Delwyn G. Fredlund1
Abstract: Unsaturated soil mechanics has rapidly become a part of geotechnical engineering practice as a result of solutions that have
emerged to a number of key problems 共or challenges兲. The solutions have emerged from numerous research studies focusing on issues that
have a hindrance to the usage of unsaturated soil mechanics. The primary challenges to the implementation of unsaturated soil mechanics
can be stated as follows: 共1兲 The need to understand the fundamental, theoretical behavior of an unsaturated soil; 共2兲 the formulation of
suitable constitutive equations and the testing for uniqueness of proposed constitutive relationships; 共3兲 the ability to formulate and solve
one or more nonlinear partial differential equations using numerical methods; 共4兲 the determination of indirect techniques for the
estimation of unsaturated soil property functions, and 共5兲 in situ and laboratory devices for the measurement of a wide range of soil
suctions. This paper explains the nature of each of the previous challenges and describes the solutions that have emerged from research
studies. Computer technology has played a major role in achieving practical geotechnical engineering solutions. Computer technology has
played an important role with regard to the estimation of unsaturated soil property functions and the solution of nonlinear partial
differential equations. Breakthroughs in the in situ and laboratory measurement of soil suction are allowing unsaturated soil theories and
formulations to be verified through use of the “observational method.”
DOI: 10.1061/共ASCE兲1090-0241共2006兲132:3共286兲
CE Database subject headings: Unsaturated soils; Soil mechanics; Geotechnical engineering; Research.
Preamble unsaturated soil behavior are truly commendable and still worthy
of study.
Karl Terzaghi is remembered most for providing the “effective Subsequent reference to the textbook Theoretical Soil Mechan-
stress” variable, 共 − uw兲, that became the key to describing the ics during my career, has caused me to ask the question, “Why
mechanical behavior of saturated soils; where = total stress and did unsaturated soil mechanics not emerge simultaneously with
uw = pore–water pressure. The effective stress variable became the saturated soil mechanics?” Pondering this question has led me to
unifying discovery that elevated geotechnical engineering to a realize that there were several theoretical and practical challenges
science basis and context. associated with unsaturated soil behavior that needed further re-
As a graduate student I was asked to purchase and study the search. Unsaturated soil mechanics would need to wait several
textbook, Theoretical Soil Mechanics, by Karl Terzaghi 共1943兲. I decades before it would take on the character of a science that
had already selected the subject of unsaturated soil behavior as could be used in routine geotechnical engineering practice.
I am not aware that Karl Terzaghi ever proposed a special
my field of research and was surprised to find considerable infor-
description of the stress state in an unsaturated soil; however,
mation on this subject in this textbook. Two of the 19 chapters of
his contemporary, Biot 共1941兲, was one of the first to suggest the
the textbook contribute extensively toward understanding unsat-
use of two independent stress state variables when formulating
urated soil behavior; namely, Chapter 14 on “Capillary Forces,”
the theory of consolidation for an unsaturated soil. This paper
and Chapter 15, on “Mechanics of Drainage” 共with special atten-
will review a series of key theoretical extensions that were
tion to drainage by desiccation兲. These chapters emphasize the
required for a more thorough representation and formulation of
importance of the unsaturated soil portion of the profile and in
unsaturated soil behavior.
particular provide an insight into the fundamental nature and Research within the agriculture-related disciplines strongly
importance of the air–water interface 共i.e., contractile skin兲. influenced the physical and hydraulic model that Terzaghi
Considerable attention was given to soils with negative pore– developed for soil mechanics 共Baver 1940兲. With time, further
water pressures. Fig. 1 shows an earth dam illustrating how significant contributions have come from the agriculture-related
water flowed above the phreatic line through the capillary zone disciplines 共i.e., soil science, soil physics, and agronomy兲 to geo-
共Terzaghi 1943兲. The contributions of Karl Terzaghi toward technical engineering. It can be said that geotechnical engineers
tended to test soils by applying total stresses to soils through
1
Professor Emeritus, Dept. of Civil and Geological Engineering, Univ. the use of oedometers and triaxial cells. On the other hand,
of Saskatchewan, Saskatoon SK, Canada S7N 5A9. agriculture-related counterparts tended to apply stresses to the
Note. Discussion open until August 1, 2006. Separate discussions water phase 共i.e., tensions兲 through use of pressure plate cells.
must be submitted for individual papers. To extend the closing date by Eventually, geotechnical engineers would realize the wealth
one month, a written request must be filed with the ASCE Managing
of information that had accumulated in the agriculture-related
Editor. The manuscript for this paper was submitted for review and pos-
sible publication on February 16, 2005; approved on May 1, 2005. This disciplines; information of value to geotechnical engineering.
paper is part of the Journal of Geotechnical and Geoenvironmental Careful consideration would need to be given to the test proce-
Engineering, Vol. 132, No. 3, March 1, 2006. ©ASCE, ISSN 1090-0241/ dures and testing techniques when transferring the technology
2006/3-286–321/$25.00. into geotechnical engineering.
The zone between the ground surface and the water table swelling soils, collapsible soils, and residual soils. Any of the
is generally referred to as the unsaturated soil zone. This is some- above-mentioned soils, as well as other soil types, can also be
what of a misnomer since the capillary fringe is essentially compacted, once again giving rise to a material with negative
saturated. A more correct term for the entire zone above the water pore–water pressures.
table is the vadose zone 共Bouwer 1978兲. The entire zone sub-
jected to negative pore–water pressures is commonly referred
to as the unsaturated zone in geotechnical engineering. The Unsaturated Soil as a Four-Phase Mixture
unsaturated zone becomes the transition between the water in
the atmosphere and the groundwater 共i.e., positive pore–water An unsaturated soil is commonly referred to as a three-phase
pressure zone兲. mixture 共i.e., solids, air, and water兲 but there is strong justification
The pore–water pressures in the unsaturated soil zone can for including a fourth independent phase called the contractile
range from zero at the water table to a maximum tension of skin or the air–water interface. The contractile skin acts like a thin
approximately 1,000,000 kPa 共i.e., soil suction of 1,000,000 kPa兲 membrane interwoven throughout the voids of the soil, acting as a
under dry soil conditions 共Croney et al. 1958兲. The water degree partition between the air and water phases. It is the interaction of
of saturation of the soil can range from 100% to zero. The the contractile skin with the soil structure that causes an unsatur-
changes in soil suction result in distinct zones of saturation. The ated soil to change in volume and shear strength. The unsaturated
zones of saturation have been defined in situ as well as in the soil properties change in response to the position of the contrac-
laboratory 关i.e., through the soil–water characteristic curve 共Fig. tile skin 共i.e., water degree of saturation兲. It is important to view
3兲兴. Table 1 illustrates the terminologies commonly used to
describe saturation conditions in situ and in the laboratory. Soils
in situ start at saturation at the water table and tend to become
unsaturated toward the ground surface.
