Flexible slope displacement models for probabilistic seismic landslide hazard assessment incorporating near-fault ground-motion velocity pulse and directionality

Abstract

Ground-motion characteristics and the dynamic response of slopes are important factors contributing to earthquake-induced landslides. In near-fault regions, seismic landslide displacement analyses are often simplified using Newmark rigid-block models and directionally-averaged displacement indices. Some prior studies developed near-fault slope displacement models for the strong-motion velocity-pulse direction. However, it remains unclear how ground-motion pulses and directionality affect the dynamic slope response in sliding displacement analyses. In this study, a probabilistic framework is presented for regional seismic landslide hazard assessment by incorporating near-fault ground-motion pulse and directionality effects. The framework generates slope displacement hazard curves through convolution of seismic hazard with flexible-slope displacement models, which are developed to statistically characterize the maximum (D₁₀₀) and median (D₅₀) slope displacements over horizontal orientations under near-fault pulse-like and non-pulse-like ground motions based on over 35 million nonlinear coupled sliding analyses. Pulse-like motions generally yield larger amplitude and stronger polarization of slope displacements than non-pulse-like motions. Strong-motion pulses and D₁₀₀ are likely to occur within ±45° of fault-normal direction, while slope displacements in the fault-normal direction or the strongest-pulse direction could be significantly (over 25 %) smaller than D₁₀₀. The new displacement models reduce predction errors by more than 30 % compared with existing models. A hypothetical example and a real case of landslides triggered by the 2008 Wenchuan earthquake are used to illustrate the proposed framework. Compared with the traditional rigid-block analysis, an appropriate selection of D₁₀₀ and D₅₀ from the proposed flexible-slope models in regional analysis achieves better consistency with the observed landslide distribution.