ERDC/GSL SR-17-4

Flood and Coastal Storm Damage Reduction Program

SWCC Prediction: Seep/W Add-In Functions

Lucas Walshire and Bryant Robbins June 2017
Geotechnical and Structures Laboratory

Approved for public release; distribution is unlimited.

The U.S. Army Engineer Research and Development Center (ERDC) solves
the nation’s toughest engineering and environmental challenges. ERDC develops
innovative solutions in civil and military engineering, geospatial sciences, water
resources, and environmental sciences for the Army, the Department of Defense,
civilian agencies, and our nation’s public good. Find out more at www.erdc.usace.army.mil.

To search for other technical reports published by ERDC, visit the ERDC online library
at http://acwc.sdp.sirsi.net/client/default.
Flood and Coastal Storm Damage ERDC/GSL SR-17-4
Reduction Program June 2017

SWCC Prediction: Seep/W Add-In Functions

Lucas Walshire and Bryant Robbins

Geotechnical and Structures Laboratory
U.S. Army Engineer Research and Development Center
3909 Halls Ferry Road
Vicksburg, MS 39180-6199

Final report
Approved for public release; distribution is unlimited.

Prepared for U.S. Army Corps of Engineers

Washington, DC 20314-1000
Under Project Number HFK2F2 “Development of best practices for use of transient
seepage analysis in geotechnical engineering practice”
ERDC/GSL SR-17-4 ii

Abstract
The soil water characteristic curve (SWCC) defines a constitutive
relationship between the negative pressure that develops when a soils
saturation level is less than fully saturated, and the corresponding volume
of water held in the pore space of the soil matrix. As this relationship is not
commonly measured in geotechnical laboratories, practitioners often
attempt to predict this relationship based on other commonly measured
material properties using empirical prediction methods. The performance
of five SWCC empirical predictors was evaluated through comparisons to
independently measured SWCC data for four soils. SWCC prediction
methods were selected for this investigation if they incorporated
commonly measured soil properties to predict the SWCC. The error in the
SWCC prediction was assessed in terms of both the mean squared error on
the SWCC prediction and the impact of the error on a numerical analysis
of the Green and Ampt infiltration problem. The results of the numerical
analysis were assessed in terms of a normalized saturation coefficient. The
normalized saturation coefficient provided a clear means of monitoring a
transient seepage analysis through a single measure. Results indicate that
the SWCC prediction methods yielding the lowest mean squared error did
not necessarily yield the smallest error in the transient seepage analysis.
Further, only the Rawls method consistently yielded conservative analysis
results for all soil types investigated.

DISCLAIMER: The contents of this report are not to be used for advertising, publication, or promotional purposes.
Citation of trade names does not constitute an official endorsement or approval of the use of such commercial products.
All product names and trademarks cited are the property of their respective owners. The findings of this report are not to
be construed as an official Department of the Army position unless so designated by other authorized documents.

DESTROY THIS REPORT WHEN NO LONGER NEEDED. DO NOT RETURN IT TO THE ORIGINATOR.
ERDC/GSL SR-17-4 iii

Contents
Abstract .......................................................................................................................................................... ii

Figures and Tables ........................................................................................................................................iv

Preface .............................................................................................................................................................v

1 Introduction ............................................................................................................................................ 1

2 General Characteristics of the SWCC ............................................................................................... 2

3 SWCC Prediction Methods .................................................................................................................. 6

Method 1: (Zapata et al. 2000) ............................................................................................... 6
Method 2: (Perera et al. 2005)................................................................................................ 7
Method 3: (Sleep 2011)........................................................................................................... 9
Method 4: (Tomasella and Hodnett 1998) ........................................................................... 11
Method 5: (Rawls et al. 1991) ............................................................................................... 12

4 Evaluation of Prediction Methods ...................................................................................................14

Comparison of the Prediction Methods ................................................................................ 14
Impacts of SWCC Prediction Error on Transient Seepage Analyses .................................... 18
Saturation Coefficient ............................................................................................................ 23

5 How to Use Seep/W SWCC Prediction Add-in .............................................................................32

Installing Function (Add-In) in Seep/W.................................................................................. 32
Method 1: (Zapata et al., 2000) ............................................................................................ 33
Method 2: (Perera et al., 2005)............................................................................................. 33
Method 3: (Sleep 2011)......................................................................................................... 34
Method 4: (Tomasella and Hodnett 1998) ........................................................................... 34
Method 5: (Rawls et al. 1991) ............................................................................................... 35

6 Conclusion ............................................................................................................................................36

References ...................................................................................................................................................38

Report Documentation Page

ERDC/GSL SR-17-4 iv

Figures and Tables

Figures
Figure 1. Soil water characteristic curve (SWCC). ...................................................................................... 3
Figure 2. Wetting and drying SWCC (data from Li et al. 2005). ............................................................... 5
Figure 3. SWCC for coarse sand, n=0.35, Sleep (2011). .......................................................................11
Figure 4. Predicted and laboratory SWCC for ML material. ....................................................................15
Figure 5. Predicted and laboratory SWCC for compacted CH. ...............................................................16
Figure 6. Predicted and laboratory SWCC for SM material..................................................................... 17
Figure 7. Predicted and laboratory SWCC for SP material. ..................................................................... 17
Figure 8. Finite element mesh with boundary conditions. .....................................................................18
Figure 9. Total head contours, laboratory-measured ML SWCC. ...........................................................20
Figure 10. Profile of pressure head values after five days for ML. ........................................................ 21
Figure 11. Profile of pressure head values after five days for compacted CH. ................................ 21
Figure 12. Profile of pressure head values after five days for SM. .......................................................22
Figure 13. Profile of pressure head values after five days for SP. .........................................................22
Figure 14. Hydraulic conductivity function for ML. ..................................................................................23
Figure 15. Elevation versus pressure head at different times............................................................... 24
Figure 16. Comparison of predicted method and methods fitted with van Genuchten
equation (1980). ..........................................................................................................................................26
Figure 17. Saturation versus time for the ML material............................................................................28
Figure 18. Saturation versus time for the CH material. ..........................................................................29
Figure 19. Saturation versus time for the SM material...........................................................................30
Figure 20. Saturation versus time for the SP material............................................................................ 31

Tables
Table 1. Soil classification based on hydraulic conductivity (Terzaghi 1996). ....................................... 9
Table 2. SWCC data used to estimate appropriate curve (Sleep 2011). ..............................................10
Table 3. SWCC soil data set. .......................................................................................................................14
Table 4. Material properties of the soil data set. .....................................................................................14
Table 5. Results of transient seepage analysis MSE between predicted SWCC and
laboratory-measured SWCC........................................................................................................................23
Table 6. Material properties used in the FLAC analysis. ......................................................................... 27
ERDC/GSL SR-17-4 v

Preface
This study was conducted under the Flood and Coastal Storm Damage
Reduction Program. The technical monitor was Dr. Cary Talbot, Associate
Technical Director, Coastal and Hydraulics Laboratory.

The work was performed by the Geotechnical and Structures Laboratory

under the supervision of the Technical Directors Office (GV-T), U.S. Army
Engineer Research and Development Center, Geotechnical and Structures
Laboratory (ERDC-GSL). At the time of publication, Dr. Maureen K.
Corcoran, CEERD-GV-T was the acting Technical Director for Water
Resources Infrastructure. The Deputy Director of ERDC-GSL was
Dr. William P. Grogan, and the Director was Mr. Bartley P. Durst.

COL Bryan S. Green was the Commander of ERDC, and Dr. David W.
Pittman was the Director.
ERDC/GSL SR-17-4 1

1 Introduction
The soil water characteristic curve (SWCC) is the relationship between the
soil-water suction and the water content of the soil. This relationship helps
define the magnitude of matric suction that occurs in a soil deposit when
the water content is less than saturated. The quantity of water retained in
a soil at a certain magnitude of suction depends on many factors: particle
shape, particle size, distribution of pore spaces, mineralogy, surface
activity of solid grain particles, and chemical composition of interstitial
water (Aubertin et al. 2003). The SWCC is important in geotechnical
engineering, as the degree of saturation and corresponding matric suction
greatly influences the shear strength and hydraulic conductivity of soils. A
partially saturated soil will have a decreased hydraulic conductivity and
increased shear strength as compared to a saturated soil.

