NUMERICAL ANALYSES AND MONITORING PERFORMANCE OF RESIDUAL SOIL SLOPES

SOILS AND FOUNDATIONS Vol. 51, No. 3, 471 482, June 2011 Japanese Geotechnical Society NUMERICAL ANALYSES AND MONITORING PERFORMANCE OF RESIDUAL SOIL SLOPES H. RAHARDJO i),v.a.santoso ii),e.c.leong iii),y.s.ng iv) andc.j.hua v) ABSTRACT Changes in the pore-water pressure of unsaturated soil slopes due to rainwater inˆltration comprise a crucial factor which ašects the shear strength of soils and may trigger slope failures. Two residual soil slopes in two main geological formations in Singapore, the Bukit Timah Granite and the sedimentary Jurong Formation, were fully instrumented. Real-time monitoring systems were developed to examine the pore-water pressure, the rainfall and the groundwater level of the slopes for over a year. The characteristics of the pore-water pressure distributions in both slopes during rainfall were highlighted and compared. The monitoring results indicate that the residual soil slope of the Bukit Timah Granite has a thicker unsaturated zone due to a deeper groundwater table than the residual soil slope of the sedimentary Jurong Formation. A higher permeability of the residual soil of the Bukit Timah Granite results in more rapid changes in negative pore-water pressure due to rainwater inˆltration than the residual soil of the sedimentary Jurong Formation. Therefore, the residual soil slope of the Bukit Timah Granite dišers from the residual soil slope of the sedimentary Jurong Formation in the changes in shear strength and the variations in the factor of safety during rainfall. The dišering characteristics of the pore-water pressure responses during rainfall for these two residual soil slopes are analyzed in the present study based on the results of ˆeld measurements and numerical analyses. Key words: matric suction, pore-water pressure, residual soil, slope instrumentation (IGC: E6/E7/E13) INTRODUCTION Rainfall-induced slope failures are a concern in many tropical areas. Global warming is causing climatic changes which are ašecting rainfall patterns in many parts of the world. Proper assessment and management strategies for the stability of residual soil slopes due to these climatic changes and a reduction in the impact of landslides on the environment can be carried out by characterizing the response of various soils during rainwater inˆltration. Numerous research works have been conducted in the ˆelds of geotechnical and geo-environmental engineering to study the importance of rainwater inˆltration, which is the most signiˆcant factor in triggering slope failures in tropical regions (Anderson, 1983; Brand, 1984, 1992; Ching et al., 1984; Tan et al., 1987; De Campos et al., 1988; Krahn et al., 1989; Anderson and Zhu, 1996; Anderson and Thallapally, 1996; Lim et al., 1996; Cheng, 1997; Au, 1998; Franks, 1999; Toll et al., 1999; Gasmo et al., 2000; Tsaparas et al., 2003; Rahardjo et al., 2005, 2006, 2007, 2010). Soils in tropical areas generally consist of residual soils that have negative pore-water pressure above the groundwater table. Changes in pore-water pressure comprise a critical factor that contributes to landslides in Singapore. Rainfall inˆltration results in the decrease in matric suction, and subsequently, reduces the shear strength of the soils. The reduction in shear strength may in turn cause slopes to become unstable and prone to failure. The relationship of the occurrence of landslides to rainfall patterns has been studied over the past 40 years (Lumb, 1962, 1975; Brand, 1984; Finlay et al,. 1997; Dai and Lee, 2001). Signiˆcant progress has been made in laboratory and ˆeld measurements of suction (Ridley and Burland, 1993, 1995; Ridley and Wray, 1996; Delage, 2004; Melani et al., 2002, 2005), physical modeling (Alonso et al., 1990; Gapollipoli et al., 2003; Schnellmann et al., 2010), parametric studies (Tsaparas et al., 2002; Rahardjo et al., 2007) and numerical modeling (Olivella et al., 1996; Gatmiri and Delage, 1997; Gatmiri et al., 1997; Alonso et al., 1998; Gasmo et al., 1999; Lee et al., 2009; Rahardjo et al., 2001, 2010) associated with quantifying the ešects of rainfall on slope stability. Studies on the ešects of rainfall on residual soil slopes using pore-water pressure i) ii) iii) iv) v) Professor, School of Civil and Environmental Engineering, Nanyang Technological University, Singapore (chrahardjo@ntu.edu.sg). ProjectO cer,ditto. Associate Professor, ditto. Deputy Director and Professional Engineer, Building and Infrastructure Department, Housing and Development Board, Singapore. Engineer, ditto. The manuscript for this paper was received for review on April 14, 2010; approved on April 12, 2011. Written discussions on this paper should be submitted before January 1, 2012 to the Japanese Geotechnical Society, 4-38-2, Sengoku, Bunkyo-ku, Tokyo 112-0011, Japan. Upon request the closing date may be extended one month. 471

472 RAHARDJO ET AL. measuring devices have been carried out by Sweeney (1982), Pitts (1985), Fredlund and Rahardjo (1993), ÄOberg (1995), Lim et al. (1996), Ridley et al. (1998), Gasmo et al. (1999), Rezaur et al. (2003), Tsaparas et al. (2003), Li et al. (2005), Cui et al. (2008), Mendes et al. (2008), Rahardjo et al. (2008) and Tarantino et al. (2008). Despite the existence of abundant ˆeld and laboratory data, there are still many uncertainties in accurately estimating the soil moisture, the matric suction and the groundwater level in relation to rainfall inˆltration. There is still a great need for quality ˆeld data to shed light on these uncertainties, and it will be valuable to utilize the measurements obtained from fully instrumented slopes to clarify some of the important issues in slope performance. This study attempts to present the continuous ˆeld measurements of rainfall, inˆltration and pore-water pressure to evaluate the magnitude, the time-dependent behavior and the contours of these variables and to provide explanations for the variations in pore-water pressure responses in relation to rainfall events. This information can then be incorporated into slope stability analyses that account for changes in the shear strength of soils due to inˆltration. The results