Parsimonious landslide susceptibility modeling at continental scales: a new high-resolution national landslides map

Landslide susceptibility maps are a fundamental tool for risk reduction. We used the USGS's landslide inventory compilation (N=613,724), high-resolution 3DEP digital elevation model (~10-m), and high-performance computing resources (YETI), to develop a new nation-wide susceptibility map for the Continental U.S., Hawaii, Alaska, and Puerto Rico. We used a parsimonious slope-relief threshold approach to distinguish between areas with some landslide potential from those with none, then down-sample our results to a 90-meter grid to account for uncertainty in the DEM, landslide position, and other factors. In addition to our rigorous 70/30 split-sample calibration and evaluation with multiple iterations, an independent validation with four statewide inventories (N=172,367) further reinforces the robustness of our threshold model, but also highlights spatially variable performance. Therefore, we propose a novel approach to susceptibility classification using the concentration of landslide-prone terrain within each down-sampled grid. While landslides are possible within any cells containing susceptible terrain, those with the highest concentration capture the majority of landslides. Our new map is more accurate than prior models; it also provides a transparent and flexible approach to susceptibility classification for implementing different tolerances for landslide risk reduction.