A series of SWCC prediction methods was investigated to ascertain which

would be appropriate for use on preliminary analyses of U.S. Army Corps
of Engineers (USACE) projects, such as dams and levees. SWCC prediction
methods were selected for this investigation if they incorporated
commonly measured soil properties to predict the SWCC. It is more
desirable to directly measure the SWCC in the laboratory, but this type of
testing is not commonplace in practice.

In the literature, five SWCC prediction methods that use common soil
properties for curve estimation were found. The soil properties used by the
selected prediction methods were saturated hydraulic conductivity, grain
size, plasticity index, and porosity. The finite element seepage program
Seep/W©, distributed by GeoStudio, was used as the platform for the
developed Add-In functions. Each prediction method was programmed as
a function and compiled as a digital library (*.dll). The performances of
the five SWCC prediction methods were evaluated by comparing the
predicted curves to independent laboratory measurements. Finally, the
significance of the errors associated with the SWCC predictions was
assessed through a numerical transient seepage analysis.
ERDC/GSL SR-17-4 2

2 General Characteristics of the SWCC

The difference between steady state and transient seepage analysis is that
no parameters vary with time during a steady state analysis. In a transient
seepage analysis, the hydraulic boundary conditions, volumetric water
content, and hydraulic conductivity all vary temporally. Hydraulic
conductivity is related to the volumetric water content (and corresponding
matric suction) through the hydraulic conductivity function, which is
typically predicted from the SWCC. Increasing suction (decreasing
saturation) decreases soil permeability. Therefore, it is important to
understand how to accurately estimate the SWCC for a transient seepage
analysis. The following sections will focus solely on the influence of SWCC
in the context of transient seepage analysis.

The governing differential equation for three-dimensional seepage is

shown in Equation 1:

  H    H    H  θ
 kx θ    ky θ     kz θ   Q  (1)
x  x  y  y  z  z  t

where Q is a boundary flux, H is total head, kn is hydraulic conductivity in

the n direction (n is either x, y, or z in Cartesian coordinates), θ is the
volumetric water content, and t is the time (Lu and Likos 2004). This
equation states that the difference between the flow (flux) entering and the
flow leaving a volume of soil at a point in time is equal to the change in
storage of the soil volume. The change in storage for an unsaturated soil
system is related to the volumetric water content, which quantifies how
much water is stored in the pore space. The SWCC relates the magnitude
of suction a soil system experiences with the corresponding water content.

SWCCs are typically plotted on a semi-log plot with the ordinate axis
showing either saturation, gravimetric water content, or volumetric water
content; the abscissa contains the matric suction in units of pressure (kPa,
psf, or cm of water). Figure 1 shows an example SWCC.
ERDC/GSL SR-17-4 3

Figure 1. Soil water characteristic curve (SWCC).

When the saturation level in a soil is less than fully saturated, negative pore
water pressures develop. The term soil suction is defined as the negative
pore water pressure. Total suction consists of two primary components:
matric suction and osmotic suction (Fredlund et al. 2012). Matric suction
(also known as capillary pressure) is the mathematical difference between
the air and water pressures (ua-uw) in the soil. The air pressure is usually
zero (gauge pressure), and the water pressure is negative due to surface
tension. Osmotic suction is associated with both saturated and unsaturated
soils and is related to the salt content in the pore water. If the salt concen-
tration in the pore fluid changes, there is a corresponding change in the
volume and shear strength of the soil (Fredlund et al. 2012). For the
purpose of this report, osmotic suction will not be considered because it is
considered to play a minor role in embankment performance when
compared to the influence of matric suction.

The air entry value (AEV) is defined as the differential pressure between
air and water that is required to cause desaturation of the largest pores.
The AEV can be defined graphically as the intersection of the line tangent
to the straight line portion of the SWCC and a horizontal line through the
saturated water content. The intersection point of these two lines is the
matric suction AEV. The residual water content is the point at high
suctions at which very little water is retained and at which pore water is
generally in the form of thin films surrounding the surface of soil grains
ERDC/GSL SR-17-4 4

(Lu and Likos 2004). The residual point can be defined as the intersection
of a tangent line along the SWCC where the curve starts to drop linearly in
the high suction range (Fredlund and Xing 1994) and the tangent line used
for determining the AEV. The AEV and the residual point (ψr,θr) are shown
in Figure 1. Some fine-grained soil water curves do not exhibit a residual
suction point. Typically, a value between 1500 to 3000 kPa is an
appropriate approximation (Fredlund and Xing 1994).

One of the more common methods of fitting SWCCs to laboratory data is

the equation proposed by van Genuchten (1980) shown in Equation 2.

θs  θr
θ  θr  m (2)
 n
1   ψ  
   
 a 

where 𝜃𝜃𝑠𝑠 and 𝜃𝜃𝑟𝑟 are the saturated and residual volumetric water contents;
𝜓𝜓 is the matric suction; and 𝑎𝑎, 𝑛𝑛 and 𝑚𝑚 are curve-fitting parameters. The
parameter m is often approximated as (1-1/n). Van Genuchten (1980)
gives a detailed explanation of the derivation of the fitting parameters. The
van Genuchten equation is used to represent the SWCC analytically; the
result can then be used to define the hydraulic conductivity function
(HCF) (van Genuchten 1980).

The SWCC exhibits hysteresis as a soil cycles through wetting and drying
processes. A soil undergoing a drying process will typically have a larger
water content than a soil undergoing a wetting process at the same value of
matric suction. Figure 2 shows an example of the wetting and drying curves.

Typically when an SWCC is obtained in the laboratory, the drying curve is

measured. With regard to dams and levees, the process being modeled is
undergoing a wetting process (i.e., the modeling of a flood load on an
embankment). One way to address this discrepancy is to simply accept the
error and use the drying curve for both wetting and drying processes.
Another approach is to scale the SWCC fitting parameters to transform a
drying curve into a wetting curve. Likos et al. (2013) performed a study
over a wide range of soil types to assess which scaling factors were
appropriate to adjust a drying curve to a wetting curve. That study found
the following conversions to be appropriate:

α w  2α d (3)
ERDC/GSL SR-17-4 5

nw  nd (4)

θsw  θsd (5)

where 𝛼𝛼 and 𝑛𝑛 are fitting parameters used in the van Genuchten SWCC
fitting model, 𝜃𝜃𝑠𝑠 is the saturated volumetric water content, and the
superscripts w and d denote wetting and drying respectively. The van
Genuchten equation is shown in Equation 5.

Figure 2. Wetting and drying SWCC (data from Li et al. 2005).

SWCC
0.5

SM Hysteresis
drying
wetting

0.4
Volumetric Water Content

0.3

0.2

0.1

0.1 1 10 100 1000

Matric Suction (kPa)
ERDC/GSL SR-17-4 6

3 SWCC Prediction Methods

Two prediction methods are already incorporated into Seep/W. These are
referred to in Seep/W as the Modified Kovacs method and the sample
functions based on soil type. For further discussion of these two methods, it
is recommended that the Seep/W user’s manual be reviewed. The manual
includes two closed form options as well: (1) the van Genuchten and (2) the
Fredlund and Xing methods. The five methods discussed below are
investigated to supplement those methods already included in the program.

Method 1: (Zapata et al. 2000)

This procedure uses the Fredlund and Xing (1994) equation (see Seep/W
user’s manual closed form option 1). The parameters a, b, c, and hr are
correlated to percent finer than the #200 sieve (0.074 mm) and PI
(plasticity index) for plastic soils (PI>0) or D60 for nonplastic soils. A
parameter wPI (defined in Equation 25) is used to derive a, b, c, and hr
parameters for plastic soils (PI>0).

wPI  P200 *PI (6)

where P200 is the percentage of material passing the #200 sieve, expressed
as a decimal, and PI is the plasticity index, expressed as a percentage. The
parameters for plastic soils (PI>0) are defined in Equations 7 to 10:

3.35
a  0.00364wPI   4wPI   11 (7)

b 0.14
2.313wPI   5 (8)
c

0.465
c  0.0514wPI   0. 5 (9)

hr 0.0186wPI 
 32.44e (10)
a

The parameters for granular soils (PI=0) are defined in Equations 11 to 14:

0.751
a  0.8627 D60  (11)
ERDC/GSL SR-17-4 7

b  7. 5 (12)

c  0.1772 ln D60   0.7734 (13)

hr 1
 (14)
a D60  9.7e4

where D60 is the grain size at which 60% of the material is finer and 𝑏𝑏� is
the average value of the fitting parameter b. This method was derived from
measurements on 190 soils. The samples consisted of 70 plastic soils
(PI>0) and 120 nonplastic soils (PI=0). This method was included in the
analysis because the input parameters are common index properties used
to classify soils according to the Unified Soil Classification System (USCS),
making the method easily applicable in practice.