Soils near to the ground surface are often classified as “prob-
lematic” soils. It is the changes in the negative pore–water
pressures that can result in adverse changes in shear strength and
volume change. Common problematic soils are: expansive or
Fig. 4. Density distribution across the contractile skin reprinted from Designation of the Stress State
Liquid–Fluid Interface, Vol. 3 of Fundamental of Interface and
Colloid Science, J. Lyklema 共2000兲, with permission from Elsevier State variables can be defined within the context of continuum
mechanics as variables independent of soil properties required
for the characterization of a system 共Fung 1965兲. The stress
an unsaturated soil as a four-phase mixture for purposes of stress state variables associated with an unsaturated soil are related
analysis, within the context of multiphase continuum mechanics. to equilibrium considerations 共i.e., conservation of energy兲 of a
Consequently, an unsaturated soil has two phases that flow under multiphase system. The stress state variables form one or more
the influence of a stress gradient 共i.e., air and water兲 and two tensors 共i.e., 3 ⫻ 3 matrix兲 because of the three-dimensional
phases that come to equilibrium under the influence of a stress Cartesian coordinate system generally used for the formulation of
gradient 共i.e., soil particles forming a structural arrangement and engineering problems 共i.e., a three-dimensional world兲.
the contractile skin forming a partition between the fluid phases兲 The description of the state variables for an unsaturated
共Fredlund and Rahardjo 1993兲. soil becomes the fundamental building block for an applied
The contractile skin has physical properties differing from engineering science. The universal acceptance of unsaturated soil
the contiguous air and water phases and interacts with the soil mechanics depends largely upon how satisfactorily the stress state
structure to influence soil behavior. The contractile skin can be variables can be defined, justified, and measured. Historically,
considered as part of the water phase with regard to changes it has been the lack of certainty regarding the description of
in volume–mass soil properties but must be considered as an the stress state for an unsaturated soil that has been largely
independent phase when describing the stress state and phenom- responsible for the slow emergence of unsaturated soil mechanics.
enological behavior of an unsaturated soil. Terzaghi 共1943兲 Biot 共1941兲 was probably the first to suggest the need for two
emphasized the important role played by surface tension effects independent stress state variables for an unsaturated soil. This is
associated with the air–water interface 共i.e., contractile skin兲. evidenced from the stress versus deformation relations used in the
Distinctive Features of the Contractile Skin: Numerous derivation of the consolidation theory for unsaturated soils. Other
research studies on the nature of the contractile skin point toward researchers began recognizing the need to use two independent
its important, independent role in unsaturated soil mechanics. stress state variables for an unsaturated soil as early as the 1950s.
Terzaghi 共1943兲 suggested that the contractile skin might be in the This realization can be observed through the three-dimensional
order of 10−6 mm in thickness. More recent studies suggest that plots of the volume change constitutive surfaces for an unsatur-
the thickness of the contractile skin is in the order of 1.5–2 water ated soil 共Bishop and Blight 1963; Matyas and Radakrishna
molecular diameters 共i.e., 5 Å兲 共Israelachvili 1991; Townsend and 1968兲. It was during the 1970s that a theoretical basis and justi-
Rice 1991兲. fication was provided for the use of two independent stress state
A surface tension of approximately 75 mN/ m translates into a variables 共Fredlund and Morgenstern 1977兲. The justification was
unit stress in the order of 140,000 kPa. Lyklema 共2000兲 showed based on the superposition of coincident equilibrium stress fields
that the distribution of water molecules across the contractile skin for each of the phases of a multiphase system, within the context
takes the form of a hyperbolic tangent function as shown in Fig. of continuum mechanics. From a continuum mechanics stand-
4. Properties of the contractile skin are different from that of point, the representative element volume 共REV兲 must be suffi-
ordinary water and have a water molecular structure similar to ciently large such that the density function associated with each
that of ice 共Derjaguin and Churaev 1981; Matsumoto and Kataoka phase is a constant. It should be noted that it is not necessary for
1988兲. all phases to be continuous but rather that the REV statistically
The Young–Laplace and Kelvin equations describe fundamen- represents the multiphase system. Although the stress analysis had
tal behavioral aspects of the contractile skin but both equations little direct application in solving practical problems, it helped
have limitations. The Young–Laplace equation is not able to unite researchers on how best to describe the stress state of an
explain why an air bubble can gradually dissolve in water without unsaturated soil.
any apparent difference between the air pressure and the water Three possible combinations of independent stress state vari-
pressure. The validity of the Kelvin equation becomes suspect as ables were shown to be justifiable from the theoretical continuum
the radius of curvature reduces to the molecular scale 共Adamson mechanics analysis. However, it was the net normal stress 关i.e.,
and Gast 1997; Christenson 1988兲. − ua, where = total net normal stress and ua = pore–air pressure兴
Terzaghi 共1943兲 recognized the limitations of the Kelvin equa- and the matric suction 共i.e., ua − uw, where uw = pore–water pres-
tion and stated that if the radius of a gas bubble “approaches zero, sure兲 combination of stress state variables that proved to be the
the gas pressure … approaches infinity. However, within the easiest to apply in engineering practice. The net normal stress
range of molecular dimensions,” the equation “loses its validity.” primarily embraces the activities of humans which are dominated
冤 冥
共x − ua兲 yx zx an unsaturated soil. Morgenstern 共1979兲 explained the limitations
of Bishop’s effective stress equation as follows:
xy 共y − ua兲 zy 共1兲
• Bishop’s effective stress equation “… proved to have little
xz yz 共z − ua兲 impact on practice. The parameter, , when determined for
and volume change behavior was found to differ when determined
for shear strength. While originally thought to be a function of
冤 冥
共ua − uw兲 0 0 degree of saturation and hence bounded by 0 and 1, experi-
0 共ua − uw兲 0 共2兲 ments were conducted in which was found to go beyond
these bounds.
