10:00 AM - 10:15 AM
[HDS09-05] Estimation of major seismic factors controlling landslides using decision tree analysis
Keywords:Earthquake, Landslide, Decision tree analysis
Large earthquakes often cause many landslides. Therefore, it is important to grasp the current situation as soon as possible for rapid initial response to disasters. The factors of landslide can be divided into geofactors such as geomorphology or geology (e.g. Chigira, 2006), and seismic factors such as acceleration (e.g. Uchida et al., 2004) or intensity (e.g. Itoh et al., 2009). However, the major seismic factors have not yet been sufficiently examined. In this presentation, we focused on the number of landslides and estimated the major seismic factors that have a high contribution to it.
We picked out 26 earthquakes with a seismic intensity of 6- or higher that occurred after 2000. We used decision tree analysis to estimate major seismic factors. As explanatory variables, we used maximum intensity, maximum acceleration, maximum horizontal displacement, JMA magnitude, epicenter depth, location of epicenter (inland fault or subduction zone) and earthquake type (reverse fault, normal fault or strike-slip fault). As the explained variables, we used three classifications of the number of landslides by MLIT and CAO. We set the threshold to10 and 100.
As a result, it was showed that the maximum horizontal displacement, epicenter depth and location of epicenter were significant in that order. This means that getting these factors as soon as possible after earthquakes can lead to rapid initial response to disasters.
We picked out 26 earthquakes with a seismic intensity of 6- or higher that occurred after 2000. We used decision tree analysis to estimate major seismic factors. As explanatory variables, we used maximum intensity, maximum acceleration, maximum horizontal displacement, JMA magnitude, epicenter depth, location of epicenter (inland fault or subduction zone) and earthquake type (reverse fault, normal fault or strike-slip fault). As the explained variables, we used three classifications of the number of landslides by MLIT and CAO. We set the threshold to10 and 100.
As a result, it was showed that the maximum horizontal displacement, epicenter depth and location of epicenter were significant in that order. This means that getting these factors as soon as possible after earthquakes can lead to rapid initial response to disasters.