering behavior of an unsaturated soil is com-
monly viewed as being more complex and more difficult to
understand than that of a saturated soil. Unsaturated soils
have negative pore-water pressures, but it is the wide range
of associated degrees of saturation that produces a broad
spectrum of soil behavior. Figure 1 shows that an unsatu-
rated soil can be near to 100% saturation in the capillary
Can. Geotech. J. 37: 963986 (2000) 2000 NRC Canada
963
Received November 2, 1998. Accepted January 10, 2000.
Published on the NRC Research Press website on October 6,
2000.
D.G. Fredlund. Department of Civil Engineering, University
of Saskatchewan, 57 Campus Drive, Saskatoon, SK S7N 5A9,
Canada.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:21 AM
Color profile: Generic CMYK printer profile
Composite Default screen
zone and completely dry near the ground surface. The be-
havioral science for an unsaturated soil has been primarily
developed for the case where the air and water phases are
continuous (i.e., two-phase zone). The degree of saturation
for the two-phase zone generally ranges from about 20 to
80%. However, it has been found that the proposed theories
can be extended throughout the entire unsaturated soil spec-
trum (Fredlund and Rahardjo 1993).
A cursory review of research into the behavior of unsatu-
rated soils shows that the 1970s (and earlier years) were a
period when the fundamental theories and concepts for un-
saturated soil mechanics were formulated. To have a scien-
tific basis for unsaturated soil mechanics, it was necessary
that the state variables (in particular, the stress state vari-
ables) be defined for an unsaturated soil. The use of two sets
of independent stress state variables was firmly established
with a supporting analysis based on multiphase continuum
mechanics during the 1970s (Fredlund and Morgenstern 1977).
The development of theories for unsaturated soil behavior
can be viewed from the standpoint of the classic constitutive
relationships established in saturated soil mechanics, namely
seepage, shear strength, and volume change. Although ther-
mal and chemical constitutive relations are of importance,
this paper will use the above-mentioned classic areas to il-
lustrate the steps required in moving towards the implemen-
tation of unsaturated soil mechanics into geotechnical
engineering practice.
The 1980s were a period when boundary conditions were
assumed for a variety of geotechnical engineering problems
and the emphasis was on solving problems. It became appar-
ent that the soil properties took the form of nonlinear mathe-
matical functions that added to the difficulty in obtaining
solutions. Iterative, numerical models became a common,
useful tool for solving most geotechnical engineering prob-
lems. The numerical solutions were of particular value for
unsaturated soil problems because of the difficulty in obtain-
ing closed-form solutions. One example is the solution for
saturatedunsaturated seepage problems where flow-net so-
lutions became no longer relevant. Fortunately, our compu-
tational abilities were increasing at a pace similar to that of
our understanding of unsaturated soil behavior (Fredlund
1996). The ability to produce more realistic solutions for
various classes of soil mechanics problems might lead to the
inference that unsaturated soil mechanics was ready for ap-
plication in standard engineering practice. Unfortunately,
that was not the case and much more research was required.
The 1990s were a period in which the emphasis was more
focused on the implementation of unsaturated soil mechan-
ics. Implementation has proven to be difficult and it is appar-
ent that a paradigm shift in methodology is required to
facilitate the implementation of unsaturated soil mechanics.
The primary drawback to implementation has been the ex-
cessive time (and therefore costs) required to experimentally
2000 NRC Canada
964 Can. Geotech. J. Vol. 37, 2000
Fig. 1. Classification of the regions within a saturated
unsaturated soil profile. S, degree of saturation.
Fig. 2. Primary stages leading towards the successful implemen-
tation of unsaturated soil mechanics.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:23 AM
Color profile: Generic CMYK printer profile
Composite Default screen
measure unsaturated soil properties. The implementation
process for unsaturated soil mechanics will extend well into
the next millennium.
What is the implementation of unsaturated
soil mechanics?
Implementation can be described as bringing a particular
science or technology into standard engineering practice.
There are a series of stages involved in the process leading
to implementation. Failure to accomplish or achieve a satis-
factory solution at any stage in the development of an ap-
plied science or technology may result in the failure to
achieve full implementation into engineering practice.
The primary stages involved in bringing unsaturated soil
mechanics from a basic science to the implementation phase
are summarized in Fig. 2. The stages leading towards imple-
mentation progress from the most fundamental theoretical
stage to the most practical stage. Research is required at all
stages leading towards implementation. Research studies di-
rected at all stages must lead towards an appropriate technol-
ogy for implementation. An appropriate technology must
meet requirements of being sufficiently accurate while at the
same time being practical and cost effective.
Stages leading towards implementation
The desired end result is to have an appropriate technol-
ogy that produces sufficiently accurate solutions such that it
can readily become part of standard, prudent engineering
practice. A survey of the research literature shows that there
has been a slow but continual development of a science-
based technology emerging for unsaturated soil mechanics
(Clifton et al. 1999). The following sections briefly illustrate
how each of the stages leads towards a more general accep-
tance of unsaturated soil mechanics in engineering practice.
State variable stage
The state variable stage is the most basic and fundamental
level at which a science for unsaturated soil behavior can be
initiated. It is most basic and fundamental because the vari-
ables so defined are generally embedded within the conser-
vative laws of mass and energy. As such, these variables are
independent of the physical properties of the material.
The most important state variables for an unsaturated soil
are the stress state variables. The net normal stress, ( u
a
),
where is the total stress and u
a
is the pore-air pressure, and
the matric suction, (u
a
u
w
), where u
w
is the pore-water
pressure, have become widely accepted stress state designa-
tions for an unsaturated soil. Figure 3 illustrates the need to
separate the effects of total stress and pore-water pressure
when the pore-water pressures are negative. There is also a
smooth transition between the saturated and unsaturated
states. The complete stress state acting in three dimensions,
at a point, is illustrated in Fig. 4. The matrix form of the
stress tensors is shown in the following:
[1]
( )
( )
( )
(

x
xy
xz
yx
y
yz
zx
zy
z
u
u
u

1
]
1
1
1
a
a
a
u u
u u
u u
a w
a w
a w

1
]
1
1
1
)
( )
( )
0
0
0
0
0
0
where
xy
is the shear stress on the x plane in the y direction.
There are two components of soil suction, , namely
matric suction, (u
a
u
w
), and osmotic suction, . Both com-
ponents are important to unsaturated soil mechanics. Matric
suction is of greater relevance for low suctions (i.e., less
than residual water content conditions) and total suction is
of greater relevance at high suctions (i.e., greater than resid-
ual water content conditions).
The stress state variables are to unsaturated soils what ef-
fective stress variables are to saturated soils. The description
of the stress state provides an important tool for sharing and
comparing engineering experiences from around the world.
A description of the stress state changes soil mechanics from
an empirical discipline to an engineering science.
The state variables used to map the relative movement
of a point associated with any phase, due to a change in
stress state, are called the deformation state variables.
Changes in void ratio, water content, or degree of saturation
are generally used in soil mechanics as deformation state
variables. Normal strains associated with the soil structure
are also part of the deformation state variables (Fredlund and
Morgenstern 1976).
Constitutive stage
The constitutive stage becomes the point at which empirical,
semiempirical, and possibly theoretical relationships be-
tween state variables are proposed and verified. The verifica-
tion of proposed constitutive relations must be conducted for
a wide range of soils to ensure uniqueness, and subsequently
confidence on the part of the practicing engineer.
Several constitutive relationships for unsaturated soils
were studied during the 1970s (Fredlund and Morgenstern
2000 NRC Canada
Fredlund 965
Fig. 3. Separation of saturated and unsaturated soil mechanics
based on the stress state description.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:25 AM
Color profile: Generic CMYK printer profile
Composite Default screen
1976; Fredlund et al. 1978; Fredlund 1979). As well, earlier
and later developments have also contributed to our under-
standing of unsaturated soil behavior (Blight 1965; Matyas
and Radhakrishna 1968).
Constitutive relations are generally proposed on the basis
of an understanding of the phenomenological behavior of the
soil. Experimental programs are then undertaken in an at-
tempt to verify the constitutive relations. The verification
process may be extremely demanding and require the testing
of several types of soil. Verification studies are generally un-
dertaken with little regard for how demanding and costly it
may be to obtain the soil parameters.
Demanding laboratory test procedures have proven to be a
significant hurdle in bringing unsaturated soil mechanics
through the implementation stage. The constitutive stage is
important and must stand the tests of rigor and careful ex-
perimental analysis. However, it may be necessary to use al-
ternate procedures to more economically quantify the soil
properties in engineering practice. The later section on the
implementation stage discusses a number of alternate proce-
dures that are feasible for the indirect quantification of
unsaturated soil properties.
Volume change constitutive relationship
The overall volume change of an unsaturated soil can be
defined as a change in void ratio in response to a change in
the stress state:
[2] d d d
a
a
a w
a w
e
e
u
u
e
u u
u u

( )
( ) +

( )
( )
where
e is the void ratio; and
is the total normal confining stress (e.g., isotropic confin-
ing pressure).
Equation [2] can be viewed as having two parts, namely a
part that is the designation of the stress state (i.e., ( u
a
)
and (u
a
u
w
)), and a part that is a designation of the soil
properties (i.e., ( e/ ( u
a
)) and ( e/ (u
a
u
w
))). The soil
properties can be viewed as the slope of the void ratio con-
stitutive surfaces as shown in Fig. 5. The soil properties are
2000 NRC Canada
966 Can. Geotech. J. Vol. 37, 2000
Fig. 4. Elements showing the stress state at a point in a saturated soil and an unsaturated soil.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:30 AM
Color profile: Generic CMYK printer profile
Composite Default screen
moduli that vary as a function of the stress state. The soil
moduli associated with the net normal stress, ( u
a
), can be
written in a general functional form:
[3]

