Influence of river incision on landslides triggered in Nepal by the Gorkha earthquake: Results from a pixel-based susceptibility model using inlabru

This study presents a comprehensive Bayesian framework for modelling earthquake-induced landslide (EQILs) susceptibility applied to the Gorkha earthquake, using the computationally efficient inlabru package. It combines diverse spatial covariates, with USGS models of ground shaking, to describe the centroid location and size distribution of the landslides. A key innovation is the projection of the normalised channel steepness index (ksn) to the upslope hillsides to test the hypothesis that higher rates of localised river incision will tend to destabilise coupled hillslopes. Our hillslope analysis constructs nodal networks from the river channel to the ridge crest. This allows us to not only propagate metrics of channel incision to hillslope, but also derive metrics related to hillslope steepness that geometrically respect irregular geometries, and that are also less noisy than local slopes derived directly from a Digital Elevation Model, making them more relevant to landslide susceptibility forecasting. To address spatial data misalignment between covariates and landslide observations, we leverage the inlabru framework, which enables coherent integration through mesh-based disaggregation, thereby overcoming challenges associated with spatially misaligned data integration. Our modelling strategy prioritises prospective transferability to unseen geographical regions and future earthquakes. It is extendable beyond EQILs provided explanatory covariates are available. By jointly modelling landslide locations and sizes and comparing with traditional covariates across four spatial metrics, we find that elevated ksn is strongly associated with greater landslide susceptibility and larger sizes. The best-fitting Bayesian model, validated through cross-validation, offers a scalable and interpretable solution for predicting EQILs in complex terrain.