Land cover mapping from RADARSAT stereo images in a mountainous area of southern Argentina

This study evaluates the capability of land cover mapping of a high-relief mountainous area of southern Argentina using a stereo pair of standard mode RADARSAT data. Land cover types in the study area include forest, glacial ice, grass, rock, sandy soil, shrub, snow, and water. A digital elevation model (DEM) is first derived from the stereo pair of RADARSAT images. The images are orthorectified using the derived DEM. Then, several textural features are computed from the original images and various textural channel combinations are evaluated for land cover mapping. Supervised maximum likelihood classification is performed on the radar tonal channels alone and the radar textural channels. The results show that the use of textures has improved the classification accuracy compared with that obtained using radar tonal data alone. However, there is confusion among water, snow, glacial ice, and grass. To reduce the confusion, a stratified land cover mapping is performed. The stratification is established accord...