Contrasting Rainfall Thresholds for Landslide Initiation: A Multi-Temporal Analysis of the 2014 and 2018 Disasters in Hiroshima, Japan

Rainfall is the mainly trigger of landslides in Japan, especially in mountainous regions with steep slopes and highly weathered terrain. This study examines rainfall conditions linked to landslides in Hiroshima Prefecture by analyzing multi-temporal precipitation data and establishing thresholds for the 2014 and 2018 disasters. Rainfall events were identified from satellite data using an inter-event dry-period. Intensity–duration (I–D) relationships, percentile thresholds, and Receiver Operating Characteristic analysis were used to assess rainfall thresholds. Additionally, topographic threshold analysis was preformed to evaluate how slope gradient and lithology influence rainfall-induced landslides. The results reveal notable differences in the triggering mechanisms between the two disasters. The 2014 landslides were caused by short-duration, intense rainfall (376 mm over about 40 hours), while the 2018 event resulted from prolonged rainfall exceeding 1000 mm over more than 11 days. I–D thresholds suggest that the 2014 event was primarily triggered by high intensity, whereas the 2018 event was driven by long duration. Landslides occurred most frequently on moderate to steep slopes, and lower triggering thresholds were observed in granitic terrain. These findings highlight that both rainfall features and terrain conditions jointly influence landslide occurrence, emphasizing the need to incorporate both factors into rainfall-based early-warning systems.