Accurate representation of the physical features of the land covering the watershed is required to understand hydrologic processes, since the extent and type of watershed cover affects the movement of water in the hydrologic cycle. Satellite imagery from Landsat and other satellites provide land cover information with high temporal frequency and spatial accuracy. The utility of these data for understanding hydrologic processes depends on how accurately they are interpreted and mapped. This paper deals with the use of the surface temperatures derived from the thermal band of Landsat images in combination with Normalized Difference Vegetation Index (NDVI) to improve the land cover mapping beyond conventional unsupervised classification of the visible/short wave bands. The study was done on Econlockhatchee River basin, part of the Upper St. Johns River basin in central Florida. Landsat images from 1984, 1994 and 2000 were obtained, and unsupervised classification was conducted. Surface temperatures and NDVI were computed and scatter diagrams were drawn to identify the spectral classes resulting from the unsupervised classification. The results show that spectral signatures in the thermal and near-infrared spectrum are very helpful to improve land cover mapping and discriminate land cover classes. In particular, the surface temperatures and NDVI were useful for delineating boundaries between wetlands and water bodies and between upland and wetland forests. Copyright ASCE 2004.
