Comparison of split window algorithms to derive land surface temperature from satellite TIRS data

Remote sensing data can be used as the basis for meteorological data. Due to the limitations of meteorological stations on the Earth, derivation of land surface temperature is one of the most important aspects of the remote sensing application in climatology studies. In the present study, Landsat-8 thermal infrared sensor data of the scene located over Khuzestan province with row/path of 165/38 were used to derive land surface temperature (LST). Normalized difference vegetation index (NDVI), fraction of vegetation cover, satellite brightness temperature, and land surface emissivity were calculated as the vital criteria to derive LSTs using the split window algorithms. LST determination was performed by nine different split window algorithms. Eventually, LST products were evaluated using ground-based measurements at the meteorological stations of the study area. The results showed that algorithm of Coll and Casselles had a highest accuracy with RMSE 1.97 °C, and Vidal’s method presented the lowest accuracy to derive LST with RMSE 4.11 °C. According to the results, regions with high density of vegetation and water resources have lowest diurnal temperature and regions with bare soils and low density of vegetation have a highest diurnal temperature. Results of the study indicated that LST algorithm accuracy is an important factor in the environmental and climate change studies.