392

Field study of rainfall infiltration into a grassed

unsaturated expansive soil slope
Tony L.T. Zhan, Charles W.W. Ng, and Del G. Fredlund

Abstract: A full-scale field study was conducted to investigate the effects of rainfall infiltration on a natural grassed
expansive soil slope in China. A 16 m wide × 28 m long area was selected for instrumentation. The instrumentation
included jet-filled tensiometers, moisture probes, a tipping bucket rain gauge, and a vee-notch flow meter. One artificial
rainfall event amounting to about 370 mm rain depth in total was applied to the slope. The monitored results suggested
that there was about a 3 day delay in the response of surface runoff, pore-water pressure, and water content to the
commencement of the simulated rainfall. The depth of influence of the rainfall, depending on the elevation along the
slope, ranged from 2.8 to 3.5 m. Positive pore-water pressures were measured within the influence depth, and there
existed significant subsurface downslope flow at the end of the simulated rainfall, particularly near the lower part of
the slope. A comparison of infiltration rates between the grassed area and a bare area nearby indicated that the pres-
ence of grass significantly increased the infiltration rate and reduced surface runoff. The cracks and fissures developed
in the unsaturated expansive soil played an important role in the hydrological process.
Key words: expansive soil, slope instability, infiltration, vegetation cover, grass, soil suction, water content, unsaturated soil.
Résumé : On a réalisé une étude à pleine échelle sur le terrain pour évaluer les effets de l’infiltration de la pluie dans
un talus naturel constitué d’un sol gonflant couvert de gazon en Chine. On a choisi pour l’instrumentation une surface
de 16 m de largeur par 28 m de longueur. L’instrumentation incluait des tensiomètres avec purge à jet d’eau, des sondes
d’humidité, une jauge de pluie à seau renversant, et un débitmètre à coche en V. Un événement de pluie artificielle
d’un total d’environ 370 mm d’épaisseur a été produite sur le talus. Les résultats des mesures ont suggéré qu’il y avait
un délai d’environ trois jours dans la réponse du ruissellement, de la pression interstitielle et de la teneur en eau par
rapport au début de la pluie arificielle. La profondeur de l’influence de la pluie dépendant de l’élévation le long du talus
variait de 2,8 à 3,5 m. Des pressions interstitielles positives ont été mesurées à l’intérieur de la zone d’influence, et il
existait une écoulement souterrain significatif vers le bas du talus à la fin de la pluie simulée, particulièrement près de
la partie inférieure du talus. Une comparaison des taux d’infiltration entre la surface gazonnée et une surface nue à
proximité a indiqué que la présence de gazon a augmenté appréciablement le taux d’infiltration et réduit le ruissellement
de surface. Les fentes et les fissures qui se sont développées dans le sol gonflant non saturé ont joué un rôle important
dans le processus hydrologique.
Mots-clés : sol gonflant, instabilité de talus, infiltration, couvert de végétation, gazon, succion du sol, teneur en eau,
sol non saturé.
[Traduit par la Rédaction] Zhan et al. 408

Introduction
Landslides in unsaturated soils are often induced by rain-
fall infiltration during wet seasons. Field studies on the ef-
fect of rainfall infiltration on slope instability have been
carried out by numerous researchers (Johnson and Sitar
Received 18 November 2005. Accepted 13 July 2006.
Published on the NRC Research Press Web site at cgj.nrc.ca
on 17 May 2007.
T.L.T. Zhan. Department of Civil Engineering, Zhejiang
University, Hangzhou, China (former research student, Hong
Kong University of Science and Technology).
C.W.W. Ng.1 Department of Civil Engineering, Hong Kong
University of Science and Technology, Clear Water Bay,
Kowloon, Hong Kong, China.
D.G. Fredlund. Department of Civil Engineering, Zhejiang
University, Hangzhou, China.
1
Corresponding author (e-mail: cecwwng@ust.hk).
1990; Affendi and Faisal 1994; Lim et al. 1996; Zhang et al.
2000). Almost all of these field studies have been carried out
on residual soils that are nonexpansive soils. An expansive
soil with active clay minerals exhibits significant swelling
and shrinking upon wetting and drying. Consequently, an
abundance of cracks and fissures develop in the upper part
of the soil profile. The hydrological response of the expan-
sive soil to rainfall can be significantly different from that of
a less active residual soil. The rainfall infiltration into a
crack-rich expansive soil results in a complex process of
rainfall infiltration and volume change (Flury et al. 1994;
Hillel 1998). The field study conducted by Johnson and Sitar
(1990) suggested that the hydrological response of a hill-
slope to rainfall was extremely complex and could not be
fully explained using traditional hydrological models. The
infiltration process and its effect on slope instability are also
influenced by the vegetation cover on the slope surface
(Coppin et al. 1990; Gray 1995). Most previous studies on
relationships amongst vegetation, rainfall infiltration, and