( )
( ) , ( ) ]
e
u
u u u
a
a a w
= func[
where func means that the soil property is a function of the
stress state. At a particular stress state, the compressibility
modulus, m
1
s
, for the void ratio constitutive surface with re-
spect to ( u
a
), can be designated as a constant:
[4]

( )
e
u
m
a
1
s
=
Similarly, the soil moduli associated with soil suction,
(u
a
u
w
), can be written in a general functional form:
[5]

( )
( ) , ( ) ]
e
u u
u u u
a w
a a w
= func[
At a particular stress state, the compressibility modulus,
m
2
s
, for the void ratio constitutive surface with respect to
(u
a
u
w
) can be designated as a constant:
[6]

( )
e
u u
m
a w
2
s
=
Each of the soil moduli is a function of both stress state
variables. To define the magnitude of the soil moduli corre-
sponding to any stress state, there needs to be a constitutive
equation describing the entire void ratio constitutive surface.
The equation then needs to be differentiated with respect to
each of the stress state variables to obtain the compressibil-
ity moduli. At present, no equations have been published
2000 NRC Canada
Fredlund 967
Fig. 5. Void ratio and water content constitutive surfaces for an unsaturated soil and a saturated soil. a
t
, coefficient of compressibility
with respect to a change in the net normal stress, d
mean a
( ); u a
m
, coefficient of compressibility with respect to a change in matric
suction, d(u
a
u
w
); b
t
, coefficient of water content change with respect to a change in the net normal stress, d
mean a
( ); u b
m
, coeffi-
cient of water content change with respect to a change in matric suction, d(u
a
u
w
); and a
v
, coefficient of compressibility for a satu-
rated soil.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:37 AM
Color profile: Generic CMYK printer profile
Composite Default screen
to represent the entire void ratio constitutive surface in
terms of the stress state variables.
The volume-change behavior of an unsaturated soil has
been the focus of numerous research studies. To date, how-
ever, volume-change behavior remains the most difficult un-
saturated soil behavior to characterize.
Unsaturated soil property functions
The soil moduli on the void ratio constitutive surface can
be referred to as an unsaturated soil property function. Es-
sentially all soil properties related to the behavior of an un-
saturated soil become a function of the stress state and are
therefore nonlinear in nature (Fredlund 1995, 1998). Satu-
rated soil properties are also a function of the stress state;
however, in general it has been possible to either linearize
the soil property or use a constant parameter to characterize
soil behavior.
Once the compressibility moduli values corresponding to
any stress state can be predicted, it is then necessary to con-
vert these values into soil parameters acceptable to numeri-
cal computer models. Generally this means converting the
compressibility moduli values into incremental elastic pa-
rameters (i.e., an elastic parameter functional).
The practice of saturated soil mechanics requires the
characterization of a number of soil parameters, but the
practice of unsaturated soil mechanics requires the charac-
terization of a similar number of unsaturated soil property
functions. The increased difficulty in experimentally measur-
ing unsaturated soil property functions becomes a primary
challenge in the implementation of unsaturated soil mechan-
ics.
Water content constitutive relationship
Two constitutive relationships are required to define the
volumemass variables in terms of the stress state variables.
The need for two independent constitutive relations for an
unsaturated soil can be demonstrated through the differentia-
tion of the basic volumemass relationship (i.e., Se = wD
r
):
[7] S e e S D w
e
e
S
S
w
w
d d = d
o
f
o
f
o
f
r
+

where
w is the water content;
S is the degree of saturation;
D
r
is the relative density of the soil solids; and
subscripts o and f represent the initial and final states, re-
spectively.
The water content constitutive surface can be used as a
second relationship for defining the volumemass behavior
of an unsaturated soil (Fig. 5). The water content constitu-
tive relationship can be written in the following general
form:
[8] d d d
a
a
a w
a w
w
w
u
u
w
u u
u u

( )
( ) +

( )
( )
Once again, eq. [8] has a part that designates the stress
state and a part that designates an unsaturated soil property
that is a function of the stress state. The soil moduli associ-
ated with the net normal stress variable, ( u
a
), can be
written as a general function:
[9]

( )
( ) , ( ) ]
w
u
u u u
a
a a w
= func[
At a particular stress state, the compressibility modulus,
m
1
w
, for the water content constitutive surface, with respect
to ( u
a
), can be designated as a constant:
[10]

( )
w
u u
m
a w
1
w
=
Similarly, the soil moduli associated with the soil suction,
(u
a
u
w
), can be written as a general function of the stress
state:
[11]

( )
( ) , ( ) ]
w
u u
u u u
a w
a a w
= func[
At a particular stress state, the compressibility modulus,
m
2
w
, for the water content constitutive surface, with respect
to (u
a
u
w
), can be designated as a constant:
[12]

( )
w
u u
m
a w
2
w
=
At present, there is no published equation to represent the
entire water content constitutive surface. Once an appropri-
ate equation is formulated, the derivatives will provide the
soil moduli values corresponding to any stress state.
A portion of the entire water content constitutive surface
has emerged to have of central importance in the develop-
ment of unsaturated soil mechanics. This portion is the soil-
water characteristic curve that relates water content to the
applied soil suction under conditions where the net normal
stress, ( u
a
), is zero or a small value (Fig. 6). The soil-
water characteristic curve becomes a special case of the en-
tire water content constitutive surface. Numerous mathemati-
cal equations have been proposed to represent the soil-water
characteristic curve. The primary application for the soil-
water characteristic curve has been in the estimation of un-
saturated soil property functions. The most commonly used
unsaturated soil property functions in engineering practice
are those related to seepage and shear strength.
2000 NRC Canada
968 Can. Geotech. J. Vol. 37, 2000
Fig. 6. Definition of variables associated with the soil-water
characteristic curve.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:39 AM
Color profile: Generic CMYK printer profile
Composite Default screen
Nature, characterization, and theory of the soil-water
characteristic curve
The soil-water characteristic curve has played a dominant
role in the study of unsaturated soils in disciplines such as
soil science, soil physics, agronomy, and agriculture
(Barbour 1998). The soil-water characteristic curve is a rela-
tionship between the amount of water in the soil and soil
suction. The amount of water in the soil is generally quanti-
fied in terms of gravimetric water content w, degree of
saturation S, or volumetric water content . The results are
plotted as matric suction in the lower suction range and total
suction in the higher suction range, and usually the term soil
suction is used as the abscissa of the plot. Typical features of
the drying and wetting portions of the soil-water characteris-
tic curves are defined in Fig. 6. The hysteresis loop associ-
ated with the wetting or drying of a soil is the first indication
that the soil-water characteristic curve is not unique.
An initially saturated soil specimen begins to desaturate
when it is subjected to soil suction beyond the air-entry
value. Figure 7 shows the desaturation stages along the
desorption branch of a soil-water characteristic curve (White
et al. 1970). Similar stages apply to the adsorptive branch.
There are three identifiable stages of desaturation, namely
the boundary-effect stage, the transition stage (i.e., with pri-
mary and secondary transition stages), and the residual stage
of desaturation. Typical desorption branches of the soil-water
characteristic curves for several soils are shown in Fig. 8.
Shear strength constitutive relationship
The shear strength equation is a constitutive relationship
defining the shear strength of a soil in terms of the stress
state variables and soil properties. Fredlund et al. (1978)
proposed a linear form for the shear strength of an unsatu-
rated soil (Fig. 9a):
2000 NRC Canada
Fredlund 969
Fig. 7. A soil-water characteristic curve showing different stages
of desaturation (modified after White et al. 1970).
Fig. 8. Typical soil-water characteristic curves (i.e., desorption
branches) for four soils from Saskatchewan, Canada.
Fig. 9. Extended MohrCoulomb shear strength envelope showing the strength parameters for a saturatedunsaturated soil.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:42 AM
Color profile: Generic CMYK printer profile
Composite Default screen
[13] = c + (
n
u
a
) tan + (u
a
u
w
) tan
b
where

n
is the total normal stress on the failure plane at failure;
c is the effective cohesion intercept;
is the effective angle of internal friction; and

b
is the angle defining the rate of increase in shear strength
with respect to soil suction.
The angle of friction associated with the soil suction vari-
able, (u
a
u
w
), was originally assumed to be a constant soil
parameter (Fredlund et al. 1978). However, further labora-
tory studies over a wide range of soil suctions have revealed
that the friction angle should be written as an unsaturated
soil property function (Fredlund et al. 1987, Gan et al.
1988):
[14] tan
b
= func[tan , (
n
u
a
), (u
a
u
w
)]
The dimensionless water content of an unsaturated soil,
(i.e., w(u
a
u
w
)/w
s
), is a function of the complete stress state
but is generally simplified as a function of soil suction.
Therefore, the increase in strength with respect to soil suc-
tion can be written in terms of the normalized water content
used to describe the soil-water characteristic curve. It is also
necessary to include an additional fitting parameter, p, to ac-
count for deficiencies in the one-to-one fit between water
content and shear strength:
[15] tan
b
= func[tan , (w(u
a
u
w
)/w
s
), p]
where
w(u
a
u
w
) is the water content at a particular soil suction;
w
s
is the water content under saturated conditions; and
p is a soil-fitting parameter.
The shear strength surface becomes curvilinear in shape
because the increase in strength changes nonlinearly with re-
spect to suction. There are, however, mathematical equations
that have been proposed and verified for the soil-water char-
acteristic curve and as a result it is possible to write a
closed-form equation for the shear strength constitutive sur-
face for an unsaturated soil, as shown later in this paper.
Seepage constitutive relationship
The driving potential for the conductive flow of water
through a saturated or unsaturated soil is the hydraulic head
gradient, dh/dy (Childs and Collis-George 1950):
[16] h
u
g
Y +
w
w

