Evaluating the exceedance probability of the runout distance of rainfall-induced landslides using a two-stage FEM-MPM approach

Evaluating the exceedance probability within a time period (EPT) of the runout distance of rainfall-induced landslides is important for the quantitative risk assessment (QRA) of rainfall-induced landslides. However, assessing the EPT of the runout distance of rainfall-induced landslides using a mechanics-based method remains a challenging problem since it requires considering uncertainties in both soil properties and rainfall. This paper proposes a novel mechanics-based method is to assess the EPT of the runout distance of rainfall-induced landslides with explicit consideration of the above two types of uncertainties. A two-stage numerical approach, which combines the finite element method (FEM) and the material point method (MPM), is first developed for the large deformation analysis to obtain runout distances of landslides under given rainfalls. To further enhance the computational efficiency, a machine learning-based surrogate model is built to predict the exceedance of the runout distance, and the EPT of the runout distance is finally estimated via Monte Carlo simulation. The proposed method is applied to a sandy slope under rainfall. The results show that the EPT increases as the time period becomes longer, and the runout distance of the landslide is controlled by the first failure of the slope caused by rainfall. This study contributes to the development of a general and efficient tool to support the QRA of rainfall-induced landslides.