A comparative approach of data-driven split-window algorithms and MODIS products for land surface temperature retrieval

The main objective of this paper is to compare eighteen data-driven split-window (SW) algorithms to derive land surface temperature (LST) and identify the most efficient techniques. We also aim to compare the results of the SW algorithms against the MODIS LST products (MOD11_L2/MYD11_L2). To archive these goals, the radiometric correction was accomplished to the MODIS satellite image, and then, the LST was derived using the SW algorithms for the year 2019. This LST was validated using RMSE, MSE, MAPE, and MAD based on the observed data in local meteorological stations. Validation analysis reveals that the Sobrino 1993 algorithm with a RMSE value of 1.79 and Qin algorithm with a RMSE value of 5.28 have respectively high and low accuracy to calculate LST in the study area. Furthermore, we compared the MODIS LST products against the local meteorological station data and achieved respectively the RMSE values of 13.3, 13.96, 18.83, 10.84, 13.91, 13.51, and 5.2 in Ahar, Tabriz, Jolfa, Sarab, Maraghe, Ligvan, and Kalibar stations. These findings demonstrated that the SW algorithms have better performance in comparison with MODIS LST products to estimate LST. Results of this research are of great importance for applying and comparing different data-driven approaches and identifying the most efficient techniques. The obtained results support future research by means of exploring the capability of each method and deriving validate results through efficient techniques.