Can. Geotech. J. 44: 392–408 (2007) doi:10.1139/T07-001 © 2007 NRC Canada

Zhan et al.
slope instability have been on a qualitative level. As far as
the authors are aware, few quantitative field studies have
been carried out to investigate the influence of vegetation on
rainfall infiltration and slope instability involving unsatu-
rated expansive soil.
A full-scale field study was carried out on an unsaturated
expansive soil slope in the Hubei province of China (i.e., in
west-north China). The field study was meant to assist in the
engineering design of the 180 km portion of the canal to be
excavated in unsaturated expansive soils along the middle
route of the South-to-North Water Transfer Project
(SNWTP) (Ng et al. 2003). The field study consisted of two
neighboring monitoring sloping areas (16 m wide × about
30 m long each): a bare soil slope and a natural slope with
grass surface. Rainfall simulation tests and in situ monitor-
ing were conducted on both areas. The observed perfor-
mance of the bare soil slope has been reported by Ng et al.
(2003). The major findings in the previous paper are the
complex interaction among the changes in suction, horizon-
tal stress, and horizontal and vertical displacements in the
bare soil slope subjected to two artificial rainfall events. In
this current paper, monitored results from the natural grassed
sloping area are reported with an emphasis on the effect of
the natural grass cover on responses of the ground to rainfall
infiltration. The data collected include rainfall intensity, in-
filtration rate, surface runoff, changes of soil suction, and
water content. The influence of the natural grass cover on
rainfall infiltration into the slope is discussed by comparing
the observed differences in the infiltration rate between the
bare soil area and the natural grassed area.
The test site
The test site is located in Zaoyang, Hubei, China, about
230 km northwest of Wuhan and about 70 km south of the
intake canal for the SNWTP in Nanyang, Henan. The site is
a semiarid area with an average annual rainfall of about
800 mm. About 70% of the annual rainfall occurs between
May and September.
The test site was selected on a cut slope that was prepared
as part of an excavation canal (Fig. 1). The canal was exca-
vated in 1970 with an mean excavation depth of 13 m. The
slope had an inclination angle of 22° (i.e., 1:2.5) and a
height of 11 m (measured from the top of the retaining
wall). The surface of the slope was covered with naturally
occurring grass but no trees were present. The primary vege-
tation species consisted of couch grass and weed. The height
of the vegetation ranged from about 100 to 500 mm. The
root depth observed on excavated surfaces ranged from
about 100 to 300 mm. The adjacent bare soil previously
mentioned was obtained by removing the top soil to a depth
of about 10 mm. As shown in Fig. 1, there are a number of
typical shallow retrogressive slope failures, immediately to
the west of the selected monitoring area.
Soil profile and properties
Prior to installing instrumentation on the slope, a borehole
investigation was conducted around the monitoring area to
investigate the ground conditions. Soil sampling, standard
penetration tests (SPT), and dilatometer tests (DMT) were
393
conducted in the boreholes (Ng et al. 2003; Zhan 2003). The
soil profiles and geotechnical parameters obtained from the
boreholes near mid-slope are shown in Fig. 2. The geo-
technical measurements include water content, dry density
(ρ d ), SPT N value, the undrained shear strength (cu), and the
lateral earth pressure coefficient (K0), deduced from the
DMT measurements. The predominant stratum below the
slope surface was a brown-yellow mottled gray clay. The
clay layer was sometimes interlayered with thin layers of
gray clay or iron concretions. X-ray diffraction analyses in-
dicated that the predominant clay minerals for the samples
taken from the top 2 m soil layer were illite (16%–25%) and
smectite (11%–21%), with a small percentage of kaolinite
(2%–5%). Some typical properties of the natural (so-called
undisturbed) expansive clay are given in Table 1. The soil is
classified as a silty clay with a medium expansivity in accor-
dance with the criterion proposed by Sridharan and Prakash
(2000). The upper soil layer with a thickness varying from
1.0 to 1.5 m was rich in cracks and fissures, particularly at
the upper part of the overall slope. Figure 3 shows pictures
of cracks and fissures observed on the wall of a two-step ex-
cavation pit near the monitoring area. The maximum depth
of the open cracks was estimated to be approximately 1.2 m.
The maximum width of the open cracks was about 10 mm.
The soil also contained a variety of narrower cracks. The
abundance of cracks and fissures is likely related to the
swelling and shrinking phenomenon associated with expan-
sive clay soils.
Double-ring infiltration tests were carried out around the
monitoring area to investigate the infiltration characteristics
of the soil layer near the ground surface. These tests were
performed prior to the installation of the instrumentation.
The double-ring infiltrometer consisted of two stainless steel
rings with different diameters (i.e., inner and outer rings), a
Marriotte supply bottle mounted on the inner ring, and a
constant-head water supply device mounted on the outer
ring. The diameters of the inner and outer rings were 0.5 and
1 m, respectively.
Figure 4 shows the changes of infiltration rate with time
measured for two different ground conditions, (i.e., one infil-
tration test in an area abundant with open cracks and the
other without obvious open cracks). The measured infiltra-
tion rate in the ground without obvious open cracks was
distinctly low (i.e., in the order of 10–7 m/s or less). The
gradual decrease of infiltration rate during the first 2 days of
the test appears to be primarily due to a decrease in hydrau-
lic gradient resulting from a loss of soil suction upon wet-
ting. In the cracked ground, the infiltration rate measured
within the first 1–2 h was quite large (i.e., in the order of
10–4 m/s). This was attributed to the rapid ingress of water
into the opened cracks. The infiltration rates decreased dra-
matically with test duration during the first half day and then
tended towards a steady value. The main reason for the
sharp decrease in infiltration rate appears to be related to the
water storage capacity of the open crack channels. These
cracks tended to close with time because of soil swelling
upon wetting. Favre et al. (1997) also reported that high in-
tensity rain falling onto a dry and cracked soil resulted in
rapid crack closure at the soil surface within 4.5 h because