where
h is the hydraulic head;

w
is the density of water;
g is the acceleration due to gravity; and
Y is the elevation head.
The constitutive relationship to describe flow through a
saturated or unsaturated soil is Darcys law:
[17] v k
h
y

w
d
d
where
v is the flow velocity over the discharge area;
k
w
is the coefficient of permeability; and
y is the depth in the y direction.
The proportionality variable between velocity and hydrau-
lic gradient is assumed to be a constant for saturated soils,
k
s
, but becomes a permeability function for an unsaturated
soil. The coefficient of permeability of an unsaturated soil is
a function of the amount of water in the soil which, in turn,
can be written in terms of the stress state of the soil (Huang
et al. 1998):
[18] k
w
= func[k
s
, ( u
a
), (u
a
u
w
)]
It is generally considered sufficient, however, to quantify
the amount of water in the soil as a function of matric suc-
tion, (u
a
u
w
). The unsaturated coefficient of permeability
can then be written as a function of the saturated coefficient
of permeability and the normalized water content (Fig. 10).
It is again necessary to include an additional fitting parame-
ter, q, to complete the functional relationship:
[19] k
w
(u
a
u
w
) = func[(w(u
a
u
w
)/w
s
), q, k
s
]
Numerous analyses have been proposed for the estimation
of the permeability function for unsaturated soils (Fredlund
et al. 1994; Leong and Rahardjo 1997b). Common to all
methods is the existence of a mathematical relationship be-
tween the coefficient of permeability and the soil-water char-
acteristic curve.
Formulation stage
The formulation stage involves combining the constitutive
behavior of a material with the conservation laws of physics
applied to an elemental volume. The result is generally a
partial differential equation that describes a designated pro-
cess for an element of the continuum. A two-dimensional,
transient seepage partial differential equation for a saturated
unsaturated soil is shown to illustrate the formulation stage:
[20] k
h
x
k
h
y
k
x
h
x
k
y
h
y
m
g
h
t
x y
x
y

2
2
2
2
+ + +
2
w
w
where
m
2
w
is the water storage coefficient;
2000 NRC Canada
970 Can. Geotech. J. Vol. 37, 2000
Fig. 10. Character of the permeability function defined in terms
of soil suction.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:44 AM
Color profile: Generic CMYK printer profile
Composite Default screen
k
x
and k
y
are permeability functions in the x and y directions,
respectively; and
t is the time.
The partial differential equation is nonlinear because the
coefficient of permeability in the flow law is a function of
the pore-water pressure, which in turn is part of the hydraulic
head. Therefore, an unsaturated soil is similar to a saturated
soil with a constantly changing coefficient of permeability.
Appropriate boundary conditions must be placed on the
region under consideration with the result that a boundary-
value problem is defined. The physics at an elemental level
is solved for all elements within the boundaries while at the
same time satisfying the designated boundary conditions.
The ground surface moisture flux boundary condition has
proven to be an important but challenging boundary to de-
fine for engineering purposes. The ground surface becomes
the plane of interaction between the soil and the atmosphere.
The local climatic conditions need to be converted into a net
moisture flux at the ground surface. This problem has
proven to be a particular challenge, but recent developments
in this area have resulted in methodologies for new applica-
tions in geotechnical engineering (Wilson 1990). The design
of soil covers is an example that illustrates the importance
and use of ground surface moisture flux conditions.
Solution stage
The solution stage involves solving specific examples rep-
resentative of a class of problems. At the solution stage, the
partial differential equations are converted to a numerical so-
lution that becomes known as a software package. An exam-
ple of the solution stage is the analysis of seepage through
an earth-fill dam using the finite element method (Fig. 11).
Essentially the same type of solution can be used for a wide
variety of seepage problems. The classic reference solutions
have given rise to the general area of saturatedunsaturated
seepage modeling.
There are generally accepted procedures for modeling a
wide variety of partial differential equations. The most com-
mon procedure involves writing the partial differential equa-
tion in the weighted residual form and using the Galerkin
method to solve a series of linear equations. The weighted
residual Galerkin solution to the saturatedunsaturated,
partial differential seepage equation can be written as an in-
tegral over the area and boundary surface of an element:
[21] [ ] d [ ] [ ]d
{ }
T
w wn
T wn
B k A h L L A
h
t
A A

