Landslide susceptibility mapping using artificial neural network and maximum likelihood method

Abstract

The purpose of this study is to investigate which method is more available for landslide susceptibility mapping, artificial neural network or maximum likelihood method. The study area is 1km² on Tama Hills near Tokyo. In advance, aerial photographs interpretation gave past landslide, which was used as the supervised information. Then, to map landslide and non-landslide areas, seven explanatory variables concerning landform and vegetation were used in the two methods. In view of the mapping result and correct ratio, which is the ratio of mapping result to actual landslide and non-landslide areas distribution, it was found that the artificial neural network gives better result than the maximum likelihood method.