A Semi-Empirical Split-Window Algorithm for Retrieving near Surface Air Temperature from MODIS Data

This study attempts to develop an effective algorithm to directly derive near surface air temperature from EOS/MODIS data. From a theoretical viewpoint, the split-window algorithm for retrieving near surface air temperature was developed based on the radiative transfer equation, which includes the calculation of atmospheric thermal radiance, the linearization of Planck functions, the transformation from effective atmospheric mean temperature to near surface air temperature and other derivation processes. Considering that the coefficients of the theoretical algorithm are highly dependent on the atmospheric profile, which is difficult to acquire in practical applications, a semi-empirical split-window algorithm is generated on the basis of the theoretical algorithm to improve the practicality. The semi-empirical algorithm was applied and validated in the Jing-Jin-Ji (JJJ) Region and the Jiang-Zhe-Hu-Wan (JZHW) Region in China. Results indicate that the algorithm achieves an MAE of 2.11 °C in the JJJ Region and an MAE of 2.22 °C in the JZHW Region. The semi-empirical split-window algorithm also shows better stability than linear regression and machine learning methods when being applied to other data periods. Due to its accuracy and simplicity, the semi-empirical split-window algorithm is a novel method for retrieving near surface air temperature from MODIS thermal bands.