Improved method of Land Surface Emissivity retrieval from Landsat TM/ETM+ data

A comparative study has been carried out on the most recent methods for Land Surface Emissivity (LSE) estimation using Landsat TM/ETM+ data. The popularly used method, the integrating NDVI and classification is chosen and analyzed further. The result shows that the estimation model is not accurate enough for LSE is underestimated at either end of the fractional vegetation cover (Pv) range which would lead to overestimating of corresponding Land Surface Temperature (LST). The drawback is modified based on the threshold method, that is, for natural surface and when Pv <; 0.165, ε= εsoil; for town surface and when Pv <; 0.1375, ε= εm; and for Pv >; 0.5, ε= εv. Finally, a processing way adapting for the improved model in large area is presented and the emissivity model before modification and after improvement is applied to Beijing, China to identify emissivity and further to retrieve LST using image based method. The results show that the large difference between LSE and corresponding LST is located in the town surface and soil with low Pv and that a decrease of emissivity by 0.011058 at 318K will increase LST by about 1K. Thus a promising improvement of comparative accuracy can be expected.