IISER MOHALI

Abstract: Rainfall‐induced landslides are among the most pervasive natural hazards in mountainous regions, often resulting in significant loss of life and infrastructure damage. In Taiwan, where steep slopes and a monsoonal–typhoon climate converge, understanding how regional precipitation regimes control landslide occurrence is critical for effective hazard mitigation. Here, we analyzed 994 rainfall‐induced landslides documented between 2011 and 2018 alongside hourly, 2 km‐resolution precipitation records from the Taiwan ReAnalysis Downscaling (TReAD) data for 1979–2019. The island was partitioned into 40 km × 40 km grid cells, and the 17 cells containing at least 20 landslides were selected for detailed study. For each event, we extracted maximum cumulative rainfall over 1, 24, and 72 hours and fitted Generalized Extreme Value distributions to estimate the return period of triggering storms. We then explored three key relationships: how absolute triggering rainfall scales with mean wet‐ season (May–October) precipitation, how storm return periods vary spatially across different climate zones, and how landslide volume correlates with rainfall rarity. Our results reveal that wetter regions require both larger rainfall totals and rarer storms to initiate slope failures (R2 up to 0.75 for 72 h extremes) and that landslide volume increases nonlinearly with storm return period, following a sublinear power‐law scaling. These findings underscore the importance of incorporating regional climate statistics and probabilistic rainfall metrics into landslide hazard assessments and suggest that early‐warning protocols based on return‐period thresholds can more accurately capture local risk under evolving climate conditions.