0 0 共ua − uw兲
• The effective stress is a stress variable and hence related to
where x, y, and z = total stresses in the x, y, and z directions, equilibrium considerations alone.”
respectively; uw = pore–water pressure; and ua = pore–air pressure. Morgenstern 共1979兲 went on to explain:
The stress tensors contain surface tractions that can be placed • Bishop’s effective stress equation “… contains the parameter,
on a cube to represent the stress state at a point 共Fig. 5兲. The stress , that bears on constitutive behavior. This parameter is found
tensors provide a fundamental description of the stress state for an by assuming that the behavior of a soil can be expressed
unsaturated soil. It has also been shown 共Barbour and Fredlund uniquely in terms of a single effective stress variable and
1989兲 that osmotic suction forms another independent stress by matching unsaturated soil behavior with saturated soil be-
tensor when there are changes in salt content of either a saturated havior in order to calculate . Normally, we link equilibrium
or unsaturated soil. All the stress state variables are independent considerations to deformations through constitutive behavior
of soil properties and become the “keys” to describing physical and do not introduce constitutive behavior into the stress state.
phenomenological behavior, as well as defining functional Another form of Bishop’s equation has been used by several
relationships for unsaturated soil properties. The inclusion of soil researchers in the development of elastoplastic models 共Jommi
parameters at the stress state level is unacceptable within the 2000; Wheeler et al. 2003; Gallipoli et al. 2003兲.
context of continuum mechanics.
As a soil approaches saturation, the pore–air pressure, ua, *ij = ij − 关Swuw + 共1 − Sw兲ua兴␦ij 共5兲
becomes equal to the pore–water pressure, uw. At this point, the
two independent stress tensors revert to a single stress tensor that where ij = total stress tensor; ␦ij = Kroneker delta or substitution
can be used to describe the behavior of saturated soils: tensor; *ij = Bishop’s average soil skeleton stress; and Sw = water
冤 冥
degree of saturation.
共x − uw兲 yx zx
In this case, the water degree of saturation has been substituted
xy 共y − uw兲 zy 共3兲 for the soil parameter. The above-mentioned equation is once
xz yz 共z − uw兲 again empirical and constitutive in character. Consequently, the
冤 冥
xx ␥yx ␥zx techniques.
␥xy yy ␥zy 共7兲 The use of the axis-translation technique 共Hilf 1956兲 remains
the primary reference measurement for the direct measurement
␥xz ␥yz zz
of matric suction in the laboratory. However, this method of mea-
where xx, yy, and zz = longitudinal strain in the x, y, and z surement cannot be applied to in situ conditions. Conventional
directions, respectively; and ␥yx, ␥zx, ␥xy, ␥zy, ␥xz, and ␥yz = shear tensiometers can provide direct measurements of matric suction
strains on the x, y, and z planes. in the range below one atmosphere but are seldom of practical
The trace of the strain tensor 共i.e., xx + yy + zz兲 yields the value in geotechnical engineering because of the requirement
volumetric strain, v. The amount of air and water in the same for daily servicing. Matric suctions encountered in geotechnical
REV can be designated on a volume basis, a 共volumetric air engineering applications commonly range from 0 to 1,500 kPa.
content兲, and w 共volumetric water content兲. Recent developments have made it possible to obtain indirect
A variety of constitutive models can be derived to relate the measurements of matric suction up to several atmospheres over
state variables. The primary constitutive models required are indefinite periods of time 共Marjerison et al. 2001兲. The most com-
those that relate the volume–mass state variables to the stress mon indirect measurement technique involves the use of thermal
state variables, and those that relate normal stresses to shear conductivity or heat dissipation suction sensors 共Fredlund et al.
stresses. It is only natural that the volume–mass models should 2000a, b, c兲. Indirect measurements of matric suction have proven
range from those consistent with historical soil mechanics 共i.e., to provide stable, long-term in situ measurements.
using coefficient of compressibility and coefficient of volume Monitoring for verification purposes: The “Observational
change兲, to equivalent, incremental elasticity models 共i.e., using Approach” 共Peck 1969兲, has provided an excellent verification
Young’s modulus and Poisson’s ratio兲, 共Fredlund and Rahardjo framework for geotechnical engineering and the same approach
1993兲, to more recent elastoplastic models using and to rep- needs to be applied to unsaturated soil problems. Only in this way
resent the initial compression and the rebound-reloading compres- is it possible to develop confidence in the application of unsatur-
sion, respectively, on a semilog scale 共Alonso et al. 1990; ated soil theories. Matric suction is the primary measurement that
Wheeler and Sivakumar 1995; Blatz and Graham 2003兲. As well, can provide verification information on unsaturated soil behavior.
the shear strength models would be expected to range from Engineers must have access to devices that can be used to monitor
extensions of a Mohr–Coulomb representation of shear strength and thereby evaluate the adequacy of an engineering design.
共Fredlund et al. 1978兲 to critical state models within the context The measurement of vertical and lateral movements can also
of elastoplastic models 共Wheeler and Sivakumar 1995兲. be used for the verification of unsaturated soils theories. Most
dua
max = − Dax
dx
dua
may = − Day
dy
dua
maz = − Daz 共11兲
az
where max, may, and maz = mass flow rate in the x, y, and z
directions, respectively; and Dax, Day, and Daz = air diffusivity in
the x, y, and z directions, respectively.
The assumption is generally made that changes in atmospheric
air pressure are negligible. The air flow law can also be written as
a velocity of flow, va, similar to Darcy’s law 共Blight 1971兲,
thereby taking on the following form
dua
vax = − kax
dx
Fig. 12. Example of permeability and water storage functions
for an anisotropic soil dua
vay = − kay
dy
dua
Ability of an Unsaturated Soil for Storage vaz = − kaz 共12兲
dz
The simulation of transient flow processes 共i.e., water, air, or heat兲 The air coefficient of permeability, ka 共and air diffusivity兲 is
requires a characterization of a storage property that changes also a mathematical function in the sense that the transmission of
with the water degree of saturation of the soil. The storage soil air varies with the amount of air which, in turn, is controlled by
property is part of the partial differential equation describing a the soil–water characteristic curve. Fig. 13 shows a soil–water
transient process. characteristic curve for sand and illustrates the form of the air
The water storage soil property associated with water flow permeability function 共Ba-Te et al. 2005兲. The air permeability
through an unsaturated soil, mw2 , is equal to the arithmetic slope function takes on an inverse form to that of the water permeability
of the SWCC. The differentiation of any mathematical equation function. The air coefficient of permeability tends toward the
proposed for the SWCC can serve as a measure of the water diffusion of air through water below the air entry value of the
storage soil property. Fig. 12 shows the form of the water storage soil. Once the air entry value is exceeded, the air coefficient of
function for a sandy soil. There is strong nonlinearity correspond- permeability increases by several orders of magnitude. The low
ing to the point of inflection along the SWCC. The nonlinearity viscosity of air indicates that air can flow through a soil with
of the water storage soil property can give rise to numerical much greater ease than water.
instability and errors in computing water balances, if not properly
taken into account during the solution of the seepage partial
Heat Flow Constitutive Relations
differential equation. Since there is hysteresis in the SWCC, there
will also be independent water storage curves for the drying and The driving potential for heat flow is a temperature or thermal
wetting processes. gradient. Heat flow, qt, can be described using Fourier’s law
The air phase also has a storage term as well as a compress- which takes a similar form to Fick’s law. The soil property
ibility component in the partial differential equation. The air controlling conductive heat flow is thermal conductivity, t:
storage term takes the same form as the water storage function.