Method 2: (Perera et al. 2005)

This method is an update of Method 1. This method defines a plastic soil as
a soil that has a wPI greater than or equal to 1.0. The parameters were
derived from multiple regressions correlating grain size distribution data
and index properties. The Fredlund and Xing (1994) parameters are
defined in Equations 15 through 25 for nonplastic soils:

a f  1.14a  0.5 (15)

where:

a  2.79 14.1 log  D20 1.9106  P200 

4.34
 7 log  D30   0.055 D100 (16)

 40 
 
 m  logD60 
D100  10  1 
(17)

30
m1  (18)
  D 
 log  90 
  D 
 60 

b f  0.936b  3.8 (19)

ERDC/GSL SR-17-4 8

where:

   D  
1.19 
b  5.39  0.29 ln  P200  90   3 D00.57  0.021  P200   m10.1 (20)
   D10  
  

 30 
 
 m  log D30 
Do  10  2 
(21)

20
m2  (22)
D 
log  30 
 D10 

c f  0.26e 0.758 c  1.4 D10 (23)

where:

 1 

c  log m2 
1.15
 1  
 b f 
(24)

hrf  100 (25)

where Dx is the grain size diameter corresponding to x percent of the

material being finer. The parameters for plastic soils (wPI ≥ 1.0) are
defined in Equations 26 through 30:

a f  32.835 lnwPI   32.438 (26)

0.3185
b f  1.421 wPI  (27)

c f  0.2154 lnwPI   0.7145 (28)

hrf  500 (29)

where:

wPI  P200 * PI (30)

ERDC/GSL SR-17-4 9

where P200 is the percent passing the #200 sieve (0.074 mm), expressed as
a decimal, and PI is the plasticity index, expressed as a percentage. There
were 154 nonplastic and 63 plastic soils used in the regression analysis.
This method was included because the parameters are dependent on soil
properties used in the USCS and because a relatively large number of
samples were used to derive them.

Method 3: (Sleep 2011)

This technique uses trends among the SWCCs for different soil types in
order to predict the most likely SWCC. The Unsaturated Soil Data
Hydraulic Database (UNSODA), compiled and distributed by the United
States Department of Agriculture (USDA), was used to create SWCC charts
based on soil type. The UNSODA is based on the textural classification
system; therefore, it does not contain soil index properties common in
geotechnical practice. In order to relate the textural and the USCS, the
SWCC data were separated by hydraulic conductivity and classified using
the saturated hydraulic conductivity. Soil classifications according to
saturated hydraulic conductivity (ksat) are shown in Table 1.

Table 1. Soil classification based on hydraulic conductivity

(Terzaghi 1996).
Soil Category ksat (cm/s)
Coarse Sand >10-1
Fine Sand 10-1 - 10-3
Silty Sand 10-3 - 10-5
Silt 10-5 – 10-7
Clay <10-7

The method involves the following steps (Sleep 2011):

1. Obtain the saturated hydraulic conductivity of the soil.

2. Use the saturated hydraulic conductivity value to select the appropriate
range of values from the provided figures (Table 2 presents the figures in
tabular form).
3. Select the appropriate SWCC based on whether the soil is wetting or
drying.
ERDC/GSL SR-17-4 10

Table 2. SWCC data used to estimate appropriate curve (Sleep 2011).

Soil Suction (kPa)

Wetting Drying
ksat (cm/s) Average Average Boundary Boundary Soil
Range S Drying Wetting (90% conf.) (90% conf.) Category
10 to 1 0.7 0.1 0.009 8 Coarse
1.0E-01 0.1 4 0.6 0.05 13.5 Sand

1.0E-1 to 1 4 0.8 0.15 30

Fine Sand
1.0E-3 0.1 150 25 3.5 85

1.0E-3 to 1 4.5 0.95 0.19 25

Silty Sand
1.0E-5 0.1 4500 1000 200 28000
1 6 0.9 0.09 70
0.14 X X 100000 X
1.0E-5 to
0.28 X 100000 X X Silt
1.0E-7
0.4 100000 X X X
0.55 X X X 100000

4. Input the two point values into Seep/W, Data Point Function, by using
Equation 31 or 32 to convert saturation to volumetric water content:

θ100%  np (31)

θ10%  np * S10% (32)

where θ is the volumetric water content, np is the porosity, and S is

saturation.

5. Repeat steps 1-4 to obtain any other curves deemed necessary.

Figure 3 shows how the curves would look for coarse sand following the
process outlined above. The wetting and drying boundaries are based on a
90% confidence interval. The wetting curve was constructed based on an
approximation that the wetting curve is usually one order of magnitude
smaller than the drying curve. Sleep (2011) provides a more complete
discussion on the formulation of the method.
ERDC/GSL SR-17-4 11

Figure 3. SWCC for coarse sand, n=0.35, Sleep (2011).

SWCC
Coarse Sand
0.4

Sleep Coarse Sand

Average Wetting
Average Drying
Wetting Boundary, 90% confidence
Drying Boundary, 90% confidence

0.3
Volumetric Water Content

0.2

0.1

0.01 0.1 1 10 100 1000 10000 100000

Matric Suction (kPa)

The limitation of this procedure is that there is no method to predict the

SWCC for clay materials. This method was selected for comparison
because it is easy to use and it provides the upper and lower bound SWCCs
for use in sensitivity analyses.

Method 4: (Tomasella and Hodnett 1998)

This method uses a pedotransfer function (PTF) to derive the SWCC. This
method was derived for use with soils from the Brazilian Amazon using
1,162 soils from the region. The Brooks and Corey model was used to
model the SWCC, and the coefficients were derived using grain size
according to the textural classification system. Equation 33 shows the
Brooks and Corey model.


 1

   B
θ  θ  θ  ψB      ψ  ψB
θ  r s r
 ψ  (33)



 θs             
0  ψ  ψB


ERDC/GSL SR-17-4 12

where θ is the volumetric water content, ψ is the matric potential (kPa),

the subscripts r and s respectively denote residual and saturated water
contents, 1/B is an empirical constant, and ψB is the matric potential (kPa)
corresponding to the air entry pressure. For this method, the Brooks and
Corey parameters are estimated solely from the grain size distribution
using Equations 34 to 37:

ψB  0.285  7.33104  Psi 2 1.3104  Psi * Pcl  3.6106  Psi 2 Pcl (34)

1.1974.1710 P 4.510 P 8.9410 P P 1.010 P

3 3 4 5 2
Pcl 
B e
si cl si cl si
(35)

θs  40.61  0.165 Psi  0.162 Pcl  1.37103  Psi 2  1.8105  Psi 2 Pcl (36)

θr  2.094  0.047Psi  0.431 Pcl  8.27103  Psi Pcl (37)

where Psi is the percent silt and Pcl is the percent clay. Some ranges of
material type may produce a small negative value for θr; in these cases, θr
should be assigned a value of 0. This method was intended for use with
tropical fine-grained soils and is based on the textural classification system.