of swelling of unsaturated soils, and that the bypass flow
processes were a matter of several hours only. As shown in
© 2007 NRC Canada
394
Fig. 1. Overall view of the research slope.
Fig. 2. Soil profile and geotechnical parameters from the boreholes at mid-slope.
Fig. 4, the infiltration rate at equilibrium in the ground with
open cracks is always greater than that in the ground without
open cracks. The final infiltration rate in the ground with
open cracks was close to the initial infiltration rate in the
ground without open cracks. The observed difference in the
infiltration rate between the two different ground surface
conditions is basically consistent with the field measure-
ments obtained by Øygarden et al. (1997). The authors
found that measured values of hydraulic conductivity on site
varied by three orders of magnitude from 10–4 to 10–7 m/s
because of the variability of cracks.
Can. Geotech. J. Vol. 44, 2007
Instrumentation
Instrumentation layout
A 16 m × 28 m naturally grassed area was selected on the
studied slope for the installation of instrumentation. The in-
struments included jet-filled tensiometers, moisture probes
(Thetaprobe 1999), a tipping bucket rain gauge, and a vee-
notch flow meter. The layout of the instruments is shown in
Figs. 5 and 6. There were three rows of instrumentation for
monitoring soil suction and water content (i.e., R1 at the up-
per part, R2 at the middle part, and R3 at the lower part of
© 2007 NRC Canada
Zhan et al.
Table 1. Typical soil properties of the natural (undisturbed) expansive clay.
Soil property
Classification Percentage of sand, silt, clay
USCS classification
Density Specific gravity, Gs
Dry density (Mg/m3)
Consistency limit Liquid limit, LL
Plasticity index, PI
Linear shrinkage limit, ws
Flow characteristic Saturated permeability, ks (m/s)
Air entry value (kPa)
Compressibility Compressibility index, Cc
Swelling index, Cs
Swelling and shrinkage One-dimension free swelling
Constant volume swelling pressure (kPa)
Volumetric strain due to shrinkage
Natural shear strength parameters Effective cohesion, c′ (kPa)
Effective angle of friction, φ′
the slope). In total, 17 jet-filled tensiometers were installed
for suction measurements. The embedment depth for all the
tensiometers was greater than 0.9 m. This depth was se-
lected because the initial soil suctions near the ground sur-
face exceeded the measuring range of tensiometers based on
data obtained from thermal conductivity suction sensor mea-
surements on the adjacent bare soil area (Ng et al. 2003). In
total, six moisture probes were installed between a depth of
0.9 and 1.8 m. The sensors at each row were spaced 1 m
apart. At each depth, there were generally two tensiometers
and one moisture probe. The measured data from the three
sensors could be compared. After installing each of the ten-
siometers or moisture probes, the installation holes were
backfilled with moist soil. The backfill soil was compacted
to a dry density close to the in situ dry density. The adopted
procedure was meant to prevent rainwater from leaking into
the installation holes through open cracks near the ground
surface (Zhan 2003). The readings from all the tensiometers
were manually collected. The data from the moisture probes
were collected using a data logger.
Moisture probes (called Thetaprobes) adopt a standing
wave technique to measure an apparent dielectric constant of
a soil. The measured apparent dielectric constant is corre-
lated with volumetric water content of the soil. The accuracy
of the measurements is affected by many factors such as
variations in soil composition, dry density, and cracks (Li et
al. 2002). For the field monitoring reported in this paper, in
situ calibrations were carried out. Results of the calibration
indicated that an error in measurement could be up to 5% of
volumetric water content. Therefore, measured volumetric
water contents by Thetaprobes should be interpreted with
caution.
Disturbed samples were also taken by auger from near the
three rows of instrumentation. These samples were used for
the determination of the gravimetric water content (GWC)
profiles. Water content sampling was undertaken every sec-
ond day during the rainfall simulation period. All auger
holes were backfilled immediately after sampling.
395
Value Remark
3%, 58%, 39%
Silty clay
2.67
1.50–1.62
50.5
31
13
10–10 to 10–7 Depending on number of cracks
5
0.13
0.025
6%–8% Air-dry to saturated specimen
300–400 Air-dry to saturated specimen
9%–11% Saturated to oven-dry specimen
16.7
28.7° Peak strength
The groundwater level shown in Fig. 6 was estimated
from the borehole investigations. The water levels in the
boreholes were monitored periodically during the field study
period. It was found that the influence of the artificial rain-
fall simulation on the groundwater level was negligible. In
other words, the depth of influence by the simulated rainfall
was confined primarily to the top 3 m.
A tipping-bucket rain gauge was installed to record the in-
tensity and duration of rainfall. Flow meters were installed
on the main water-supply line to the sprinkler system to re-
cord the total amount of water applied to the entire slope
area at any given time interval. A water collection trench
was constructed along the toe of the slope to measure sur-
face runoff using an automatic vee-notch flow meter system
installed at the end of the trench (Zhan 2003).
Artificial rainfall simulation
Rainfall was produced artificially using a sprinkler system
specifically designed for this test site. The sprinkler system
was comprised of a pump, a main water-supply line, three
groups of distribution tubes (numbered 1, 2, and 3, respec-
tively), and 35 sprinkler heads (see Fig. 7). The system
could produce three levels of rainfall intensity (i.e., approxi-
mately 3, 5, and 9 mm/h). The three levels of rainfall were
regulated via three flow-rate controlling valves. The lowest,
medium, and highest levels of rainfall intensity were con-
trolled by opening three distribution tubes (i.e., A, E, and I),
five distribution tubes (i.e., A, C, E, G, and I), and all 10 dis-
tribution tubes, respectively. The sprinkler heads for the
three groups of distribution tubes were arranged in such a
way that a relatively uniform rainfall could be produced
(Zhan 2003). The artificial rainfall intensity applied in this
study was approximately 2.9 mm/h.
The monitoring period for the grassed area lasted for
2 weeks (i.e., from 14 September to 28 September 2001),
during which time one rainfall event was simulated (see
Fig. 8). No natural precipitation occurred during the moni-
© 2007 NRC Canada
396 Can. Geotech. J. Vol. 44, 2007