+ [ ] { }

[ ] L v s
s
T
w
d 0
where
[B] is the matrix of the derivatives of the area coordinates of
a finite element;
A is the area of the element;
[k
w
] is the tensor of the water coefficients of permeability;
{h
wn
} is the matrix of hydraulic heads at the nodal points;
[L] is the matrix of the element area coordinates;
=
w
gm
2
w
;
v
w
is the external water flow rate in a direction perpendicular
to the boundary;
s is the perimeter of the element; and
T is the transpose of matrix.
The finite element procedure has become routine and there
are computer programs dedicated to solving specific partial
differential equations and general partial differential equa-
tion solvers (e.g., PDEase (MacSyma Inc. 1996) and
FlexPDE (PDE Solutions Inc. 1999)). Linear or nonlinear
partial differential equations can also be solved in a coupled
or uncoupled manner using the general partial differential
equation solver. These capabilities are particularly attractive
for solving problems involving saturatedunsaturated soil
systems that require mathematical functions to describe the
soil properties.
The formulation and solution stages were the focus of
considerable research during the 1980s. Numerical methods
such as the finite element technique have become a necessary
and routine tool for solving saturatedunsaturated problems
in engineering practice.
2000 NRC Canada
Fredlund 971
Fig. 11. Two-dimensional, steady state solution of a saturatedunsaturated formulation of seepage through an earthfill dam.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:46 AM
Color profile: Generic CMYK printer profile
Composite Default screen
Design stage
There is a gradual increase in engineering confidence as
research progresses from the formulation stage to the solution
stage and on to the design stage. The design stage focuses
on the primary unknowns that must be quantified from a
practical engineering standpoint. The design stage generally
involves a quantification of geometric and soil property vari-
ables that become part of an engineering design.
The computer has an important role in the design of earth
systems and has changed the way in which geotechnical de-
signs are conducted. The design stage generally takes the
form of a parametric-type study. There are many variables,
as well as ranges of variables, related to soil behavior which
might need to be estimated or approximated. Each set of
variables means another series of analyses. Eventually there
is a matrix of solutions from which the engineer must select
a design solution. The engineer repeatedly asks the question,
How would the design be affected if certain soil parameters
were changed in the following manner? Each What if ?
scenario can be studied in a matter of seconds through an-
other run on the computer. In the final analysis, engineering
judgement and experience must be used to decide upon the
most suitable and preferable engineering design.
The design stage has taken the solution stage, along with
other information, and determined the geometric (and other)
variables required for the engineered structure. Figure 12 il-
lustrates how a variety of possible pore-water pressure pro-
files can be analyzed to show the relationship between the
stability of a slope and the amount of suction in the soil in
the upper portion of the profile. Repeated studies for various
slope geometries become part of the design stage.
There may be significant benefits associated with new en-
gineering design procedures, but the momentum related to
the use of past procedures is often hard to overcome. This is
true even when the new design procedures are more reliable
and accurate.
Verification and monitoring stage
There is need to observe the behavior of any infrastruc-
ture during and subsequent to construction to provide feed-
back to the designer. Only through field monitoring and
feedback can confidence be firmly established in the design
procedures. The observational method as defined by Peck
(1969) goes even beyond the verification of design and is
considered to be a part of the design process.
Case histories play an important part in the practice of
geotechnical engineering. Many conferences have been held
where geotechnical engineers report on investigative studies,
soil testing programs, and the design procedures that have
been used, along with an evaluation of the performance of
the structure. These case histories are particularly necessary
for situations involving unsaturated soils, just as they have
proven necessary for saturated-soils cases. The limited num-
ber of case histories involving unsaturated soil conditions is
one of the factors contributing to the slow implementation of
unsaturated soil design procedures.
The engineer needs to have techniques that can be used to
monitor, evaluate, and ensure the adequacy of the engineering
design. The measurement of positive pore-water pressures is
often used for this purpose when monitoring saturated soils.
There is a similar need to measure in situ negative pore-
water pressures when monitoring unsaturated soils. Negative
2000 NRC Canada
972 Can. Geotech. J. Vol. 37, 2000
Fig. 12. Illustration of What if ? scenario with respect to the stability of a slope subjected to various magnitudes of steady state seepage.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:47 AM
Color profile: Generic CMYK printer profile
Composite Default screen
pore-water pressure measurements have proven to be a chal-
lenge to geotechnical engineers. There have been several
new technological advances related to devices that can mea-
sure highly negative pore-water pressures. The thermal con-
ductivity soil suction sensor has been improved in recent
years and shows promise for use in engineering practice
(Fredlund 1992).
Figure 13 shows a cross section illustrating the compo-
nents of a thermal conductivity soil suction sensor. The ther-
mal conductivity measurement on a standard ceramic that
contains air and water is calibrated against applied matric
suction. Recent thermal conductivity soil suction sensors
have been shown to have durable ceramics and reliable elec-
tronics for the measurement of soil suctions as high as
1000 kPa over an extended period. Sensors can be initially
wet or dry, and then installed into a sample of soil, or in situ.
In Fig. 13, sensor 16 was initially dry and then inserted into
the soil, while sensor 13 was initially water saturated. The fi-
nal equilibrium suctions differed by about 45 kPa because
one sensor (i.e., sensor 16) needed to absorb water from the
soil when coming to equilibrium and the other sensor (i.e.,
sensor 13) underwent desorption in coming to equilibrium. It
is possible, however, to measure the hysteresis associated
with the drying and wetting of the ceramic sensor and to
take the hysteretic effects into consideration in the calibra-
tion of the sensor. Figure 13 shows that the suction values
become essentially equal once the effect of hysteresis is
taken into account. Thermal conductivity sensors are receiv-
ing increasing usage in monitoring the performance of engi-
neered structures involving unsaturated soils.
Implementation stage
The unsaturated soil mechanics area requires research on
the subject of implementation. Failure to give serious study
to the implementation stage will mean the loss of an oppor-
tunity to expand the scope of geotechnical engineering.
The implementation stage may not be realized in engi-
neering practice even when the theoretical formulations and
related design procedures have been fully studied and
verified. Implementation is the final stage in bringing an en-
gineering science into standard engineering practice. Other
factors that need to be addressed at the implementation level
are (i) the cost of undertaking any special site investigations,
soil testing, and engineering analyses; (ii) the human resis-
tance to change; and (iii) the political, regulatory, and litiga-
tion factors that may be involved.
The slowness in the implementation of unsaturated soil
mechanics appears to be related to the cost of soil testing for
the quantification of soil properties. The old soil mechanics
paradigm involving the direct measurement of soil properties
becomes extremely costly when measuring unsaturated soil
property functions. However, there are a number of other
procedures that provide a new paradigm for evaluating un-
saturated soil property functions. These procedures differ in
some respects from classical saturated soil mechanics proce-
dures but provide the necessary accuracy required for ana-
lyzing most unsaturated soil mechanics problems (Fredlund
1996). Figure 14 provides a qualitative illustration of the
benefits derived from using estimated unsaturated soil prop-
erty functions. Estimates of the unsaturated soil property
functions are shown to provide a significant increase in the
accuracy of the engineered design, for a nominal increase
at the soil investigation and testing stage. The accuracy of
the output from an analysis depends strongly upon the
independent variable being computed. The proposed proce-
dures for the estimation of unsaturated soil property func-
tions should not be considered as sufficient and satisfactory
for all modeling situations.
Moving towards the implementation of
unsaturated soil mechanics
The quantification of unsaturated soil property functions,
more than any other single factor, becomes the key to the
implementation of unsaturated soil mechanics in
geotechnical engineering practice. As shown in the previous
2000 NRC Canada
Fredlund 973
Fig. 13. Suction measurements using thermal conductivity sensors installed in the test track, Regina, Saskatchewan.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:50 AM
Color profile: Generic CMYK printer profile
Composite Default screen
section, all of the stages related to bringing an engineering
science into practice have been successfully addressed for
unsaturated soil mechanics. The main remaining challenge is
to determine economically viable procedures for the assess-
ment of unsaturated soil property functions.
The estimation of unsaturated soil property functions pro-
vides a new philosophical framework that could greatly as-
sist in the implementation of unsaturated soil mechanics. It
is important not to remain in a fixed soil mechanics para-
digm that is going to deter the implementation of unsatu-
rated soil mechanics. The challenge is to find new
procedures that can result in a more sound engineering ap-
proach with respect to the unsaturated soil portion of the
profile.
The upper portion of the soils profile plays a dominant
role in water storage and the transmission of water to under-
lying soil strata. The shear strength and volume-change be-
havior of the unsaturated soil portion of the profile also
change significantly in response to the ground surface mois-
ture flux. To model the behavior of the upper portion of the
soil strata it is necessary to be able to compute or estimate
relevant unsaturated soil property functions.
Assessment of unsaturated soil property functions
One of several approaches can be taken to determine un-
saturated soil property functions as shown in Fig. 15. Labo-
ratory tests can be used as a direct measure of the required
unsaturated soil property. Let us consider the determination
of the shear strength properties of an unsaturated soil. For
example, a modified direct shear test can be used to measure
the relationship between matric suction and shear strength.
These tests are costly to perform and the necessary equip-
ment may not be available. However, it may be sufficient to
perform an indirect laboratory test (i.e., pressure plate test)
to obtain the unsaturated soil property function or it may be
sufficient to estimate the function from the results of another
test or from classification soil properties.
Measurement of the soil-water characteristic curve for a
soil can be used as an indirect laboratory test to compute an
unsaturated soil property function. The soil-water character-
istic curve can then be used in conjunction with the saturated
shear strength properties of the soil to estimate the relation-
ship between shear strength and soil suction to an acceptable
level for most engineering projects.
Figure 15 also suggests the use of a classification test for
the prediction of the desired unsaturated soil property func-
tion. A grain-size analysis can be used to estimate the soil-
water characteristic curve, which is then used to determine
the unsaturated soil property function (Fredlund et al. 1997).
There may be a reduction in the accuracy of the estimated
unsaturated soil property function when using this proce-
dure. The engineer must assess whether or not the approxi-
mated unsaturated soil property function is satisfactory for
the analyses to be performed.
Mathematical form for the soil-water characteristic
curve
Several mathematical equations have been proposed to de-
scribe the soil-water characteristic curve. The equation of
Gardner (1958) was originally proposed for defining the un-
saturated coefficient of permeability function, and its appli-
cation to the soil-water characteristic curve is inferred. The
mathematical equations proposed by Burdine (1953) and
Maulem (1976) are two-parameter equations that become
special cases of the more general three-parameter equation
proposed by van Genuchten (1980). These equations are as-
ymptotic to horizontal lines in the low soil suction range and
a suction beyond residual conditions. As such, these equa-
tions are not forced through zero water content at
1 000 000 kPa of suction. A correction factor, C
r
, has been
applied to the mathematical equation proposed by Fredlund
and Xing (1994). The correction factor forces the soil-water
characteristic curve function through a suction of
1 000 000 kPa at a water content of zero. Some of the com-
mon equations proposed for the soil-water characteristic
curve are summarized in Table 1. A more complete sum-
mary of proposed equations can be found in Sillers (1997).
All of the proposed equations provide a reasonable fit of
soil-water characteristic data in the low- and intermediate-
suction ranges (Leong and Rahardjo 1997a). In all cases, the
a parameter bears a relationship to the air-entry value of the
soil and usually refers to the inflection point along the curve.
The n parameter corresponds to the slope of the straight-line
portion of the main desorption (or adsorption) portion of the
soil-water characteristic curve.
2000 NRC Canada
974 Can. Geotech. J. Vol. 37, 2000
Fig. 14. Qualitative representation of the benefits derived from
using estimated unsaturated soil property functions.
Fig. 15. Approaches that can be used in the laboratory to deter-
mine the unsaturated soil property functions.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:55 AM
Color profile: Generic CMYK printer profile
Composite Default screen
The Fredlund and Xing (1994) mathematical function ap-
plies over the entire range of soil suctions from 0 to
1 000 000 kPa. The relationship is essentially empirical and,
similar to earlier models, is based on the assumption that the
soil consists of a set of interconnected pores that are ran-
domly distributed. Subsequent discussions on the soil-water
characteristic curve are restricted to the Fredlund and Xing
equation.
The Fredlund and Xing (1994) equation, written in terms
of gravimetric water content, w, is as follows:
[22] w C
w
a
n
m

_
,

_
,

1
]
1
1
( )
ln

s
e
where
w
s
is the saturated gravimetric water content;
a is a suction value corresponding to the inflection point on
the curve and is somewhat greater than the air-entry value;
n is a soil parameter related to the slope of the soil-water
characteristic curve at the inflection point;
is the soil suction (i.e., matric suction at low suctions and
total suction at high suctions);
m is a fitting parameter related to the results near to residual
water content;
e is the natural number, 2.71828 ; and
C() is the correction function that causes the soil-water
characteristic curve to pass through a suction of
1 000 000 kPa at zero water content.
The correction factor is defined as
[23] C( )
ln
ln

_
,

_
,

1
1
1
r
r
1 000 000

1
]
1
1
1
1
1
1
where
r
is the suction value corresponding to residual water
content, w
r
. Residual suction can be estimated as 1500 kPa
for most soils, unless the actual value is known.
Equation [23] can be written in a dimensionless form by
dividing both sides of the equation by the saturated
gravimetric water content (i.e., = w/w
s
, where is the
dimensionless water content):
[24]
+

_
,

_
,

1
]
1
1
C
a
n
m
( )
ln

1
e
Equation [24] can be used to best-fit the desorption or ad-
sorption branches of soil-water characteristic curve data over
the entire range of suctions. The fitting parameters (i.e., a, n,
and m values) can be determined using a nonlinear regres-
sion procedure such as the one proposed by Fredlund and
Xing (1994).
The character of the Fredlund and Xing (1994) equation
(eq. [24]) can be observed by varying each of the curve-
fitting parameters (i.e., a, n, and m). Figure 16 illustrates the
lateral translation of the soil-water characteristic curves as a
result of varying the a parameter with the n parameter fixed
at 1.5. Figure 17 illustrates the change in slope of the soil-
water characteristic curve as a result of changing the n pa-
rameter with the a parameter fixed at 25 kPa. Figure 18 il-
lustrates the rise in the soil-water characteristic curve as the
2000 NRC Canada
Fredlund 975
Author(s) Equation Soil parameter
Gardner 1958
w
w
a
n

