Substratum virtualization in three-dimensional landslide modeling with the material point method

The material point method is a first-line choice in modeling large-deformation problems and has been widely utilized to investigate landslide dynamics. However, topographic complexity and computational demand have often hindered three-dimensional implementations. Here, we propose an algorithm for the virtualization of the substratum, in which the latter can be defined on a regular background grid as a non-zero kinematic boundary condition permitting frictional contact. In this setup, the computational power can be fully employed to model the moving mass. To evaluate the algorithm, we successfully simulated a rolling ball and flume experiments. Then, we took the Yanyuan landslide (27°29′25”N, 101°01′38″E) as an example to test the performance in a real case. By speeding up computation by ∼90 times, we argue that the proposed algorithm can promote accurate three-dimensional modeling of landslides and facilitate physically-based hazard assessments. •We proposed an algorithm of substratum virtualization in MPM simulations replacing the Bardenhagen contact algorithm.•We obtained good performances in simulating lab-scale problems as well as a real mudflow runout.•The proposed algorithm accelerates 3D MPM modeling in physically-based hazard assessments.