You are on page 1of 8

Available online at www.sciencedirect.

com

Computers and Geotechnics 35 (2008) 223230


www.elsevier.com/locate/compgeo

A simple method to analyze inltration into unsaturated soil slopes


Kenneth Gavin 1, Jianfeng Xue

School of Architecture, Landscape and Civil Engineering, Earlsfort Terrace, University College Dublin, Dublin 2, Ireland
Received 25 August 2006; received in revised form 3 April 2007; accepted 4 April 2007
Available online 21 May 2007

Abstract
Assessment of the stability of embankment and cut slopes over the life of a project are critical issues for railway and motorway infrastructure projects. Experience has shown that many slope failures occur during or shortly after rainfall. Analyses show that failure is
initiated by the reduction in near surface suction over some critical depth in the slope. A simple method is proposed in this paper to
estimate the time needed for a wetting front to develop. The method which is a modication of the traditional GreenAmpt inltration
model assumes that ponding of water cannot occur on soil slopes and as a consequence soil in the wetted zone remains partially saturated
at the point of slope failure. It dierentiates between cases where the initial suction in the slope is high and the rate of inltration is controlled by the rainfall intensity (supply controlled) and, situations where the suction is low, and the rate of inltration is controlled by the
inltration capacity of the soil (demand controlled). When applied to a case history where eld measurements of inltration into a slope
were available the new method provided a reasonable approximation of the measured inltration time.
 2007 Elsevier Ltd. All rights reserved.
Keywords: Inltration; Unsaturated soil slopes; Wetting front; Inltration capacity

1. Introduction
In tropical and sub-tropical areas, the water table is at
great depth and the near surface soils experience negative
pore-water pressure (suction). Similar conditions exist in
embankments and other engineered lls placed above the
water table level even in temperate climates. Experience
has shown that many slope failures occur during or shortly
after rainfall, as water inltrates into the slope reducing
near surface suction. Given changing climatic conditions
(increasing and more intense rainfall events) and deforestation which removes the natural suction eects provided by
tree roots, landslides are becoming much more prevalent
throughout the world. In addition to the risk posed to life
by rapid landslide events, additional problems are caused

Corresponding author. Tel.: +353 (0) 1 7167374; fax: +353 (0) 1


7167399.
E-mail addresses: kenneth.gavin@ucd.ie (K. Gavin), jianfeng.xue@
ucd.ie (J. Xue).
1
Tel.: +353 (0) 1 7167292; fax: +353 (0) 1 7167399.
0266-352X/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2007.04.002

by the interruption of transport routes and contamination


of water supplies.
Careful eld measurements [1] have identied the key
role of suction in maintaining the stability of steep slopes,
and in providing a rational failure mechanism for rainfall
induced landslides. Failure generally takes the form of
shallow (typically <2 m deep), translational slides, which
form parallel to the original slope surface. The mechanism
of slope failure described by Fourie et al. [1], Lumb [2] and
others comprises the downward migration of a wetting
front from the surface of the slope. The advance of water
causes a reduction of near surface soil suction, reducing
the soil strength and leading to failure. Fredlund and Rahardjo [3] noted that traditional saturated soil mechanics
approaches cannot model the problem successfully as the
soil is unsaturated before and in many cases after failure.
Fredlund et al. [4] expanded the MohrCoulomb model to
incorporate negative pore-water pressure (matric suction):
s c0 rn  ua tan /0 ua  uw tan /b
C rn  ua tan /0