_
,

s
g
g
1

a
g
, n
g
Van Genuchten 1980
w
w
a
n
m

_
,

1
]
1
1
1
s
vg
vg
vg
1

a
vg
, n
vg
, m
vg
Maulem 1976
w
w
a
n
m

_
,

1
]
1
1
s
m
m
m
1

a
m
, n
m
, m
m
= 1/(1 n
m
)
Burdine 1953
w
w
a
n
m

_
,

1
]
1
1
s
b
b
b
1

a
b
, n
b
, m
b
= 2/(1 n
b
)
Fredlund and Xing 1994
w C
w
a
n
m

_
,

_
,

1
]
1
1
(
ln
)

s
f
e
f
f
a
f
, n
f
, m
f
, C()
Table 1. Summary some of the mathematical equations proposed for the soil-water characteristic curve.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:56 AM
Color profile: Generic CMYK printer profile
Composite Default screen
m parameter is varied. The correction factor, C(), was set
at 1.0 for each of the three illustrative examples. If the cor-
rection factor, C(), is computed using eq. [23], each of the
soil-water characteristic curves would pass through a soil
suction of 1 000 000 kPa at zero water content.
Laboratory measurement of the soil-water
characteristic curve
The soil-water characteristic curve can be measured in the
laboratory with relative ease. The test equipment is com-
monly found in soil science laboratories and has also found
its way into some soil mechanics laboratories. The cost of
performing the tests is somewhat less than the costs associ-
ated with performing a one-dimensional consolidation test.
The experimental measurement of the soil-water charac-
teristic curve can be divided into two parts, namely the re-
gions where the suctions are less than approximately
1500 kPa, and the regions where the suctions are greater
than 1500 kPa. Suctions greater than 1500 kPa are generally
established using an osmotic desiccator. The water content
corresponding to high suction values (i.e., suctions greater
than 1500 kPa) is determined by allowing small soil speci-
mens to come to equilibrium in an osmotic desiccator con-
taining a salt solution.
Water contents corresponding to low suction values are
usually determined using an acrylic pressure plate device
(often referred to as a Tempe Cell, Soilmoisture Equipment
Inc., Santa Barbara, Calif.), with a 1 bar (100 kPa) high-
air-entry disk. A second commercially available pressure
plate device is the Volumetric Pressure Plate (Soilmoisture
Equipment Inc. Santa Barbara, Calif.), which has a 2 bar
(200 kPa) high-air-entry disk. A third pressure plate device
is available (Soilmoisture Equipment Inc.) with a 15 bar
(1500 kPa) ceramic or a pressure membrane. Each of the
above pieces of equipment was designed for use in areas
other than geotechnical engineering. As such, each apparatus
has certain limitations, and several attempts have been made
to develop an apparatus more suitable for geotechnical engi-
neering. One such device is the pressure plate cell developed
at the University of Saskatchewan, Saskatoon (Fig. 19).
Test procedure for measuring the soil-water
characteristic curve
The test procedure for measuring the soil-water character-
istic curve was originally developed in soil science and
agronomy, but the test procedures adopted within
geotechnical engineering have remained similar. The atten-
tion given to the initial preparation of soil specimens is quite
different between geotechnical engineering and soil science.
The agriculture-related disciplines have generally not paid
much attention to the initial state (or structure) of the soil
specimens. On the other hand, many of the soil mechanics
theories used in geotechnical engineering are predicated on
the assumption that it is possible to obtain undisturbed soil
samples that can be tested to measure in situ physical soil
properties. Soil-water characteristic curve data collected
from a variety of sources will generally have used several
different specimen preparation procedures.
Initial preparation states for a soil specimen can be cate-
gorized as follows: (i) undisturbed samples that retain the in
situ soil structure, (ii) completely remolded specimens where
the soil is mixed with water to form a semiliquid paste, and
(iii) remolded and compacted samples where the initial wa-
ter content is near to the plastic limit. Regardless of which
of the above procedures is used, the soil specimens are
placed into the pressure plate apparatus, covered with water,
and allowed to saturate. Therefore, the soil suction is re-
duced to zero prior to commencing the formal test.
The difference between the various specimen preparation
procedures may not be of great concern for a sandy soil but
the opposite is true for a clayey soil. The soil structure and
secondary macrostructure become of increasing concern for
2000 NRC Canada
976 Can. Geotech. J. Vol. 37, 2000
Fig. 18. Plot of the Fredlund and Xing (1994) equation with a
and n constant and m varying.
Fig. 16. Plot of the Fredlund and Xing (1994) equation with m
and n constant and a varying.
Fig. 17. Plot of the Fredlund and Xing (1994) equation with m
and a constant and n varying.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:02:59 AM
Color profile: Generic CMYK printer profile
Composite Default screen
clay-rich soils. Figure 20 shows the effect that the above ini-
tial states can have on the desorption curves for a clayey
soil. It is well recognized that there is no single or unique
soil-water characteristic curve for a particular soil.
Interpretation of soil-water characteristic
curve data
The hysteresis associated with the wetting and drying of a
soil also illustrates that there is no unique soil-water charac-
teristic curve. The wetting and drying branches form the ex-
treme bounds for the soil-water characteristic curve. In
addition, there are an infinite number of intermediate (drying
or wetting) scanning curves. The drying and wetting scan-
ning curves become asymptotic to the bounding curves as
shown in Fig. 21. An undisturbed soil sample from the field
will have a soil suction that lies somewhere between the
bounding curves. Let us assume that the effect of total con-
fining stress on the wetting and drying curves is negligible.
For interpretation purposes, the stress state can be on any
one of the scanning curves or on either of the bounding
curves.
Once the soil specimen is placed in the pressure plate ap-
paratus, it is immersed in water and time is allowed for the
2000 NRC Canada
Fredlund 977
Fig. 19. A single-specimen, pressure plate cell developed at the University of Saskatchewan, Saskatoon, Canada.
Fig. 21. Description of wetting and drying scanning curves and
initial stress state.
Fig. 20. Illustration of the influence of initial state on the soil-
water characteristic curve.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:06 AM
Color profile: Generic CMYK printer profile
Composite Default screen
soil suction to go to zero. The stress state of the soil has now
been altered and the stress path followed is shown in
Fig. 22. If the soil is clayey, it is likely that the specimen
will undergo a change in volume. Once the specimen is satu-
rated, the test data for the soil-water characteristic curve is
obtained by applying a series of suction values while allow-
ing the specimen to come to equilibrium. In so doing, the
desorption (bounding) branch of the water content versus
soil suction relationship is measured. Figure 23 illustrates
the change in state experienced by a sandy soil as it is wet-
ted prior to testing. The effects of initial wetting have the
appearance of being quite dramatic.
The changes in state experienced by the soil being tested
have been outlined to assist in the interpretation of the test
data. For example, it is obvious that the laboratory data do
not represent a soil being dried from the in situ stress state.
Rather, the laboratory-measured soil-water characteristic
curve data correspond to a particular test procedure that does
not accurately represent in situ conditions and may not rep-
resent conditions applied later in an engineering analysis.
The fact that soil-water characteristic curve data have proven
to be of significant value in engineering applications would
indicate that a high level of accuracy is not required for the
characterization of unsaturated soil property functions. The
above description of the pressure plate test procedure also
helps explain why the soil-water characteristic curve cannot
be used to provide an indication of in situ suction.
It would appear to be difficult to propose a new test proce-
dure for obtaining the soil-water characteristic curve data.
Possibly there would be some merit in more accurately rep-
resenting the in situ total stresses when performing the test.
There appears to have been considerable success in using
data obtained from the above-mentioned test procedure. It
would appear reasonable to make as much use as possible of
existing data sets and the present test procedure (Fredlund et
al. 1996). It must be recognized that most of the existing
data represent the desorption curve, subsequent to saturating
a soil specimen. It would be of value to always record the
details regarding specimen preparation.
Estimation of the soil-water characteristic curve
One of several approaches can be used to estimate a soil-
water characteristic curve. Some attempts have been made to
correlate the fitting parameters for a soil-water characteristic
curve with the plasticity and (or) grain-size distribution clas-
sification properties of a soil (Ahuja et al. 1985). These cor-
relations are based on limited data, but further studies may
prove these relationships to be of value in engineering prac-
tice.
The use of operations research techniques holds promise
in the search for a suitable soil-water characteristic curve. A
large volume of soil-water characteristic curve data has been
collected in several disciplines (e.g., soil science, agronomy,
agriculture, and engineering) and in many countries. A com-
piled database can be used to select an approximate soil-
water characteristic curve. The grain-size distribution curves
for a soil can be matched to other grain-size curves to select
an approximate soil-water characteristic curve.
Figure 24 illustrates several approaches that can be used
to obtain a soil-water characteristic curve that can subse-
quently be used for the determination of unsaturated soil
property functions. The classification and (or) soil-water
characteristic curves can be used in conjunction with a
knowledge-based database to assist the user in arriving at a
reasonable soil-water characteristic curve (Fredlund et al.
1996; Fredlund 1997).
The first suggested procedure compares a measured soil-
water characteristic curves with soil-water characteristic
curves already in the database. The measured soil-water
characteristic curve can be used either to compute unsatu-
rated soil property functions or to select unsaturated soil
property functions already in the database.
The second suggested procedure involves matching mea-
sured classification properties (i.e., grain-size curves) with
classification properties already in the database. Once one or
more similar soils have been found, corresponding soil-water
characteristic curves can be retrieved from the database. The
soil-water characteristic curve data can be used to compute
suitable unsaturated soil property functions, or existing
2000 NRC Canada
978 Can. Geotech. J. Vol. 37, 2000
Fig. 22. Stress path followed at the start of a pressure plate test
of a clayey soil.
Fig. 23. Stress path followed during the initial stages of pressure
plate testing of a sandy soil.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:10 AM
Color profile: Generic CMYK printer profile
Composite Default screen
unsaturated soil property functions can be retrieved from the
database.
The third suggested procedure makes direct use of the
measured grain-size distribution curve. The grain-size
distribution curve for a given soil is compared to grain-size
curves already in the database. Soil-water characteristic curves
can then be computed from the grain-size curves and com-
pared to soil-water characteristic curves in the database. An
engineering decision must be made regarding a reasonable
soil-water characteristic curve and then the unsaturated soil
property functions can be computed. In general, each of the
above procedures becomes increasingly less precise in the
estimation of the unsaturated soil property function.
The third procedure involves computing the soil-water
characteristic curve directly from the grain-size distribution
curve (Arya and Paria 1981). The mathematical equation
that is used for describing the soil-water characteristic curve
can also be used for fitting the grain-size curve. The form of
the Fredlund and Xing (1994) equation for the soil-water
characteristic curve can be modified to fit the grain-size dis-
tribution curve, since it has the ability to independently char-
acterize the two extremes of the function (Fredlund et al.
1997). Three parameters and a correction factor are used to
define the mathematical function:
[25] P C d
d
a
n
m