Method 5: (Rawls et al. 1991)

This method is based on previous work by Rawls et al. (1982) in which a
regression analysis was performed on existing SWCC data to derive the
Brooks and Corey SWCC parameters; the Brooks and Corey model is
defined in Equation 33. The equations for the Brooks and Corey
parameters resulting from the regression analyses are

 5.339670.1845 P 2.4839 n 0.00214 P 2 0.04356 P n 0.61745 P n 

 cl p cl sa p cl p 
 
0.001436 Psa2n p2 0.008554 Psa2n p2 0.00001282 Psa2 Pcl 0.008954 Psa2n p 
 
 2 2 2

  0. 0007247 P n  0. 0000054 P P  0. 5 n P 
ψB  e  sa p cl sa p cl 
(38)

 0.7840.0177 P 1.062 n 0.000053 P 2 0.00273 P 2 1.111 n 2 0.03088 P n 

 sa p sa cl p sa p 
 
0.000266 P 2n 2 0.0061 P 2n 2 0.00000235 P 2 P 0.007987 P 2n 0.00674 n 2 P 
λ e  sa p cl p sa cl cl p p cl 
(39)

2
θr  0.0182  0.000873 Psa  0.005135 Pcl  0.02939n p  0.000154 Pcl
2 2 2 2
(40)
0.00108 Psa n p  0.000182 Pcl n p  0.000307 Pcl n p  0.00236n p Pcl
ERDC/GSL SR-17-4 13

where Psa is percent sand, Pcl is percent clay, and np is the porosity. The
bubbling pressure, otherwise known as the air entry value, ψB, is in
centimeters of water; λ is the pore size index (λ=1/B); and θr is the
residual water content. These equations (Equations 38-40) are reported to
be valid for 5%<PSa<70% and 5%<Pcl<60%.
ERDC/GSL SR-17-4 14

4 Evaluation of Prediction Methods

An evaluation of the prediction methods was performed by comparing
laboratory-measured SWCCs obtained from the literature for four soil
types to the curves predicted by the five methods above. This evaluation
was conducted to illustrate how the prediction methods compared across a
range of soil types. By using independent data (data not included in the
method development), the performance of the prediction methods was
also evaluated. The predicted SWCCs obtained from each method for the
four soils were also carried through a simple transient seepage analysis to
quantify the influence of the differences in SWCC in terms of the time
required for a wetting front to propagate. Table 3 shows the soils used,
classified according to the USCS, and each reference from which the data
were obtained.

Table 3. SWCC soil data set.

USCS Reference
ML Askerinejad et al. 2011
SM Li et al. 2005
CH Tinjum et al. 1997
SP Song et al. 2012

Table 4 shows the material properties necessary to predict the SWCC using
each of the five prediction methods. The CH sample was compacted at
optimum water to maximum dry density by the modified Proctor method.

Table 4. Material properties of the soil data set.

D10 D20 D30 D60 D90 %Passing ksat
USCS (mm) (mm) (mm) (mm) (mm) %Sand %Silt %Clay #200* LL PL PI Porosity (cm/s)
ML 0.0075 0.06 0.085 0.28 0.53 78.87 16.45 3.55 0.27 31 20 11 0.46 1.00E-05
SM 0.0019 0.013 0.045 0.59 2.6 53 20 12 0.35 57 40 17 0.43 1.00E-04
CH - - - - - 6 41 53 0.94 67 21 46 0.345 1.00E-09
SP 0.43 0.47 0.5 0.6 0.75 100 0 0 0 NP 0.42 1.00E-01
*Expressed as a decimal, necessary for Zapata et al. (2000) and Perera et al. (2005) methods.

Comparison of the Prediction Methods

The predicted SWCCs obtained from each method for the soils in Table 4
were graphically compared to the laboratory-measured curve obtained
ERDC/GSL SR-17-4 15

from the literature. These analyses were performed on a very limited

dataset (four samples), making only general observations regarding the
predictive methods possible. All of the sample SWCCs were drying curves.
Figure 4 shows the various predicted SWCCs and the laboratory-measured
curve for the ML material.

Figure 4. Predicted and laboratory SWCC for ML material.

SWCC
ML
0.45

ML SWCC
ML-Lab
0.4
ML-Tomasella
ML-Rawls
ML-Zapata
0.35
ML-Perera
ML-Sleep

0.3
Volumetric Water Content

0.25

0.2

0.15

0.1

0.05

0.01 0.1 1 10 100 1000 1E+004 1E+005 1E+006 1E+007 1E+008

Matric Suction (psf)

The closest approximation of the laboratory measured SWCC for the ML

material was obtained using Sleep’s method (2011). However, the
measured data are slightly concave with respect to the vertical axis, and
the two methods outlined by Rawls et al. (1991) and Tomasella and
Hodnett (1998) seem to approximate this shape the best.

Figure 5 shows the predicted and laboratory SWCC for the compacted CH.
Tomasella and Hodnett’s method (1998) overpredicts the SWCC by a large
margin. This overprediction is likely due to the reliance of this method on
predicting the saturated volumetric water content while the other methods
rely on user input of the saturated volumetric water content (estimated as
porosity). For this sample the closest approximation of the laboratory-
measured SWCC data was acquired by Sleep’s method (2011), which uses
the silt curve due to the absence of a clay curve for this method.
ERDC/GSL SR-17-4 16

Figure 5. Predicted and laboratory SWCC for compacted CH.

SWCC
CH
0.6

CH SWCC
0.55
CH-Lab
CH-Tomasella
0.5 CH-Rawls
CH-Zapata
CH-Perera
0.45
CH-Sleep

0.4
Volumetric Water Content

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0.01 0.1 1 10 100 1000 1E+004 1E+005 1E+006 1E+007 1E+008

Matric Suction (psf)

Figure 6 shows the predicted and the laboratory-measured SWCCs for the
SM sample. The measured SWCC data are convex with regard to the
vertical axis. This shape is matched accurately by the Brooks and Corey
function utilized by Tomasella and Hodnett (1998) and Rawls et al. (1991).
The closest approximation of the laboratory data with regard to shape and
accuracy was obtained by Zapata et al. (2000).

Figure 7 shows the predicted and laboratory SWCC for the SP material.
The shape of the measured SWCC is nearly vertical after it reaches the
AEV, occurring at approximately 35 kPa. Zapata et al.’s (2000) method
most nearly approximates the measured SWCC data in shape and
accuracy.

Considering the limited data set (four samples) analyzed, Zapata et al.’s
(2000) and Sleep’s (2011) methods seem to approximate more accurately
the measured SWCC with regard to both shape and accuracy across the
range of soils.
ERDC/GSL SR-17-4 17

Figure 6. Predicted and laboratory SWCC for SM material.

SWCC
SM
0.5

SM SWCC
0.45 SM-Lab
SM-Tomasella
SM-Rawls
0.4 SM-Zapata
SM-Perera
SM-Sleep
0.35
Volumetric Water Content

0.3

0.25

0.2

0.15

0.1

0.05

0.01 0.1 1 10 100 1000 1E+004 1E+005 1E+006 1E+007 1E+008

Matric Suction (psf)

Figure 7. Predicted and laboratory SWCC for SP material.

SWCC
SP
0.45

SP SWCC
SP-Lab
0.4
SP-Tomasella
SP-Rawls
SP-Zapata
0.35
SP-Perera
SP-Sleep

0.3
Volumetric Water Content

0.25

0.2

0.15

0.1

0.05

0.01 0.1 1 10 100 1000 1E+004 1E+005 1E+006 1E+007 1E+008

Matric Suction (psf)
ERDC/GSL SR-17-4 18

Impacts of SWCC Prediction Error on Transient Seepage Analyses

The second analysis was conducted to illustrate how the error associated
with each respective prediction method impacts the results of a transient
seepage analysis. The analysis was conducted by using the finite element
software Seep/W (2007) along with the SWCC prediction Add-In
developed for this study. The problem geometry will be briefly
summarized below; Tracy et al. (2015) gives a more detailed definition of
the problem. The problem consists of a column 10 ft wide and 50 ft tall.
The grid size is 2 ft by 2 ft, and the time step is set to 0.001 days (86.4 s).
Figure 8 shows the column geometry and total head boundary conditions.
The model was set to an arbitrary nearly dry state (ht = -30 ft), and at time
greater than zero the total head at the upper boundary was set to 50 ft.
This simulates infiltration from the upper surface of the column, similar to
the Green-Ampt infiltration problem, as defined in Tracy et al. (2016).

Figure 8. Finite element mesh with boundary conditions.

ERDC/GSL SR-17-4 19

The SWCC developed for each of the samples shown in Table 4 and
Figures 4-8 was used in this analysis. The process modeled was a wetting
process, and all of the SWCCs were drying curves; but for comparison
purposes, these SWCCs were used with no adjustment. The hydraulic
conductivity ratio was assumed to be 1, and the van Genuchten hydraulic
conductivity function (1980) was used. Each analysis was run for a
simulated five days; and at the end of the simulated time, the total heads
were collected from the nodes that lie along the dashed line shown in
Figure 8.

