Susceptibility modelling of rainfall induced landslides

Abstract

Occurrence of rainfall induced landslides coincides in time with the period of high persistence of heavy wet spells and in spatial locations where vegetative cover has been denuded. As the rain water percolates into the slope it increases the soil water content and reduces in situ suction, followed by built-up of pore water pressures resulting in a decrease in the effective stress in the slope soils, and may induce to slope failure. The state of soil moisture content is a function of soil type, land use, soil thickness and rainfall intensity. The highest infiltration capacities are observed in loose, sandy soils while heavy clay or loamy soils have considerable smaller infiltration capacities. Vegetation cover is known to stabilize slopes majorly by roots reinforcement of soils. In this study a model was formulated to simulate previous landslide occurrence and then used in scenario analysis to establish the effects of wet spell persistence and land use changes on the stability of slope in the western stretch of Kerio escarpment. The April and May rainfall data for region’s rainfall stations were applied in the analysis whereby HYDRUS 1D model was used in soil moisture analysis, formulated Markov chain/event based models was used in stochastic rainfall generation, formulated Ms-EXCEL infinite slope model was used in sensitivity and factor of safety analysis, ArcGIS tool was use in maps preparation and ILWIS GIS tool was used in mapping spatial distribution of slope stability. Results indicated that rainsum of wet spell greatly influences soil moisture condition and weekly antecedent rainfall best describes expectation of unstable areas when two rainfall patterns are being compared. Statistical assessment of unstable areas under different land use revealed the importance of vegetation cover in slope stabilization since the scenario that the study area was completely forested, the percentage of unstable and critical areas were 0.05% and 1.81% respectively while for the scenario that it was completely denuded (agriculture), the percentage of unstable and critical areas were 0.69% and 5.35% respectively. It was concluded that landslide prediction model was successfully developed and should be adopted in generation of landslide hazard maps when heavy rainfall is anticipated. The generated hazard maps was recommended as a basis tool for decision making to residents (in identifying habitable areas), environmentalists (in showing importance of tress) rescuers (e.g. Red Cross, in marshal resources) and engineers (in establishing major projects) so as to reducing, preventing and mitigating losses caused by landslides.