Micro‐scale modelling of terrain susceptibility to landsliding from a DEM: A GIS approach

Abstract

Aerial photographs of Nelson County, Virginia were manually interpreted to delineate the outlines of eighty‐seven landslide scars within a 28.5 km2 area. A Digital Elevation Model (DEM) was constructed from a topographic map of the same area at 24 m resolution. Four topographic variables: elevation, slope gradient, slope aspect, and configuration, were extracted from the DEM for incorporation in a raster GIS model. A linear polynomial equation was developed which incorporated the four variables and their weights to model the landslide potential of each grid cell in the study area. Methods of deriving these weights were developed. Analysis of the modelled results revealed that in the study area the upper sections of slopes facing west, northwest, and north were the most susceptible to landsliding, while the bottoms of lowland valleys and the tops of tall mountains with convex slopes and gentle gradients were the least susceptible. A preliminary evaluation of the model indicated an accuracy of about 83 per cent. It was concluded that terrain susceptibility to landsliding could be realistically modelled from a DEM in a raster GIS environment in a small area where geology, soil, and rainfall remained constant spatially.