Developing hydrometeorological landslide thresholds in California using WRF-Hydro hindcast simulations and radar-constrained precipitation data

Statistical thresholds based on rainfall intensity and duration are widely used to identify landslide triggering events. However, these estimates are limited by the scarcity of rain gage observations and often lack hydrological information such as soil moisture and/or overland flow that more directly link to landslide triggering. In this study, we incorporate a full suite of hydrological information - soil moisture, soil saturation degree, water table depth, and surface runoff - from Weather Research and Forecasting hydrological modeling system (WRF-Hydro) hindcast simulations forced with Multi Radar Multi Sensor (MRMS) precipitation to identify hydrometeorological thresholds for landslide triggering in California. WRF-Hydro is a 3-D, physics-based, and fully distributed hydrologic model. MRMS provides radar-constrained, gauge-corrected precipitation at 1-km continuous spatial coverage over the U.S. We use a landslide inventory compiled by the California Geologic Survey that documents 115 landslide events triggered by 80 rainfall events in California during 2016-2020. We begin by cataloging the local intensity and duration of rainfall events in MRMS that have triggered observed landslides. We then employ the full-suite of hydrological information from WRF-Hydro simulations to develop comprehensive hydrometeorological thresholds for landslide triggering. Preliminary findings show that multiple self-organizing clusters of the investigated hydrometeorological variables (i.e., rainfall intensity and duration, soil moisture, soil saturation degree, water table depth, and surface runoff) exist for different landslides. Future work will apply a machine learning algorithm to parse hydrometeorological characteristics and thresholds to more formally classify distinct landslide types and regions.