Computers and Geotechnics 23 (1998) 255±275

E€ects of horizontal drains on slope stability

under rainfall by three-dimensional ®nite
element analysis
F. Cai*, K. Ugai, A. Wakai, Q. Li
Civil Engineering Department, Gunma University, Kiryu, Gunma 376-8515, Japan

Received 2 December 1997; received in revised form 6 May 1998; accepted 29 September 1998

Abstract

The e€ects of the horizontal drains on the ground water level during rainfall are predicted
using a three-dimensional ®nite element analyses of transient water ¯ow through unsaturated±
saturated soils. The slope stability is evaluated by the global safety factor, based on the three-
dimensional elasto-plastic shear strength reduction ®nite element method. The following
conclusions are obtained: (1) the horizontal drains can e€ectively lower the ground water level
and increase the slope stability under rainfall, but the rate of the increase in the safety factor
becomes smaller and smaller when the drains is extended beyond a critical length; (2) length-
ening the horizontal drains is more e€ective than making the spacing smaller and increasing
the number of the drains in a group in order to lower the ground water level and increase the
slope stability; (3) the e€ects of the drains are mainly in¯uenced by the ratio of the rainfall
intensity to the saturated hydraulic conductivity. # 1999 Elsevier Science Ltd. All rights
reserved.

1. Introduction

Rainfall, especially the heavy rain in summer and the long-term in®ltration of
melting snow, has been causing many landslides and slope failures. For instance, the
annular number of landslides and slope failure caused by rainfall is beyond 2700 in
recent years [1]. The in®ltration of rainfall results in the rise of the ground water

* Corresponding author. E-mail: cai@ce.gunma-u.ac.jp

0266-352X/99/$Ðsee front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S026 6-352X(98)0002 1-4
256 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

level, and the decrease in the matric suction and the shear strength of soils. This, in
turn, maybe leads to landslides and slope failures.
The horizontal drains is an e€ective measure to lower the ground water level and
to increase the slope stability during rainfall, especially for the dangerous slopes and
embankments [2]. The e€ects of the following parameters, i.e. the length, the spa-
cing, and the direction angle of the horizontal drains, on the ground water level and
the slope stability are the factors concerned in the design of the horizontal drains.
In the present paper, the e€ects of the horizontal drains on the ground water level
are analyzed with a three-dimensional ®nite element analysis of transient water ¯ow
through unsaturated±saturated soils. The slope stability is evaluated with the global
safety factor, obtained with the three-dimensional elasto-plastic shear strength
reduction ®nite element method. The pore water pressure is obtained from the
above-mentioned analysis of transient water ¯ow through unsaturated±saturated
soils. The conventional elasto-plastic FEM is modi®ed to predict the global safety
factor of slopes, whose de®nition is identical to the one in the conventional limit
equilibrium methods [3,4]. The shear strength of the unsaturated soils is expressed
with the Bishop's e€ective stress equation [5]. The e€ects of the length, the spacing,
and the direction angle of the horizontal drains on the ground water level and the
slope stability are numerically analyzed for a typical slope. Three sets of hydraulic
characteristic parameters of van Genuchten model are used to investigate their
in¯uences on the ground water level and the slope stability.

2. Modeling water ¯ow in soils

2.1. Fundamental ¯ow equation

The Darcy's law has been shown to be valid for the water ¯ow through unsatu-
rated soils as well as the ¯ow through saturated soils [6]. The main di€erence is that
the hydraulic conductivity is assumed to be constant for saturated soils, while it
depends on the pore volume occupied by water for unsaturated soils. Based on the
mass conservation and the Darcy's law, the di€erential equation governing the water
¯ow through unsaturated±saturated soils is given by [6]

@
r
…K…†r… ‡ z†† ˆ c…† …1†
@t

where K() is the hydraulic conductivity,  the volumetric moisture content,  the
pressure head, z the elevation head, t time, and c()(=@/@) the speci®c moisture
capacity.