Fig. 3. Cracks and fissures in the excavation pit near the monitoring area (the maximum depth of cracks ≅ d1 + d2 = 1.2 m). (a) Upper
part; (b) lower part.

toring period. The simulated rainfall event lasted for 7 days
(from 09:00 of 18 September to 16:00 of 24 September
2001) with an mean rainfall intensity of 53 mm per day. Fig-
ure 8 also shows the natural precipitation recorded during
the 3.5 months prior to the commencement of the simulated
rainfall (i.e., from 1 June to 18 September 2001). The total
precipitation during the period was about 250 mm. The natu-
ral rainfall intensity ranged from 1 to 26 mm/h. Before
1 June 2001, the test site had experienced about 7 months of
drought with a total rainfall of less than 100 mm.

© 2007 NRC Canada

Zhan et al. 397

Fig. 4. Changes of infiltration rate with time from the double-ring infiltration tests.

Fig. 5. Layout of instruments on the grassed area.

© 2007 NRC Canada

398
Fig. 6. Cross section of the instrumented slope.
A more detailed plot of the simulated rainfall intensity is
shown in Fig. 9. It can be seen that the rainfall intensity was
maintained at approximately 2.9 mm/h except for three in-
termittent periods. The first intermittent period was between
8:20 pm of 18 September to 9:40 am of 19 September. This
first stoppage of rainfall was due to a malfunction of the
pumping system. During the other two intermittent periods,
artificial rainfall was stopped to obtain auger samples for the
determination of water contents.
Observed performance
Surface runoff
Figure 9 shows the changes in percentage of surface run-
off with time along with the intensity of the simulated rain-
fall. The percentage of runoff is equal to measured surface
runoff divided by rainfall intensity. During the first 3 days
after the beginning of the artificial rainfall, no surface runoff
was observed on the slope surface (i.e., 100% of rainfall in-
filtrated). It was like a supply controlled infiltration (Hillel
1998). Three days later, surface runoff started to occur and
the percentage of runoff increased progressively with the du-
ration of rainfall. After 5 days of rainfall, the percentage of
runoff tended towards about 45%. With the measured sur-
face runoff, the infiltration rate could be deduced. The infil-
tration rate was initially equal to the rainfall intensity (i.e.,
2.9 mm/h), and it decreased to a value of 1.6 mm/h after
5 days of rainfall. The decrease in infiltration rate was pri-
marily related to the open cracks and fissures that had be-
come filled and gradually closed up because of the swelling
of the soil upon wetting. Another reason may be due to the
decrease of the hydraulic gradient as a result of reduction in
Can. Geotech. J. Vol. 44, 2007
soil suction. It is understandable that the infiltration rate at
the end of the simulated rainfall period (1.6 mm/h = 4.7 ×
10–7 m/s) was close to the final infiltration rate measured
from the double-ring infiltration test on the ground with
open cracks (see Fig. 4). The results indicate that the final
infiltrability of the soil essentially attained a constant hy-
draulic conductivity value regardless of the significant dif-
ferences in the initial infiltrability resulting from different
ground surface conditions.
Pore-water pressures or soil suctions
Response of pore-water pressure
Figures 10a–10c show the changes of pore-water pressure
with time in response to the simulated rainfall recorded at
sections R1, R2, and R3, respectively (i.e., at the upper, mid-
dle, and lower parts of the slope). At each section, if the
measurements from the two tensiometers installed at the
same depth were consistent with each other, only the results
from one of the two tensiometers were presented in Fig. 10.
The number in the legend of Fig. 10 denotes the installation
depth of the tensiometers and “W or E” denotes the west or
east tensiometers installed at the same depth (see Fig. 5).
Figure 10a shows that just prior to the beginning of the rain-
fall on 18 September 2001, negative pore-water pressures rang-
ing from 58 to 68 kPa were recorded by the four tensiometers
within the top 1.5 m of soil. The values of negative pore-water
pressure measured by the two tensiometers installed at the same
depth but spaced 2 m apart (i.e., R1-T-1.5(W) and R1-T-1.5(E))
were reasonably close to each other. The variation in negative
pore-water pressure with depth appears to be relatively insignifi-
cant.
© 2007 NRC Canada
Zhan et al.
Fig. 7. Layout of a specially designed sprinkler system.
After the commencement of the simulated rainfall, nega-
tive pore-water pressures recorded by the tensiometers began
to decrease only after 3 or more days of rainfall. In other
words, the pore-water pressure responses showed a delay of
3 days or more. The duration of delay was consistent with
the delay observed in the onset of surface runoff. Tensi-
ometer R1-T-1.5(E), showed a significantly longer delay
399
than the other tensiometer installed at the same depth (i.e.,
R1-T-1.5(W)). The delay appeared to have no relationship to
the installation depth. The observation may be related to the
preferential flow through the open cracks in the unsaturated
expansive soil. As previously mentioned, an abundance of
cracks and fissures were found in the field and some open
cracks were observed to extend as deep as 1 m. The ingress
© 2007 NRC Canada
400
Fig. 8. Natural precipitation and simulated rainfall event.
Fig. 9. Intensity of the simulated rainfall and percentage of runoff.
of water through the open cracks was much more rapid than
the seepage through the intact expansive clay that had a dis-
tinctly lower water coefficient of permeability (see Fig. 4).
As a result, the tensiometer tip that was located near an open
crack tended to show a more rapid response. On the other
hand, a tip located in an intact soil mass could record a
delayed response (e.g., R1-T-0.9). After the tensiometers
started to respond, the recorded negative pore-water pres-
sures decreased gradually towards zero. Positive pore-water
pressures were recorded at the end of the simulated rainfall
by all the tensiometers installed within the top 1.5 m of soil.
After the end of the simulated rainfall (on 25 September),
positive pore-water pressures quickly dissipated, and a slight
recovery of suction was observed. The fluctuation of pore-
water pressure observed on 25 September may be attributed
Can. Geotech. J. Vol. 44, 2007
to the difference in barometric pressure between the day
time and the night time.
Figures 10b and 10c show the pore-water pressure re-
sponses measured at the other two sections (i.e., R2 and R3).
Each of these sections also showed a delay in pore-pressure
response of about 3 days, except that the tensiometer installed
at a depth of 0.9 m at the R3 section showed shorter delay. It
is unexpected that the tensiometers installed in the deep soil
layers at R3 (i.e., R3-T-1.8 and R3-T-2.4) responded as rap-
idly as the shallower tensiometer (i.e., R3-T-1.2). The unex-
pected observation can be explained in terms of the
downslope water flow along the deep uncracked layer or
possibly through preferential flow through a deep, extended
crack. Johnson and Sitar (1990) observed a similar phenom-
enon during the field monitoring of a hillslope. Further de-
© 2007 NRC Canada
Zhan et al. 401

