Risk evaluation of rainfall-triggered landslides on multiple scales of Japan

Landslide risk is the product of landslide hazards, exposure, and vulnerability. Spatial and temporal variations in risk and its three components of rainfall-triggered landslides were examined on multiple scales in Japan. Landslide fatalities in Japan decreased between the 1940s and the 1990s. The factors affecting the decrease changed the decrease in household members, increase in people evacuated, and change in the structure of houses to the increase in forest maturity and implementation of structural measures. Similar trends were also found in Kure City with three destructive landslide events in 1945, 1967, and 2018. However, the timing of the main contributions was different from that in Japan overall. In Japan, landslide frequency (i.e., landslide hazards) also decreased with time. Based on a model estimating landslide frequency from the forest age components and rainfall, a larger contribution of the increase in forest maturity to landslide frequency than rainfall was demonstrated on the national scale. Factors determining the number of landslide disasters were examined using generalized linear models on prefectural scales. The factor differed among the three landslide types (i.e., steep-slope failure, deep-seated landslide, and debris flow). For all types, rainfall and the number of landslide-prone areas were selected with positive coefficients: the accretionary complexes geological type with negative coefficients. In addition, forests and land for buildings were selected for steep-slope failures with negative and positive coefficients, respectively, which were not selected for deep-seated landslides and debris flows. The historical and future populations in landslide-affected areas (i.e., landslide exposure) were examined in all municipalities of Japan. The population in the landslide-affected areas continuously decreased during the analysis period. The decrease was gentler than those in landslide risk, hazards, and vulnerability, suggesting that the effects of landslide exposure on temporal changes in landslide risk were less than those of landslide hazards and vulnerability, on the national scale. Finally, the mortality rate in collapsed-houses by landslides was examined from 2014 to 2027. The database for victims and survivors in collapsed houses was developed mainly based on newspapers. The floor number, gender, and type of trigger affected the mortality of landslides. These evaluations can be used to develop strategies for the mitigation of landslide disasters.