2.2. Hydraulic characteristics

Eq. (1) includes two soil parameters that must be determined: the hydraulic con-
ductivity and the speci®c moisture capacity. These parameters under unsaturated
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 257

conditions are dependent on the volumetric moisture content, which is in turn rela-
ted to the pressure head. A widely used representation of the hydraulic character-
istics of unsaturated soils is the set of closed-form equations formulated by van
Genuchten [7], which is based on the capillary model of Mualem [8]. The soil-
moisture retention, the speci®c moisture capacity and the hydraulic conductivity
functions are given by
 ÿm
Se ˆ … ÿ r †=…s ÿ r † ˆ 1 ‡ jjn ; …2†

1=m m
c…† ˆ …n ÿ 1†…s ÿ r †S1=m
e …1 ÿ Se † ; …3†

 
m 2
K…† ˆ Ks Kr ˆ Ks S1=2
e 1 ÿ …1 ÿ S1=m
e † ; …4†

respectively, where

m ˆ 1 ÿ 1=n; n>1 …5†

and Se is the relative degree of saturation, and r and s denote the residual and
saturated volumetric moisture contents, respectively. Ks and Kr are the saturated
and the relative hydraulic conductivity, respectively. , n, and m are empirical
parameters of the hydraulic characteristics. The hydraulic functions are determined
by a set of ®ve parameters, r, s, , n, and Ks. The van Genuchten model is con-
sidered to provide a better match to the experimental data [9], although there are
some alternative models with parameters obtained more easily [10]. Leong and
Rahardjo comprehensively evaluated the models of the hydraulic characteristics of
soils [11,12].

2.3. Numerical approach

The ®nite element formulation for the transient water ¯ow through unsaturated±
saturated soils can be derived by the Galerkin principle of weighted residual [13].
The numerical integration of the Galerkin solution to the governing equation is
given by
@
D‡E ˆQ …6†
@t
where, D is the seepage matrix, E the capacitance matrix, and Q the ¯ux vector
re¯ecting the gravitational ¯ow and the ¯ux boundary conditions. The two-dimen-
sional approach has been adopted for some practical problems of interest [14±18].
The time derivative can be approximated with the following ®nite di€erence
algorithm
   
E E
lD ‡ ˆ Q ÿ …1 ÿ l†D ÿ t …7†

t t‡t t
258 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

When the parameter  is selected di€erently, the alternate method can be yielded
[13,19]. In the present analysis, the parameter =1/2 is used due to the second order
of accuracy. Because the hydraulic conductivity and the speci®c water capacity are
functions of the volumetric moisture content, Eq. (7) is highly nonlinear and is
solved by an iterative method.

3. Finite element analysis of slope stability

3.1. Shear strength reduction ®nite element method

The slope stability is commonly assessed using a limit-equilibrium analyses [20±

23]. Leshchinsky et al. [24] proposed the variational limit equilibrium method to
conduct the three dimensional analysis of slope stability. However, it is dicult to
extend any two dimensional limit-equilibrium analysis to the three dimensional, and
inadequate static assumptions often lead to paradoxical results [25]. When the water
pressure is considered, the three limit-equilibrium analysis is more dicult. How-
ever, the shear strength reduction ®nite element method can analyze the slope sta-
bility under a general frame. The numerical comparison shows that the elasto-plastic
shear strength reduction ®nite element method, in which the global safety factor is
identical to that in the conventional limit equilibrium methods, can yield nearly the
same safety factor and the corresponding critical sliding surface as the conventional
limit equilibrium methods under either two-dimensional and three-dimensional
condition [3,4].
The global safety factor of slopes, de®ned in the shear strength reduction ®nite
element method, is identical to the one in the limit equilibrium methods. The
0 0
reduced strength parameters cF and F are de®ned as
0
0 c
cF ˆ ; …8†
F

 0 
0 tan 
F ˆ tanÿ1 …9†
F

and replace the shear strength parameters c0 and 0 of the Mohr±Coulomb's failure
criterion. Stresses and strains are then calculated in the slope by the elasto-plastic
®nite element method. The initial F is selected to be so small that the soils of
the slope is under elastic condition. Then the value of F is increased step by step
until the global failure of the slope ®nally develops, which means that the ®nite
element calculation diverges under a physically real convergence criterion. The
global safety factor at failure lies between the F, at which the iteration limit is
reached, and the immediately previous value. The detailed procedure can be seen
elsewhere [3].
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 259

