Statistical Analysis of Deep-Seated Landslides Induced by Recent Strong Earthquakes in Eastern Japan: An Approach Based on Pre-Existing Landslide Topography

摘要

This paper conducted logistic regression analysis for earthquake-induced landslides within pre-existing landslide topographies. We constructed logistic regression model by using landslides data set of the Chuetsu earthquake, Niigata Prefecture, 2004. The results showed that surface roughness and marginal erosion ratio are the most influential factors for the landslide occurrence. Themodel was validated by the part of landslides data set of the Chuetsu earthquake and the Iwate-Miyagi inland earthquake, 2008. Based on ROC curves, the AUC is larger than 0.8, showed a good prediction accuracy and suggested the method is useful.