Probabilistic hazard analysis of rainfall-induced landslides at a specific slope considering rainfall uncertainty and soil spatial variability

A key index for assessing landslide risk quantitatively is annual slope failure probability, P_FA , induced by rainfalls at a specific slope. Although P_FA is expected to be significantly influenced by both spatial variability of soil properties and rainfall uncertainty, previous studies seldom estimate P_FA  with simultaneous consideration of both soil spatial variability and rainfall uncertainty. Most previous studies focused on the effect of soil spatial variability and ignored rainfall uncertainty, and the estimated slope failure probability is not associated with a reference time period (e.g., a year for P_FA). This study aims to assess P_FA  considering both soil spatial variability and rainfall uncertainty by developing a fully Monte Carlo simulation (MCS)-based method. To bypass the high-dimensional integration in estimation of P_FA  and high computational costs of MCS, a semi-analytical method is further developed for enhancing computational efficiency in estimating a small failure probability. Results show that the semi-analytical and fully MCS-based methods produce consistent P_FA . Using the proposed semi-analytical method, a small number (e.g., 30) of random samples is sufficient for estimating a small failure probability (e.g., 10⁻⁴) accurately. When the variability of soil properties is negligible, P_FA  is dominated by rainfall uncertainty and converges to a constant failure probability. © 2023 Elsevier Ltd.