3.2. Shear strength of unsaturated soils

For unsaturated soils, the water phase occupies only parts of the pore volume,
while the remainder is covered by air. Bishop [5] introduced the -factor to account
for the fact, and suggested an equation for the e€ective stress of unsaturated soils.
The shear strength of unsaturated soils is then calculated as

0 0
f ˆ c ‡ ‰… ÿ ua † ‡ …ua ÿ uw †Š tan  ; …10†

where  f is the shear stress at failure, c0 the e€ective cohesion,  the total normal
stress, ua the pore-air pressure, uw the pore-water pressure, 0 the e€ective friction
angle, and  a parameter with the value between zero and unity, depending on soil
type and the degree of saturation.
Since the shear strength of unsaturated soils is strongly related to the amount of
water in the voids of soils, therefore, to the matric suction, OÈberg and SaÈllfors [26],
and Vanapalli et al. [27] proposed that the -factor can roughly be replaced by the
degree of saturation or the relative degree of saturation. The shear strength obtained
from the hypotheses are in good agreement with the experimental results, although
the suggested hypotheses have some limitations [26,27]. When the degree of satura-
tion is comparatively high, say, larger than 50%, there is a better correlation
between the predicted and measured values of the shear strength for unsaturated
soils.
The e€ect of the matric suction on the slope stability becomes smaller and smaller
with the in®ltration of rainfall, and disappears when the soil is saturated [28].
Therefore, the above-mentioned hypothesis, i.e. =Se, is used to consider the e€ect
of the matric suction on the shear strength for unsaturated soils.

4. E€ects of horizontal drains

4.1. Model slope

An idealized slope with the height of 10 m and the gradient of 1 V:1.5 H is ana-
lyzed with a three-dimensional ®nite element mesh, as shown in Fig. 1. According to
the spacing between the horizontal drains, the number of cross sections of the ®nite
element mesh is adjusted to keep the interval of 1.5±2.0 m between two cross sec-
tions. The initial ground water level is supposed to be horizontal and at the lower
ground surface. The slope and the ground are assumed to be composed of the same
soil. Three sets of the van Genuchten model parameters of the hydraulic character-
istics for the Glendale clayey loam (GCL) [4], the Uplands silty sand (USS) [29], and
the Bet Degan loamy sand (BLS) [30], as shown in Table 1, are used to investigate
their e€ects on the ground water level and the slope stability during rainfall with the
horizontal drains installed. The soil-moisture retention and the hydraulic con-
ductivity characteristics for each type of soil are shown in Fig. 2.
260 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

Fig. 1. Model slope and ®nite element mesh: (a) 3D ®nite element mesh, (b) ®nite element mesh on the
horizontal plane with horizontal drains when the direction angle is 45 , and L=15 m.

Table 1
Hydraulic properties of soils

Soil type  (mÿ1) n r s Ks (10ÿ4 cm/s)

GCL 1.0601 1.3954 0.106 0.469 1.516

USS 7.0870 1.8103 0.049 0.304 18.292
BLS 2.7610 3.0224 0.044 0.375 63.832

The e€ects of the horizontal drains on the slope stability during rainfall are pre-
dicted with the shear strength reduction ®nite element method, in which the pore
water pressure is obtained from the ®nite element analysis of transient water ¯ow
through unsaturated±saturated soils. In order to compare the e€ects of hydraulic
characteristics on the slope stability when the horizontal drains is installed, it is
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 261

Fig. 2. Hydraulic characteristics.

assumed that material parameters of GCL, USS and BLS are the same, as shown
in Table 2, in spite of these soils with di€erent grain characteristics and micro-
structures.
The initial relative degree of saturation is assumed to be the same, i.e. Se=0.617,
at the slope crest, and linearly increases to unity at the height of the initial water
level for each type of soil. The assumed initial relative degree of saturation is com-
paratively high, but it has been shown that the time histories of pressure head are
similar and the ®nal steady ground water level, which is also the most dangerous
ground water level, is the same for di€erent initial moisture contents [28].
The rainfall of the uniform intensity 10 mm/h is assumed to last 72 h. The calcu-
lated results show that the ground water level reaches the steady state after the
262 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