224

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

where s is the shear strength of unsaturated soils, c 0 is the


eective cohesion, rn is the total normal strength on the
failure plane, ua is the pore-air pressure on the failure
plane, / 0 is the angle of internal friction associated with
the net normal stress state variable rn  ua, uw is the
pore-water pressure on the failure plane, ua  uw is the
matric suction on the failure plane, /b is the angle indicating the rate of increase in shear strength relative to the matric suction and C is the total cohesion of the soil.
It is common to consider the total cohesion (C) of unsaturated soil as been composed of two parts, the rst of
which is the eective cohesion and the second represents
the contribution of suction to strength. Water ow through
soil generates seepage (drag) forces, which change the eective stress. The eect is considered to be relatively minor
and is ignored in the following.
Fourie et al. [1] and Pradel and Raad [5] show that when
slope failure is caused by inltration, the failure plane
forms parallel to the existing slope surface. The authors
suggest using an innite slope model in which the soil
strength is described by an expression of the form given
in Eq. (1). The factor of safety (F) is given by:
F

C ch cos2 a tan /0
C
cos a tan /0

ch cos a sin a
ch cos a sin a
sin a

in which C is the total cohesion (incorporating the eect of


suction and cementation), c is the unit weight of the soil, a
is the angle of the slope and h is the depth of slip surface,
which is normally related to the depth of wetting front
depth Zf.
The second part of Eq. (2) remains constant during rainfall, therefore failure of an unsaturated soil slope is caused
by the total cohesion reducing as inltration reduces the
in situ suction and increases the wetting front depth. This
results in an increase in the disturbing forces and a decrease
in the resisting force. Eventually the soil strength will
reduce suciently that in combination with the increasing
load (weight of soil), failure will occur.
The loss of matric suction and the advance of the wetting front are controlled by the inltration characteristics
of the slope. An ability to model this process is critical to
an accurate assessment of slope stability problems. Because
of the importance of suction in maintaining the stability of
slopes, the variation of the strength of unsaturated soil as a
consequence of changes in water content is the topic of
intensive research [4]. Estimating the change of water content that will occur during a given rainfall event is an extremely complex problem. It is usually solved by either:
(i) Simple physical models such as GreenAmpt [6] and
Horton [7] equations.
(ii) Finite element analysis (FEA), which usually employ
Richards equation [8] and incorporate input data
from the soil water characteristic curve (SWCC)
which accounts for the variation in soil permeability
as the matric suction (or water content) varies.

Although the FEA approach is theoretically rigorous,


unsaturated soil permeability is highly variable during inltration and a rigorous analytical technique may still not yield
correct results if the input parameters are uncertain. For this
reason models such as GreenAmpt remain in widespread
use due to their relative simplicity. This paper reviews several
of the issues involved in analysing inltration into unsaturated soil slopes using the GreenAmpt model. Some of
the well-known limitations of the model are examined and
a simple alternative approach which overcomes some of
these limitations is proposed. The new model should be used
for preliminary design and to identify critical slopes where
FEA analyses are required. In the nal section predictions
of the inltration response of slopes made with the new
model are compared to both the GreenAmpt model and
numerical analyses using FEA.
2. Inltration analysis
During rainfall water inltrates the soil from the surface
and redistributes in the unsaturated zone. The distribution
process depends upon the soil moisture conditions, water
pressure, and unsaturated permeability. The inltration
capacity (i), which is a measure of the maximum rate at
which water can enter the soil, varies throughout a rainfall
event. It is controlled mainly by the permeability and water
content of the soil and the topography of the slope. Generally in unsaturated slopes the inltration capacity is initially high as large suction pressures are present which
compensate for the relatively low unsaturated permeability
(K) of the soil. As inltration continues the in situ suction
and i reduce. The inltration rate (I) is the rate of water
supply to the slope, and generally speaking is equal to or
less than the rainfall intensity. Cracking of low permeability soils can cause preferential ow paths to develop, in
which the ow of water is independent of the mass permeability of the soil matrix. Such eects are not usually critical and are ignored in simple inltration models.
The complexity of the inltration response of natural
slopes is illustrated by eld measurements made by Rahardjo et al. [9]. These show that during the initial stages
of low intensity rainfall events or when the initial i was
high, all water inltrated the slope. As the storage capacity
of the soil is approached, run-o begins. The time at which
run-o commenced was aected by antecedent rainfall or
initial water content (suction). The response varied with
rainfall intensity, with higher rainfall intensities resulting
in earlier occurrence, and larger quantities of run-o. An
additional factor aecting the inltration rate was location
on the slope, with instrumentation showing preferential,
sub-horizontal drainage causing suction to be lowest (in a
given cross-section) near the toe of the slope. Given the
complexities involved, the common assumption that all
rainfall will inltrate the slope when the rainfall intensity
is lower than the permeability of the soil [10], appears to
be a reasonable, yet conservative assumption.