The air compressibility and air storage terms have similar effects dT
qtx = − t
on a transient air flow process. dx
Fig. 17. Shear strength results showing the relationship between the
SWCC and the peak shear strength of a soil 关data adapted from equations attempt to describe the shear strength of an unsaturated
Melinda et al. 共2004兲, ASCE兴 soil under critical state condition, in terms of q – p – s space. A set
of shear strength results for compacted kaolin is shown in Fig. 18
共Wheeler and Sivakumar 1995兲. The results show an increase
The linear form is more appropriate for limited ranges of in shear strength as the matric suction of the soil is increased.
matric suction. Some of the earlier unsaturated soils shear Specific volumes corresponding to critical state conditions were
strength data sets 共e.g., Bishop et al. 1960兲 show a close fit to the also presented. The stress state variables used to present the
linear equation 共Fredlund and Rahardjo 1993兲. The linear form is results are as follows:
also more convenient to use for shear strength solutions.
Most research programs related to the shear strength of unsat- q = 1 − 3
urated soils appear to have been undertaken on soils that were
initially compacted to an initial water content and density and 共1 + 2 + 3兲
then wetted such that the initial suction was allowed to come p=
3
toward a zero value 共Gan and Fredlund 1996兲. The soil specimens
are then subjected to a series of increasing matric suctions along
the desorption branch of the SWCC. Since there is hysteresis s = 共ua − uw兲 共16兲
between the drying and wetting curve it would be anticipated that The challenge has been to find a consistent means of incor-
soils may exhibit a different shear strength envelope if first porating the effect of matric suction, ua − uw into the shear
subjected to high matric suction conditions and then reduced to strength equation. A general form for the shear strength equation
a series of suction values along the wetting curve. Melinda et al. under critical state failure conditions can be written as follows:
共2004兲 reported the shear strength results on a residual soil from
Singapore tested along both the drying curve and the wetting q = M f 1关p − ua,ua − uw兴 共17兲
curve 共Fig. 17兲. The results showed that the measured shear
strengths along the drying curve are higher than those measured where M = a material characteristic independent of stresses; and
along the wetting curve. These results can be explained on the f 1 = an independent function of p − ua and ua − uw.
basis of the hysteresis of the SWCC that shows the matric suction The critical state shear strength model proposed by Alonso
having a greater cross-sectional area over which to act along the et al. 共1990兲 has the following form:
drying curve, for a specific suction. The difference in shear
strength between drying and wetting conditions appears to be q = M共p − ua兲 + 共ua − uw兲 共18兲
related to the magnitude of the drying and wetting hysteresis loop. where = a soil constant.
Shear strength equations formulated for saturated soils, within Jommi 共2000兲 combined the net mean stress and matric suc-
the context of critical state models, have also been extended to tion using the water degree of saturation and suggested a similar
unsaturated soil conditions. Several models have been proposed equation:
共Alonso et al. 1990; Wheeler and Sivakumar 1995; Toll 1990;
Maatouk et al. 1995; Blatz and Graham 2003兲. The proposed q = M关共p − ua兲 + Sr共ua − uw兲兴 共19兲
Fig. 30. Drying and wetting void ratio curves for a sand soil using a
pressure plate apparatus with volume change measurements 关Lins and
Fig. 28. Fifteen bar pressure plate apparatus for the measurement of Schanz 共2004兲; with kind permission of Springer Science and Busi-
volume–mass relations for unsaturated soils ness Media兴
再 冋 册冎
cohesionless soils and it is possible to obtain satisfactory test w共兲 = C共兲 共28兲
results using a simple column test where the distance above the 共ua − uw兲 n m
water level is converted into an equivalent matric suction value. ln e +
共ua − uw兲aev
Columns approximately 1 m in height have proven satisfactorily
when the air entry value of the soil is less than about 7 kPa and where w共兲 = water content at any soil suction; C共兲 = correction
residual conditions are above the height of the column. Both the factor directing all SWCC curves to 1,000,000 at zero water con-
drying and wetting curves can be obtained using column tests. tent; 共ua − uw兲aev = soil parameter indicating the inflection point
Fig. 32 shows the SWCC for three coarse sands tested using a that bears a relationship to the air entry value; n = soil parameter
column test for the wetting curve and a Tempe cell test for the related to the rate of desaturation; and m = soil parameter related
drying curve 共Yang et al. 2004兲. to the curvature near residual conditions.
Brooks and Corey 共1964兲 ⌰n = 1 , 艋 aev aev = air entry value of the soil and bc = pore size distribution
冉 冊
index.
−bc
⌰n = , ⬎ aev
aev
冉冊
⌰n = entry value of the soil and nb = soil parameter which is
nb
primarily a function of the rate of water extraction from the
1+
ab soil, once the air entry value has been exceeded.
Laliberte 共1969兲 The parameters al, bl, and cl are assumed to be unique
冤 冉 冊冥
1 bl
⌰n = erfc al − functions of the pore-size distribution index, .
2
cl +
aev
冉 冊
Campbell 共1974兲 aev = air entry value of the soil and bc = a constant.
−1/bc
w = ws , 艌 aev
aev
w = ws , ⬍ aev
Van Genuchten 共1980兲 av = soil parameter which is primarily a function of air entry
1
冋 冉 冊册
⌰n = value of the soil 共1 / kPa兲; nv = soil parameter which is
nv mv
primarily a function of the rate of water extraction from the
1+
av soil, once the air entry value has been exceeded; and mv = soil
parameter which is primarily a function of the residual water
content.
