By David Toll, School of Engineering, Durham University, currently academic visitor at the National University of Singapore.
Landslides are often triggered by rainfall, particularly in Southeast Asia, where rain storms can be very intense. Major events, such as the recent tragic mudslide on 17 February in Leyte in the eastern Philippines, occur all too frequently (GE March 06).
However, minor landslides are even more common. Although they may not lead to loss of life, they still have economic and social impacts.
Figure 1 shows a slide that happened on the campus of the National University of Singapore (NUS) on 11 January this year. With typical Singaporean efficiency, remediation was carried out very quickly, but it still caused safety concerns and incurred repair costs.
Such minor, shallow landslides have occurred frequently on the island of Singapore, particularly as urban development has greatly increased since the 1970s (Tan et al, 1987). However, there have been very few major landslides; where slides have happened, the volumes of material involved have generally not been large and serious damage has been rare (Brand, 1984; Toll et al, 1999).
Rainfall has been the dominant triggering event in Singapore (Ramaswamy & Aziz, 1980; Toll, 2001).
Studies of minor landslides on the Nanyang Technological University (NTU) and NUS campuses show spates of landslides occurring after unusually wet periods.
It may not be a single rainfall event that is the cause. In low permeability clayey soils (typical of the residual soils of Singapore) porewater pressures may build up over a number of days (due to a series of rainstorms), culminating in the final rainfall event that precipitates a failure.
Figure 2 shows rainfall data for a large number of Singaporean landslides (Toll, 2001). It shows the rainfall on the day of the landslide (triggering rainfall) plotted against that in the fi ve days preceding it (antecedent rainfall).
Some minor landslides have occurred after heavy one-day rainfalls with little antecedent rainfall (eg slides at NUS and NTU in February to March 1984). In the case of February 1984, the daily rainfall inducing failures was almost 100mm, whereas those in March 1984 were higher. However, it can also be seen that other minor slides took place with low one-day rainfall but where the five-day antecedent rainfall was significant.
An example of this is the event on 28 December 1984, where a slide occurred with only 18mm of daily rainfall, but after a five-day antecedent rainfall of 85mm.
This suggests conditions for failure are dictated by total rainfall, since either daily or antecedent rainfall can induce failures.
The diagonal line drawn in Figure 2, representing a total rainfall of 100mm in a six-day period, appears to defi e the minimum rainfall conditions that have led to minor failures. The data suggest that a total rainfall of 100mm within a six-day period (equivalent to a sustained 1520mm/day for six days) is sufficient for minor landslides to take place.
Although these empirical observations on rainfall patterns can be useful in identifying the minimum conditions likely to precipitate a landslide, they do not explain why landslides occur.
To properly comprehend such failures, an understanding of unsaturated soil behaviour needs to be applied. In many tropical regions, water tables are at signifi cant depth (over 10m). This means porewater pressures can be negative (suctions).
Therefore, there is a need to understand the role of suction in supporting the slope (increasing the strength of the soil) and how infi ation of rainwater causes changes in porewater pressures (or suctions).
Four research sites in Singapore were instrumented as part of a major study of rainfall-induced landslides in Singapore (Rahardjo et al, 2000).
Measurements were made of rainfall, run-off and pore-water pressures/suctions. Only one site is presented here; a site on the NTU campus next to the School of Civil and Structural Engineering (CSE).
Full details of the instrumentation regime are given by Rahardjo et al (2000).
Rainfall gauges were installed on each slope to provide rainfall data specific to that site. Negative porewater pressures in the unsaturated zone above the water table were measured using jet-fill tensiometers. These were installed at depths of 0.5, 1.1, 1.4, 2.3 and 3.2m on the NTU-CSE site. Piezometers were also installed at greater depths to monitor changes in groundwater levels. At this site, piezometer data indicated that the groundwater table was 10m below the ground surface (Rahardjo et al, 2000).
The porewater pressures within the NTU-CSE slope were monitored from August 1999 until August 2000 (Toll et al, 2001; Tsaparas et al, 2003). Figure 3 shows the porewater pressures at the various measuring depths for a row of tensiometers installed near the mid-point of the slope (6m downslope from the crest).
The daily rainfall is also shown as a bar graph.
It can be seen from Figure 3 that the porewater pressures within the NTU-CSE slope were, for a large part of the monitoring period, only slightly negative, and at 3.2m depth were generally positive.
However, there were six periods during the year when porewater pressures reduced significantly following drier weather. During March 2000, porewater pressures dropped to as low as -70kPa near the surface (0.5m depth). But piezometer data shows there was little change in groundwater table level. Therefore these suction changes are the result of infiltration and evapotranspiration occurring at the surface rather than changes in the water table.
Figure 4 shows porewater pressure profiles within the slope, during and after two rainfall events in December 1999 and March 2000 that are described in detail by Toll et al (2001). These dates represent a 'wet period' (with high initial porewater pressures) and a 'dry period' (with low initial porewater pressures).
