The Effect of DEM on the Land Use/Cover Classification Accuracy of Landsat OLI Images

Abstract

Land use provides crucial data for earth science research and its accuracy has always been a hot topic. Various auxiliary data are used to improve the classification accuracy of land use. The digital elevation model (DEM) is one of the common auxiliary data since topography directly affects the surface landscape. However, few researches focus on the impact of the DEM on the classification accuracy of images. In this study, the terrain steepness index (TSI) was initially put forward to study the effect of the DEM on the classification accuracy of the land use/cover. Seven areas with different terrain were taken as the studied areas and the classification and regression tree was utilized to derive thematic land use/cover maps. It indicated that the DEM had an impact on classification accuracy in a certain TSI extent. The developed quadratic model accurately described the correlation between the DEM and the TSI. Based on the model, the DEM with 30 m resolution had a positive impact on the classification accuracy, when the TSI varies from 9.106 to 34.014. It was concluded that the TSI index for the DEM could be effectively used in the land use/cover classification.