Fig. 10. Pore-water pressure responses from tensiometers at (a) R1, (b) R2, and (c) R3.

tails related to downslope water flow are presented later in
this paper. After the tensiometers started to respond, the re-
corded changes of pore-water pressure with time at R2 and
R3 became more rapid than those measured at R1. This may
be attributable to the difference in the water storage capacity
of the soil profile above the monitoring level in the three
sections. The soil profile on the lower part of the slope is
closer to the long-term groundwater table and had less water
storage volume available for infiltration. Consequently, it
appears that the pore-water pressure response can be more
rapid. At the end of the simulated rainfall, all the tensiom-
eters installed on the three sections recorded a positive pore-

© 2007 NRC Canada

402
water pressure. The dissipation of positive pore-water pres-
sure appeared to be most rapid on the R1 section, followed
by the R2 section, and most gradual on the R3 section.
Variations of pore-water pressure profiles
Figures 11a–11c show the variations in pore-water pres-
sure profiles at sections R1, R2, and R3, respectively. Prior
to the commencement of the simulated rainfall (17 Septem-
ber), the negative pore-water pressures could be seen to gen-
erally decrease with depth. The hydraulic gradient at all
three sections within a depth of 1.5 m deviated from the hy-
drostatic no-flow profile, suggesting an upward movement
of moisture via evaporation or evapotranspiration.
After 3 days of rainfall (i.e., on 21 September), changes in
the pore-water pressure profiles due to rainfall infiltration at
sections R1 and R2 (see Figs. 11a and 11b) were relatively
insignificant. The soil layers between a depth of 0.9 and
1.5 m still maintained the original soil suction. At the R3
section (see Fig. 11c), the changes in the pore-water pres-
sure profile appeared to be relatively significant. On 22 Sep-
tember (i.e., 4 days after the commencement of rainfall), a
significant increase in pore-water pressures was observed at
all three sections. Positive pore-water pressures eventually
appeared at other locations.
At the end of the simulated rainfall (i.e., on 25 Septem-
ber), a positive pore-water pressure profile was observed at
all the three sections. The buildup of positive pore-water
pressure at section R1 (see Fig. 11a) was relatively insignifi-
cant when compared with those at sections R2 and R3 (see
Figs. 11b and 11c). This may be related to a greater defi-
ciency in water content in the soil at the upper part of the
slope prior to rainfall and also possibly because of the loss
of water content resulting from a downslope subsurface flow
after rainfall. The positive pore-water pressure profiles at
sections R2 and R3 were close to the hydrostatic line, which
was constructed by assuming that the groundwater table was
at the surface of the slope (see Figs. 11b and 11c). The data
was consistent with a flooding of the sampling holes ob-
served when the auger drill holes extended to a depth of
about 2.5 m at sections R2 and R3. The positive pore-water
pressures can be explained by considering water columns
that extend through the deep, extended cracks or possibly
there is an inclined perched water table that developed at the
middle and lower parts of the slope. The development of
perched water is most likely related to the presence of less
permeable soil layers below the ground surface. The dry
density profile shown in Fig. 2 suggests that there existed
two relatively dense soil layers within the top 4 m (i.e., near
1.2 and 2.8 m). The borehole investigation results showed
that the dense layer generally had a higher clay content and
may possess a lower water coefficient of permeability. On
the other hand, the amount of open cracks tended to de-
crease with an increase in the depth. The deeper dense layer
(i.e., near 2.8 m) was likely to be relatively impermeable,
creating conditions for the development of perched water
above the dense layer.
Variations of piezometric level
Piezometric levels at each monitoring point can be deter-
mined from the pore-water pressure measurement and the el-
evation. Contours of piezometric level within the monitoring
Can. Geotech. J. Vol. 44, 2007
zone were constructed using the commercial software called
SURFER. Figures 12a and 12b show the contours of piezo-
metric level on 17 September 2001 (i.e., prior to the simu-
lated rainfall) and 25 September 2001 (i.e., at the end of the
simulated rainfall), respectively. Contours of the piezometric
level prior to the simulated rainfall indicated an upward flux
in the vertical direction and a downslope flux along the in-
clined slope. The simulated rainfall resulted in a directional
change in moisture movement in the vertical direction. The
contours of piezometric level in Fig. 12b show that in the
upper part of the slope there is a downward flux in the verti-
cal direction and a downslope flux along the inclined slope.
However, in the lower part of the slope, the contours of
piezometric level are nearly perpendicular to the slope sur-
face, indicating that the downslope flux is dominant over the
flux in the vertical direction. This appeared to be consistent
with the significant surface runoff and the exiting of ground-
water observed on the lower part of the slope. The data fur-
ther supports the postulation that there exists a subsurface
downslope water flow following a heavy rainfall.
Water content
Variations of gravimetric water content profiles
Figure 13 shows the changes in gravimetric water content
profiles in response to the simulated rainfall. The measure-
ments were obtained from direct sampling of auger cuttings.
More data points were obtained for the gravimetric water
content profiles than from the pore-water pressures profiles
as shown in Fig. 11. For the gravimetric water content pro-
files on 22 and 25 September, data points were not obtained
below a depth of about 2.5 m at both R2 and R3. This was
due to flooding of the sampling holes when the auger drilled
down to a depth of 2.5 m.
Figure 13a shows the changes in the gravimetric water
content profiles at the R1 section. Just prior to the com-
mencement of the simulated rainfall (on 17 September), the
initial gravimetric water contents generally increased with
depth, particularly within the top 1.5 m. The slight variation
in gravimetric water content with depth might be due to the
natural variations in soil density and composition. A com-
parison between the initial gravimetric water content profile
and the profile corresponding to 20 September indicated that
the influence of the first 2 day rainfall was confined to the
top 0.6 m of soil. After 4 days of rainfall (i.e., on 22 Sep-
tember), a significant increase in gravimetric water content
was measured, except in the dense soil layer near 1.2 m and
in the soil layer deeper than 3.5 m. The significant increase
in gravimetric water content between depths of 1.5 and
3.5 m indicated that a considerable amount of rainwater had
infiltrated the relatively deep soil layer. This may be due to
some deep, extended open cracks at the upper part of the
slope. After 1 week of rainfall (i.e., on 25 September), a
substantial increase in gravimetric water content was ob-
served in all of the soil layers above a depth of 3.5 m when
compared with the initial gravimetric water content profiles.
Figures 13b and 13c show the changes of gravimetric
water content profiles at R2 and R3, respectively. A signifi-
cant increase in the gravimetric water content profile was
observed within the top 1.5 m. However, the increase in
gravimetric water content below 1.5 m was not as significant
© 2007 NRC Canada
Zhan et al. 403