are expected to have relevance to other geographical locations with similar climatic conditions and to contribute to the limited literature presently available on the hydrologic responses of slopes in tropical regions. To investigate the mechanism of rainfall-induced slope failure, two instrumented residual soil slopes were studied. The monitoring results of the pore-water pressure responses were analyzed to gain a quantitative understanding of the changes in the factor of safety of the slopes due to rainwater inˆltration. Other monitoring results and interpretations from both sites were presented in Rahardjo et al. (2009). In this paper, numerical analyses were conducted to simulate the ˆeld conditions. Given the rainfall input data, the pore-water pressure distribution was obtained by performing a transient seepage analysis using the ˆnite element program SEEP/W (Geo- Slope International Ltd., 2004a). The computed porewater pressure distribution was subsequently used to calculate the factor of safety for each of the two slopes by means of the limit equilibrium method of slices using the computer program SLOPE/W (Geo-Slope International Ltd., 2004b). FIELD STUDY The geology of Singapore (Fig. 1) can be classiˆed into three main formations: (a) the sedimentary rock of the Jurong Formation in the west, (b) the igneous rock of the Bukit Timah Granite in the northwest and center and (c) the semi-hardened alluvium or Old Alluvium in the east, as described in PWD (1976), Pitts (1984) and Leong et al. (2002). Two-thirds of Singapore contain residual soils from the granitic and the sedimentary rock formations. The two instrumented slopes studied in this paper consist of residual soils from the sedimentary Jurong Formation Fig. 1. Generalized geological map of Singapore and schematic diagram of relative positions and arrangement of ˆeld instrumentation for the Jalan Kukoh and Marsiling Road slopes and the Bukit Timah Granite. FIELD INSTRUMENTATIONS Real-time monitoring systems were used in the soil slopes to study the responses of the slopes to environmental changes. The measuring devices installed in the slopes consist of jet-ˆll tensiometers, tipping-bucket rain gauges and piezometers to measure the negative pore-water pressure, the rainfall and the groundwater level, respectively. At each site, sixteen tensiometers, three piezometers and one rain gauge were installed in four rows, as illustrated in the schematic diagram of the relative positions and the arrangement of the ˆeld instrumentation for the Jalan Kukoh and the Marsiling Road sites in Fig. 1. Each row consisted of four tensiometers installed at depths of 0.64, 1.31, 1.66 and 2.08 m below the ground surface. Casagrande piezometers, with depth transmitters connected to a data acquisition system, were used to monitor changes in the groundwater table in response to rainfall. The three piezometers at the Jalan Kukoh site with piezometer depths of 13, 10 and 16 m were installed at the crest, the middle and the toe of the slope, respectively. The three piezometers at the Marsiling Road site with piezometer depths of 20, 14 and 11 m were installed at the crest, the middle and the toe of the slope, respectively. Data from these instruments were captured by the data acquisition system and subsequently transmitted to a secured website using the General Packet Radio Service (GPRS). Data acquisition systems were set up for fullscale ˆeld experiments. Instrumentation huts were constructed at the crest of the slopes to house the data loggers and the power supplies. The data acquisition systems consisted of sensors, cables, data loggers and power supplies. Pressure transducers were used in all the tensiometers and piezometers for automated data collection. The data loggers were conˆgured to collect data at 10-minute intervals if there was rainfall and at 30-minute intervals if there was no rainfall. Real-time information was processed and accessed through a secured website.

MONITORING PERFORMANCE OF SLOPES 473 RESIDUAL SOIL SLOPES OF SEDIMENTARY JURONG FORMATION AND BUKIT TIMAH GRANITE Site Descriptions The Jalan Kukoh site is a natural slope covered with grass at a height of 11.74 m and an inclination of 339. Fig. 2. Slope geometry, ˆnite element mesh and boundary conditions of the Jalan Kukoh slope Fig. 3. Slope geometry, ˆnite element mesh and boundary conditions of the Marsiling Road slope This site is underlain by residual soils from the sedimentary Jurong Formation. Borehole logs obtained from the Jalan Kukoh site indicate that the subsoils generally consist of clayey sand. The ˆll material observed from the borehole data contains ˆrm sandy silt, soft clay with sand and rock fragments, loose silty sand with rock and wood and some foreign materials. The Marsiling Road site is a natural slope located in the northwest region of Singapore. The slope is approximately 17 m high with an average inclination of about 26.79. This site consists of residual soils from the Bukit Timah Granite, mainly silty sand and sandy silt. The slope geometries and the properties of the soils at the Jalan Kukoh and the Marsiling Road slopes are shown in Figs. 2 and 3, respectively. Soil Properties The basic soil properties of the residual soils at the Jalan Kukoh and the Marsiling Road slopes are presented in Table 1. Soil-water characteristic curves (SWCC) for the soils were ˆtted using Fredlund and Xing SWCC equations (1994), as presented in Fig. 4. In general, the SWCC for the residual soils from the Bukit Timah Granite at the Marsiling Road site have a higher saturated volumetric water content, a lower air-entry value, a steeper slope and a higher residual water content than the SWCC for the residual soils from the sedimentary Jurong Formation at the Jalan Kukoh site. The permeability functions for the three types of residual soils were estimated using the statistical method and are presented in Fig. 5. The SWCC parameters and the saturated permeability obtained from laboratory tests, k s,onthese materials are summarized in Table 2. The shear strength parameters of the clayey sand, the silty sand and the sandy silt layers were required for the slope stability analyses. The shear strength parameters Table 1. Basic properties of soils Parameters Sedimentary Jurong Formation (JF) Bukit Timah Granite (BT) Clayey sand I Clayey sand II Silty sand Sandy silt EŠective cohesion c? (kpa) 4 0 9 0 EŠective angle of friction q? (9) 33 36 34 33 q b angle (9) 25 26.5 21 26.1 Saturated coe cient of permeability k s (m/s) 8.21 10-6 8.21 10-6 6 10-6 3.3 10-5 Total density r t (Mg/m 3 ) 1.84 1.92 2.03 1.88 Saturated water content w sat (z) 19.65 16.24 38.43 35.58 Speciˆc gravity G s 2.53 2.64 2.66 2.58 Void ratio e= Ø Gsrw r t 1+w sat» -1 0.60 0.65 0.8 0.86 Liquid limit LL (z) 47 30 52 54 Plasticity index PI 33 13 21 15 Gravel (z) 4 0 1 0 Sand (z) 69 73 53 40 Silt (z) 12 14 37 57 Clay (z) 16 13 9 3 Fines (z) 46 60 USCS group symbol SC SC SM MH USCS: Uniˆed Soil Classiˆcation System