Figure 9 shows the total head contours for the laboratory-measured ML

SWCC. The infiltration wetting front had not advanced more than a foot
after five days. This was typical for all of the analysis. Of interest for this
particular analysis was how each of the predicted SWCCs compared to the
laboratory-measured SWCC in terms of loading response. Figures 10
through 13 show the profiles of pressure head values for each prediction
method and the laboratory-measured SWCC for each soil.

The results shown in Figure 10 show that the SWCC predicted by the Rawls
et al. method (1991) closely matched the results of the laboratory SWCC.
This is interesting because, even though the Rawls method (as well as the
Tomasella method [Tomasella and Hodnett 1998]) matched the shape of
the lab SWCC, they were rather inaccurate. An investigation of the hydraulic
conductivity functions (HCF) (Figure 14) shows the Rawls et al. method
(1991) more closely matches the laboratory-measured SWCC HCF.

Figures 11 and 12 show that all prediction methods matched the laboratory
data more closely for the SM and CH soils than for the ML material. In the
case of the CH material, this is in part due to the limited development of
the wetting front over the period of five days. Table 5 shows the mean
squared error (MSE), which is a measure of the difference between the
pore pressures modeled using the laboratory-measured SWCC and the
pore pressures obtained using the various predicted SWCCs. Equation 42
shows how the MSE was calculated:

1 n ˆ
 
2
MSE   Yi  Yi
n i1
(42)

where n is the quantity of data points; for this case there were 26 data
points, one for each node. 𝑌𝑌𝑖𝑖 is the total head values obtained at the
26 different nodes for the analysis that used the predicted SWCC, and 𝑌𝑌�𝑖𝑖 is
ERDC/GSL SR-17-4 20

the total head values from the laboratory-measured SWCC analysis. The
smaller the MSE value, the more accurately the results of the predicted
SWCC analysis matched the results of the analysis performed using the
laboratory-measured SWCC. The MSE values for the compacted CH are all
less than ten, but Sleep’s (2011) and Tomasella and Hodnett’s (1998)
methods were much smaller than one, indicating that these two prediction
methods closely matched the results of the laboratory-measured SWCC
analysis. This is an interesting result, considering that Sleep’s (2011)
prediction was derived from the silt data as a result of no clay SWCC curve’s
being available in the UNSODA database (Nemes et al. 2001). Tomasella
and Hodnett’s (1998) method was derived from mainly fine-grained soil, so
its adequate prediction of the clay SWCC should be expected.

Figure 9. Total head contours, laboratory-

measured ML SWCC.
ERDC/GSL SR-17-4 21

Figure 10. Profile of pressure head values after five days for ML.
Pressure Head vs. Elevation
ML
50
ML
Lab
Tomasella
Rawls

40 Zapata
Perera
Sleep

30
Elevation (ft)

20

10

-30 -20 -10 0 10

Pressure Head (ft)

Figure 11. Profile of pressure head values after five days for compacted CH.
ERDC/GSL SR-17-4 22

Figure 12. Profile of pressure head values after five days for SM.
Pressure Head vs. Elevation
SM
50

40

30
Elevation (ft)

SM
Lab
20
Tomasella
Rawls
Zapata
Perera
Sleep
10

-40 -30 -20 -10 0

Pressure Head (ft)

Figure 13. Profile of pressure head values after five days for SP.
Pressure Head vs. Elevation
SP
50

SP
Lab
Tomasella
40
Rawls
Zapata
Perera
Sleep

30
Elevation (ft)

20

10

-30 -20 -10 0 10

Pressure Head (ft)
ERDC/GSL SR-17-4 23

Figure 14. Hydraulic conductivity function for ML.

Table 5. Results of transient seepage analysis MSE between predicted SWCC and laboratory-
measured SWCC.
Tomasella
Material and Hodnett Rawls Zapata Perera Sleep
ML 8.06E+02 9.34E+00 9.83E+02 8.08E+02 7.38E+02
SM 2.08E+01 5.58E-02 1.08E+01 1.27E+02 7.82E+01
CH 2.90E-06 2.02E+00 6.05E-01 3.46E+00 7.18E-03
SP 8.13E+02 8.75E+00 2.70E+00 2.71E+00 4.60E+02

Comparing the results from this rather limited analysis, Rawls et al.’s
(1991) method consistently matches the laboratory-measured SWCC
analysis’s results. Given the size of the data set, these results are very
preliminary; but each prediction method seems to give reasonable results,
with the method proposed by Rawls et al. (1991) being applicable across
the full range of soil types investigated, based on the pore pressure results.

Saturation Coefficient
Evaluating the relative significance of the differences in the SWCC
predictions on the results of a transient seepage analysis from the above
pressure profiles is rather difficult. To truly evaluate the significance of the
different SWCC curves, the solution must be described over the entire time
ERDC/GSL SR-17-4 24

domain. This has been done in the past either by plotting the evolution of
the pore pressure at certain locations over time or by providing profiles of
the pore pressure at discrete points in time (Lam and Fredlund [1984],
Cedergren [1997]). Using either of these approaches for fine-grained soils is
still rather inconvenient due to the sharp transition between unsaturated
and saturated soils at the wetting front. As seen by plots of pressure profiles
at different times for the CH column in Figure 15, the response throughout
much of the sample is essentially zero until the wetting front approaches, at
which point the pressure rapidly increases to a near constant value. Plotting
numerous profiles or pressure versus time curves on the same plot to
compare the various methods would produce a figure quite cluttered and
difficult to compare. Therefore, the concept of a normalized saturation
coefficient was developed to allow a continuous comparison of the transient
seepage solutions for each of the predictive methods.

Figure 15. Elevation versus pressure head at different times.

The normalized saturation coefficient (𝜔𝜔) is a dimensionless parameter

that varies from a value of zero at time zero to a value of 1.0,
corresponding to steady state seepage conditions. The normalized
saturation coefficient is computed as
ERDC/GSL SR-17-4 25

ω
 S dA   S dA
t i
(43)
 S dA   S dA
ss i

where 𝑆𝑆𝑡𝑡 is the saturation at time t , 𝑆𝑆𝑖𝑖 is the saturation corresponding to

the initial conditions, and 𝑆𝑆𝑠𝑠𝑠𝑠 is the saturation corresponding to the steady
state seepage solution. The integrals are evaluated over the entire seepage
domain of interest (the full column, in this particular case.) As seepage is a
diffusion process, the change in 𝜔𝜔 over time can be thought of as an
analogue to the average degree of consolidation. Just as the average degree
of consolidation describes the progress towards complete consolidation, 𝜔𝜔
describes the progress a transient wetting front has made towards the
steady state solution. The value of 𝜔𝜔 is the percent of available pore space
(pore space that would be saturated at steady state but is initially empty)
that has become saturated at any point in time. Using the concept of the
normalized saturation coefficient, a more thorough evaluation of the
differences in the SWCC predictions on the transient column analysis was
able to be performed.

A final numerical analysis was conducted to see how the predicted SWCC
and the associated prediction error impacted a numerical analysis by using
the finite difference program FLAC (Fast Lagrangian Analysis of Continua).
The same geometry used in the SEEP/W analysis was used in the FLAC
analysis. The two-phase-flow option was used in FLAC and, with this
option, only van Genuchten’s (1980) fitting parameters are available for use.
Not all of the SWCC prediction methods use van Genuchten’s fitting
equation, so each of the predicted SWCCs had to be fitted with this
equation. It was recognized that some additional error might have been
incorporated with this procedure, but the error was assumed to be small
when compared to error already present. Figure 16 shows the SM-predicted
SWCCs fitted with van Genuchten’s (1980) equation (the solid line
represents the predicted curve, and the dashed line represents the van
Genuchten fitted curve). Zapata et al.’s (2000) and Rawls et al.’s (1991)
SWCCs fit with van Genuchten’s equation very well while Sleep’s (2011)
equation did not. The likely reason for the poor fit to Sleep’s SWCC was its
odd shape.
ERDC/GSL SR-17-4 26

Figure 16. Comparison of predicted method and methods fitted with van Genuchten equation (1980).

The material properties used in the FLAC analysis are presented in Table 6.