Fig. 11. Variations of pore-water pressure profile at (a) R1, (b) R2, and (c) R3.

as that observed in the upper part of the slope (i.e., R1). The
smaller changes in water content can be explained by a de-
creased amount of desiccation and crack formation in the
soil profile in the lower part of the slope. This part of the
slope is closer to the long-term groundwater table. The de-
crease in crack formation means that less water from infil-
tration can be stored, and hence, there is a smaller increase
in water content. On the basis of the water content profiles
and the pore-water pressure profiles (see Fig. 11), it is esti-
mated that the depth of influence of the simulated rainfall
was about 3.0 m for the R2 section and at the R3 section it
was about 2.8 m. The depths are smaller than the depth of

© 2007 NRC Canada

404
Fig. 12. Contour of piezometer level within the monitoring zone obtained on (a) 17 September 2001 and (b) 25 September 2001.
influence at the R1 section (i.e., 3.5 m). The greater depth of
influence at R1 may be related to the greater depth of open
cracks in the upper part of the slope and the deeper ground-
water level. The depth of influence found in this study
seems to be greater than that found in the field study per-
formed by Lim et al. (1996) on a residual soil slope. The
greater depth of influence would appear to be attributable to
the extensive network of cracks existing in the unsaturated
expansive soil.
Response of volumetric water content recorded by
Thetaprobes
Figure 14 shows the changes of volumetric water content
measured by the six Thetaprobes together with the changes
of pore-water pressure recorded by the neighboring tensiom-
eters installed at the same depth as those of the Thetaprobes.
Similar to the pore-water pressure response (see Fig. 10), a
delayed response was observed for the changes of volumet-
ric water content recorded by the Thetaprobes. The length of
the delay for the six Thetaprobes varied from 2 to 4 days.
The duration of the delay recorded by each of the six Theta-
probes was fairly consistent with that recorded by each cor-
responding tensiometer, particularly at sections R2 and R3.
After the Thetaprobes began to respond, a significant in-
crease in volumetric water content was recorded by each of
the Thetaprobes, corresponding to an increase in the pore-
water pressure recorded by each corresponding tensiometer.
Two different types of response could be identified for the
Can. Geotech. J. Vol. 44, 2007
Thetaprobes after the delay. Four of the six Thetaprobes
showed a rapid increase in volumetric water content up to a
steady value. However, the other two Thetaprobes at a depth
of 1.2 m (i.e., R2-θ-1.2 and R3-θ-1.2) showed a gradual in-
crease in volumetric water content to the maximum value at
the end of the rainfall and then showed a prompt decrease. It
is quite possible that the four Thetaprobes with a quick re-
sponse were embedded near cracks and fractures, while the
other two were located in a slower responding matrix of the
soil. It should be noted that the slower responding The-
taprobes happened to be embedded at a depth of 1.2 m
where a dense soil existed as shown in Fig. 2. The maximum
volumetric water contents recorded by R2-θ-1.2 and R3-θ-
1.2 were significantly lower than those by the other Theta-
probes. This might indicate that the soil wrapping around
these two Thetaprobes remained unsaturated at the end of
the simulated rainfall, even though the corresponding tensi-
ometers recorded positive pore pressures at the same depth
but at different locations, indicating the heterogeneous na-
ture of the ground.
Influence of grass cover on rainfall
infiltration
As previously discussed, the field program consisted of
two neighboring monitoring areas (both 16 m wide × about
30 m long); namely, a bare area and a grassed area. The du-
ration and intensity of the rainfall event simulated in the
© 2007 NRC Canada
Zhan et al. 405