474 RAHARDJO ET AL. for unsaturated soil, such as ešective cohesion c?, ešective angle of internal friction q? and angle q b, were determined from consolidated drained triaxial tests, as presented in Table 1. In the numerical analyses, the low retaining wall at the toe of the Jalan Kukoh slope was assumed to have a unit weight of g=24 kn/m 3 and a low permeability of k s =10-11 m/s. The retaining wall was assumed as a soil layer with an ešective cohesion of c?=5 kpa and an ešective internal friction angle of q?=409. Fig. 4. SWCCs of soils at the Jalan Kukoh and Marsiling Road slopes Fig. 5. Permeability functions of soils at the Jalan Kukoh and Marsiling Road slopes Initial and Boundary Conditions The ˆnite element mesh and the boundary conditions used for the seepage analysis are presented in Figs. 2 and 3 for the Jalan Kukoh and the Marsiling Road slopes, respectively. The left and right edges were located at a distance of 3H s (H s =slope height) from the crest and the toe, respectively, in order to avoid any in uence from the boundary conditions on the seepage process within the slope area. A transient seepage analysis was performed by applying a small uniform inˆltration to the slope surface for a long duration to represent the net ux of water at the site in order to obtain the initial proˆle of the porewater pressure. DiŠerent amounts of rainfall with dišerent rainfall durations were applied to the slope for dišerent soil types. For instance, a rainfall intensity of 1 10-7 m/s was applied to the Jalan Kukoh slope for 23 days and a rainfall intensity of 1 10-7 m/s was applied to the Marsiling Road slope for 62 days. Subsequently, another transient seepage analysis was carried out for actual rainfall. Flux boundary conditions were applied on the seepage face along the slope surface by considering the potential seepage face review. Flux boundary conditions were assigned to the ground surface which allowed inˆltration at a rate that depended on the soil permeability and the hydraulic gradient, and the remaining amount of rainwater was ˆxed to runoš. In addition, a non-ponding condition was applied to the seepage analyses. This means that the software will check to be sure that every node at the ground surface has a zero or a negative hydraulic head. The side boundaries below the groundwater table were speciˆed as constant total head boundaries, while the side boundaries above the groundwater table were speciˆed as no- ow zones or nodal ux Q, equal to zero. The total head applied corresponded to the hydrostatic condition. Tensiometer measurements in Singapore have been found to have a limit value of 75 kpa for negative pore-water pressure (Rahardjo, 2000). Therefore, a hydrostatic porewater pressure distribution with a limiting negative porewaterpressureof75kpawassetastheinitialcondition for the slope model in this paper. Seepage analyses of the Jalan Kukoh slope were performed using SEEP/W and the actual rainfall patterns on 26 September 2008 with a total rainfall of 155.2 mm Table 2. Fredlund and Xing parameters for residual soils Description Sedimentary Jurong Formation (JF) Bukit Timah Granite (BT) Clayey sand I Clayey sand II Silty sand Sandy silt Saturated coe cient of permeability k s (m/s) 8.21 10-6 8.21 10-6 6 10-6 3.3 10-5 Saturated volumetric water content u s 0.4020 0.5800 0.6000 0.5306 Fitting parameter a (kpa) 298 65 101 7 Fitting parameter n 0.62 1.27 0.55 5 Fitting parameter m 1.08 1.54 1.33 0.70 Air-entry value (AEV) 50 20 15 5 Residual volumetric water content u r 0.15 0.05 0.15 0.15 Residual suction c r (kpa) 7000 600 6000 22 C(c)=1 as suggested by Leong and Rahardjo (1997)

MONITORING PERFORMANCE OF SLOPES 475 Fig. 6. Rainfall and inˆltration rate on 26 September 2008 used in the analyses of the Jalan Kukoh slope Fig. 8. Total head contours obtained from the ˆeld measurements for the Jalan Kukoh slope on 26 September 2008: (a) before rainfall and (b) at the end of rainfall Fig. 7. Rainfall and inˆltration rate on 11 March 2008 used in the analyses of the Marsiling Road slope for 5 h, as depicted in Fig. 6. On the other hand, seepage analyses of the Marsiling Road slope were carried out using the actual rainfall pattern of 11 March 2008 with a total rainfall of 74.4 mm over 6 h, as illustrated in Fig. 7. Seepage Analyses Total head contours were generated using Surfer (Golden Software, Inc., 1997), based on the ˆeld measurements of negative pore-water pressure in the Jalan Kukoh residual soil slope of the sedimentary Jurong Formation before and at the end of rainfall, as presented in Fig. 8. The total head is equal to the sum of elevation head h e and pore-water pressure head h p. Water ow occurs as a result of the dišerence in total head. A comparison of total head contours obtained from the numerical analyses and the total head contours based on the ˆeld measurements of the Jalan Kukoh residual soil slope before and at the end of rainfall is shown in Fig. 9. The total head contours obtained from the ˆeld measurements in the Marsiling Road residual soil slope from the Bukit Timah Granite before and at the end of rainfall were generated using Surfer, as presented in Fig. 10. A comparison of the total head contours obtained from the numerical analyses and the total head contours based on the ˆeld measurements in the residual soil slope from the Fig. 9. Total head contours obtained from numerical analyses versus ˆeld data for the Jalan Kukoh slope on 26 September 2008: (a) before rainfall and (b) at the end of rainfall Bukit Timah Granite before and at the end of rainfall are shown in Fig. 11. Generally, it can be seen that the total head contours