The rainfall event in December 1999 was very large (86mm) whereas that in March 2000 was small (1mm). Yet, it can be seen from Figure 4 that the small rainfall in March during the dry period produced a signifi cant change in the porewater pressure near the surface.
After a period of equalisation (24 hours after the rain) the porewater pressure near the surface dropped back and porewater pressures at 1m to 1.5m depth increased. This was due to the infiltrated water draining down from the surface to lower depths. However, it can be seen that at 2.5m to 3m depth there was no significant change in porewater pressure.
In contrast, the very large rainstorm in December 1999 produced only a small change in porewater pressure near the surface, but the porewater pressure did approach a hydrostatic condition (defi ned by a porewater pressure of zero at the ground surface). Again, after the storm, porewater pressures dropped back near the surface and increased slightly at depth.
In both cases, the field measurements suggest porewater pressures do approach the hydrostatic condition near the surface due to infi ltration. However at 2.5m to 3m depth there is little change. Porewater pressures remain significantly below the hydrostatic line, even at the wettest time of the year. Therefore, assuming porewater pressures were hydrostatic throughout the slope (as would often be assumed in a saturated soil analysis) would be over-conservative.
The porewater pressure responses can be better understood by numerical modelling (Tsaparas and Toll, 2002; 2003). A major factor in controlling the response is the change in water permeability that occurs in an unsaturated soil as a result of changes in degree of saturation (the change in permeability can be 4-5 orders of magnitude).
When water infiltrates at the surface, a near-surface zone with a high degree of saturation is produced. This produces a zone of much higher permeability.
Further down (2m to 3m below the ground surface) the unsaturated permeability remains low, so water is not encouraged to fl ow to greater depths; instead flow tends to take place down the slope within the near-saturated surface zone. This means the porewater pressures at depth are not changed significantly since water does not fl ow down to this region of the slope.
This data shows that, for a scenario where the water table is at significant depth (deeper than 10m), most porewater pressure changes take place near the surface (less than 2m depth). This is consistent with the observation that many minor landslides in Singapore are quite shallow in nature (Toll et al, 1999). Failures tend to occur within the near-surface zone where porewater pressures increase close to hydrostatic levels.
It is clear that rainfall has been the dominant triggering event for landslides in Singapore. Studies show spates of landslides occurring after unusually wet periods. Observations of past landslides suggest that a total rainfall of 100mm within a six-day period (equivalent to a sustained 1520 mm/day for six days) is sufficient for minor landslides to take place in Singapore.
Measurements of porewater pressures in slopes in Singapore show that rainfall infi ltration produces changes in porewater pressure near the surface. However, at greater depths (around 3m) porewater pressures do not change significantly.
This is because water tends to flow down the slope within the zone of higher saturation that develops near the surface. As a result, failures tend to occur within the near surface zone and are not usually deep-seated.
Brand EW (1984). Landslides in Southeast Asia: A state-of-the-art report, Proc 4th International Symposium on Landslides, Toronto, Vol 1, pp17-59.
Rahardjo H, Leong EC, Deutcher MS, Gasmo JM and Tang SK (2000). Rainfall-induced slope failures, Geotechnical engineering monograph No 3, NTU-PWD Geotechnical Research Centre, Singapore.
Ramaswamy SD & Aziz MA (1980). Rain induced landslides of Singapore, Proc.
International symposium on landslides, New Delhi, Vol 1, pp403-306.
Tan SB, Tan SL, Lim TL and Yang KS (1987). Landslide problems and their control in Singapore, Proc 9th Southeast Asian Geotechnical Conference, Bangkok, pp1:25-1:36.
Toll DG (2001). Rainfall-induced landslides in Singapore, Proc Institution of Civil Engineers: Geotechnical Engineering, Vol 149, No 4, pp211216.
Toll DG, Rahardjo H and Leong EC (1999).
Landslides in Singapore, Proc 2nd International Conference on landslides, slope stability and the safety of infrastructures, Singapore, pp269-276.
Toll DG, Tsaparas I and Rahardjo H (2001).
The influence of rainfall sequences on negative porewater pressures within slopes, Proc 15th International conference on soil mechanics and geotechnical engineering, Istanbul, Rotterdam: Balkema, Vol 2, pp1269-1272.
Tsaparas I, Rahardjo H, Toll D and Leong EC (2003). Infi ration characteristics of two instrumented residual soil slopes, Canadian Geotechnical Journal, Vol 40, No 5, pp10121032.
Tsaparas I and Toll DG (2002). Numerical analysis of infi tration into unsaturated residual soil slopes, in unsaturated soils (eds Jucá JFT, de Campos TMP and Marinho FAM). Proc 3rd International conference on unsaturated soils, Recife, Brazil, Lisse: Swets & Zeitlinger, Vol 2, pp755-762.
Tsaparas I and Toll DG (2003). Factors affecting infi tration into an unsaturated soil slope.
Proc 2nd Asian conference on unsaturated soils (UNSAT-ASIA 2003), Osaka, Japan, pp463-468.