Fig. 13. Variations of water content profile at (a) R1, (b) R2, and (c) R3.

grassed area were similar to those of the first rainfall event
simulated in the bare area (Ng et al. 2003). Figure 15 shows
a comparison of infiltration rates between the bare and the
grassed areas. The infiltration rate is assumed to be equal to
the difference between the measured rainfall intensity and
surface runoff. It is recognized that this calculation may not
be very accurate for the grassed area since the grass cover
may intercept and adsorb some rainwater. However,
Branson et al. (1972) estimated that the storage capacity of
grass ranges from 1 to 5 mm of rainfall, which is a relatively
small portion of the simulated rainfall (370 mm in total).
Consequently, the deduced infiltration rates appear to be suf-

© 2007 NRC Canada

406 Can. Geotech. J. Vol. 44, 2007

Fig. 14. Changes of volumetric water content from Thetaprobes and corresponding pore-water pressure changes at (a) R1, (b) R2, and
(c) R3.

ficiently accurate. The intensity of the artificial rainfall sim-
ulated in the two areas is similar (i.e., 2.9 mm/h). No surface
runoff was measured after the commencement of rainfall;
therefore, the infiltration rate was equal to the rainfall inten-
sity. In the bare area, the infiltration rate started to decrease
after 1.5 days of rainfall. The infiltration rate decreased dra-
matically as the duration of the rainfall continued. In the
grassed area, the infiltration rate started to decrease after
about 3 days of rainfall. Therefore, the surface runoff in the
grassed area showed a longer delay. The infiltration rate in