476 RAHARDJO ET AL. Fig. 10. Total head contours obtained from the ˆeld measurements for the Marsiling Road slope on 11 March 2008: (a) before rainfall and (b) at the end of rainfall Fig. 11. Total head contours obtained from numerical analyses versus ˆeld data for the Marsiling Road slope on 11 March 2008: (a) before rainfall and (b) at the end of rainfall obtained from the ˆeld measurements and the numerical analyses of the residual soil slope from the sedimentary Jurong Formation before and at the end of rainfall were in reasonably good agreement. The presence of signiˆcant matric suction near the ground surface during the dry period contributed to the stability of the slope. It appears that water ow occurred within the slope during the dry period when water moved upward, due to evaporation, and also across the slope, as shown in Figs. 8(a) and 9(a). Matric suction near the ground surface decreased during rainfall due to water inˆltrating the slope, as illustrated in the total head distributions in Figs. 8(b) and 9(b). The total head contours in Fig. 8(b) indicate that water inˆltrated the slope and owed from the crest to the toe of the slope. The total head contours obtained from the ˆeld measurements and the numerical analyses of the residual soil slope from the Bukit Timah Granite before and at the end of rainfall were in good agreement. Figures 10(a) and 11(a) indicate the existence of a hydraulic head gradient between the slope crest and the slope toe even during the dry period. In other words, there was a movement of water from the crest to the toe of the slope in addition to the upward movement of water due to evaporation. As a result, the lower slope area remained wetter than the upper slope area. Figures 10(b) and 11(b) show that the hydraulic gradient during the wet period caused a downward movement of rainwater into and across the slope, increasing the pore-water pressure and reducing the matric suction during the wet period. The contour lines in Figs. 11(a) and (b) represent the total heads of 134 m and 135 m and indicate more drainage downslope due to the hydraulic head gradient during the wet period and the rising of the groundwater table at the toe of the slope. A comparison of the pore-water pressure proˆles obtained from the numerical analyses and the ˆeld measurements near the crest, the middle and the toe of the residual soil slope from the sedimentary Jurong Formation are presented in Fig. 12. It can be seen from Figs. 12(a) and (b) that the numerical analyses were able to simulate the process in the ˆeld with reasonably good accuracy. The pore-water pressure proˆles obtained from the numerical analyses near the crest and the middle of the slope were very close to the data measured in the ˆeld. The negative pore-water pressure approached zero near the ground surface due to rainfall. It was suspected that cavitation occurred in the tensiometers in Row A (at the crest of the slope), and therefore, there was no proper measurement of the negative pore-water pressure available at the crest. The results of the numerical analyses did not quite agree with the data measured near the toe of the slope (Fig. 12(c)). In the numerical analyses, the initial porewater pressure proˆle near the toe of the slope did not match that of the ˆeld data that had positive pore-water pressure of about 18 kpa at depths of 1.66 and 2.08 m from the ground surface. The actual pore-water pressure near the toe of the slope (below an elevation of 104.4 m) built up faster than the computed pore-water pressure from the numerical analyses, causing discrepancies below this elevation. The discrepancies were caused by the ponding of water near the toe of the slope at the beginning of the rainfall, and thus, the pore-water pressure did not change signiˆcantly when the rainfall actually oc-

MONITORING PERFORMANCE OF SLOPES 477 Fig. 12. Pore-water pressure proˆles obtained from numerical analyses versus ˆeld data in Row: (a) B, (b) C and (c) D, for the Jalan Kukoh slope from the beginning of rainfall (t=0) until the end of rainfall (t=5 h) curred. A comparison of the pore-water pressure proˆles obtained from the numerical analyses and the ˆeld measurements near the crest, the middle and the toe of the residual soil slope of the Bukit Timah Granite is presented in Fig. 13. It can be seen that generally the numerical analyses were able to simulate the process in the ˆeld with reasonably good accuracy. The pore-water pressure pro- ˆles obtained from the numerical analyses near the crest and the middle of the slope were very close to the ˆeld measurements. The measured negative pore-water pressure in the residual soil slope of the Bukit Timah Granite increased to a value of zero after rainfall. At the toe of the slope, positive pore-water pressure up to 12 kpa was measured after rainfall. Figures 13(a) and (b) appear to have drastic changes in the pore-water pressure distribution at a depth of 141 m due to a contrast in the unsaturated coe cient of permeability between the silty sand (k sat =6 10-6 m/s) and the sandy silt (k sat =3.3 10-5 m/s) layers that limits the downward movement of the water. Since there is only silty sand (the 1st layer) above the groundwater table in Row D, such a drastic change is not observed in Fig. 13(d). The inˆltration rates at the slope face, the top and the toe due to rainfall on 26 September 2008 at the Jalan Kukoh slope, as obtained from the numerical analyses, are also shown in Fig. 6. The inˆltration rates at the slope face, the top and the toe (i.e., about 20 mm/h) were much lower than the rainfall rates with a maximum rate of 148 mm/h. Figure 14 shows the rate of increase in pore-water pressure and the rainfall rate against time as obtained from the ˆeld measurements. As soon as the rainwater inˆltrated, an immediate increase in pore-water pressure took place with a maximum rate of increase of 24 kpa/h at an elapsed time of 0.5 hours at a depth of 0.64 m from the ground surface. At this depth, the rate of increase in pore-water pressure gradually slowed down to 3.4 kpa/h and eventually reached zero after an elapsed time of 2.83 hours. Meanwhile, a less signiˆcant rate of increase in pore-water pressure due to rainfall was shown at the deeper depths of tensiometers, e.g., 1.31, 1.66 and 2.08 m depths from the ground surface. Figure 15 shows the percentage of increase in porewater pressure and the accumulated rainfall against time as obtained from the ˆeld measurements for the Jalan Kukoh slope. It can be seen that the response of the increase in pore-water pressure due to rainfall was the fastest at the depth of 0.64 m below the ground surface, as demonstrated by the steepest increment in gradient

