Better Satellite Precipitation Algorithms Slightly Improved Landslide Hazard Assessment

Abstract Recent upgrades to satellite precipitation algorithms have the potential to improve landslide hazard assessment. Algorithm changes were designed to improve precipitation detection, reduce systematic and random bias, enhance orographic precipitation estimation, and better represent precipitation across coastal and frozen surfaces. This study compares versions 6 and 7 of the Integrated Multi-satellitE Retrievals for Global Precipitation Measurement algorithm, focusing on their performance in landslide-prone terrain. We compared satellite-based precipitation estimates to ground measurements from 670 rain gauges in areas highly susceptible to landslides and then tested their effectiveness within a global landslide prediction model. Version 7 showed improved correlation with ground measurements (at 82% of gauge locations), reduced bias, and better root-mean-square error (at 73% of locations). However, both versions underestimate heavy rainfall events (>100 mm day −1 ), with false negatives more prevalent in cold and arid climates. When implemented in the landslide model, version 7 produced statistically significant improvements in landslide prediction, showing higher probabilities at known landslide locations while maintaining false positive rates. Despite these positive changes, the magnitude of improvement was modest. Satellite precipitation algorithms continue to advance, but significant challenges remain in accurately capturing intense rainfall events that often trigger landslides, particularly in mountainous terrain.