Evaluating the performances of satellite-based rainfall data for global rainfall-induced landslide warnings

Satellite-based precipitation estimates (SPEs) show great promise for promoting landslide warning and mitigating landslide disaster risk with quasi-global coverage, near real-time monitoring, increasing spatial-temporal resolution, and accuracy. In this study, we evaluated the performances of four SPE products in detecting the initiation of rainfall-induced landslides globally using Hanssen-Kuiper (HK) skill score based on rainfall frequentist thresholds. The results show that SPEs can distinguish rainfall events responsible for landslides from those not related to landslides, suggesting that SPEs can capture rainfall conditions corresponding to landslide occurrence well and are of great use for landslide detecting. Further investigation indicates that performances at the global scale vary with products. CMORPH-3h V1 (HK = 0.43) and TMPA-3B42RT V7 (HK = 0.42) are superior to two other rainfall products with high HK values. Rainfall threshold establishment and evaluation for specific landslide types can improve SPEs' performances in landslide modeling with higher HK values compared to results based on all landslide records. Performances also vary spatially with HK values ranging from 0.1 to 0.9 at a spatial grid of 5 degrees x 5 degrees. Linear relationship analysis reveals the variation in mean annual precipitation can partially explain the heterogeneous spatial distribution of rainfall threshold parameters. These findings serve to promote the application of satellite-based rainfall data in landslide warnings.