崩壊発生時刻予測手法による2013年7月山口・島根豪雨の再現と長期間雨量データによる検証

抄録

Previously, we proposed a model called idH-SLIDER for assessing the time and location of landslides. The method was shown to predict landslides triggered by heavy rainfall and to demonstrate the influence of each modeled parameter. As the validity of this approach has not been confirmed, this study applied the method to areas in northern Yamaguchi prefecture, and attempted to reproduce 38 years of landslide occurrences using long-term successive hourly rainfall data, including multiple rainfall events. Many shallow landslides occurred in the study region in July 2013 and July 1983. The distribution of soil thickness was estimated using simple penetration tests and LiDAR data obtained after the 2013 rainfall event. Parameters were determined from soil tests, except for the saturated hydraulic conductivity, which was scaled in increasing order of magnitude from the default value of 1 to 10,000 times. It is well known that soil conductivity varies greatly. Calculations show that short-term rainfall data from 2013 were not distinguishable between saturated hydraulic conductivities of 1, 10, and 100 times. At 1,000 and 10,000 times, a few elements were unstable, preventing the modeling of landslides. We demonstrated the prediction of landslide occurrences using data recorded over 37 years, and identified saturated hydraulic conductivity of 100 to 10,000 times as an appropriate default. This method was more accurate than previous models designed for long-term datasets of longer than 10 years. The model is applicable to various rainfall patterns and geological environments.