Improved landslide runout prediction by integrating the pore pressure response to dilatancy

Water plays a crucial role in the initiation and runout patterns of most landslides. Variations in the degree of saturation and porewater pressure influence the strength of landslide materials, thereby determining the final runout for a given topography. These properties can vary within a single landslide body, leading to different movement patterns and mobility across different sections, with associated implications for the landslide hazard and risk. This complexity poses challenges for numerical modelling aimed at accurately predicting landslide runout.

In this study, we used Material Point Method—a hybrid Lagrangian-Eulerian method—coupled with mixture theory (Tampubolon et al., 2017) to simulate elastoplastic deformation and runout behaviour of the landslide body. To better capture the evolution of movement patterns with minimal manual constraints and to enhance the accuracy of runout predictions, we integrated an excess pore pressure generation curve (Wang, 1999) into the computational workflow. This allowed us to simulate the excess pore pressure induced by the negative dilatancy of the solid phase under conditions of rapid motion or low permeability. The integration of this mechanism captures the effects of dilatancy, which arise from compaction and grain crushing in the sliding zone during the runout process. We show that by accounting for this localised material strength loss and the pore pressure dissipation, the evolution of landslide movement and landslide runout may be more accurately simulated.

The model was validated against a two-dimensional cross-sectional slope failure scenario with varying permeability conditions. Subsequently, it was applied to two typical multi-pattern landslide cases: a giant loess landslide on the Qinghai-Tibet Plateau and another one in London Clay on the northern shore of the Isle of Sheppey. The initial state of the slope was reconstructed based on pre-landslide digital elevation model data, while the groundwater variations, driven by either rainfall or tidal influences, were modelled as the triggering factors. This approach effectively captures the localised pore pressure effects, thereby improving the accuracy of runout distance and area predictions. We expect our model to be broadly applicable to improve runout simulation and associated hazard assessment for a broad range of hydrologically modulated landslides.