Table 2
Mechanical properties of soils

Parameter Value

Young's modulus, E (MPa) 98.1

Poisson's ratio,  (ÿ) 0.3
Unit weight, (kN/m3) 17.66
E€ective cohesion, c0 (kPa) 7.85
E€ective friction angle, 0 ( ) 25
Dilatancy angle,  ( ) 0

rainfall lasts a certain hours less than 72 h, and the following results is of the steady
state.

4.2. E€ects of horizontal drains length

The horizontal drains are installed at the height of the lower ground surface, as
shown in Fig. 1. The spacing of the horizontal drains is assumed to be 10 m, i.e.
S=5 m in Fig. 1, when the e€ects of the length are analyzed. The pressure head
along the horizontal drains is speci®ed as zero. The length of the horizontal drains,
L, from the slope toe is changed to clarify its e€ects on the ground water level and
on the slope stability.
The time histories of the pressure head without the horizontal drains and with the
drains of L=15 m for the appointed points in Fig. 1 show that the ground water
level during rainfall is e€ectively lowered for each type of soil with the horizontal
drains installed (Fig. 3). The time histories of the pressure head are similar for dif-
ferent length of the horizontal drains, although the pressure heads decrease with the
horizontal drains extended. There is an unsaturated zone with negative pressure
head in the middle of the GCL slope, i.e. there is a perched water zone with positive
pressure head above the unsaturated zone, due to the comparatively low hydraulic
conductivity. The unsaturated zone becomes smaller and smaller with in®ltration of
rainfall, and disappears at last. This results in that the ground water level suddenly
rises, i.e. the perched water is connected with the main ground water level. In the
contrast, the ground water level gradually rises in the USS and BLS slopes, until the
steady ground water level reaches (Fig. 4).
The contours of the pressure head for the longitudinal sections at the appointed
points in Fig. 1 show that the three-dimensional e€ect of the horizontal drains exists
in the slopes and it is more remarkable for the soils with lower hydraulic con-
ductivity (Fig. 5). However, the pressure head at the bottom of any longitudinal
section is about the same for each type of soil, so that the e€ects of the horizontal
drains on the ground water level can be con®rmed by the pressure head at the
appointed points in Fig. 1.
Fig. 6 indicates the relationships between the ratio of the pressure head with the
horizontal drains to that without the horizontal drains at the appointed points in
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 263

Fig. 3. Time histories of pressure head.

Fig. 1 and the length of the horizontal drains. The pressure head at the Points B, C,
and E without the horizontal drains is shown in Table 3. These relationships show
that the ground water level is e€ectively lowered for all three types of soil, but the
rate of the decrease in the ground water level at the point B and Point C becomes
smaller and smaller when the horizontal drains are extended beyond a critical
length, i.e. the distance between the slope toe and the slope shoulder.
The slope stability increases with the increase in the length of the horizontal
drains, but it does not increase further when the horizontal drains are extended
beyond the critical length, as shown in Fig. 7, where the increase in the safety factor,

F, means the di€erence between the safety factor with the drains and that without
the drains. This is because only the pressure head of the zones along the slip surface
264 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

Fig. 4. Contours of pressure head for cross section with horizontal drains.

in¯uences the slope stability. As shown in Fig. 8, the safety factor is almost linearly
dependent on the pressure heads at the Points B and C, and is not in¯uenced by the
pressure head at Point E, which is far from the slip surface. The results for various
spacing and direction angle of the horizontal drains are also included in Fig. 8.
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 265

Fig. 5. Contours of pressure head for the longitudinal sections at points: (a) Point B; (b) Point C; (c)
Point E.