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

The GreenAmpt model [6] is described in more detail in


the following section.
2.1. GreenAmpt model
The GreenAmpt model [6] was originally derived to
predict the inltration of ponded water into underlying
soil. The basic assumption underpinning the model is that
inltration causes the development of well-dened wetting
front (see Fig. 1). The soil above the wetting front is fully
saturated, whilst below the wetting front it remains at the
initial (pre-inltration) water content. Gravity and matric
suction eects control the movement of water in the saturated zone, and the hydraulic gradient (hwf) at the wetting
front is:
hwf

hp Z f S
Zf

where hp is the depth of ponded water, Zf is the wetting


front depth and S is the suction at the wetting front.
Although the model was originally developed to analyse
inltration into homogenous soils with ponded water,
extensive development has included extensions to account
for inltration under steady [10] and unsteady [11] rainfall
events. However, the basic assumption of the two layer
model (with the wetting front forming a boundary between
the saturated and unsaturated zones) is ubiquitous.
Using the GreenAmpt model in Fig. 1 and neglecting
ponded water, Brakensiek [12] showed that by applying
Darcys Law, the inltration capacity of the soil under
non-ponded conditions at time t can be calculated with:


Zf S
i Ks
4
Zf
in which i is the inltration capacity, Ks is the permeability
of saturated soil.
In this model the inltration capacity converges to Ks as
the soil approaches the fully saturated condition. Therefore, run-o will not occur when the rainfall intensity Ir
is 6Ks [10]. However, laboratory and eld measurements
of run-o have shown that this is not the case. Li et al.

225

[13] demonstrated that run-o began before the near surface soils become fully saturated, at Ir values between
0.2Ks and 0.4Ks. Rahardjo et al. [9] reported in situ inltration measurements on a residual slope with Ks of
5.18 106 m/s. A test conducted by applying a rainfall
intensity of 13 106 m/s to the initially unsaturated slope
revealed that the inltration capacity of the slope converged to 2 106 m/s (0.4 Ks), after 70 min.
3. New model considering suction variation
In this section a simple method (based on the Green
Ampt model) is developed to allow the assessment of inltration into unsaturated soil slopes using basic soil
properties.
3.1. Inltration mechanism in unsaturated soils
One of the main assumptions in the GreenAmpt model
is that the soil in the wetted zone is fully saturated. However, eld measurements [1,14] and others show that failure
occurs before the soil above the wetting front becomes fully
saturated. If the soil is partially saturated, the water phase
is not continuous (see Fig. 2), and the hydraulic head in
this zone is controlled exclusively by matric suction. The
hydraulic head Zf in Eq. (4) is therefore not applicable to
unsaturated soils under these conditions.
Making the assumption that for a slope which remains
partly saturated at failure:
1. The soil is continuously supplied with water but not
fully saturated within the wetted zone (see Fig. 2). The
assumption is generally restricted to the soils slopes,
where ponding cannot occur and the supply of water
to the soil is therefore limited.
2. After rainfall the nal suction prole in the wetted zone
is linearly distributed within the wetting front (see
Fig. 3). Although both the initial and nal suction pro-

Slope surface
wetted soil

hpH

Infiltrating water

wetting front

X
Saturated soil

Wetting front;
matric suction pulls
water into dry soil

drier soil

Zf

soil particle

Wetting front
moves down
into dry soil

Z
Fig. 1. Wetting front development in the GreenAmpt model.

water
air
Z
Fig. 2. Wetting front developed in an unsaturated soil slope.

