ERVISED LANDFORM CLASSIFICATION METHOD USING NEURAL NETWORK AND ITS APPLICATION TO ESTIMATION OF SEISMIC GROUND MOTION

Abstract

This paper presents a supervised classification method using a neural network to classify typical landforms based on a land cover map and a Digital Elevation Model (DEM). The proposed method classified the landform of Kobe city in Japan into hill, plateau, fan and reclaimed land. As a result, a Self-Organizing Map(SOM) produces the higher classification accuracy than Back Propagation method. Furthermore, we adopted these classified landforms for a ground motion estimation in Kobe during the 1995 Hyogoken Nanbu earthquake, and could obtain detailed ground motion distribution compared with the one based on the Digital National Land Information (DNLI).