The following define the properties shown in Table 6: a, b, and c are

constant parameters; and P0 is equivalent to van Genuchten’s α (1980) (a
in Equation 2). The constant a is equal to van Genuchten’s (1980) m, and b
and c are the l parameters from the hydraulic conductivity function as
shown in Schaap and Leij (2000). Instead of using hydraulic conductivity,
FLAC uses the mobility coefficient, which is simply the ratio of the
hydraulic conductivity to the unit weight of water.
ERDC/GSL SR-17-4 27

Table 6. Material properties used in the FLAC analysis.

FLAC Properties

SWCC origin Material Porosity a b c P0 (psf) Sr 4

κ (ft /lb*s)
Lab 0.44 0.11 0.5 0.5 17.30 0.00
Tomasella 0.45 0.26 0.5 0.5 18.51 0.02
Rawls 0.45 0.40 0.5 0.5 23.46 0.11
ML 5.26E-09
Zapata 0.45 0.22 0.5 0.5 431.00 0.00
Perera 0.45 0.23 0.5 0.5 2089.02 0.00
Sleep 0.45 0.27 0.5 0.5 519.61 0.00
Lab 0.41 0.16 0.5 0.5 355.05 0.00
Tomasella 0.465 0.29 0.5 0.5 26.82 0.17
Rawls 0.43 0.32 0.5 0.5 51.13 0.15
SM 1.60E-08
Zapata 0.43 0.23 0.5 0.5 705.95 0.00
Perera 0.43 0.23 0.5 0.5 5686.79 0.00
Sleep 0.43 0.15 0.5 0.5 2437.20 0.00
Lab 0.35 0.17 0.5 0.5 2155.00 0.00
Tomasella 0.60 0.12 0.5 0.5 47.81 0.14
Rawls 0.35 0.10 0.5 0.5 1987.62 0.54
CH 6.84E-13
Zapata 0.34 0.26 0.5 0.5 29246.34 0.00
Perera 0.34 0.28 0.5 0.5 68241.47 0.00
Sleep 0.34 0.15 0.5 0.5 2437.20 0.00
Lab 0.41 0.88 0.5 0.5 51.58 0.05
Tomasella 0.41 0.26 0.5 0.5 8.89 0.01
Rawls 0.42 0.43 0.5 0.5 11.33 0.10
SP 5.26E-05
Zapata 0.42 0.64 0.5 0.5 40.48 0.03
Perera 0.42 0.88 0.5 0.5 21.69 0.00
Sleep 0.42 0.68 0.5 0.5 28.85 0.04

The normalized saturation coefficient was computed at each time step in

FLAC as

ω
S tn An  Si n An
(44)
S ssn An  Si n An

where 𝜔𝜔 is the saturation coefficient, S is the degree of saturation for each

zone, A is the zone area, and n is the number of zones in the finite difference
grid. The saturation coefficient is calculated by first allowing the solution to
come to equilibrium with the initial conditions applied, which are the same
as described in the SEEP/W analysis. The initial saturation (Si from
ERDC/GSL SR-17-4 28

Equation 42) and the area were multiplied and summed over the entire
grid, and this value was stored for later use. The next step was to apply the
infiltration boundary conditions and run the analysis until steady state
conditions were reached. Once steady state conditions were reached, the
steady state zone saturations (𝑆𝑆𝑠𝑠𝑠𝑠 from Equation 42) and the areas were
multiplied and summed over the entire grid, and this value was stored. After
these two steps were completed, the program was reset to the initial condi-
tions and the infiltration boundary conditions were reapplied; this allowed
the saturation coefficient to be calculated and plotted against flow time.

Figure 17 shows the levee saturation coefficient plotted against time for the
ML material. The solid line represents the analysis using the laboratory-
measured SWCC data. The results that most closely matched the analysis
using the lab data were attained by the SWCCs predicted by the Rawls et
al. (1991) and Tomasella and Hodnett methods. The Perera, Sleep, and
Zapata methods all reached 90% of steady state much faster than the
laboratory analysis did. The times to reach 90% steady state for the Perera,
Sleep, and Zapata methods were 315.75, 807.87 and 873.29 days,
respectively, while at 1,000 days the analysis using lab data was at a
saturation coefficient value of 57.9%. From observation of Figure 17, it is
readily seen that all of the predictive methods except the Tomasella and
Hodnett method lead to conservative results.

Figure 17. Saturation versus time for the ML material.

Saturation Coeficient vs. Time
ML
1

0.8

0.6
Saturation Coeficient

0.4

ML
Lab
Perera
0.2
Rawls
Zapata
Sleep
Tomasella

0 200 400 600 800 1000

time (day)
ERDC/GSL SR-17-4 29

Figure 18 shows the saturation coefficient versus time for the CH material.
The results of this analysis show that Sleep’s (2011) method most closely
approximates the results of the numerical analysis using the laboratory-
measured SWCC. Tomasella’s predicted SWCC leads to a solution that
greatly underpredicts the wetting front when compared to the laboratory
analysis. The analyses performed using Perera’s, Rawls’s, and Zapata’s
predicted SWCC overpredict the location of the wetting front. At a time of
500,000 days, the saturation coefficient for the laboratory analysis was
44.1%, while for Perera’s, Rawls’s and Zapata’s analyses the coefficient
values were 79.6%, 86.5% and 86.4% respectively. The low values for
Tomasella and Hodnett’s predicted SWCC are likely due to the higher-
than-measured saturated volumetric water content, which is nearly double
the measured value. Also, the AEV for this SWCC is extremely low
compared to the other curves shown in Figure 5. Once again, all of the
predictive methods except for the Tomasella and Hodnett method lead to a
conservative estimate of the wetting front location.

Figure 18. Saturation versus time for the CH material.

The results of the analyses performed on the SM material are presented in

Figure 19. The analyses performed using Rawls’ and Tomasella and
Hodnett’s predicted SWCC both underpredict the saturation coefficient. At
ERDC/GSL SR-17-4 30

a flow time of 291.2 days, the laboratory analysis value of the saturation
coefficient was at 90% while the values for Rawls’ and Tomasella and
Hodnett’s analyses were 64% and 56%, respectively. The rest of the analyses
overpredicted the saturation coefficient, with Zapata’s most closely
matching the laboratory results. At a flow time of 291.2 days Perera’s,
Sleep’s, and Zapata’s saturation coefficients were 98.6%, 99.6%, and 99%,
respectively.

Figure 19. Saturation versus time for the SM material.

Saturation Coeficient vs. Time
SM
1.2

0.8
Saturation Coeficient

0.6

0.4
SM
Lab
Perera
Rawls
0.2
Zapata
Sleep
Tomasella

0 200 400 600

time (day)

The saturation coefficient versus time obtained for analysis of the SP

material is shown in Figure 20. The results of these analyses show that
Perera’s method matches the analysis using the laboratory data the best.
An MSE of 6.18E-4 was calculated for Perera’s predicted SWCC compared
to an MSE of 0.275, the poorest fit, by Tomasella’s predicted SWCC. A
saturation coefficient of 90% occurred for the laboratory analysis at
43.2 min, while at this same time the saturation coefficients for Rawls,
Sleep, Tomasella and Zapata were 23.6%, 63.9%, 24.6%, and 57.1%,
respectively. It should be noted that all predictive methods yielded non-
conservative results for the SP soil.
ERDC/GSL SR-17-4 31

Figure 20. Saturation versus time for the SP material.

Saturation Coeficient vs. Time
SP
1.2

0.8
Saturation Coeficient

0.6

SP
Lab
0.4 Perera
Rawls
Zapata
Sleep
Tomasella
0.2

0 0.04 0.08 0.12 0.16 0.2

time (day)

The results of this analysis showed that the normalized saturation

coefficient provides a useful way to characterize the results of a transient
analysis. Portraying the transient solutions in terms of 𝜔𝜔 allowed the
entire transient seepage process to be represented by a single numeric
value such that the transient solutions from using the various SWCC
predictive methods could readily be compared on a single graph. The
results show that the Rawls method of predicting the SWCC performed
well for the ML and SM materials but not for the SP and CH materials.
While Sleep’s method closely matched the laboratory CH and SP results, it
did not perform as well with the silty materials. The use of 𝜔𝜔 not only
evaluated the performance of the predictive methods over the entire time
domain of a transient solution but also clearly revealed which SWCC
predictive methods yielded conservative seepage analysis results. While
general statements regarding the appropriateness of each SWCC predictive
method cannot be made from the results of four comparisons alone, these
results can aid in the selection of an SWCC predictive method.
ERDC/GSL SR-17-4 32

5 How to Use Seep/W SWCC Prediction

Add-in
This section describes where to place the SWCC prediction functions and
in what format to input the various methods for successful use.