The discharge ratio is de®ned as the ratio of the discharge rate through the hor-
izontal drains to the total rainfall above the slope surface and the slope crest. As
shown in Fig. 9, it is readily known that the discharge ratio increases with the length
of the horizontal drains. The ®nal in®ltration capacity is equal to the saturated
266 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

Fig. 6. Pressure head ratio versus length of horizontal drains.

Table 3
Pressure head at appointed points and safety factor of slopes without drains

Soil i=Ks Pressure head (m) Safety factor

Point B Point C Point E

GCL 1.83 9.75 11.84 14.46 0.84

USS 0.152 9.52 11.45 14.24 0.90
BLS 0.0435 7.43 8.32 9.72 1.10
*
Ratio of rainfall intensity to saturated hydraulic conductivity.

hydraulic conductivity because there is no matric suction. For the GCL slope, the
rainfall intensity is greatly larger than the saturated conductivity, the most part of
the rainfall forms the surface ¯ow and does not in®ltrate into the slope.
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 267

Fig. 7. Safety factor increase versus length of horizontal drains.

Fig. 8. Safety factor versus pressure head at di€erent points.

268 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

Fig. 9. Discharge ratio versus length of horizontal drains.

Fig. 10. Pressure head ratio versus spacing.

 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 269

The calculated results show that the e€ects of the horizontal drains on the ground
water level and on the stability of the typical homogeneous slope almost completely
depend on the ratio of the rainfall intensity to the saturated hydraulic conductivity
when the steady ground water level is reached, although the other of the hydraulic
characteristics of soils have in¯uences on the time-histories of the ground water.
Therefore, the foregoing and the following results under the steady state are e€ective
for the ratio of the rainfall intensity to the saturated hydraulic conductivity in the
used extent.

4.3. E€ects of horizontal drains spacing

The relationships between the pressure head ratio and the spacing of the hor-
izontal drains (L=7.5 and 15 m) are shown in Fig. 10. For all three types of soil the
ground water level during rainfall is de®nitely lowered only in the zones in the extent
of the horizontal drains with the spacing becoming smaller. When L=7.5 m, the
slope stability increases a little with the spacing becoming smaller. In contrast, when
L=15 m, the slope stability comparatively more increases with the spacing becom-
ing smaller (Fig. 11) except for the BLS slope, due to its very low ground water level

Fig. 11. Safety factor increase versus spacing.

270 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

even when the spacing of the horizontal drains is 20 m. By comparing the rate of the
increase in the safety factor in Fig. 11, it is shown that lengthening the horizontal
drains is more e€ective to lower the ground water level and increase the slope sta-
bility during rainfall when the length of the horizontal drains is shorter than the
critical length.
The discharge ratio decreases with the increase in the spacing of the drains
(Fig. 12). The discharge ratio mainly depends on the ratio of the rainfall intensity to
the saturated hydraulic conductivity. For the USS and BLS slopes, the ratio is less
than unity, thus the most part of the rainfall is discharged by the drains, and the
part, discharged by seepage ¯ow, increases with the increase in the spacing. For the
GCL slope, only small part of the rainfall is discharged by the horizontal drains.
The discharge ratio almost linearly increases with the spacing becoming smaller for
all three types of soils.

4.4. E€ects of horizontal drains direction

The horizontal drains are installed in a group in the horizontal plane for easy
construction. It is assumed that the group consists of three horizontal drains, and
only half of the group is analyzed due to its symmetry (Fig. 1). The direction angle
of the horizontal drains, as shown in Fig. 1, is assumed to be 0 , 30 , 45 , 60 , and

Fig. 12. Discharge ratio versus spacing.

 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 271

90 , respectively. When the length of every horizontal drains in the group L=7.5 m,
the spacing is assumed to be 15 m, i.e. S=7.5 m. When L=15 m, the spacing is
assumed to be 30 m, i.e. S=15 m.
The relationships between the pressure head ratio and the direction angle of the
horizontal drains (Fig. 13) shows that for all three types of soil, the direction angle
in¯uences slightly the pressure head ratio. When the direction angle is 30 , and the
length reaches the critical length, the ground water level is the lowest, and the pres-
sure head decreases only 10% further. This implies that the ground water level is
slightly lowered further during rainfall with the increase in the number of the hor-
izontal drains in a group if the length of the drains in the group is the same. This is
because the ground water level has been lowered with installing the horizontal
drains with the direction angle of zero degree and the other drains in the group are
installed in the region with the lowered ground water level.

