Three-dimensional simulations of large-scale long run-out landslides with a GPU-accelerated elasto-plastic SPH model

Smoothed particle hydrodynamics (SPH) is now becoming more and more welcome in modeling large deformation, soil failure and landslide in geotechnical and geological community. Aiming to solve the low efficiency problem in modeling large-scale long run-out landslide, this study implemented a GPU-accelerated SPH model based on the Drucker–Prager yield criterion with a non-associated flow rule on the open source platform DualSPHysics (v4.0). Two typical three-dimensional (3D) cases with simple boundary conditions were simulated to validate and demonstrate the high effectiveness and capacity of the GPU based elasto-plastic SPH model in modelling large deformation, failure, runout and deposit in a unified framework. Furthermore, we used the GPU-based elasto-plastic SPH model to simulate the Yigong avalanche and Shenzhen landslide to reproduce complete landslide dynamic processes and deposit detail, where a comparable deposit distribution with field survey was found. It is concluded that the implemented GPU-based elasto-plastic SPH model is able to simulate large scale long runout landslides on complex terrain to provide full process and deposit, which could be helpful for hazard assessment and mitigation.