Abstract
Land surface temperature (LST) is important factor in global change studies, in estimating radiation budgets in heat balance studies and as a control for climate models. The knowledge of surface temperature is important to a range of issues and themes in earth sciences central to urban climatology, global environmental change, and human-environment interactions. In the study an attempt has been made to estimate surface temperature over Delhi area using Landsat-7 ETM+ satellite data. The variability of these retrieved LSTs has been investigated with respect to different land use / land cover (LU/LC) types determined from the Landsat visible and NIR channels. The classification uncertainties over different land use land cover were assessed and it has been observed that the classification uncertainties were found to be lowest using Minimum Noise Fraction (MNF) components. The emissivity per pixel is retrieved directly from satellite data and has been estimated as narrow band emissivity at the satellite sensor channel in order to have least error in the surface temperature estimation. Strong correlation is observed between surface temperature with Normalized Difference Vegetation Index (NDVI) over different LU/LC classes and the relationship is moderate with fractional vegetation cover (FVC). A regression relation between these parameters has also been estimated indicating that surface temperatures can be predicted if NDVI values are known. The results suggest that the methodology is feasible to estimate NDVI, surface emissivity and surface temperature with reasonable accuracy over heterogeneous urban areas.