A probabilistic approach to predicting landslide runout based on an inventory of snowmelt-induced landslide disasters in Japan

摘要

In this study, a probabilistic analysis to predict landslide runout based on an inventory of snowmelt-induced landslide disasters with consideration of differences in the topographic features of travel paths was conducted. The inventory included 76 landslides that occurred during snowfall/snowmelt periods between 1947 and 2012. These landslides were divided into three groups based on a topographic classification using GIS analysis. The probability of a landslide traveling for a specific distance and its traveling ratio were examined for each group by fitting the empirical cumulative distributions to exponential distribution functions. The fitted exponential distribution functions and outcomes of the estimations were very similar for landslides traveling along floodplain (FP) and slope (SL), despite differences in the slope gradients of the travel paths. The probability that landslide runout would exceed the criteria of the Sediement Disaster Prone Areas (SDPAs) was <1% for the FP and SL groups. In contrast, landslides traveling along headwater channel (HC) were likely to travel further than the criteria, particularly in landslide-initiated debris flows. Even if their travel distances was restricted to the movement of displaced mass by initial landslides, HC landslides tended to travel further than the other landslide groups.