226

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

60

e
fac

afte

tion

suc

ll

St

ore

bef

fa
rain

ng
etit

nt

fro

w
H
Sb

suction profile

(Ua-Uw)r

Volumetric metric content (%)

slo

ll

fa
rain

r
e su

Air entry value

40

20

Residual suction

0
0.1

Fig. 3. Suction prole assumed in the new model.

les are often non-linear as illustrated by Zhan and Ng


[15], the analysis is greatly simplied by modelling complex non-linear suction proles with several discrete linear functions.
3. The permeability of the soil above the wetting front is
uniform with depth and time. In the class of simple inltration models considered, the soil permeability is often
assigned a constant value (usually the saturated permeability because of the assumption that the soil becomes
fully saturated during inltration). If the soil is partially
saturated the permeability depends on the negative porewater pressure (or the degree of saturation) [21]. The
variation of permeability (or water content) with change
in suction can be measured using a soil water characteristic curve (SWCC). Whilst the exact form of the SWCC
will depend on the soil type and whether the soil is experiencing wetting or drying, Zhan and Ng [15] describe
the general form of an SWCC shown in Fig. 4, which
shows the water content varying from the fully saturated
condition at zero suction, to a residual water content at
very high suction. It is clear that the SWCC relationship
is highly non-linear at suctions close to the saturated
and residual values, whilst at intermediate suctions it
is relatively linear (noting that suction is plotted on a
log scale). When suction is large (at the start of a rainfall
event), the inltration capacity of the soil can exceed the
rainfall intensity (water supply rate) and the rate of ow
of water into the soil is controlled by the rainfall intensity rather than the permeability. As the wetting front
forms and the near surface suctions reduce, the inltration capacity may become lower than the rainfall intensity and the inltration rate will then be controlled by
the permeability. Therefore, an appropriate average permeability, applicable over the range of suction values
where inltration is controlled by the soil permeability
can be chosen from the SWCC curve for the soil. Whilst
it is strongly recommended that the permeability of the
soil over the appropriate range of suction values is measured using the SWCC, in the absence of specic information the recommendation by Bouwer [16] that
K = 0.5Ks is adopted in this paper.

10
100
Ua-Uw (kPa)

(Ua-Uw)r

1000

10000

Fig. 4. Typical soil-water characteristic curve (after Zhan and Ng [15]).

Due to the continuous supply of water at the ground


surface (during rainfall), the matric suction at the ground
surface is zero. Setting the ground surface as the reference
elevation (where the total hydraulic head is zero) and given
the suction value at a depth y is Sy (see Fig. 3), the hydraulic gradient (hi) between the surface and the depth of y is:
hi

S y  0 S y

y
y

The inltration capacity at the depth y can be expressed


as:
iK

Sy
y

in which K is the permeability of the unsaturated soil in the


wetted zone.
In Eq. (6), the inltration capacity is controlled by the
permeability of the soil and the hydraulic gradient due to
matric suction. The inltration capacity will be greater
than the permeability of the soil when Sy/y > 1. This situation occurs in dry soils with high initial matric suction.
Under these conditions, an initially high hydraulic gradient
can compensate for the low unsaturated soil permeability
and result in a large inltration capacity [17]. If the inltration capacity is larger than or equal to the rainfall intensity
(i > Ir), all rainfall will inltrate into the slope. Therefore,
the inltration rate is controlled by the rainfall intensity
Ir. As the suction values in the wetted zone decrease and
the wetting front depth increases, the hydraulic gradient
and inltration capacity decrease and runo will start once
the inltration capacity is lower than the rainfall intensity.
3.2. Numerical model
During the initial stage of inltration when suction values are at a maximum and the slope has a large inltration
capacity, which is greater than the rainfall intensity. From
Eq. (6) we have:

