Coseismic Shallow Landslide Distributions Reveal Patterns of Regolith Thickness in High Mountains

Abstract

Earthquake-triggered landslides are a major source of sediment mobilization in steep mountain landscapes, yet the controls that organize regional patterns of coseismic shallow failure remain incompletely understood. Because shallow landslides occur within regolith, their initiation and magnitude depend strongly on near-surface material thickness, which is rarely measured directly in mountainous terrain. Here we introduce ShallowLandslider, a hybrid physical–statistical framework that combines critical-acceleration slope-stability calculations with stochastic realization of discrete landslide populations. We apply the model to two geomorphically distinct subregions affected by the 2015 $M_{w}7.8$ Gorkha earthquake in Nepal and compare modeled and observed distributions of landslide area, elevation, slope, and aspect. Parameterizations incorporating spatially variable regolith thickness consistently outperform a uniform-thickness baseline, and model performance is highest for moderate cohesion values ($c^{\prime }\approx 10$–15 kPa), characteristic of shallow colluvium. The two subregions favor contrasting regolith-organization hypotheses. In the high-relief, bedrock-dominated terrain of Subregion 1, landslide occurrence is reproduced most successfully by parameterizations that impose broad elevation-dependent thinning of regolith while landslide magnitude frequencies are best represented through a curvature-informed representation. In the lower-relief, more extensively regolith-mantled terrain of Subregion 2, curvature-informed parameterizations that represent local topographic convergence provide the closest agreement with the mapped inventory. These results indicate that landscape-scale organization of regolith thickness exerts a strong influence on coseismic shallow-landslide populations within the tested model framework. More broadly, they demonstrate how landslide inventories can be used to evaluate competing hypotheses of regolith distribution where direct observations remain sparse.