Land-Cover Classification

Land-cover classification is an important application area of satellite remote sensing. However, deriving thematic map from satellite imagery through classification approaches is not a straightforward task, especially from high-resolution satellite imagery. In this study, Terra/Moderate Resolution Imaging Spectroradiometer (MODIS) multispectral composite image is successfully used to land-cover classification for the Lhasa area located at central Tibetan Plateau (TP) using maximum likelihood classifier. Accuracy assessment for final results is also made using quantitative approaches. Study shows that there is a good agreement between classification results and reference data for defined land-cover classes in central TP. The overall classification accuracy is 87.68%. Reference and ancillary data are increasingly available and are very useful for refining accuracy of classification results during postclassification process. The integration of digital elevation model (DEM) into land-cover classification is particularly important in mountain region since land-cover distribution in mountain region is spatially topography-dependent. Study also suggests that with increase of spatial resolution, how to effectively use the spatial information inherent in satellite remote sensing images to extract thematic maps for various applications remains a challenge and is an important task to be fulfilled in the future.