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

S y =yK > I r

Rewriting:
S y > I r =Ky

Thus suction values larger than (Ir/K)y, indicate that the


inltration capacity is greater than the rainfall intensity.
The ratio Ir/K, the relative rainfall intensity (Ri), describes
the ratio between the actual rainfall intensity and the permeability of the soil.
If we divide the suction prole of the soil into two zones
with the line Sy = (Ir/K)y (see Fig. 5), in which the suction
values are denoted in meters of water. In zone 1, we have:
S y > I r =Ky
In this zone, the inltration rate is controlled by the
rainfall intensity Ir and all the water will inltrate into
the soil.
Considering zone 2 in Fig. 5, we have:
S y < I r =Ky
and:
i S y =yK < I r

In this zone the inltration capacity is lower than the


rainfall intensity. The inltration rate is thus controlled
by the actual inltration capacity, which is determined by
the suction head and the permeability of the soil. At this
stage, as the rainfall intensity exceeds the inltration capacity, run-o occurs.
According to the law of mass conservation, in zone 1, we
have:
I r dt Dh dy
Rewriting the equation and integrating with depth (y),
we have the time required to form the wetting front to
depth H1 in zone 1 (Fig. 5):
Dh1 H 1
T1
Ir

10

227

in which Dh1 is the change of the volumetric water content


(based on the total volume of the soil) in zone 1.
As inltration continues, suction values in the soil
decrease and eventually fall into zone 2. In this zone, the
inltration capacity is lower than the rainfall intensity
and the actual inltration rate (I) is given by:
Sy
I iK
y
Therefore:
 
Sy
K
dt Dh2 dy
y
By integration we get:
T2

Dh2 H 2  H 22
2KS b

11

where Sb is the suction value at wetting front. Noting that


at this stage the inltration capacity is controlled by the
unsaturated soil permeability and hydraulic gradient.
Then the total time needed to form the wetting front is
the sum of T1 and T2:
T

Dh1 H 1 Dh2 H 2  H 22

2KS b
Ir

12

In order to analyze a given slope, it is necessary to divide


the suction prole into two zones using the line: Sy = (Ir/
K)y (see Fig. 5). The calculation accuracy can be improved
if the wetted zone is divided into several segments with
depth. This procedure also allows complex suction proles
to be modeled. The initial and nal volumetric water content and the corresponding matric suction, and the permeability of the soil are required as input parameters in the
model and some guidance to the choice of these parameters
is appropriate:
(i) The initial water content or suction prole can be
measured with relative ease using moisture probes
or tensiometers.
(ii) The nal water content or suction prole (which
results in a failure surface developing) can be assessed
based on a stability analysis performed by assuming
the advancing wetting front causes a planar failure
surface to develop [1,21].
(iii) With respect to the permeability value to be adopted,
a representative value should be chosen from values
measured over the appropriate range of suction from
the SWCC of the soil. Where no such information is
available the assumption that K = 0.5Ks should give a
reasonably conservative estimate for cases where
ponding cannot occur.
4. Application of the inltration model
4.1. Comparison with nite element method

Fig. 5. Two zones in the suction prole within wetting front.

Tami et al. [18] reported numerical simulations using


SEEP/W (Geoslope Int.) performed to investigate bound-

228

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

ary eects in their 2.45 m long, 2 m high and 0.4 m wide


laboratory based slope experiment. They modelled a
400 mm deep ne sand layer overlying a 200 mm thick
gravely sand at a slope angle of 30. The ne sand had a
saturated permeability (Ks) of 2.4 104 m/s, and the variation of suction with water content is shown Fig. 6. In
order to study the eect of rainfall duration on the inltration response the slope was subjected to a xed rainfall
intensity of 86.4 mm/h (which corresponded to 10% of
Ks) for periods of 15 and 42 min.
The initial and nal suction proles for the 15 min inltration event is shown in Fig. 7a. It is apparent that after
15 min rainfall there was a signicant reduction in suction
at the slope surface, a small reduction at a depth of 200 mm
and no change at 400 mm, suggesting that the wetting front
depth just exceeded 200 mm. When the duration was
increased to 42 min the wetting front depth evidently penetrated through the entire 400 mm depth of ne sand (see
Fig. 7b). Calculation of the time needed to form these wetting front depths (Case 1 200 mm and Case 2 400 mm)
was carried out using the GreenAmpt method and the
new model and the results are shown in Table 1. Consideration of Fig. 7 shows that in both cases, with the relative