© 2007 NRC Canada

Zhan et al.
Fig. 15. Comparison of infiltration rates between the bare and well-grassed areas.
the grassed area was equal to or significantly greater than
that in the bare area for 1 week of rainfall.
The difference described above can be primarily attributed
to the difference in the surface conditions between the two
areas. In the bare soil area, the impact of raindrops tended to
result in a thin, less-permeable crust being formed. The bare
slope seemed to undergo spontaneous slaking and break-
down of clay aggregates upon wetting (Hillel 1960; Bridge
and Collis-George 1973; Morel-Seytoux 1983). Once the
crust was formed, water ingress into the soil could be greatly
impeded (Hillel 1998). The formation of the surface crust
was observed during examination of the ground surface after
the simulated rainfalls. After the rainfall, the surface of the
bare soil area appeared to be more uniform than the initial
ground surface that consisted of soil aggregates networked
with a number of open cracks. On the other hand, the grass
cover in the grassed area appeared to protect the soil on the
slope surface from being struck directly, and hence, mini-
mized the formation of a less permeable crust on the surface
of the slope (Coppin et al. 1990; Hillel 1998).
The soil particles on the bare soil slope surface were eas-
ily detached and transported down the slope by surface run-
off (i.e., hydraulic erosion). Soil particles also tended to
move into the open cracks near the slope surface, gradually
filling the open cracks. This process may also account for
the decrease in infiltration rate of the soil layer near the
ground surface. In the grassed area, the grass tended to in-
crease the roughness of the surface of the slope and the root
system of the grass provided more channels for water infil-
tration (Holtz 1983; Greenway 1987; Gray 1995). The root
system of grass also had a reinforcing effect on the soil mass
near the slope surface. The reinforcing effect appeared to
help to stabilize the soil aggregates so that the relatively
large interaggregate pores could be maintained during rain-
fall. It was also postulated that the reinforcing effect may
hinder the swelling of an expansive soil upon wetting and
hence delay the closing of the open cracks and fissures. The
soil layer near the ground surface in the grassed area showed
407
a greater infiltration rate than that of the bare area. The
surface runoff in the grassed area consequently showed a
longer delay than that in the bare soil area.
After the termination of the simulated rainfall, a substan-
tial amount of sediment was observed in the water collection
trench at the toe of the slope in the bare area. On the other
hand, the sedimentation was found to be negligible in the
grass-covered area. This observation suggests that grass may
provide more effective erosion control on the slope surface.
Conclusions
A full-scale, field-monitoring program was conducted to
investigate the effect of natural grass cover on the rainfall in-
filtration characteristics of an initially unsaturated expansive
soil slope. An artificial rainfall lasting for 7 days (about
370 mm in rainfall depth in total) was simulated and im-
posed on the slope. The observed responses in surface run-
off, pore-water pressure, and water content generally showed
a delay of about 3 days relative to the initiation of the simu-
lated rainfall. Due to the randomly distributed crack patterns,
the responses in pore-water pressure and water content were
spatial and temporal variables. However, the measurements
by Thetaprobes and tensiometers were qualitatively consis-
tent.
The depth of influence by the simulated rainfall, depending
on its elevation along the slope, varied from 2.8 to 3.5 m. At
the end of the simulated rainfall, a significant buildup of posi-
tive pore-water pressures was observed within the influence
depth, particularly at the lower part of the slope. Results of
field monitoring indicated a significant subsurface downslope
flow along the surface of the deep uncracked soil layer.
A comparison of infiltration rates between the grassed
area and the nearby bare area indicated that the grass could
significantly increase the infiltration rate in the shallow soil
layer and retard surface runoff by about 1.5 days. After the
7 day artificial rainfall period, the infiltration rate was esti-
mated to be in the order of 10–7 m/s, which was similar to
© 2007 NRC Canada
408
the stabilized infiltration rate measured from the double-ring
infiltration test on the ground with open cracks.
Acknowledgements
The authors would like to acknowledge the financial support
from research grants HKUST6108/99E and HKUST6087/00E
provided by the Research Grants Council (RGC) of the
Hong Kong Special Administrative Region, and in-kind sup-
port provided by the Yangtse River Scientific Research Insti-
tute, Wuhan, China. The authors would also like to thank
Mr. J.P. Yuan and Professor C.L. Liu of Hohai University,
Nanjing, P.R. China, for their participation in the field tests.
References
Affendi, A.A., and Faisal, A. 1994. Field measurement of soil suc-
tion. In Proceedings of the 13th International Conference on
Soil Mechanics and Foundation Engineering, New Delhi, India,
5–10 January 1994. A.A. Balkema Publishers, Lisse, the Nether-
lands. pp. 1013–1016.
Branson, F.A., Gifford, G.F., and Owen, J.R. 1972. Rangeland hy-
drology. Society for Range Management, Wheat Ridge, Colo-
rado. Range Science Series No. 1.
Bridge, B.J., and Collis-George, N. 1973. An experimental study
of vertical infiltration into a structurally unstable swelling soil
with particular reference to the infiltration throttle. Australia
Journal of Soil Research, 11: 121–132.
Coppin, N.J., Barker, D.H., and Richards, I. 1990. Use of vegeta-
tion in civil engineering. Butterworths, Sevenoaks, Kent, UK.
Favre, F., Boivin, P., and Wopereis, M.C.S. 1997. Water movement
and soil swelling in a dry, cracked vertisol. Geoderma, 78: 113–
123.
Flury, M., Fluhler, W., Jury, W.A., and Leuenberger, J. 1994. Sus-
ceptibility of soils to preferential flow of water: a field study.
Water Resource Research, 30: 1945–1954.
Gray, D.H. 1995. Influence of vegetation on the stability of slopes.
In Vegetation and slopes. Edited by D.H. Barker. Thomas Tel-
ford, London. pp. 2–25.
Greenway, D.R. 1987. Vegetation and slope stability. In Slope sta-
bility. Edited by M.G. Anderson and K.S. Richards. Wiley, Chi-
chester. pp. 187–230.
Can. Geotech. J. Vol. 44, 2007
Hillel, D. 1960. Crust formation in loessial soils. In Transactions
of International Congress of Soil Science, 7th, Madison, Wis-
consin, 14–24 August 1960. Edited by F.A. van Baren. Elsevier,
Amsterdam. Vol. 1, pp. 330–339.
Hillel, D. 1998. Environmental soil physics. Academic Press, San
Diego, California.
Holtz, W.G. 1983. The influence of vegetation on the swelling and
shrinkage of clays in the United States of America. Géo-
technique, 33(2): 159–163.
Johnson, K.A., and Sitar, N. 1990. Hydrologic conditions leading
to debris-flow initiation. Canadian Geotechnical Journal, 27(6):
789–801.
Li, J., Smith, D.W., Fityus, S.G., and Sheng, D.C. 2002. Quantita-
tive analysis of moisture content determination in expansive
soils using neutron probes. In Unsaturated soils: Proceedings of
the Third International Conference, Recife, Brazil, 10–13 March
2002. A.A. Balkema Publishers, Lisse, the Netherlands. Vol. 1,
pp. 363–368.
Lim, T.T., Rahardjo, H., Chang, M.F., and Fredlund, D.G. 1996.
Effect of rainfall on matric suction in a residual soil slope. Ca-
nadian Geotechnical Journal, 33(2): 618–628.
Morel-Seytoux, H.J. 1983. Infiltration affected by air, seai, crust,
ice, and various sources of heterogeneity. In Advances in Infil-
tration: Proceedings of theNational Conference on Advances in
Infiltration, Chicago, Illinois, 12–13 December 1983. American
Society of Agriculture Engineering, St. Joseph, Michigan. Publ.
11-83.
Ng, C.W.W., Zhan, L.T., Bao, C.G., Fredlund, D.G., and Gong,
B.W. 2003. Performance of an unsaturated expansive soil slope
subjected to artificial rainfall infiltration. Géotechnique, 53(2):
143–157.
Øygarden, L., Kværner, J., and Jenssen, P.D. 1997. Soil erosion via
preferential flow to drainage systems in clay soils. Geoderma,
76: 65–86.
Sridharan, A., and Prakash, K. 2000. Classification procedures for
expansive soils. Proceedings of Institution of Civil Engineers:
Geotechnical Engineering, 143: 235–240.
Thetaprobe. 1999. Thetaprobe soil moisture sensor user manual.
Delta-T Device Ltd., Cambridge, UK.
Zhan, L.T. 2003. Field and laboratory study of an unsaturated ex-
pansive soil associated with rain-induced slope instability. Ph.D.
thesis, The Hong Kong University of Science and Technology.
Zhang, J., Jiao, J.J., and Yang, J. 2000. In situ rainfall infiltration
studies at a hillside in Hubei Province, China. Engineering Ge-
ology, 57: 31–38.
© 2007 NRC Canada