Installing Function (Add-In) in Seep/W

There are two directories that GeoStudio scans for Add-Ins. The first is
located where the GeoStudio binaries are installed, usually the C drive; this
is the location intended for core Add-Ins installation as part of the
GeoStudio software product. The second location is where custom Add-Ins
should be placed (Geo-Slope International 2012). The default location is in
the “Application Data” directory, but it can be set to any directory specified.
It is recommended that a directory be created in the “My Documents” folder
as the new default location for GeoStudio Add-Ins. Specifying a new
directory can be accomplished by navigating to Tools>Options in Geo-
Studio and then browsing and selecting the location desired.

The file titled “SWCCprediction.dll” should be placed in the location set as

the new default directory for Add-Ins in GeoStudio. Once this task has
been completed, the five prediction methods are ready to be accessed. To
perform a transient seepage analysis in Seep/W in the “KeyIn Analysis”
screen, “Transient Analysis” must be selected and the settings, control,
convergence, and time parameters adjusted as needed. Once the geometry
and grid have been defined, the user should navigate to the “KeyIn
Materials” screen, add a material, and select the “Saturated/Unsaturated
Material Model.” Once this is completed, the “Hydraulic Properties”
parameters will appear. Selection of the KeyIn Volumetric Water Content
button (labeled with three periods next to the volumetric water content
pull-down menu) will bring up the “KeyIn Vol. Water Content Functions”
screen. In the “Vol. Water Content Functions” screen, “Add a new
function” should be selected and the function named accordingly. Once a
new “Vol. Water Content Function” has been added, under the “Types”
pull-down menu, “Add-In Function” should be chosen. In the Add-In field,
“SWCC Prediction Functions” should be selected and the desired function
chosen. Once the desired function has been chosen, the desired
parameters must be entered, and “Close” selected. From the “Vol. Water
ERDC/GSL SR-17-4 33

Content Fn” pull-down menu, the function that was given a unique name
by the user should be selected. The following sections give guidance on the
format of each of the prediction method’s (function’s) input parameters.

Note: Each of the five methods is programmed to use imperial units and
will work as intended only if imperial units are used.

Method 1: (Zapata et al., 2000)

The SWCC prediction method outlined by Zapata et al. (2000) predicts the
SWCC based on the following parameters:

• porosity: volume of voids divided by the total sample volume, input as

a decimal
• D60: grain size for which 60% of the material is finer, required to be
input in millimeters (mm)
• PIpercent: the plasticity index, input as a percent
• percentPassing: the percentage of material finer than the #200
sieve, expressed as a decimal

The default data correspond to a material with the following parameters:

porosity is 0.5, D60 is 0.1 mm, PIpercent is 1.5, and percentPassing is 0.3.
If the wPI value is greater than 0, then the D60 parameter is not needed.

The method as outlined in the main report makes a distinction between

the fine-grained and the coarse-grained soils. A coarse-grained soil is
defined as a soil with a wPI value equal to zero, where wPI is the product
of the percent passing the #200 (expressed as a decimal) sieve and the
plasticity index (expressed as a percent). This method predicts the drying
curve of the SWCC, and this function’s name is “Zapata_Imp.”

Method 2: (Perera et al., 2005)

The SWCC prediction method outlined by Perera et al. (2005) predicts the
SWCC based on the following parameters:

• percentPassing: the percentage of material finer than the #200

sieve, expressed as a decimal
• D10, D20, D30, D60, D90: the diameter (mm) of material at each
percent passing. For example D10 is the grain size diameter for which
10% of the material is finer.
ERDC/GSL SR-17-4 34

• porosity: the volume of voids divided by the total sample volume,

input as a decimal.

Perera’s method of SWCC prediction is an expansion of the work performed

by Zapata et al. (2000). The coarse-grained soil parameters were expanded
to better enhance the prediction of SWCC for these types of materials. A
coarse-grained soil is defined as a soil whose wPI value is less than 1.0 and
has the same parameter described in Zapata’s method (Method 1). The
function name is “Perera_Imp.” The default material has the following
parameters: porosity is 0.5, PIpercent is 15, and percentPassing is 0.3. If a
material has a wPI value greater than 1.0, it is considered a fine-grained
material; and the D10 through D90 values need not be assigned.

Method 3: (Sleep 2011)

Sleep’s method of SWCC prediction is based on data extracted from the
UNSODA database. Due to the use of the textural classification system
used in the UNSODA database, the USCS classification is estimated by
using the saturated hydraulic conductivity values. Unlike the other
methods, this method allows the user to determine whether or not the
function returns one of four curves: the wetting, the 90% confidence
wetting, the drying, or the 90% confidence drying curves. The following
parameters must be defined in order to predict the desired curve:

• Curve: a value of 1 to 4 corresponding to the desired curve:

1. Average wetting curve
2. 90% confidence wetting curve
3. Average drying curve
4. 90% confidence drying curve
• ksat: the saturated hydraulic conductivity value expressed in cm/s
• porosity: the volume of voids divided by the total sample, input as a
decimal.

Clay materials with a hydraulic conductivity value less than 10-7 cm/s will
be assigned the range of SWCCs used for the silt material. This is due to
the absence of clay material in the UNSODA database. This function’s
name is “Sleep_Imp.”

Method 4: (Tomasella and Hodnett 1998)

The Tomasella and Hodnett method predicts the SWCC based on the
following parameters:
ERDC/GSL SR-17-4 35

• perClay: the percent clay of the material, based on the textural

classification (percent of particles finer than 0.002 mm,) expressed as
a percent
• perSilt: the percent silt of the material, based on the textural
classification ( percent of particles finer than 0.05 mm but coarser than
0.002 mm)

The porosity is not an input parameter: the function predicts the porosity
based on the two function input parameters, percent silt and clay. This
function’s name is “Tomasella_Imp”; this function predicts the drying
SWCC.

Method 5: (Rawls et al. 1991)

Rawls’s method of predicting the SWCC uses the following input parameters:

• Porosity: volume of voids divided by the total sample volume; it

corresponds to the saturated volumetric water content, input as a
decimal
• perSand: the percent sand of the material, based on the textural
classification (percent particles finer the 2 mm but coarser than
0.05 mm), expressed as a percent
• perClay: the percent clay of the material, based on the textural
classification (percent particles finer the 0.002 mm), expressed as a
percent

This function’s name is “Rawls_Imp”; this function predicts the drying

SWCC.
ERDC/GSL SR-17-4 36

6 Conclusion
Five soil water characteristic curve prediction methods were reviewed and
compared to independent laboratory data over a range of soil types. The
significance of the error in the predictive methods was assessed through
transient seepage analyses. The relationship between matric suction and
volumetric water content is of central importance when performing a
transient seepage analysis. This relationship significantly impacts the soil
hydraulic conductivity. The results of the numerical analysis show that, in
general, the Rawls predictive method gives an adequate prediction of the
SWCC over a range of material types. The results of the analyses also
demonstrate that the closest SWCC prediction does not guarantee the best
match to the transient analysis using the laboratory data curve: the shape
of the curve and the AEV are also important.

The concept of a normalized saturation coefficient was developed to

quantify the effects of parameter variation on transient seepage. The
coefficient provides a convenient means of displaying the response of any
geometry to a transient hydraulic loading. The saturation coefficient
illustrated more of a difference between the measured SWCC results and the
predicted SWCC results than was obtained by comparing pressure profiles
at a fixed time. This is in part due to the short time duration for which
pressure plots were generated. Nevertheless, the saturation coefficient
curves show the propagation of the wetting front through the model in one
continuous curve, reflecting the transient response of the whole geometry
instead of just one location. From the results of these analyses, it became
evident that only the method by Perrera et al. (2005) was conservative (or
nearly so) for all four soil types. Further comparisons should be conducted
on a larger data set of both natural and compacted soils to evaluate which
methods are consistently conservative for various soil types.