Volumetric water cotent (%)

0.24
0.22
0.20
0.18
0.16
0.14

rainfall intensity being 0.2 (Ir/0.5Ks), the soil falls completely in zone 1, and therefore only T1 needs to be calculated. Whilst the predictions made using the new model are
compatible with the FEA analyses, the GreenAmpt model
severely under-predicts the inltration time. This is presumably due to the combined eect of the use of Ks and
the role of the static water head (imparted by the wetting
front) increasing the hydraulic gradient.
4.2. Comparison with other numerical methods
McDougall and Pyrah [17] reported numerical simulations of inltration in a 2 m deep soil sample. The experiments investigated the eect of rainfall intensity and
initial suction, on the distribution of inltrating water in
the sample. Numerical simulations were performed based
on the method proposed by Kisch [19]. The complete Soil
Water Characteristic Curve (SWCC) was used to describe
the change of matric suction with water content. Two simulations (test 22 and 23 in the reference) were chosen to test
the new model. The saturated permeability of the soil in the
numerical simulations was 864 mm/day. A 24 hour rainfall
event was applied at an intensity of 43 mm/day in test 23.
In test 22, the intensity was increased by a factor of 10
432 mm/day for equal duration. The initial and nal suction proles of these two simulations are shown in Fig. 8.
Dividing the suction prole in Fig. 8 with the line (Ir/
K)y, we see that in test 22 (Fig. 8a), only a small part of
the nal suction prole falls into zone 2, whilst in test 23
(Fig. 8b), zone 2 does not occur. Thus the eect of zone
2 can be (conservatively) neglected. The time needed for
those wetting fronts to shape can be calculated with equa-

0.12
0.10
0.08

Table 1
Comparison of predicted inltration times for 30 model slope

0.06

Case

2
3
4
Matric suction Ua-Uw (kPa)

150
Final

250 350

GreenAmpt

Fig. 6. Soil water characteristic curve of concrete sand (after Tami et al.
[18]).

Time (min)

540 S (mm)

1
2

2.2
10.2

100

Zone 1

15
42

250

New model
T1

T2

Total

10.4
37.5

0
0

10.4
37.5

540 S (mm)

Zone 1
Final

Initial

200

FEA result

Initial

400

400

S=0.2y
y (mm)

S=0.2y
y (mm)

Fig. 7. Initial and nal suction prole in the slope with the 15 min (a) and 42 min (b) precipitation of 86.4 mm/h.

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

Fig. 8. Initial and nal suction in a 2 m deep soil sample (after McDougal
and Pyrah [17]). (Note: numbers in brackets on horizontal axis are
volumetric water contents at related suction values.)

