Effect of surface emissivity and retrieval algorithms on the accuracy of Land Surface Temperature retrieved from Landsat data
Rahul Harod, Eswar Rajasekaran, Bimal K. Bhattacharya
Abstract
One of the important variables in the retrieval of Land Surface Temperature (LST) is surface emissivity. The main aim of this study is to test how surface emissivity estimated using different methods affect the accuracy of the retrieved LST. The emissivity was estimated from Landsat data using different variants of the vegetation index (VI) based model and also obtained from the Advanced Spaceborne Thermal Emission and Reflection Radiometer Global Emissivity Dataset (ASTER GED). The LST was retrieved using Statistical Mono Window (SMW) and Generalized Single Channel (GSC) algorithms. The retrieved LSTs were compared with the Landsat collection-2, level-2 surface temperature (L2ST) product and also validated against ground measurements over eight sites across India. From the results, it was observed that the emissivity estimated by different methods resulted in LST with no significant differences in accuracy between them. However, the LST retrieved from the SMW algorithm performed better than the LST from the GSC algorithm. In addition, the LST from the L2ST product performed relatively poorer in comparison with the SMW algorithm. The results indicated that the accuracy of retrieved LST is highly dependent on the LST retrieval algorithm as different surface emissivity methods caused only minor variations in LST values.