478 RAHARDJO ET AL. Fig. 13. Pore-water pressure proˆles obtained from numerical analyses versus ˆeld data in Row: (a) A, (b) B, (c) C and (d) D for the Marsiling Road slope from the beginning of rainfall (t=0) until the end of rainfall (t=6 h) Fig. 14. Rate of increase in pore-water pressure at dišerent depths of tensiometers obtained from ˆeld measurements at Row C for the Jalan Kukoh slope due to rainfall Fig. 15. Response of pore-water pressure at dišerent depths of tensiometers obtained from ˆeld measurements at Row C for the Jalan Kukoh slope due to rainfall during the ˆrst half hour. Its trend resembled that of the accumulated rainfall. This could be attributed to the fact that water owed down faster near the slope surface, which consisted of ˆll materials of heterogeneous and coarser particle sizes 1.5 m in depth. The tensiometers located at depths of 1.31, 1.66 and 2.08 m below the ground level showed a slower response in reducing the negative pore-water pressure due to the rainfall events, because the rainwater required a longer time to reach these deeper depths. The fact that the rates of increase in pore-water pressure at depths of 1.66 and 2.08 m were faster than that at the depth of 1.31 m could be due to the presence of lateral ow from the upper slope to the lower slope within the 1st clayey sand layer below the ˆll material 1.5 m in thickness. The inˆltration rates at the slope face, the top and the toe due to the rainfall on 11 March 2008 at the Marsiling

MONITORING PERFORMANCE OF SLOPES 479 Road slope, as obtained from the numerical analyses, are also shown in Fig. 7. The inˆltration rate at the top of the slope appeared to be the highest, followed by those at the slope face and the slope toe. The variation in inˆltration rates of the residual soil slope from the Bukit Timah Granite at Marsiling Road appeared to closely resemble the variation in rainfall rates (Fig. 7) due to the high permeability of the top layer of the slope (i.e., the sandy silt layer). This is in contrast to the variation in inˆltration rate of the residual soil slope from the sedimentary Jurong Formation at the Jalan Kukoh site that did not resemble the variation in rainfall rate (Fig. 6) due to the low permeability of the top layer of the slope. As the soil became wet, the permeability increased. Figure 7 shows that for a rainfall rate (i.e., 1.33 10-5 m/s in the early stage of rainfall) greater than k sat =6 10-6 m/s (surface soil), the initial inˆltration rate (45.69 mm/h=9.20 10-4 m/s) was higher than the k sat value by as much as 1.8 times. In other words, rainwater could in- ˆltrate the soil at a rate higher than the saturated permeability of the soil. This was made possible due to the high hydraulic head gradient, i,availableforthe ow, resulting in a high inˆltration rate, [, even greater than k sat following Darcy's law, [=k sat i. This observation agrees with the results reported by Gasmo et al. (2000). It should be noted that a rapid inˆltration rate may result in a rapid decrease in the shear strength of the soil, and consequently, a decrease in the stability of the slope. Figure 16 presents the rate of increase in pore-water pressure and the rainfall rate against time as obtained from the ˆeld measurements at the Marsiling Road slope. It can be seen that the maximum rate of increase in porewater pressure at the depth of 0.64 m from the ground surface (i.e., 13.8 kpa/h) occurred after the third peak of the rainfall event at the elapsed time of 2.3 hours. Meanwhile, the maximum rate of increase in pore-water pressure was 9.24 kpa/h at the depth of 1.31 m below the ground surface, which occurred after the fourth peak of the rainfall event at the elapsed time of 4.67 hours. At the depth of 1.66 m below the ground surface, the rate of increase in pore-water pressure reached its maximum rate of 3.18 kpa/h during the ˆfth peak of the rainfall event at the elapsed time of 5.5 hours. At the deeper depth of 2.08 m below the ground surface, only small changes in pore-water pressure were observed, as it was not quite ašected by the rainfall event. From Fig. 16, it can be observed that the increase in pore-water pressure throughout the depth, due to the rainwater percolation in response to a rainfall event, has a time lag of about 0.3 to 0.6 hours in this case. Figure 16 indicates that as soon as the rain begins, there is a rapid response, namely, an increase in porewater pressure, due to inˆltration at depths of 0.64 and 2.08 m below the slope surface. However, while the rainwater inˆltration continues with time, the response of the increase in pore-water pressure at depths of 1.66 and 1.31 m below the ground surface is faster than the response at depths of 0.64 and 2.08 m below the ground surface. This is due to the lateral water ow from the upper slope Fig. 16. Rate of increase in pore-water pressure at dišerent depths of tensiometers obtained from ˆeld measurements at Row C for the Marsiling Road slope due to rainfall Fig. 17. Response of pore-water pressure at dišerent depths of tensiometers obtained from ˆeld measurements at Row C for the Marsiling Road slope due to rainfall towards the lower slope. In Fig. 17, the percentage of increase in pore-water pressure and the accumulated rainfall are plotted against time using data from the ˆeld measurements at the Marsiling Road slope. A rapid response of the increase in pore-water pressure due to rainfall is shown at depths of 0.64 and 2.08 m below the ground surface with the steepest gradient occurring after the second peak of the rainfall event (i.e., elapsed time of 2 to 3 hours). The tensiometers located at depths of 1.31 and 1.66 m below the ground surface showed a slower response in reducing the negative pore-water pressure due to the rainfall. After the elapsed time of 2.5 hours, the response of the increase in pore-water pressure at the depth of 2.08 m was faster than the response at depths of 1.66 and 1.31 m below the ground surface. The presence of the high water content of soil, as a result of water accumulation at the toe of the slope, was suspected to be the cause of the faster rate of increase in pore-water pressure at the depth of 2.08 m. Slope Stability Analyses Rainwater inˆltrated the unsaturated zone of the soil