Van Genuchten 共1980兲
1 1
冋 冉 冊册
⌰n = nv mv , mv = 1 −
nv
1+
av
冋 冉 冊册
⌰n = nv mv , mv = 1 −
nv
1+
av
冉 冊
⌰n =
− am
1 + exp
nm
some cases to use an average of the drying and wetting SWCCs tested in each case, showing reproducible and consistent results.
when estimating the unsaturated soil property functions. The air entry value for the sand was 2 kPa and the residual
It is the desorption curve that is easiest to measure and is suction was 13 kPa. The air entry value for the silt was 10 kPa
therefore most commonly measured in the laboratory. However, and the residual suction was 120 kPa.
desorption and adsorption curves have been measured on a num- Pham et al. 共2003a,b兲 analyzed the drying and wetting curves
ber of soils. Pham et al. 共2002兲 measured drying and wetting for 34 data sets on a variety of soils reported in the literature. The
curves on sand 共Fig. 33兲 and silt 共Fig. 34兲. Three specimens were primary intent of the study was to better understand the nature of
再 冋 冉 冊 册冎
Xing 共1994兲 w共兲 = C共兲 entry value of the soil; n f = soil parameter which is primarily
nf mf a function of the rate of water extraction from the soil, once
ln e + the air entry value has been exceeded; m f = soil parameter
af
which is primarily a function of residual water content; and
冉 冊
C共兲 = correction which is primarily a function of the suction
ln 1 + at which residual water content occurs.
r
冋 冉 冊册
C共兲 = 1 −
1,000,000
ln 1 +
r
the hysteresis loop. The difference between the hysteresis loops devices and matric suctions were measured using tensiometers.
at the inflection points was used as the primary indicator of The SWCC for each of the soils was independently measured
the magnitude of hysteresis. It was observed that the drying in the laboratory. Fig. 35 presents some of the model test results
bounding curve and the wetting bounding curves tended to be and clearly indicates that the SWCC relationships independently
approximately parallel. The distance between the main drying and measured in the laboratory provided a reasonable representation
wetting curves varied between 0.15 and 0.35 of a log cycle for of the water content versus matric suction curves followed under
sands. The spacing between the main drying and wetting curves in situ conditions 共Tami et al. 2004b兲.
for more well-graded loam soils varied between 0.35 and 0.60 of
a log cycle. The distances were found to be quite consistent for
Relationship between the Accuracy Required
individual textural soil types. Approximate values for the spacing
for the SWCC and the Engineering Analysis
between the drying and wetting curves can be assumed to be 0.25
being Performed
of a log cycle for sands and 0.50 of a log cycle for loams.
It is important to verify that the hysteresis observed in the Unsaturated soil property functions have been developed as
laboratory during wetting and drying occurs on a large scale extensions of saturated soil properties. It might seem that it would
in accordance with the estimation techniques that have been be difficult to obtain reasonable unsaturated soil property func-
proposed. Tami et al. 共2004b兲 constructed and instrumented a tions for solving unsaturated soil problems. However, this appears
2 m long model of a slope consisting of two cohesionless to not be the case for most unsaturated soil problems since it is
soils. The experiment was undertaken to study the performance only necessary to obtain an approximation of the unsaturated soil
of capillary barriers on slopes. A flux boundary condition properties 共Fredlund et al. 2003兲.
was imposed while water contents were measured using TDR
Fig. 33. Measured drying and wetting SWCCs on Beaver Creek Fig. 34. Measured drying and wetting SWCCs on Processed Silt
sand 关Pham 共2002兲; reprinted with permission of the University of 关Pham 共2002兲; reprinted with permission of the University of
Saskatchewan, Canada兴 Saskatchewan, Canada兴
Fig. 36. Approaches that can be used to obtain the soil–water characteristic curves
kr = 冉 冊
aev
−n
for ⬎ aev
Fig. 37. Soil–water characteristic curves computed from grain size
distribution curves for a variety of soil types Rijtema 共1965兲 kw = ks for 艋 aev 1 = residual soil suction
kr = exp关−␣共 − aev兲兴 and k1 = coefficient of
for 1 艋 ⬍ aev permeability at 1.
have been proposed 共Fredlund et al. 1997, 2002c兲. A mathemati- kw = k1共 / 1兲−n
cal equation similar to that used for describing a SWCC can be for ⬍ 1
best fit to a grain size distribution curve. The equation for the Note: kw = unsaturated water permeability coefficient; ks = saturated
grain size distribution curve is then used to compute a soil–water permeability coefficient; kr = kw / ks = relative permeability; = soil suction;
characteristic curve 共Fredlund et al. 1997兲. Fig. 37 shows a series aev = air entry value; and w = gravimetric soil water content.
of SWCCs that have been computed from grain size distribution
curves for several soil types. The results are encouraging for
sands and silts, but more research is required when using this
procedure for structured and clayey soils. model and the mathematical manner 共e.g., method of integration兲
Correlations between classification soil properties 共e.g., grain in which the SWCC is used in conjunction with the saturated soil
size distribution and Atterberg limits兲 can provide approximate properties. As an example, several permeability functions have
parameters for soil–water characteristic curve equations and these been proposed, each one using the soil–water characteristic curve
are satisfactory in many situations 共Aubertin et al. 2003; Zapata in a somewhat different manner.
et al. 2000兲. The fundamental constitutive relations previously presented in
this paper can be rewritten incorporating a soil–water character-
istic curve equation. The SWCC is also dependent upon the total
Incorporation of the SWCC into the Constitutive stress state; however, the constitutive equations for an unsaturated
Relations for Unsaturated Soils soil are sufficiently accurate when written simply as a function of
soil suction.
Comprehensive experimental research studies can be undertaken
to verify the uniqueness of proposed constitutive relations; how-
Water Seepage Constitutive Relations
ever, it is not always practical to follow the same test procedures
Written in Terms of SWCC
in routine geotechnical engineering practice for economic
reasons. This is particularly true of unsaturated soil behavior and Numerous mathematical procedures have been proposed for the
consequently, a variety of practical, indirect methodologies estimation of the liquid water permeability function, kw共兲. These
and test procedures have emerged. These indirect procedures models can be categorized as empirical equations and theoretical
have proven to be adequate for most geotechnical engineering equations derived as macroscopic and microscopic 共statistical兲
problems 共Fredlund 2000兲. models 共Mualem 1986兲.