Fig. 13. Pressure head ratio versus direction angle.

272 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

Fig. 14. Safety factor increase versus direction angle.

The direction angle of 30 obtains the maximum increase in the safety factor
whether L=7.5 or 15 m (Fig. 14). The increase in the safety factor for the spacing of
30 m and the length of 15 m is larger than that for the spacing of 15 m and the
length of 7.5 m. Under such condition, the total length of the horizontal length is the
same. This implies that the lengthening the horizontal drains in the extent of the
critical length is more e€ective than increasing the number of the horizontal drains
in the group in order to lower the ground water level and to increase the slope sta-
bility.
The discharge ratio increases a little with the number of drains in a group, and the
direction angle of the drains almost does not in¯uence the discharge ratio, as shown
in Fig. 15, where the discharge ratio with the direction angle of 90 includes the
seepage ¯ow through the toe of the slope.

5. Conclusions

The three-dimensional ®nite element analyses have been conducted to investigate

the e€ects of the horizontal drains on the ground water level and the slope stability,
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 273

Fig. 15. Discharge ratio versus direction angle.

where the e€ects of the length, the spacing and the direction angle on the ground
water level in slopes under rainfall are predicted with the three-dimensional ®nite
element analyses of transient water ¯ow through unsaturated±saturated soils. The
slope stability is evaluated by the global safety factor, based on the elasto-plastic
shear strength reduction ®nite element method. From the calculated results, the
following conclusions are obtained:
(1). The procedure proposed in the present paper can be used to determine the
length, the spacing of the horizontal drains, which is used to stabilize dangerous
slopes under rainfall. The signi®cant three-dimensional e€ect exits when the hor-
izontal drains are installed in the slopes, so the three-dimensional analyses can give
more realistic results.
(2). The ground water level is e€ectively lowered for all three types of soils due to
the drainage e€ect of the horizontal drains. The slope stability increases with the
increase in the length of the horizontal drains, but the rate of the increase in the
safety factor of the slopes becomes smaller and smaller when the horizontal drains
are extended beyond a critical length, i.e. the distance between the slope toe and the
slope shoulder.
(3). In order to lower the ground water level and increase the slope stability under
rainfall, lengthening the horizontal drains is more e€ective than making the spacing
274 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275

smaller and increasing the number of the horizontal drains in a group when the
length of the horizontal drains is shorter than the critical length.
(4). The e€ects of the horizontal drains on the ground water level, in turn, the
slope stability depends mainly on the ratio of the rainfall intensity to the saturated
hydraulic conductivity when the steady ground water level is reached under rainfall,
although the other hydraulic characteristics of soils have in¯uences on the time-his-
tories of the ground water level.