480 RAHARDJO ET AL. Fig. 18. Changes in factor of safety for the Jalan Kukoh and Marsiling Road slopes during and after rainfall on 26 September and 11 March 2008, respectively Fig. 19. Factor of safety and critical slip surfaces for the Jalan Kukoh slope: (a) before rainfall on 26 September 2008 and (b) at elapsed time of 30 h and increased the negative pore-water pressure. When the rainwater percolated downward, the matric suction at deeper depths also decreased, causing a reduction in shear strength, and subsequently, a reduction in the factor of safety of the slope. As the inˆltrating rainwater became less than the rainwater percolating downward into the slope, the pore-water pressure above the wetting front started to decrease again, the shear strength of the soil increased, and consequently, the factor of safety of the slope started to increase. The computer program SLOPE/W (Geo-slope International Ltd., 2004b) was used to perform a limit equilibrium slope stability analysis. Bishop's simpliˆed method was selected for the analysis due to its ability to calculate the factor of safety with accuracy close to the more rigorous methods (Ching and Fredlund, 1984; Fredlund and Krahn, 1977). Slope stability analyses were performed by importing the obtained pore-water pressure distributions from the seepage analyses. The time-dependent pore-water pressure distributions were used to calculate the factor of safety with time. Figure 18 presents the factor of safety with time from the slope stability analyses using SLOPE/W. The initial factor of safety for the Jalan Kukoh residual soil slope was 1.37, which decreased gradually to a value of 1.33 when the rainfall stopped (i.e., elapsed time of 5 hours) and continued to decline to a minimum value of 1.28 25 hours after the rainfall stopped (i.e., elapsed time of 30 hours). The minimum factor of safety for the Jalan Kukoh slope did not coincide with the time when the rainfall stopped, but 25 hours after the rainfall stopped. The lower permeability of the residual soil slope from the sedimentary Jurong Formation resulted in a slower rainwater inˆltration, which delayed the occurrence of the minimum factor of safety to sometime later after the rainfall had stopped. A steeper gradient for the reduction in the factor of safety at the Marsiling Road slope during the rainfall illustrates the important role of the permeability of the soil. The high permeability of the soil allowed rainwater to inˆltrate and percolate downward quickly to deeper Fig. 20. Factor of safety and critical slip surfaces for the Marsiling Road slope: (a) before and (b) at the end of rainfall event on 26 September 2008 depths. As a result, a rapid increase in the pore-water pressure developed in the slope. As a consequence, the factor of safety at the Marsiling Road slope decreased rapidly until it reached a minimum value of 1.57 when the rainfall stopped, i.e., at an elapsed time of 6 hours. The location of the deeper initial water table at the Marsiling Road slope might contribute to the higher initial factor of safety of the slope as compared to that of the Jalan Kukoh slope. On the other hand, the variation in the factor of safety during the rainfall is controlled by the applied rainfall characteristics (including the rainfall pattern, the intensity and the duration) and the soil properties of the slope, particularly the coe cient of permeability. Figure 18 also shows that the recovery rate for the factor of safety of the Marsiling Road slope after the rainfall was faster than that of the Jalan Kukoh slope due to the higher permeability of the residual soil from the Bukit Timah Granite. Figures 19 and 20 present the critical failure surfaces of the Jalan Kukoh and the Marsiling Road slopes, respectively. The depths of the critical failure surfaces of the Jalan Kukoh slope before the rainfall and at the elapsed time of 30 h, when the minimum factor of safety occurred, were 4.6 m and 3.1 m from the slope surface, respectively. Meanwhile, the depths of the critical failure surfaces of the Marsiling Road slope before and at the end of the rainfall were 9.1 m and 5.3 m from the slope surface, respectively.

MONITORING PERFORMANCE OF SLOPES 481 CONCLUSIONS Two residual soil slopes from the Bukit Timah Granite and the sedimentary Jurong Formation have been fully instrumented with real-time monitoring systems to provide valuable information on the changes in pore-water pressure, the rainfall and the groundwater level during rainfall and dry periods. These changes have been simulated in numerical analyses in order to quantify the response characteristics of the two dišerent residual soil slopes to rainfall. The following conclusions can be drawn from the monitoring results and numerical analyses: 1. The residual soil slope of the sedimentary Jurong Formation has a shallower initial groundwater table, resulting in a lower initial factor of safety than the residual soil slope of the Bukit Timah Granite. 2. The variation in the factor of safety during rainfall is determined by the applied rainfall to the slope (rainfall pattern, intensity and rainfall duration) and the properties of the soil (coe cient of permeability and shear strength) in the slope. 