Table 2
Comparison of inltration times predicted for 2 m deep soil sample
Test

Time (day)
GreenAmpt

22
23

0.16
0.037

McDougall

1
1

New model
T1

Total

0.91
0.95

0.91
0.95

229

project was 17.5 m high with a slope angle of 30 and had a
1m wide berm at mid-height. The instrumentation included
moisture probes and tensiometers, which allowed the collection of moisture content and suction measurements
through a wet and dry season.
The response of the slope to a single rainfall event is
considered here. The event started on the 29th of August
2001 and lasted approximately 50 h. The saturated coecient of permeability of the soil Ks is 7.7 106 m/s. The
rainfall intensity was 1.4 106 m/s (0.18Ks) during this
period, and measurements show runo occurred during
precipitation. Suction and volumetric water content values
measured at a depth of 2 m started to change after approximately 26 h of rainfall. The initial and nal volumetric
water content and suction values are shown in Fig. 9.
Using the Ks value and volumetric water content measurements Li et al. [13] calculated the permeability of the
unsaturated soil with the empirical permeability function
proposed by Averjanov (see [20]). The result indicated that
K varied from 0.1Ks at the start of inltration, to 0.21Ks at
the end of the test. The average value (0.15Ks) and the
value at 0.5Ks were used to calculate the relative rainfall
intensity (Ri) in the new model. The line Sy = Riy for these
two conditions are plotted in Fig. 9. In the gure, the suction prole is divided into two regions, by this line. On the
upper right portion, the inltration capacity is larger than
the rainfall intensity and on the lower left the inltration
capacity is lower than the rainfall intensity. Therefore both
T1 and T2 need to be calculated. The nal suction prole in
Fig. 9 is divided into four segments (zones I, II, III, IV) for
ease of calculation and to improve accuracy.
The results of the analysis are shown in Table 3. Both
the calculated K value and the measured rainfall intensity
were used to predict the time required to develop a 2 m
deep wetting front. The estimate using the new model with

tion . The results of the analyses are shown in Table 2, from


which we see that the prediction with the new model is
within 9195% of the numerical analyses.
The encouraging comparison of the new models performance with two FEA approaches suggests that when the
nal suction prole is predominantly covered by zone 1
and the rainfall intensity is relatively low in comparison
to the saturated soil permeability (no larger than 0.5Ks),
i.e. the rainfall intensity controls inltration, the new model
provides comparable results to full FEA analyses.
5. Case study
Li et al. [13] describe a case study of a recently constructed instrumented slope cut through weathered Granite
in Hong Kong. The slope which formed part of a highway

Fig. 9. Matric suction prole in weathered soil slope (after Li et al. [13]).
(Note: numbers in brackets on horizontal axis are volumetric water
contents at related suction values.)

230

K. Gavin, J. Xue / Computers and Geotechnics 35 (2008) 223230

Table 3
Comparison of inltration times predicted for weathered soil slope
K

Time (h)
New model
T1

0.15Ks
0.5Ks
Ks

5.2
3.2

GreenAmpt

T2

Monitored

Acknowledgements

Total

II

III

IV

1.11

9.67

7.02
2.1

23.01
5.3

intensity) as opposed to the zone where the rainfall intensity exceeds the inltration capacity and the inltration rate
is controlled by the permeability of the soil.

8.27
2.5
1.24

26.25

K equal to 0.15Ks (i.e. the Ks value derived from the empirical equation by Li et al. [13]) is around 88% of the actual
time recorded. An estimate using K = 0.5Ks is seen to predict a time for the wetting front to develop of just 20% of
the actual. This is obviously because the actual supply of
water to the slope (rainfall intensity) is much lower than
0.5Ks. Also included in Table 3 are the GreenAmpt model
predictions, which are seen to signicantly underestimate
the inltration time at all assumed values of permeability.
6. Conclusion
An understanding of the wetting front development is
crucial to stability analysis of partially saturated soil
slopes. Although rigorous results can be obtained using
FEA analyses, soil permeability is a highly variable characteristic and FEA analysis will give inaccurate results if the
permeability measurements are not representative of the insitu conditions. This is one of the reasons why simple inltration models remain in widespread use. The GreenAmpt
model was seen to signicantly underestimate the time
required for a wetting front to form. The reason for the
poor performance of the model is due to the assumption
that the soil above the wetting front is fully saturated. This
leads to overestimation of the permeability and hydraulic
gradient of the soil in the wetting front.
A simple inltration model based on the GreenAmpt
model is proposed in this paper. The new model assumes
the soil in the wetting front remains partially saturated,
and the permeability of the soil in this zone is 0.5Ks (or
lower). In addition the model species two zones within
the wetting front. The rst where the inltration rate is
controlled by the supply of water (rainfall intensity), and
the second where inltration is controlled by the water content and the suction distribution within the wetted zone.
The model will provide more realistic estimates of the inltration time than the traditional GreenAmpt approach
when the soil in the wetting front does not reach full saturation prior to the initiation of slope failure.
By comparing the inltration capacity of the soil based
on the initial suction prole, to the rainfall intensity
applied at the slope surface, the new model dierentiates
between zones where the inltration rate is supply controlled (i.e. the inltration capacity exceeds the rainfall