Indirect procedures for the estimation of unsaturated soil Empirical equations describe the variation in the water coeffi-
property functions have primarily made use of the SWCC along cient of permeability with soil suction, kw共兲, 关or with volumetric
with the saturated soil properties 共Fredlund and Rahardjo 1993; water content, kw共兲兴. The parameters for the equations are
Vanapalli et al. 1996; Barbour 1998; Fredlund 2000; Fredlund generally determined using a curve-fitting procedure. Some of
et al. 2000兲. It is possible to measure the SWCC used in a number empirical equations along with an appropriate reference are given
of the commercially available apparatuses and the procedure has in Table 4. The Brooks and Corey 共1964兲 equation is considered
become a generally accepted means of estimating unsaturated soil to be both an empirical and a macroscopic model because
property functions. Various empirical procedures have been pro- elements of physics are used to relate pore size distribution to the
posed and tested for the estimation of essentially all unsaturated permeability function.
soil property functions 共e.g., permeability, shear strength, volume There are two different groups of theoretical models, 共i.e.,
change, and others兲. In each case, the estimation procedure in- macroscopic and microscopic approaches兲 based on the statistical
volves the use of the saturated soil properties in conjunction with assumptions regarding pore distributions and the interpretation
the SWCC. There are, for example, numerous estimation proce- applied to the soil–water characteristic curve. The macroscopic
dures that have been proposed for the water permeability function models provide an analytical, closed-form equation for the unsat-
共Fredlund and Xing 1994兲. The procedures differ primarily in the urated permeability function. All macroscopic models have the
basic assumptions involved in the development of the proposed following general form:
Permeability Models van Genuchten 共1980兲 Fredlund and Xing 共1994兲 Brooks and Corey 共1964兲 Campbell 共1974兲
Childs and Collis-George
共1950兲
—
冕
b
ln共兲
共ey兲 − 共兲
ey
⬘共ey兲dy
—
kr = 冉 冊
aev
−2共2/b兲
冕
kr = b
共ey兲 − s
⬘共ey兲dy
ln共aev兲
ey
2
m=1−
n
Mualem 共1976b兲 — — —
兵1 − 共␣兲n−1关1 + 共␣兲n兴−m其2
kr共兲 =
关1 + 共␣兲n兴0.5
1
m=1−
n
Note: k = unsaturated permeability coefficient; ks = saturated permeability coefficient; kr = k / ks = relative permeability; = soil suction; aev = air entry value;
= soil water content; s = saturated water content; b = ln共1,000,000兲; and y = dummy variable of integration representing the logarithm of integration.
冉 冊 冉 冊
Fredlund and Xing 共1994兲
冦冤 冥 冧 冦 冤冢 冣 冥冧
2
1/q q
ln 1 + ln 1 +
r 1 r 1
冉 冊 冋 冉 冉 冊 冊册 冉 冊 冋 冉 冉 冊 冊册
ka = kd 1− 1− 1− 1−
106 n m 106 n m
ln 1 + ln e + ln 1 + ln e +
r a r a
Note: Se = 共Sw − Sr兲 / 共1 − Sr兲; Se = water effective degree of saturation; Sr = water residual degree of saturation; q = variable determined using a least-squared
technique; ua − uw = soil suction; 共ua − uw兲b = air entry value; ka = air coefficient of premeability; and kd = dry soil–air coefficient of permeability.
Little research has been done regarding the form that the air-entry value is the point where air starts to enter the largest
permeability function should take once residual water content pores in the soil. Below this value, the airflow takes the form
conditions are reached. A recent study 共Ebrahimi-B et al. 2004兲 of air diffusion through the soil–water and the air coefficient of
has suggested that there should be a lower limit for the coefficient permeability becomes extremely small.
of water permeability and it is suggested that it be related to the The concept of a soil-air characteristic curve 共SACC兲 共i.e., the
rate of vapor diffusion. Fig. 38 shows the results of the proposed inverse of a soil–water characteristic curve兲, can be used to
method applied to a silt soil. The lower limit that has been describe the relationship between the air degree of saturation
tentatively suggested for liquid water flow is 1 ⫻ 10−14 m / s. It is and soil suction. The SACC can be used to construct the air
suggested the same lower limit for the coefficient of permeability
permeability function 共Ba-Te et al. 2005兲. Sa共兲 is the air degree
might be applied to all soils. The lower limit for the water
of saturation and it is possible to write,
coefficient of permeability is of importance from a numerical
modeling standpoint as well as from a physical behavioral
standpoint. The lack of a lower limit on the water coefficients of Sa共兲 = 1 − Sw共兲 共30兲
permeability can give rise to numerical convergence problems.
The water storage property of a soil, mw2 , is defined as the slope Table 6 summarizes air phase constitutive flow equations that
of the 共volumetric兲 water content versus soil suction relationship. have been written in terms of the soil–water characteristic curve.
The water storage variable is required whenever a transient
seepage analysis is performed. The water storage modulus can be
obtained through the differentiation of any of the equations Heat Flow Constitutive Relations
designated for the SWCC 共Fig. 12兲. Written in Terms of SWCC
Table 7. Functions for Heat Capacity and Thermal Conductivity of an Unsaturated Soil
Reference Equation Description
de Vries 共1963兲 = s共1 − n兲 + wnSw = heat capacity of the soil; s = volumetric specific heat of
共heat capacity of air phase is neglected兲 solids, 2.235⫻ 106 关J / m3 ° C兴; and w = volumetric specific
heat of water, 4.154⫻ 106 at 35° C 关J / m3 ° C兴.
de Vries 共1963兲 = thermal conductivity of the soil; s = thermal conductivity
Fss共1 − n兲 + FwwnSw + Faan共1 − Sw兲 of solids, typically around s = 6 共W / m ° C兲; w = thermal
=
Fs共1 − n兲 + FwnSw + Fan共1 − Sw兲 conductivity of water, typically around w = 0.57 共W / m ° C兲;
a = da + va, where da = thermal conductivity of dry air,
typically ⬃da = 0.025 W / m ° C and va = thermal conductivity
of water vapor, assumed as va = 共0.0736兲Sw 共W / m ° C兲;
3
Fa,s = 1 / 3兺i=1 关1 + 共a,s / w − 1兲gi兴−1; Fw = 1 共water assumed as
the continuum medium兲; g1,2 = 0.015+ 共0.333− 0.015兲Sw
共assuming spherical particles兲; and g3 = 1 − g1 − g2.
Note: n = soil porosity and Sw = water degree of saturation.
冉 冊
Khalili and Khabbaz 共1998兲 f = matrix suction in the specimens at failure conditions.
f −0.55
= c⬘ + 共n − ua兲tan ⬘ + 共tan ⬘兲
b
Shear Strength Constitutive Relations are computed based on an arithmetic scale. Sandy soils may
Written in Terms of SWCC even show a decrease in shear strength 共i.e., −10°兲 once residual
conditions are exceeded 共Donald 1956兲.