3. The variation in the inˆltration rate of the Marsiling Road residual soil slope in the Bukit Timah Granite appeared to closely resemble the variation in the rainfall rate because of the high permeability of the top layer of the slope. Meanwhile, the variation in the in- ˆltration rate of the Jalan Kukoh residual soil slope in the sedimentary Jurong Formation did not resemble the variation in the rainfall rate due to the low permeability of the top layer of the slope. 4. A faster rate of increase in negative pore-water pressure, due to the rainwater inˆltration, was observed in the residual soil slope of the Bukit Timah Granite as compared to that in the residual soil slope of the sedimentary Jurong Formation. This was caused by the higher permeability of the residual soil of the Bukit Timah granite as compared to the permeability of the residual soil of the sedimentary Jurong Formation. Consequently, the factor of safety of the residual soil slope of the Bukit Timah Granite decreased at a faster rate during rainfall and also increased at a faster rate after the rainfall stopped as compared to the factor of safety of the residual soil slope of the sedimentary Jurong Formation. ACKNOWLEDGEMENTS This study was supported by a research grant from the Housing and Development Board and Nanyang Technological University, Singapore. The authors gratefully acknowledge the assistance of the staš at Geotechnical Laboratory and Research, School of Civil and Environmental Engineering, NTU, Singapore with the data collection. REFERENCES 1) Alonso, E. E., Gens, A. and Josa, A. (1990): A constitutive model for partially saturated soils, Gáeotechnique, 40(3):405 430. 2) Alonso,E.E.,Lloret,A.,Delahaye,C.H.,Vaunat,J.,Gens,A. and Volckaert, G. (1998): Coupled analysis of a backˆll hydration test, Int. J. Numer. Anal. Methods Geomech., 22, 1 27. 3) Anderson, M. G. (1983): Prediction of soil suction for slopes in Hong Kong, Geotechnical Control O ce, Hong Kong, GCO publication, 1(84), 243. 4) Anderson, S. A. and Thallapally, L. K. (1996): Hydrologic response of steep tropical slope to heavy rainfall, In: Senneset K. (ed.), Proc. of the 7 th Int. Symposium on Landslides. Trondheim, Norway, Balkema, Rotterdam, 1489 1495. 5) Anderson, S. A. and Zhu, J. H. (1996): Assessing the stability of a tropical residual soil slope, Proc. of the 7th Int. Symposium on Landslides. (ed. by Senneset K.), Trondheim, Norway, Balkema, Rotterdam, 1073 1077. 6) Au, S. W. C. (1998): Rain-induced slope instability in Hong Kong, Engineering Geology, 51, 1 36. 7) Brand, E. W. (1984): State-of-the-art report of landslides in Southeast Asia, Proc. of the 4th Int. Symposium on Landslides, Toronto, Canada, 17 37. 8) Brand, E. W. (1992): Slope instability in tropical areas, Proc. of the 6th Int. Symposium on Landslides, Christchurch, New Zealand, 3, 2031 2051. 9) Cheng, P. F. K. (1997): Soil inˆltration prediction and its signiˆcance in slope stability, Proc. of the 3rd Asian Young Geotechnical Engineers Conf. (eds. by Tan T. S. et al.), Singapore, 661 669. 10) Ching, R. K. H. and Fredlund, D. G. (1984): Quantitative comparison of limit equilibrium methods of slices, Proc. of the 4th Int. Symposium on Landslides, Toronto, Canada. September 16 21, 373 379. 11) Ching, R. K. H., Sweeney, D. J. and Fredlund, D. G. (1984): Increase in factor of safety due to soil suction for two Hong Kong slopes, Proc. of the 4th Int. Symposium on Landslides, Toronto, Canada, 617 623. 12) Cui, Y. J., Tang, A. M., Mantho, A. T. and De Laure, E. (2008): Monitoring ˆeld suction using a miniature tensiometer, Geotech. Testing J, 31(1), 95 100. 13) Dai, F. C. and Lee, C. F. (2001): Terrain-based mapping of landslide susceptibility using a geographical information system: a case study, Can. Geotech. J. 38, 911 923. 14) De Campos, L. E. P., De Sousa Menezes, M. S. and Fonseca, E. C. (1988): Parameter selection for stability analysis, Proc. of the 2nd Int. Conf. on Geomechanics in Tropical Soils, (ed. by Publications Committee of 2 ICOTS), Singapore. Balkema, Rotterdam, 241 243. 15) Delage, P. (2004): State of the art report-experimental unsaturated soil mechanics, Proc. of 3rd Int. Conf. on Unsaturated Soils, Recife, Brazil, 3, 973 998. 16) Finlay, P. J., Fell, R. and Maguire, P. K. (1997): The relationship between the probability of landslide occurrence and rainfall, Can. Geotech. J, 34, 811 824. 17) Franks, C. A. M. (1999): Characteristics of some rainfall-induced landslides on natural slopes, Lantau Island, Hong Kong, Quarterly J. of Eng. Geology, 32, 247 259. 18) Fredlund, D. G. and Krahn, J. (1977): Comparison of slope stability methods of analysis, Can. Geotech. J., 14, 429 439. 19) Fredlund, D. G. and Rahardjo, H. (1993): Soil Mechanics for Unsaturated Soils, NewYork,JohnWileyandSons,Inc.,517. 20) Fredlund, D. G. and Xing, A. (1994): Equations for the soil-water characteristic curve, Can. Geotech. J., 31, 533 546. 21) Gapollipoli, D., Gens, A., Sharma, R. and Vaunat, J. (2003): An elastoplastic model for unsaturated soil incorporating the ešects of suction and degree of saturation on mechanical behavior, Gáeotechnique, 53(1), 123 136. 22) Gasmo, J. M., Hritzuk, J. K., Rahardjo, H. and Leong, E. C. (1999): Instrumentation of an Unsaturated residual soil slope, Geotechnical Testing Journal, ASTM International, 22(2), 134 143. 23) Gasmo, J. M., Rahardjo, H. and Leong, E. C. (2000): Inˆltration ešects on stability of a residual soil slope, Computer and Geo-