_
,

_
,

1
]
1
1
1
( )
ln
r
s
e
s
s
100
where
P is the percent passing;
100 represents 100% passing;
e is the natural number 2.71828 ;
d is the diameter of the particles in millimetres;
a
s
is a parameter corresponding to the inflection point on the
grain-size curve, related to the largest particles in the distri-
bution (on a semilogarithmic scale);
n
s
is a parameter related to the uniformity of the particle-size
distribution;
m
s
is a parameter related to residual particle sizes; and
C(d
r
) is a correction factor to ensure that the function goes
through a lower limit particle diameter (e.g., 0.00001 mm).
The grain-size distribution curve parameters (i.e., a
s
, n
s
,
and m
s
) can be obtained by performing a best-fit regression
analysis on particle-size data. The grain-size distribution
data may have been measured in the laboratory or obtained
from an existing database. The grain-size curve parameters
allow the soil to be represented as a continuous mathemati-
cal function for further analysis. The grain-size distribution
curve provides information on the pore-size distribution of
the soil. This is true only to a degree, and the volumemass
properties must also be taken into consideration in the form
of a packing factor (Fredlund et al. 1997).
Mathematically characterizing the grain-size curve pro-
vides a function that forms the basis for a theory to estimate
the pore-size distribution. Several attempts have been made
to characterize the pore-size distribution from the grain-size
curve (Gupta and Larson 1979; Ghosh 1980; Arya and Paria
1981; Ahuja et al. 1985; Haverkamp and Parlange 1986).
Fredlund et al. (1997) showed that it is possible to esti-
mate the soil-water characteristic curve from a grain-size
distribution curve provided the procedure is trained with
the assistance of a knowledge base. Figure 25a shows a best-
fit of the grain-size curve for a sand. This curve is then used
to predict the soil-water characteristic curve that can be
compared with experimental data (Fig. 25b). The results are
encouraging for sands and silts, but more research is re-
quired when using this procedure for structured and clayey
soils.
2000 NRC Canada
Fredlund 979
Fig. 24. Approaches that can be used to estimate soil-water characteristic curves for the determination of unsaturated soil property
functions when using classification tests and a data base.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:14 AM
Color profile: Generic CMYK printer profile
Composite Default screen
Application of unsaturated soil property functions
The unsaturated soil property functions take on a variety
of mathematical forms for shear strength, seepage, and vol-
ume change. The general graphical form is consistent for a
particular unsaturated soil property. However, there are a va-
riety of mathematical functions that can be used to represent
a particular unsaturated soil property. A particular class of
geotechnical problems may require several mathematical
functions.
It is beyond the scope of this paper to present all proposed
unsaturated soil property functions. Also, it is not always
possible to present the soil property functions as simple
closed-form solutions. Figure 26 illustrates some common
formats that can be used for unsaturated soil property func-
tions. The functional data is used as input when solving an
engineering problem. A closed-form, theoretically based
equation is the preferable format for unsaturated soil prop-
erty functions. It is not always possible to present the data in
this format and therefore other formats must be used.
Figure 27 illustrates the different ways in which tabular
data can be input to an analysis. In some cases, a bar-graph
format may be satisfactory; however, in general it is better to
fit the data points with a meaningful empirical equation. The
spline function may pass through the data points, but the
best-fit empirical equation may produce more meaningful re-
sults with less convergence difficulties in a subsequent nu-
merical analysis.
The remaining sections illustrate the general forms for
shear strength and seepage unsaturated soil property func-
tions. A number of theoretically based unsaturated soil prop-
erty functions have been presented for shear strength and
seepage properties, but it is beyond the scope of this paper to
summarize all the equations.
Shear strength of unsaturated soils
It is possible to mathematically represent the nonlinear
form of the unsaturated shear strength envelope through the
use of the soil-water characteristic curve and the saturated
shear strength parameters of the soil. The angle
b
begins to
deviate from the effective angle of internal friction, , as the
soil desaturates at suctions greater than the air-entry value
(Fig. 28). As the soil suction reaches a value corresponding
to the residual water content,
b
appears to approach a value
close to zero degrees (or it may even be negative).
The shear strength function can be obtained through the
use of the soil-water characteristic curve. Fredlund and Xing
(1994) proposed the following shear strength expression as a
function of matric suction and the effective shear strength
parameters:
[26] = (u
a
u
w
)
p
tan
where
is the normalized water content = w(u
a
u
w
)/w
s
;
w(u
a
u
w
) is the gravimetric water content at any suction,
which can be represented by the equation for the soil-water
of characteristic curve; and
p is a fitting parameter.
The unsaturated shear strength prediction depends on the
fitting parameter p. A value of 1 can be assumed for p for
most inactive soils such as sands, silts, and some fine-
grained soils in the soil suction range between 0 and
500 kPa. Comparisons between predictions and measured
shear strength data will provide a better understanding of the
parameter p.
The entire shear strength equation for an unsaturated soil
can be written as follows:
[27] = c + (
n
u
a
)tan + (u
a
u
w
)
p
tan
The unsaturated soil now appears to have one friction an-
gle, , but the area over which soil suction acts (i.e., water
phase) is reduced as the suction increases. The soil-water
characteristic curve quantifies the amount of water on any
section through the soil.
2000 NRC Canada
980 Can. Geotech. J. Vol. 37, 2000
Fig. 26. Summary of formats for presenting unsaturated soil
property data.
Fig. 25. Comparison between experimental and predicted soil-
water characteristic curves for sand.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:18 AM
Color profile: Generic CMYK printer profile
Composite Default screen
Seepage through unsaturated soils
The nature of the coefficient of permeability function can
be visualized by comparing its shape to that of the soil-water
characteristic curve. Figure 29 shows typical plots of the
soil-water characteristic curve and the coefficient of
permeability function for a sand and a clayey silt. The coef-
ficient of permeability for both soils remains relatively con-
stant until the air-entry value of the soil is reached. Beyond
this point, the coefficient of permeability decreases rapidly
for both soils.
The function appears to remain essentially linear until the
residual suction of the soil is reached. Beyond this point, the
coefficient of permeability appears to remain essentially
constant but there are a lack of data to confirm this portion
of the function. The relationship between soil suction and
the coefficient of permeability of an unsaturated soil (i.e.,
permeability function) can be predicted with sufficient accu-
racy for many engineering problems through a knowledge of
the saturated coefficient of permeability and the soil-water
characteristic curve.
Childs and Collis-George (1950) proposed a statistical
model to predict the coefficient of permeability based on a
random variation of pore sizes. This model was modified by
Marshall (1958) and further modified by Kunze et al. (1968).
These calculations are performed by dividing the volumetric
water content versus suction relationship into several water-
content increments. This is equivalent to integration with re-
spect to volumetric water content.
The accuracy of the prediction of the coefficient of perme-
ability function depends not only on the closeness of the
best-fit curve to the experimental soil-water characteristic
curve data, but also on the prediction model adopted.
Mualem (1986) concluded that there was no single model
that fits every soil type. The proposed models have been
found to be most satisfactory for sandy soils, whereas agree-
ment with experimental data may prove to be unsatisfactory
2000 NRC Canada
Fredlund 981
Fig. 27. Various formats for inputting tabular data when analyzing a problem.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:20 AM
Color profile: Generic CMYK printer profile
Composite Default screen
for finer grained soils. The accuracy of the permeability
function prediction can be improved by incorporating a cor-
rection factor for tortuosity,
q
(Mualem 1986). The integra-
tion form of the permeability function is as follows:
[28] k
w w
w y
w w
w
q
y
y
y
y
y
( )
( )
( ) ( )
( )
( )
ln( )

e
e
e d
e
e
r
s
( )
ln( )
e
aev
r
y

where
e is the natural number 2.71828 ;
k( ) is the coefficient of permeability at any soil suction;
y is a variable of integration representing the logarithm of
section;
w is the derivative of the soil-water characteristic curve;
is the variable of integration representing soil suction;

aev
is the soil suction corresponding to the air-entry value;

r
is the soil suction corresponding to residual water con-
tent; and
q is a correction factor to account for tortuosity and produce
close fits between experimental data and the theory.
The value of the power, q, can be assumed to be 1 unless
there is reason to assume otherwise (Kunze et. al. 1968).
The numerical integration of eq. [28] can be performed us-
ing the procedure of numerical integration presented by
Fredlund et al. (1994). A series of data points are computed
that can be placed into the form of a table of soil suction
versus coefficient of permeability. This information can be
used in one of the forms suggested in Fig. 27.
Leong and Rahardjo (1997b) showed that results similar
to the above integration procedure (i.e., eq. [28]), can be ob-
tained using the Fredlund and Xing (1994) equation for the
soil-water characteristic curve along with one additional fit-
ting parameter. The suggested extension to the soil-water
characteristic curve equation is as follows:
[29] k C
k
a
n
mq
( ) ( )
ln