The shear strength of an unsaturated soil appears to bear a close
relationship to changes in the water degree of saturation 共or water
content兲 of a soil. Therefore, it is not surprising that the shear Stress-Deformation Constitutive Relations
strength function should be related to the SWCC. Table 8 sum- Written in Terms of SWCC
marizes proposed equations that incorporate the SWCC into the
shear strength constitutive relationship. Verification and compara- The relationship between overall volume change 共i.e., defined
tive studies have been conducted on the proposed equations using void ratio, e, or specific volume, v兲, and soil suction is
using a number of soil types 共Vanapalli and Fredlund 2000兲. The considered to be the most difficult to measure when using con-
Fredlund et al. 共1996兲 form showed the closest fit to the experi- ventional soil testing equipment. It is possible however, to esti-
mental data; however, it was necessary to have an indication of mate the volume change versus soil suction relationship through
the fitting parameter. use of a SWCC and a shrinkage curve for the soil 共Fredlund et al.
The shear strength constitutive relations appear to adhere to 2002b兲.
the following limiting conditions. The shear strength of an unsat- The shrinkage limit of a soil was originally promoted as one of
urated soil increased in response to the effective angle of internal the plasticity classification properties for a soil. The shrinkage
friction, ⬘, for matric suctions up to the air entry value of the curve describes the ratio of the water content change to the void
soil. Once the air entry value is exceeded, the increase in shear ratio change for a specific change in soil suction. Typical shrink-
strength responds to matric suction at a continuously decreasing age curve data are shown in Fig. 40. The equation proposed by
rate throughout the transition region. Once residual conditions are Fredlund et al. 共2002b兲 to fit the shrinkage curves can be written
reached, there appears to be no significant increase 共or decrease兲 as follows:
in shear strength for most soils. Fig. 39 shows the change in
strength that occurs for a number of soils when soil suction is
increased beyond residual suction conditions. A logarithm scale
is used to accommodate the wide range of suctions associated
with all soil types; however, the angles recorded on the graph
Fig. 39. Increase in shear strength with respect to soil suctions Fig. 40. Typical shrinkage curve data relating the effect of matric
greater than residual water content for a variety of soil types suction changes on water content and void ratio change
冋 册
taneously solved, there will be more than one primary variable.
共1/csh兲
wcsh However, each partial differential equation can have one or more
e共w兲 = ash csh +1 共31兲 soil properties to be input and if these soil properties are a func-
bsh
tion of the primary variable being solved, the solution becomes
where e = void ratio; w = gravimetric water content; nonlinear. This is the general case for most unsaturated soils prob-
ash = minimum void ratio, emin; bsh = slope of the line of tangency; lems and as a consequence it is necessary to iterate toward a
and csh = curvature of the shrinkage curve. converged solution. When the equations are highly nonlinear
The curve for void ratio versus soil suction can be computed 共which is often the case for unsaturated soils兲, it can be difficult to
by combining an equation for the soil–water characteristic curve obtain a converged solution.
共e.g., Fredlund and Xing 1994兲 with the equation for the shrink- The ability to solve a wide range of geotechnical engineering
age curve. The relationship between void ratio and soil suction problems in a similar manner gives rise to the possibility of
represents the limiting boundary condition on the void ratio developing a special problem solving computer platform called a
constitutive surface. problem solving environment 共PSE兲. Gallopoulos et al. 共1994兲
described a PSE as “… a computer system that provides all the
computational facilities needed to solve a target class of prob-
Saturated–Unsaturated Soil Mechanics lems. PSEs use the language of the target class of problems, so
as the Solution of a Series of Partial users can run them without specialized knowledge of the under-
Differential Equations lying computer hardware or software. PSEs create a framework
that is all things to all people; they solve simple or complex
The advent of the digital computer has allowed various classes of problems, support rapid prototyping or detailed analysis, and can
soil mechanics problems to be visualized as the solution of a be used in introductory education or at the frontiers of science.”
partial differential equation. The partial differential equation Partial differential equation solvers appear to have become the
共PDE兲 is derived by applying appropriate constitutive relation- PSE for solving saturated–unsaturated soil mechanics problems.
ships to a REV while adhering to the conservative laws of physics All classic soil mechanics problems can be viewed in terms
共i.e., conservation of mass and conservation of energy兲. The of the solution of a partial differential equation. In this paper,
resulting partial differential equation satisfies the physical condi- consideration is given to a few of the problem areas commonly
tions associated with the behavior of the soil for a particular class encountered in unsaturated soil mechanics. The partial differential
of geotechnical problems. equation for water flow through a saturated–unsaturated soil
The physics of the REV can then be applied to a finite-sized system in either two or three-dimensions, is the most common
element of the continuum called a finite element. Combining solution. These solutions have found extensive applications in
the finite elements eventually allows an entire continuum to be disciplines beyond geotechnical engineering such as agriculture,
modeled. Boundaries or limits must be placed on the region environmental engineering, and water resources. Hydraulic heads
considered to make the problem manageable. This gives rise to a are the primary variable computed, opening the way for solving
“boundary value problem” such as shown in Fig. 41. For flow other variables of interest. The water coefficient of permeability
type problems, specifying the head for water flow, pressure for is dependent upon the negative pore–water pressure and this
air flow, or temperature for heat flow, results in what is called gives rise to a nonlinear equation with associated convergence
a Dirichlet boundary condition. The specification of a flow rate challenges.
across a boundary of the problem results in a Neuman type All flow processes 共i.e., water, air, and heat兲 have similar
boundary condition. Other intermediate type boundary conditions partial differential equations that can be solved using a similar
are also possible. Similar type boundary conditions can be partial differential equation solver. Heat flow problems can
specified for stress-deformation types of analyses. readily be solved but there are added challenges associated with
Partial differential equation solvers 共i.e., PDE solvers兲, have the freezing and thawing of water. Air flow problems add
been developed in mathematics and computing science disciplines challenges associated with a compressible fluid phase.
and are increasingly being used to obtain solutions for specific Analyses associated with slope stability, bearing capacity and
geotechnical engineering problems. A single partial differential earth pressure calculations, have historically been performed
equation solver can be used to solve several types of PDEs using plasticity and limit equilibrium methods of slices. However,
relevant to saturated–unsaturated soil mechanics problems. It is all of these application areas are increasingly being viewed as
also possible for more than one physical phenomenon to be “optimization” solution imposed on the results of a stress analysis
operative within an REV, resulting in the necessity to combine the 共Pham and Fredlund 2003兲. Stresses computed from linear elastic
solution of multiple partial differential equations in a coupled or analyses with approximate elastic parameters have been shown to
冉 冊
niques that have emerged that greatly assist in the solution of
e ag u a
highly nonlinear partial differential equations 共Mansell et al. =− Sa − uamw2 共33兲
2002兲. The most successful technique appears to involve auto- 1+e RT t
matic, dynamic finite element mesh refinement, as well as mesh where a⫽average molecular weight of air, 28.8 g/mol;
optimization 共Oden 1989兲, commonly referred to as adaptive grid R⫽universal gas constant, 8.314 g/共s2 cm2 mol K兲;
refinement methods. T⫽temperature 共K兲; and Sa⫽air degree of saturation 共1 − Sw兲 共—兲.