482 RAHARDJO ET AL. technics, 26, 145 165. 24) Gatmiri, B. and Delage, P. (1997): A formulation of fully coupled thermal-hydraulic-mechanical behavior of saturated porous media: numerical approach, Int. J. Anal. Numer. Methods Geomech., 21, 199 225. 25) Gatmiri, B., Delage, P. and Cerrola, M. (1997): U-Dam: a powerful ˆnite element software for the analysis of unsaturated porous media, Int I. Adv. Eng. Software, 29(1), 29 43. 26) Geoslope International Pte. Ltd. (2004a): SEEP/W User's Guide for Slope Stability Analysis, Calgary, Geoslope International Ltd. 27) Geoslope International Pte. Ltd. (2004b): SLOPE/W User's Guide for Slope Stability Analysis, Calgary, Geoslope International Ltd. 28) Golden Software, Inc. (1997): Surfer Version 6, Contouring and 3D Surface Mapping, Golden, CO, USA, User's Guide. 29) Krahn, J., Fredlund, D. G. and Klassen, M. J. (1989): EŠect of soil suction on slope stability at Notch Hill, Can. Geotech. J., 26, 269 278. 30) Lee,M.L.,Gofar,N.andRahardjo,H.(2009):Asimplemodel for preliminary evaluation of rainfall-induced slope instability, J. of Eng. Geology, 108, 272 285. 31) Leong, E. C. and Rahardjo, H. (1997): A review on soil-water characteristic curve equations, ASCE J. of Geotechnical and Geoenvironmental Eng., 123(12), 1106 1117. 32) Leong, E. C., Rahardjo, H. and Tang, S. K. (2002): Characterisation and engineering properties of Singapore residual soils, Proc. of the Int. Workshop on Characterisation and Engineering Properties of Natural Soils, Singapore, December, 2 4, 2, 1279 1304. 33)Li,A.G.,Yue,Z.Q.,Tham,L.G.,Lee,C.F.andLaw,K.T. (2005): Field-monitored variations of soil moisture and matric suction in a saprolite slope, Can. Geotech. J., 42, 13 26. 34) Lim, T. T., Rahardjo, H., Chang, M. F. and Fredlund, D. G. (1996): EŠect of rainfall on matric suction in a residual soil slope, Can. Geotech. J., 33, 618 628. 35) Lumb, P. (1962): General nature of the soil of Hong Kong, Proc. of Symposium on Hong Kong Soils, Hong Kong, 19 23. 36) Lumb, P. (1975): Slope failures in Hong Kong, Quarterly J. of Eng. Geology, 8, 31 65. 37) Meilani, I., Rahardjo, H., Leong, E. C. and Fredlund, D. G. (2002): Mini suction probe for matric suction measurement, Can. Geotech. J., 39(6), 1427 1432. 38) Meilani, I., Rahardjo, H. and Leong, E. C. (2005): Pore-water pressure and water volume change of an unsaturated soil under in- ˆltration conditions, Can. Geotech. J., 42(6), 1509 1531. 39) Mendes, J., Toll, D. G., Augarde, C. E. and Gallipoli, D. (2008): A system for ˆeld measurement of suction using high capacity tensiometers, Proc. of 1st European Conf. on Unsaturated Soils, Durham, UK, July 2008. 40) ÄOberg, A. L. (1995): Negative pore pressures-seasonal variation and importance in slope stability analysis, Proc. of 1st Int. Conf. on Unsaturated Soils, Paris, 1995. 41) Olivella,S.,Gens,A.,Carrera,J.andAlonso,E.E.(1996):Numerical formulation for a simulator (CODE_BRIGHT) for the coupled analysis of saline media, Eng. Comput., 13, 87 112. 42) Pitts, J. (1984): A review of geology and engineering geology in Singapore, QuarterlyJ.ofEng.Geology, 17, 93 101. 43) Pitts, J. (1985): An investigation of slope stability on the NTU Campus, Singapore, Applied Research Project RPI/83, Nanyang Technological Institute, Singapore, 54. 44) PWD (1976): Geology of the Republic of Singapore, Public Works Department, Singapore. 45) Rahardjo, H. (2000): Rainfall-induced slope failures, Research report: NSTB 17/6/16, Nanyang Technological University, Singapore, 407. 46) Rahardjo, H., Li, X. W., Toll, D. G. and Leong, E. C. (2001): The ešect of antecedent rainfall on slope stability, J. of Geotechnical and Geological Eng. Special Issue on ``Unsaturated and Collapsible Soils,'' 19(3 4), 371 399. 47) Rahardjo,H.,Lee,T.T.,Leong,E.C.andRezaur,R.B.(2005): Response of a residual soil slope to rainfall, Can. Geotech. J., 42(2), 340 351. 48) Rahardjo,H.,Lee,T.T.,Leong,E.C.andRezaur,R.B.(2006): Reply to Discussion by Londono A.C. on ``Response of a Residual Soil Slope to Rainfall'' by H. Rahardjo, T.T. Lee, E.C. Leong and R.B. Rezaur. [Can. Geotech. J., 42: 340 351 (2005)]. Can. Geotech. J., 43, 979 984. 49) Rahardjo, H., Ong, T. H., Rezaur, R. B. and Leong, E. C. (2007): Factors controlling instability of homogeneous soil slopes under rainfall loading, ASCE J. of Geotechnical and Geoenvironmental Eng., 133(12), 1532 1543. 50) Rahardjo, H., Leong, E. C. and Rezaur, R. B. (2008): EŠect of antecedent rainfall on pore-water pressure distribution characteristics in residual soil slopes under tropical rainfall, Hydrological Processes, Special Issue on Rainfall Induced Landslides and Debris Flow, 22, 506 523. 51) Rahardjo,H.,Santoso,V.A.,Leong,E.C.,Ng,Y.S.andTam, C. P. H. (2009): Pore-water pressure characteristics of two instrumented residual soil slopes, Proc. of 4th Asia-Paciˆc Conf. on Unsaturated Soils, Newcastle, Australia, 333 339. 52) Rahardjo,H.,Ong,T.H.,Rezaur,R.B.,Leong,E.C.andFredlund, D. G. (2010): Response parameters for characterization of in- ˆltration, Environmental Earth Sciences, 60(7), 1369 1380. 53) Rezaur,R.B.,Rahardjo,H.,Leong,E.C.andLee,T.T.(2003): Hydrologic behaviour of residual soil slopes in singapore, ASCE J. of Hydrologic Eng, 8(3), 133 144. 54) Ridley, A. M. and Burland, J. B. (1993): A new instrument for the measurement of soil moisture suction, G áeotechnique, 43(2), 321 324. 55) Ridley, A. M. and Burland, J. B. (1995): Measurement of suction in materials which swell, Appl. Mech. Rev., 48(9), 727 732. 56) Ridley, A. M. and Wray, W. K. (1996): State of the art report-suction measurement: a review of current theory and practice, Proc. of 1st Int. Conf. on Unsaturated Soils, (eds. by Alonso, E. E., Delage, P.), Paris, 3, 1293 1322. 57) Ridley, A. M., Patel, A. R. and Marsland, F. (1998): Tensiometers: their design and use for civil engineering purposes, Geotechnical Site Characterisation. Balkema Rotterdam, 851 856. 58) Schnellmann, R., Busslinger, M., Schneider, H. and Rahardjo, H. (2010): EŠect of rising water table in an unsaturated slope, J. of Eng. Geology, 114, 71 83. 59) Sweeney, D. J. (1982): Some in situ soil suction measurements in Hong Kong residual soil slopes, Proc. of the 7th South East Asian Geotechnical Conf., Hong Kong, November 22 26, 91 106. 60) Tan, S. B., Tan, S. L., Lim, T. L. and Yang, K. S. (1987): Landslides problems and their control in Singapore, Proc. 9th Southeast Asian Geotechnical Conf., Bangkok, 1:25 1:36. 61) Tarantino, A., Ridley, A. M. and Toll, D. G. (2008): Field measurement of suction, water content and water permeability, Geotech. Geol. Eng., 26, 751 782. 62) Toll, D. G., Rahardjo, H. and Leong, E. C. (1999): Landslides in Singapore, Proc. of 2nd Int. Conf. on Landslides, Slope Stability and the Safety of Infra-Structures, Singapore, July 27 28, 269 276. 63) Tsaparas, I., Rahardjo, H., Toll, D. G. and Leong, E. C. (2002): Controlling parameters for rainfall-induced landslides, Computer and Geotechnics, 29(1), 1 27. 64) Tsaparas, I., Rahardjo, H., Toll, D. G. and Leong, E. C. (2003): Inˆltration characteristics of two instrumented residual soil slopes, Can. Geotech. J., 40(5), 1012 1032.