_
,

_
,

1
]
1
1
s
e
The soil parameters in the permeability function are now
the same as those in the soil-water characteristic curve. The
fitting parameter, q, can be assumed equal to 1 for silts and
sands. Comparisons between predictions and measured data
sets will provide a little understanding of the parameter q.
The procedure suggested by Leong and Rahardjo (1997b)
provides a form similar to that used for shear strength and
allows the user to more clearly visualize the role of the soil-
water characteristic curve. In each case, the soil parameters
for the soil-water characteristic curve are used in the unsatu-
rated soil property function.
The water storage function is required in conjunction with
the coefficient of permeability function when performing a
transient flow analysis in an unsaturated soil. The water stor-
age function can be defined as the slope of the soil-water
characteristic curve. The differentiation of the equation for
2000 NRC Canada
982 Can. Geotech. J. Vol. 37, 2000
Fig. 29. Typical permeability functions for a sand and a clayey
silt with the suction plotted on a logarithmic scale.
Fig. 28. Relationship between soil-water characteristic curve and
shear strength for a sand and a clayey silt.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:26 AM
Color profile: Generic CMYK printer profile
Composite Default screen
the soil-water characteristic curve yields the water storage
function for an unsaturated soil. Figure 30 shows a
soil-water characteristic curve along with a plot of the water
storage function. The water storage function becomes highly
nonlinear as the soil desaturates.
Ground surface moisture flux boundary
conditions
Discussions regarding the implementation of unsaturated
soil mechanics would be incomplete without some mention
of the paramount role to be played by the ground surface
interaction with the atmosphere. The solutions require real-
time moisture and temperature flux boundary conditions.
The dynamic nature of climatic conditions makes the
ground surface flux conditions difficult to quantify. How-
ever, the climate must be quantified, since design conditions
require answers to questions that are strongly controlled by
weather conditions. The performance of a soil-cover system
is predominantly controlled by climatic conditions (Wilson
1998). A soil-cover system can be specifically designed to
accommodate virtually any type of climatic conditions. The
analytical tools required for the design of a soil-cover sys-
tem involve the coupling of two nonlinear partial differential
equations. The moisture-flow portion of the analysis is sub-
jected to a changing net moisture flux condition. One of the
partial differential equations is for computing moisture flux
and it is coupled with a partial differential equation for ther-
mal analysis. Design technologies for soil-cover systems that
have emerged over the past one or two decades provide an
impressive example of how the climate conditions can be
quantified and, along with unsaturated soil properties, yield
the primary information required in soil-cover design. Soil-
cover design procedures have been established for a number
of climatic conditions, namely (i) arid climates, (ii) temper-
ate regions, (iii) high-rainfall terrains, and (iv) tropical mon-
soon conditions with wet and dry seasons (Wilson 1998). In
all cases a coupled soilatmospheric numerical model (i.e.,
SoilCover; see Wilson 1997) implements the theoretical as-
pects of heat and mass flow (Wilson et al. 1994, 1997).
The theoretical analysis for the design of soil-cover sys-
tems requires two key pieces of information: (i) an assess-
ment of net moisture flux conditions, and (ii) an assessment
of the soil-water characteristic curve information for each
soil involved in the problem. In the case of a cover for a rec-
lamation of a tailings site, a soil-water characteristic curve is
required for the tailing material and for potential cover mate-
rials. The key characteristic of the soil-water characteristic
curve that must be examined are (i) the air-entry value, and
(ii) the residual conditions. These states are defined in terms
of a water content and a soil suction stress state. Figure 31
shows the soil-water characteristic curves for the tailings and
the proposed cover material to be used at an arid climatic
site (Wilson 1998). The upward rate of evaporation was to
be controlled by the colluvium, which had an air-entry value
2000 NRC Canada
Fredlund 983
Fig. 31. Cover design and associated soil-water characteristic
curves for a tailings impoundment in an arid climatic environ-
ment (Wilson 1998).
Fig. 30. Typical coefficient of water storage, m
2
w
, function com-
puted from the slope of the soil-water characteristic curve for a
coarse sand.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:30 AM
Color profile: Generic CMYK printer profile
Composite Default screen
of about 1 kPa and a residual soil suction of 10 kPa. These
proved to be excellent soil properties to minimize the
upward migration of water from the tailings in an arid envi-
ronment.
The quantification of the moisture flux boundary condi-
tions arises from a study of the water balance at the ground
surface. Precipitation enters the soil surface through the pro-
cess of infiltration. Precipitation must be quantified through
the statistical analysis of weather station data. Alternatively,
soil water leaves the soil surface as water vapor through the
process of evaporation and evapotranspiration. The actual
evaporative flux can be computed based on a soilatmo-
spheric model (e.g., SoilCover; see Wilson 1997) that re-
quires thermal data input in addition to information on the
soil-water characteristic curve for the soil. This is the pri-
mary input information required; however, other data are
also required.
Surface hydrologists have mainly focused on predicting
potential evaporation from a water surface (Hillel 1980), but
it is the actual evaporation from the soil surface that is re-
quired when assessing the ground surface moisture flux. It is
also the actual evaporative flux that controls the negative
pore-water pressure profile below the ground surface, as
shown in Fig. 32.
Wilson (1990) provided experimental evidence to confirm
that the rate of evaporation from a soil surface was uniquely
related to the suction in the soil. This relationship proved to
be essentially unique for sand, silt, and clay soils (Fig. 33).
The finding provided important confirmation for the devel-
opment of the theoretical soilatmosphere numerical model.
Figure 34 shows a plot of the components of ground sur-
face flux for a monsoon drywet climate setting (Wilson
1998). The figure shows the difference between the potential
and actual evaporative flux components. The analysis is for
the case of bare tailings. There is considerably more infor-
mation that can be obtained from the design procedure, but
the focus remains on the key elements of input for design.
In summary, it is the soil-water characteristic curves that
become the primary information required to assess the actual
evaporative moisture flux and the storage and hydraulic con-
ductivity characteristics of the cover materials. All materials
involved in the design must be analyzed as unsaturated soils
and appropriate unsaturated soil property functions need to
be assessed. It is important to be able to characterize the un-
saturated soil property functions in an economical and rea-
sonably accurate manner for engineering design purposes.
Steps towards implementation of
unsaturated soil mechanics
The primary stages leading towards the implementation of
unsaturated soil mechanics into geotechnical engineering
were outlined at the start of the paper. Substantial research
has been undertaken at all stages and the emphasis is for
engineers to take advantage of the potential for applying
unsaturated soil mechanics to all classes of geotechnical
2000 NRC Canada
984 Can. Geotech. J. Vol. 37, 2000
Fig. 32. Illustration of the effect of climate conditions on the pore-water pressure profile near the ground surface.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:35 AM
Color profile: Generic CMYK printer profile
Composite Default screen
problems. The soil-water characteristic curve, even with its
limitations and assumptions, appears to hold the key to an
economical implementation of unsaturated soil mechanics.
Further research is required on many aspects of unsatu-
rated soil mechanics. Listed below are suggested areas of re-
search and other needs that would encourage the
implementation of unsaturated soil mechanics.
(1) There are limitations in unsaturated soil theories at
various stages leading towards implementation. Research
should be directed towards better understanding and clarify-
ing the issues involved.
(2) The research community needs to encourage and facil-
itate the application of research findings in unsaturated soil
mechanics wherever possible.
(3) The observational method needs to be promoted
more often when working with unsaturated soils. Monitoring
of the performance of structures will advance our under-
standing and assume a greater confidence level in unsatu-
rated soil mechanics.
(4) There needs to be more technical training and a general
upgrade in the understanding of unsaturated soil behavior.
Along with the training should come a clearer understanding
of the array of potential applications in geotechnical engi-
neering.
(5) There needs to be a focus on research related to imple-
mentation procedures. For example, the procedures need to
be defined for the assessment of climatic conditions as it
pertains to establishing moisture flux boundary conditions.
The same is true for procedures related to the assessment of
unsaturated soil property functions and design criteria.
(6) There needs to be a focus on various classes of