References

[1] Fujita H. State of the art of the landslide control in Japan and the issues in the future. Journal of the
Japanese Geotechnical Society 1997;45(6):1±4 (in Japanese)
[2] Mallawaratchie DP, Thuraisamy M, Jayamanne J, Attanayake AML. Stage-1 remedial measures for
stabilizing landslide at Beragala Haliela Road. In: Senneset K. (ed.). Proceedings of the Seventh
International Symposium on Landslides, vol. 3, Norway: Trondheim, 1996:1759±64
[3] Ugai K. A method of calculation of global safety factor of slopes by elasto-plastic FEM. Soils and
Foundations 1989; 29(2):190±5 (in Japanese)
[4] Ugai K, Leshchinsky D. Three-dimensional limit equilibrium and ®nite element analyses: a compar-
ison of results. Soils and Foundations 1995;35(4):1±7
[5] Bishop AW. The principle of e€ective stress. Publication 32, Oslo: Norwegian Geotechnical Institute,
1959:1±4
[6] Richards LA. Capillary conduction of liquids through porous mediums. Physics 1931;1:318±33
[7] van Genuchten MT. A closed-form equation for predicting the hydraulic conductivity of unsaturated
soils. Soil Science Society of America Journal 1980;44:892±8
[8] Mualem Y. A new model for predicting the hydraulic conductivity of unsaturated porous media.
Water Resources Research 1976;12:513±22
[9] Stankovich JM, Lockinton DA. Brooks-Corey and van Genuchten soil-water-retention models.
Journal of Irrigation and drainage engineering ASCE 1995;121(1):1±7
[10] Brooks RH, Corey AT. Properties of porous media a€ecting ¯uid ¯ow. Journal of Irrigation and
Drainage Engineering ASCE 1966;92(2):61±88
[11] Leong EC, Rahardjo H. Review of soil-water characteristic curve equations. Journal of Geotechnical
and Geoenvironmental Engineering ASCE 1997;123(12):1106±17
[12] Leong EC, Rahardjo H. Permeability functions for unsaturated soils. Journal of Geotechnical and
Geoenvironmental Engineering ASCE 1997;123(12):1118±26
[13] Zienkiewicz OC, Tayor RL. The ®nite element method. 4th ed. London: McGraw-Hill, 1989
[14] Neuman SP. Saturated-unsaturated seepage by ®nite element. Journal of Hydraulic Engineering
ASCE 1973;99(12):2233±49
[15] Akai K, Ohnishi Y, Nishigaki M. Finite element analysis of saturated-unsaturated seepage in soil.
Proceedings of the Japan Society of Civil Engineers 1977;264:87±96 (in Japanese)
[16] Thomas HR, Rees SW. Modeling ®eld in®ltration into unsaturated clay. Journal of Geotechnical
Engineering ASCE 1990;116(10):1483±501
[17] Thomas HR, Rees SW. A comparison of ®eld-monitored and numerically predicted moisture
movement in unsaturated soil. International Journal for Numerical and Analytical Methods in
Geomechanics 1991;15(6):417±31
[18] Thomas HR, Rees SW. The numerical simulation of seasonal soil drying in an unsaturated clay soil.
International Journal for Numerical and Analytical Methods in Geomechanics 1993;17(2):119±32
[19] Bathe K-J. Finite element procedures. Englewood Cli€s, New Jersey: Prentice Hall, 1996
[20] Bishop AW. The use of the slip circle in the stability of slopes. Geotechnique 1955;5(1):7±17
[21] Janbu N. Stability analysis of slopes with dimensionless parameters. Harvard soil mechanics series,
No. 46: 1954
 F. Cai et al./Computers and Geotechnics 23 (1998) 255±275 275

[22] Morgenstern NR, Price VE. The analysis of the stability of generalized slip surfaces. Geotechnique
1965;15(1):79±93
[23] Spencer E. A method of analysis of the stability of embankments assuming parallel inter-slice forces.
Geotechnique 1967;17(1):11±26
[24] Leshchinsky D, Baker R, Silver ML. Three dimensional analysis of slope stability. International
Journal for Numerical and Analytical Methods in Geomechanics 1985;9(2):199±223
[25] Leshchinsky D. Discussion on three dimensional atability of concave slope in plan view. Journal of
Geotechnical Engineering ASCE 1990;116(2):342±5
[26] OÈberg A-L, SaÈllfors G. Determination of shear strength parameters of unsaturated silts and sands
based on the water retention curve. Geotechnical Testing Journal 1997;20(1):40±8
[27] Vanapalli SK, Fredlund DG, Pufahl DE, Clifton AW. Model for the prediction of shear strength
with respect to soil suction. Canadian Geotechnical Journal 1996;33(3):379±92
[28] Cai F, Ugai K, Sakajo S. Prediction of rainfall e€ects on slope stability by ®nite element methods.
Submitted to Soils and Foundations, 1997
[29] Russo D, Bresler E. Field determinations of soil hydraulic properties. Soil Science Society of Amer-
ica Journal 1980;44:697±702
[30] Staple WJ. Moisture redistribution following in®ltration. Soil Science Society of America Proceed-
ings 1969;33:840±7