The primary source of error when solving partial differential The partial differential equations for heat flow through an
equations using numerical solutions is insufficient spatial resolu- unsaturated soil can be written as follows:
tion 共Yeh 2000兲. Traditional solutions have used fixed spatial
grids that are generated prior to the onset of the numerical
solution. It is highly unlikely that the grid generated without con-
x
2T x T
x 2 +
x x
2T y T
+ y 2 +
y y y
冉
= + Lf
u T
T t
冊 共34兲
siderations of the types of errors associated with the numerical
where x and y = thermal conductivity of the soil in x and y
solution can produce a satisfactory solution for nonlinear partial
directions 关J / 共m s ° C兲兴; T = temperature 共°C兲; = volumetric
differential equations. specific heat of soil, = ␥natc = f共ua − uw兲关J / 共m3 ° C兲兴; L f = latent
There are two approaches that can be taken to ensure an accu- heat of fusion of water, 3.34⫻ 108 J / m3; = volumetric water
rate solution of partial differential equations. The first approach content at the initiation of freezing; and u / T = change in
involves the mathematical alteration of the partial differential unfrozen water content of the soil with temperature.
equation and the second approach involves the incorporation of The partial differential equations for a stress-deformation
AGR algorithms 共Mansell et al. 2002兲. Bern et al. 共1999兲 stated, analysis can be written as
“Scientific and engineering computations has become so complex
that traditional numerical computations on uniform meshes are
generally not possible or too expensive.” The use of finer and
finer meshes becomes an increasingly impossible solution as three
x
冋 u
D11 + D12
x
v
y
+
y
册 冋 冉 冊册
D44
u v
+
y x
=0 共35兲
冋 冉 冊册 冋 册
dimensional and large geographical areas are modeled. This is
particularly true when solving problems such as the infiltration of u v u v
D44 + + D12 + D11 + ␥t = 0 共36兲
water into the unsaturated soil near ground surface. The solution x y x y x y
that has received the most attention for convergence involves the where D11 = E共1 − 兲 / 关共1 + 兲共1 − 2兲兴; D12 = E / 关共1 + 兲
use of automated grid assignment based upon error estimates ⫻共1 − 2兲兴; D44 = E / 关2共1 + 兲兴; E = Young’s modulus 共kPa兲;
共Bern et al. 1999兲. Babuska 共1989兲 stated, “The main objective = Poisson ratio; and ␥t = body force acting in the y direction
for utilizing any local adaptive grid refinement approach is to 共vertical兲.
effectively achieve an approximate numerical solution that occurs If the computed stress states are optimized with respect to the
within the range of admissible accuracy and to do so with mini- failure conditions of the soil, the shape and location of the critical
mal computational cost.” It would appear that adaptive mesh slip can be determined for a slope stability analysis. If the
refinement is a necessity when solving the nonlinear partial objective of the analysis is to compute potential volume changes,
differential equations common to unsaturated soil behavior the problem becomes a stress-deformation analysis and it is
共Fredlund et al. 2002a兲. important that appropriate nonlinear soil properties be used
Fig. 45. Comparison of the results between dynamic programming Fig. 47. Matric suction values along the ground surface elapsed
and limit equilibrium methods of analyses times of 1, 3, and 5 days
prior to commencing the modeling process because of the nonlin- of the impervious slab. The soil is shown to shrink away from the
earity in the soil properties. The results of an uncoupled analysis concrete slab as evaporation takes place from the ground surface.
on a slab-on-ground problem are presented herein. Similar plots to those presented above can be produced for the
Fig. 46 shows an example problem involving a slab-on-ground case of infiltration of water at the ground surface. The boundary
subjected to the influences of a moisture flux boundary condition condition at the ground surface can be specified in terms of a
共Fredlund and Vu 2003; Vu and Fredlund 2004兲. The expansive moisture flux or a specified head 共or pressure兲 condition. The
soil layer is 3 m deep and has a swelling index of 0.15. The initial microclimatic conditions at a specific site need to be analyzed
matric suction is assumed to be 400 kPa at the base of the expan- in order that realistic boundary conditions can be specified and
sive soil layer. The swelling index is converted to an equivalent studied through use of a parametric type analysis. There are a
variable elasticity function for the unsaturated soil. The elasticity wide variety of expansive soils and collapsible soil problems that
function means that the incremental Young’s modulus for the can be studied as a result of combining a transient seepage analy-
material is a function of the stress state, rendering the stress sis for an unsaturated soil with a nonlinear stress-deformation
analysis highly nonlinear. There is nonlinearity with respect to analysis. Space does not permit for the illustration of other
both the net total stress state and the matric suction stress state. example problems involving unsaturated soils.
A permeability function is specified for the swelling soil and a
moisture flux is designated to simulate either rainfall conditions
or evaporation conditions from the soil around the flexible slab. “Challenge” for the Future
A transient analysis was first run for the case of evaporation from
the soil surrounding the impervious surface slab. Matric suction There are many “challenges” that still lie ahead before it can
values along the ground surface are shown for elapsed times of 1, be said that engineers clearly understand how best to apply un-
3, and 5 days in Fig. 47. Matric suctions for the same elapsed saturated soil mechanics in engineering practice. However, there
times, along a vertical section passed below the edge of the appears to be one challenge that is most important to address and
impervious slab are shown in Fig. 48. Figs. 47 and 48 provide that is the need to establish protocols for various geotechnical
an indication of the change in matric suction along vertical and engineering problems associated with unsaturated soil mechanics.
horizontal directions from the edge of the slab. Numerous other As long as there are no distinct standards of practice or
graphs could be plotted to depict changes in matric suction with “protocols” for addressing various unsaturated soils problems,
time, at any location in the soil mass. Fig. 49 shows the matric then total reliance on past experience and empirical practices will
suction contours throughout the soil mass after three days of be followed. The end result will be a low level of engineering
evaporation from the soil surface. practice with difficulties being encountered that will be settled
The changes in matric suction can be combined with a stress-
deformation analysis that takes into account changes in matric
suction. Fig. 50 shows the vertical displacements along the
ground surface for 1, 3, and 5 days of evaporation. Fig. 51 shows
the vertical displacements along a vertical section below the edge
Fig. 49. Matric suction contours throughout the soil mass after three Fig. 51. Vertical displacements along a vertical section below the
days of evaporation from the soil surface edge of the impervious slab