geotechnical problems and the role of unsaturated soil me-
chanics. Some classes of problems are slope stability, cou-
pled stress and seepage analyses, and the prediction of heave
and collapse. There also need to be studies devoted to spe-
cific near-ground-surface infrastructure elements, such as
pipelines, roads, and sidewalks. Unsaturated soil mechanics
has an important role to play in the engineering of all struc-
tures forming a part of the infrastructure.
Acknowledgements
It is indeed an honor and a great privilege to be chosen to
deliver the R.M. Hardy Lecture at the 52nd Canadian
Geotechnical Conference in Regina, Saskatchewan. It is co-
incidental that upon completing my Masters degree at the
University of Alberta in 1964 my first job involved working
for the geotechnical engineering consulting company, R.M.
Hardy and Associates Ltd., in Edmonton, Alberta. During
my 2 years at R.M. Hardy and Associates Ltd. I had the op-
portunity to work with Dr. Bob Hardy on several engineer-
ing projects. I have greatly admired Dr. Hardy as a
geotechnical engineer and researcher and feel honored to
have had the opportunity to work with him and to be chosen
to deliver the R.M. Hardy Lecture. I would also like to thank
Ms. Noshin Zakerzadeh for her assistance in preparing this
paper.
References
Ahuja, L.R., Naney, J.W., and Williams, R.D. 1985. Estimating
soil-water characteristics from simple properties and limited
data. Soil Science Society of America Journal, 49: 11001105.
Arya, L.M., and Paria, 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.
Barbour, S.L. 1998. Nineteenth Canadian Geotechnical Collo-
quium: The soil-water characteristic curve: a historical perspec-
tive. Canadian Geotechnical Journal, 35: 873894.
Blight, G.E. 1965. A study of effective stresses for volume change.
In Moisture equilibria and moisture changes in soils beneath
covered areas (A symposium-in-print). Butterworths, Australia,
pp. 259269.
Burdine, N.T. 1953. Relative permeability calculations from pore
size distribution data. Journal of Petroleum Technology, 5: 71
78.
Childs, E.C., and Collis-George, N. 1950. The permeability of po-
rous materials. Proceedings of the Royal Society of London, Se-
ries A, 201: 392405.
Clifton, A.W., Wilson, G.W., and Barbour, S.L. 1999. The emer-
gence of unsaturated soil mechanics: Fredlund Volume. NRC
Research Press, National Research Council Canada, Ottawa,
Ont.
Fredlund, D.G. 1979. Second Canadian Geotechnical Colloquium:
Appropriate concepts and technology for unsaturated soils. Ca-
nadian Geotechnical Journal, 16: 121139.
Fredlund, D.G. 1992. Theory and research related to the use of
thermal conductivity sensors for matric suction measurements.
In Advances in measurement of soil physical properties; bring-
ing theory to practice. Soil Science Society of America, Special
2000 NRC Canada
Fredlund 985
Fig. 33. Relationship between soil suction and the ratio of the
actual evaporation to the potential evaporation from the ground
surface (Wilson et al. 1994).
Fig. 34. Water balance analysis using SoilCover for bare tailings
in a tropical wetdry climatic setting (Wilson 1998).
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:41 AM
Color profile: Generic CMYK printer profile
Composite Default screen
2000 NRC Canada
986 Can. Geotech. J. Vol. 37, 2000
Publication 30. Edited by G.C. Topp, W.D. Reynolds, and R.E.
Green. pp. 249262.
Fredlund, D.G. 1995. The prediction of unsaturated soil property
functions using the soil-water characteristic curve. In Proceed-
ings of the Bengt Broms Symposium in Geotechnical Engi-
neering, Singapore, 1316 Dec., pp. 113133.
Fredlund, D.G. 1996. Microcomputers and saturatedunsaturated
continuum modelling in geotechnical engineering. In Sympo-
sium on Computers in Geotechnical Engineering, INFOGEO
96, Sao Paulo, Brazil, 2830 Aug., Vol. 2, pp. 2950.
Fredlund, D.G. 1998. Bringing unsaturated soil mechanics into en-
gineering practice. Keynote address. In Second International
Conference on Unsaturated Soil Mechanics. International Aca-
demic Publishers, Beijing, China, Vol. 2, pp. 136.
Fredlund, D.G., and Morgenstern, N.R. 1976. Constitutive relations
for volume change in unsaturated soils. Canadian Geotechnical
Journal, 13: 261276.
Fredlund, D.G., and Morgenstern, N.R. 1977. Stress state variables
for unsaturated soils. Journal of the Geotechnical Engineering
Division, ASCE, 103(GT5): 447466.
Fredlund, D.G., and Rahardjo, H. 1993. Soil mechanics for unsatu-
rated soils. John Wiley & Sons Inc., New York.
Fredlund, D.G., and Xing, A. 1994. Equations for the soil-water
characteristic curve. Canadian Geotechnical Journal, 31: 521
532.
Fredlund, D.G., Morgenstern, N.R., and Widger, R.A. 1978. Shear
strength of unsaturated soils. Canadian Geotechnical Journal,
15: 313321.
Fredlund, D.G., Rahardjo, H., and Gan, J.K-M. 1987. Non-linearity
of strength envelope for unsaturated soils. In Proceedings of the
6th International Conference on Expansive Soils, New Delhi, In-
dia, 14 Dec., Vol. 1, pp. 4956.
Fredlund, D.G., Xing, A., and Huang, S. 1994. Predicting the per-
meability function for unsaturated soil using the soil-water char-
acteristic curve. Canadian Geotechnical Journal, 31: 533546.
Fredlund, M.D. 1997. SoilVision users manual. Version 1.3.
SoilVision Systems Ltd., Saskatoon, Sask.
Fredlund, M.D., Sillers, W.S., Fredlund, D.G., and Wilson, G.W.
1996. Design of a knowledge-based system for unsaturated soil
properties. In Proceedings of the 3rd Canadian Conference on
Computing in Civil Engineering, Montral, Que., 2628 Aug.,
pp. 659677.
Fredlund, M.D., Fredlund, D.G., and Wilson, G.W. 1997. Predic-
tion of the soil-water characteristic curve from grain size distri-
bution and volume-mass properties. In Proceedings of the 3rd
Brazilian Symposium on Unsaturated Soils, NONSAT 97, Rio
de Janeiro, 2225 Apr., Vol. 1, pp. 1323.
Gan, J.K.M., Fredlund, D.G., and Rahardjo, H. 1988. Determina-
tion of the shear strength parameters of an unsaturated soil using
the direct shear test. Canadian Geotechnical Journal, 25: 500
510.
Gardner, W.R. 1958. Some steady state solutions of the unsaturated
moisture flow equation with application to evaporation from a
water table. Soil Science, 85(4): 228232.
Gupta, S.C., and Larson, W.E. 1979. A model for predicting pack-
ing density of soils using particle-size distribution. Soil Science
Society of Americal Journal, 43: 758764.
Haverkamp, R., and Parlange, J.Y. 1986. Predicting the water-
retention curve from a particle-size distribution: 1. Sandy soil
without organic matter. Soil Science, 142(6): 325339.
Hillel, D. 1980. Applications of soil physics. Academic Press, New
York.
Huang, S.Y., Barbour, S.L., and Fredlund, D.G. 1998. Develop-
ment and verification of a coefficient of permeability function
for a deformable, unsaturated soil. Canadian Geotechnical Jour-
nal, 35: 411425.
Kunze, R.J., Mehara, G., and Graham, K. 1968. Factors important
in the calculation of hydraulic conductivity. Soil Science Society
of America Proceedings, 32: 760765.
Leong, E.C., and Rahardjo, H. 1997a. Review of soil-water charac-
teristic curve functions. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 123(12): 11061117.
Leong, E.C., and Rahardjo, H. 1997b. Permeability functions for
unsaturated soils. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 123(12): 11181126.
MacSyma Inc. 1996. PDEase users manual. 3rd ed. MacSyma
Inc., Arlington, Mass.
Marshall, T.J. 1958. A relation between permeability and size dis-
tribution of pores. Journal of Soil Science, 9: 18.
Matyas, E.L., and Radhakrishna, H.S. 1968. Volume change char-
acteristics of partially saturated soils. Gotechnique, 18(4): 432
448.
Maulem, Y. 1976. A new model for predicting the hydraulic con-
ductivity of unsaturated porous medial. Water Resources Re-
search, 12: 513522.
Mualem, Y. 1986. Hydraulic conductivity of unsaturated soils: pre-
diction and formulas. In Methods of soil analysis. No. 9, Part 1.
Physical and mineralogical methods. Edited by A. Klute. 2nd ed.
American Society of Agronomy, Madison, Wis., pp. 799823.
PDE Solutions Inc. 1999. FlexPDE users manual. Version 2.12,
June 11. PDE Solutions Inc., Fremont, Calif.
Peck, R.B. 1969. Advantages and limitations of the observational
method in applied soil mechanics. Gotechnique, 19(2): 171
187.
Sillers, W.S. 1997. The mathematical representation of the soil-
water characteristic curve. M.Sc. thesis, Department of Civil En-
gineering, University of Saskatchewan, Saskatoon, Sask.
Van Genuchten, M.T. 1980. A closed-form equation for predicting
the hydraulic conductivity of unsaturated soils. Soil Science So-
ciety of America Journal, 44: 892898.
White, N.F., Duke, H.R., Sunada, D.K., and Corey, A.T. 1970.
Physics of desaturation in porous materials. Journal of the Irri-
gation and Drainage Division, ASCE, 96(IR2): 165191.
Wilson, G.W. 1990. Soil Evaporative fluxes for geotechnical engi-
neering problems. Ph.D. dissertation, Department of Civil Engi-
neering, University of Saskatchewan, Saskatoon, Sask.
Wilson, G.W. 1997. SoilCover users manual. Unsaturated Soils
Group, University of Saskatchewan, Saskatoon, Sask.
Wilson, G.W. 1998. A review of cover design technologies for the
reclamation of tailings. In Proceedings of the Workshop on Fu-
ture Directions in Tailings Environmental Management, Perth,
Australia, Sponsored by Australian Centre for Mining Environ-
mental Research and the Australian Centre for Geomechanics,
1415 Sept. In press.
Wilson, G.W., Fredlund, D.G., and Barbour, S.L. 1994. Coupled
soilatmosphere modelling for soil evaporation. Canadian
Geotechnical Journal, 31: 151161.
Wilson, G.W., Fredlund, D.G., and Barbour, S.L. 1997. The effect
of soil suction on evaporative fluxes from soil surfaces. Cana-
dian Geotechnical Journal, 34: 145155.
I:\cgj\Cgj37\Cgj05\T00-026.vp
Monday, October 02, 2000 11:03:42 AM
Color profile: Generic CMYK printer profile
Composite Default screen