This project is funded by Iarnrod Eireann. The authors


thank Mr. Brian Garvey, Chief Civil Engineer with Iarnrod
Eireann for nancial assistance received. The second
author was the recipient of a Geotechnical Trust Fund
award from the Geotechnical Society of Ireland. The
authors acknowledge the useful comments of the reviewers.
References
[1] Fourie AB, Owe DR, Blight GE. The eect of inltration on the
stability of the slopes of a dry ash dump. Geotechnique 1999;49(1):
113.
[2] Lumb PB. Slope failures in Hong Kong. Q J Eng Geol Hydrogeol
1975;8(1):3165.
[3] Fredlund DG, Rahardjo H. Soil mechanics for unsaturated soils.
New York: Wiley; 1993.
[4] Fredlund DG, Morgenstern NR, Widger RA. The shear strength of
unsaturated soils. Can Geotech J 1978;15(3):31321.
[5] Pradel D, Raad G. Eect of permeability on surcial stability of
homogeneous slopes. J Geotech Eng 1993;119(2):31532.
[6] Green WH, Ampt CA. Studies on soil physics: ow of air and water
through soils. J Agr Sci 1911;4:124.
[7] Jury WA, Horton R. Soil physics. New Jersey: Wiley; 2004.
[8] Richards LA. Capillary conduction of liquids through porous
mediums. Physics 1931;1:31833.
[9] Rahardjo H, Lee TT, Leong EC, Rezaur RB. Response of a residual
soil slope to rainfall. Can Geotech J 2005;42:34051.
[10] Mein RG, Larson CL. Modeling inltration during a steady rain.
Water Resour Res 1973;9(2):38494.
[11] Chu ST. Inltration during an unsteady rain. Water Resour Res
1978;14(3):4616.
[12] Brakensiek DL. Empirical and simplied models of the inltration
process. Report for the USDA-ARS inltration research planning
workshop, St. Louis, MO; 1977.
[13] Li AG, Yue ZQ, Tham LG, Lee CF, Law KT. Field monitored
variations of soil moisture and matric suction in sarolite slope. Can
Geotech J 2005;42:1326.
[14] Springman SM, Jommi C, Teysseire P. Instability on moraine slopes
induced by loss of suction: a case history. Geotechnique 2003;53(1):
310.
[15] Zhan TLT, Ng CWW. Analytical analysis of rainfall inltration
mechanism in unsaturated soils. Int J Geomech ASCE 2004;4(4).
[16] Bouwer H. Rapid eld measurement of air entry value and hydraulic
conductivity of soil as signicant parameters in ow system analysis.
Water Resour Res 1966;2(4):72938.
[17] McDougall JR, Pyrah IC. Simulating transient inltration in unsaturated soils. Can Geotech J 1998;35:1093100.
[18] Tami D, Rahardjo H, Leong EC, Fredlund DG. Design and
laboratory verication of physical model of sloping capillary barrier.
Can Geotech J 2004;41(9):81430.
[19] Kisch M. The theory of seepage from clay-blanketed reservoirs.
Geotechnique 1959;9(9-21).
[20] Fredlund DG, Xing A, Huang S. Predicting the permeability function
for unsaturated soil using the soil water characteristic curve. Can
Geotech J 1994;31:53346.
[21] Ng CWW, Shi Q. A numerical investigation of the stability of
unsaturated soil slopes subjected to transient seepage. Comput
Geotech 1998;22(1):128.

You might also like