A SEEP/W Add-In function that provides four methods of approximating

the SWCC based on soil classification and index properties was developed.
It is highly recommended that these predictive methods be used only for
preliminary analysis, as they are not meant to replace laboratory-
measured SWCC, but only to give an initial estimate. The results also
indicate that when using a predictive method to derive the SWCC, it is
imperative to know which predictive method works the best for the
ERDC/GSL SR-17-4 37

particular soil that is to be analyzed. It seems that the best way to acquire
this information is to investigate from which soil data set the predictive
method was derived.
ERDC/GSL SR-17-4 38

References
Askarinejad, A., F. Casini, P. Kienzler, and S. M. Springman. 2011. Comparison
between the in situ and laboratory water retention curves for a silty sand. In
Unsaturated Soils-Proceedings of the 5th International Conference on
Unsaturated Soils, 6-8 September, Barcelona, Spain, 1:423-428.

ASTM. 2016. Standard practice for classification of soils for engineering purposes
(Unified Soil Classification System). In Annual Book of ASTM Standards.
ASTM 2487, 249-260. West Conshohocken, PA: American Society for Testing
and Materials.

Aubertin, M., M. Mbonimpa, B. Bussiere, and R. P. Chapuis. 2003. A model to

predict the water retention curve from basic geotechnical properties.
Canadian Geotechnical Journal 40(6):1104-1122.

Cedergren, H. R. 1997. Seepage, drainage, and flow nets. Vol. 16. Hoboken, NJ:
John Wiley & Sons.

Fredlund, D. G., H. Rahardjo, and M. D. Fredlund. 2012. Unsaturated soil

mechanics in engineering practice. Hoboken, NJ: John Wiley & Sons.

Fredlund, D. G., and A. Xing. 1994. Equations for the soil-water characteristic
curve. Canadian Geotechnical Journal 31(4):521-532.

Geo-Slope International Ltd. 2007. SEEP/W. Version 7.23, user’s guide. Calgary,
Alberta, Canada: GeoStudio.

Geo-Slope International Ltd. 2012. Add-ins programming guide and reference.

Calgary, Alberta, Canada: GeoStudio.

Lam, L., and D. G. Fredlund. 1984. Saturated-unsaturated transient finite

element seepage model for geotechnical engineering. Finite Elements in
Water Resources. Berlin Heidelberg: Springer.

Leij, F. J. 1996. The UNSODA unsaturated soil hydraulic database: user's

manual. Vol. 96, no. 95. National Risk Management Research Laboratory,
Office of Research and Development. Washington, DC: U.S. Environmental
Protection Agency.

Li, A. G., L. G. Tham, Z. Q. Yue, C. F. Lee, and K. T. Law. 2005. Comparison of

field and laboratory soil-water characteristic curves. Journal of Geotechnical
and Geo-environmental Engineering 131(9):1176-1180.

Likos, W. J., N. Lu, and J. W. Godt. 2013. Hysteresis and uncertainty in soil
water-retention curve parameters. Journal of Geotechnical and
Geoenvironmental Engineering 140(4), 04013050.

Lu, Ning, and W. J. Likos. 2004. Unsaturated soil mechanics. New York: John
Wiley & Sons.
ERDC/GSL SR-17-4 39

Nemes, A., M. G. Schaap, F. J. Leij, and J. H. M. Wösten. 2001. Description of the

unsaturated soil hydraulic database UNSODA version 2.0. Journal of
Hydrology 251(3):151-162.

Perera, Y. Y., C. E. Zapata, W. N. Houston, and S. L. Houston. 2005. Prediction of

the soil-water characteristic curve based on grain-size-distribution and
index properties. ASCE Geotechnical Special Publication 130. Reston, VA:
American Society of Civil Engineers.

Rawls, W. J., D. L. Brakensiek, and K. E. Saxton. 1982. Estimation of soil water

properties. Transactions of the ASAE 25(5):1316-1320.

Rawls, W. J., T. J. Gish, and D. L. Brakensiek. 1991. Estimating soil water

retention from soil physical properties and characteristics. In Advances in
soil science. New York: Springer.

Schaap, M. G., and F. J. Leij. 2000. Improved prediction of unsaturated

hydraulic conductivity with the Mualem-van Genuchten model. Soil Science
Society of America Journal 64(3):843-851.

Sleep, M. D. 2011. Analysis of transient seepage through levees. PhD diss.,

Virginia Polytechnic Institute and State University.

Song, Y. S., W. K. Hwang, S. J. Jung, and T. H. Kim. 2012. A comparative study of

suction stress between sand and silt under unsaturated conditions.
Engineering Geology, 124:90-97.

Tinjum, J. M., C. H. Benson, and L. R. Blotz. 1997. Soil-water characteristic

curves for compacted clays. Journal of Geotechnical and Geo-environmental
Engineering 123(11):1060-1069.

Terzaghi, Karl. 1996. Soil mechanics in engineering practice. New York: John
Wiley & Sons.

Tomasella, J., and M. G. Hodnett. 1998. Estimating soil water retention

characteristics from limited data in Brazilian Amazonia. Soil Science
163(3):190-202.

Tracy, F. T., T. L. Brandon, and M. K. Corcoran. 2016. Transient seepage

analyses in levee engineering practice. ERDC TR-16-8. Vicksburg, MS: U.S.
Army Engineer Research and Development Center.

van Genuchten, M. T. 1980. A closed-form equation for predicting the hydraulic

conductivity of unsaturated soils. Soil Science Society of America Journal
44(5):892-898.

Zapata, C. E., W. N. Houston, S. L. Houston, and K. D. Walsh. 2000. Soil-water

characteristic curve variability. ASCE Geotechnical Special Publication 99.
Denver, CO: American Society of Civil Engineers.
 Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining
the data needed, and completing and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for
reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington,
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not
display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS.
1. REPORT DATE (DD-MM-YYYY) 2. REPORT TYPE 3. DATES COVERED (From - To)
June 2017 Final
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER

SWCC Prediction: Seep/W Add-In Functions 5b. GRANT NUMBER

5c. PROGRAM ELEMENT NUMBER

6. AUTHOR(S) 5d. PROJECT NUMBER

HFK2F2
Lucas Walshire and Bryant Robbins 5e. TASK NUMBER

5f. WORK UNIT NUMBER

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION REPORT

NUMBER
Geotechnical and Structures Laboratory
U.S. Army Engineer Research and Development Center ERDC/GSL SR-17-4
3909 Halls Ferry Road
Vicksburg, MS 39180-6199

9. SPONSORING / MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)

Headquarters, U.S. Army Corps of Engineers HQ-USACE
Washington, DC 20314-1000
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)

12. DISTRIBUTION / AVAILABILITY STATEMENT

Approved for public release; distribution is unlimited.

13. SUPPLEMENTARY NOTES

14. ABSTRACT
The soil water characteristic curve (SWCC) defines a constitutive relationship between the negative pressure that develops when a soils
saturation level is less than fully saturated, and the corresponding volume of water held in the pore space of the soil matrix. As this
relationship is not commonly measured in geotechnical laboratories, practitioners often attempt to predict this relationship based on
other commonly measured material properties using empirical prediction methods. The performance of five SWCC empirical predictors
was evaluated through comparisons to independently measured SWCC data for four soils. SWCC prediction methods were selected for
this investigation if they incorporated commonly measured soil properties to predict the SWCC. The error in the SWCC prediction was
assessed in terms of both the mean squared error on the SWCC prediction and the impact of the error on a numerical analysis of the
Green and Ampt infiltration problem. The results of the numerical analysis were assessed in terms of a normalized saturation
coefficient. The normalized saturation coefficient provided a clear means of monitoring a transient seepage analysis through a single
measure. Results indicate that the SWCC prediction methods yielding the lowest mean squared error did not necessarily yield the
smallest error in the transient seepage analysis. Further, only the Rawls method consistently yielded conservative analysis results for all
soil types investigated.

15. SUBJECT TERMS Transient seepage Soil infiltration rate

Soil physics Soil water retention Soil matric potential
Soil moisture Soil water characteristic curve Seepage
16. SECURITY CLASSIFICATION OF: 17. LIMITATION 18. NUMBER 19a. NAME OF RESPONSIBLE
OF ABSTRACT OF PAGES PERSON
a. REPORT b. ABSTRACT c. THIS PAGE 19b. TELEPHONE NUMBER (include
area code)
Unclassified Unclassified Unclassified 45
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std. 239.18