ShallowLandslider: a physics-based component for predicting regional distributions of coseismic landslides

Suryodoy Ghoshal , Sarah J Boulton , T.C. Hales, Georgina Bennett , Amy Beswick, Josh N. Jones, Shaun Lewin, Zoe K Mildon , Martin Stokes , Michael R. Z. Whitworth , Benjamin Campforts 

Earthquake-triggered landslides are a dominant agent of sediment mobilisation in steep terrain, yet the physical controls that organise coseismic shallow-failure patterns across landscapes remain incompletely understood. In particular, the influence of regolith thickness and its spatial organisation on regional-scale failure patterns has been difficult to quantify. Here we introduce ShallowLandslider, a hybrid physical-statistical framework that couples a three-dimensional Newmark stability formulation with topographically informed regolith-thickness fields and population-level realisation of landslide patches. We evaluate the model in two geomorphically distinct subregions impacted by the 2015 Gorkha earthquake, assessing its ability to reproduce observed patterns of landslide occurrence and area across elevation, slope and aspect.

Across all configurations, the spatial structure of regolith thickness exerts the strongest control on the distribution of coseismic shallow landsliding. Parameterisations incorporating realistic regolith-thickness variability consistently outperform a uniform-depth baseline, and model performance peaks at moderate cohesion values (c'≅10--15 kPa) characteristic of shallow colluvium. The two subregions exhibit contrasting dominant controls: in the high-relief, bedrock-dominated terrain of Subregion 1, landslide patterns are best simulated when regolith organisation is represented by broad hypsometric thinning, whereas in the lower-relief, regolith-mantled terrain of Subregion 2, curvature-informed parameterisations capturing local hillslope convergence provide the closest agreement with the mapped landslide population. Our findings demonstrate that a process-based model respecting the physics controlling shallow landsliding performs reliably only when the spatial variability of regolith thickness is accurately represented. This indicates that landscape-scale patterns of regolith thickness, not uncertainty in failure mechanics, explain regional variation in landslide response.