Effect of Strength Spatial Variability on the Probabilistic Post‐Landslide Behavior of Heterogeneous Slopes Under Rainfall

ABSTRACT The strength spatial variability and the complex hydro‐mechanical interactions render the quantitative evaluation of rainfall‐induced landslides a persistent challenge. The probabilistic analysis methods face critical limitations: the traditional small‐deformation numerical methods cannot capture post‐landslide consequences, while direct large‐deformation analysis is computationally prohibitive. Therefore, this study develops an integrated random finite element method—material point method (RFEM‐MPM) approach to analyze the entire failure process, with particular emphasis on the post‐landslide stage. The approach builds upon a sequential FEM‐MPM coupling scheme, extended through a secondary development that enables the import and storage of spatially variable strength properties directly at the material point level. This allows for the integration of random field (RF) theory into the MPM for probabilistic analysis. The cohesion and internal friction angle are modeled as a bivariate cross‐correlated RF. Monte Carlo simulations (MCS) are performed considering the strain‐softening characteristics of the soil. The failure criterion is identified through the factor of safety (FoS), combining FEM's computational efficiency for pre‐landslide seepage‐stability analysis with MPM's large‐deformation capabilities through equivalent state mapping. The post‐landslide behavior is quantitatively assessed using four large‐deformation parameters: sliding volume, runout distance, retrogressive distance, and sliding depth. The results indicate that the approach achieves a 95.1% computational reduction compared to the random material point method (RMPM). The strength spatial variability significantly affects the slope stability, large‐deformation characteristics and failure modes. Increased variability in strength parameters leads to wider, right‐skewed distributions of runout distance and sliding volume, increasing the probability of extreme failure events.