Probabilistic prediction of rainfall-induced slope failure using a mechanics-based model

Abstract

This paper presents an efficient probabilistic method for predicting rainfall-induced slope failures based on Monte Carlo simulation. The suggested method can calculate the time-dependent failure probability of the slope during a rainfall infiltration process. It can also predict the probability density function of the failure time. The past proof testing information can be incorporated in the failure probability estimation. To facilitate its practical application, spreadsheet templates are developed to implement the suggested method. Parametric studies indicate that when the soil parameters are uncertain, the rainfall intensity–duration curve for landslide initiation is also uncertain. The proposed method can be used to assess the reliability of intensity–duration curves. Analysis shows that descending rainfall is more damaging than uniform rainfall and ascending rainfall for the same cumulative precipitation. For the slope studied in the paper, saturated permeability, cohesion and friction angle of the soil contribute more uncertainty to the stability of the slope than initial water content, water content in the wetting zone, and suction at the wetting front. The past performance information can significantly affect the estimated failure probability.