An urban land surface temperature and emissivity separation algorithm from ASTER TIR data and its application

Urban land surface temperature (LST) is critical for understanding urban thermal environments and expansion. Compared with natural surfaces, urban areas have complex geometric structures causing multiple scattering and adjacency effects. This study combines the XGBoost algorithm with an improved temperature and emissivity separation (TES) algorithm to retrieve urban LST and land surface emissivity (LSE) from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) thermal infrared (TIR) data. Three aspects are included: firstly, sing the XGBoost algorithm to estimate urban canopy bright temperature (BT) from 5 ASTER TIR top of atmosphere (TOA) BTs; secondly, employing an improved TES algorithm based on sky view factor (SVF) to separate LST and LSE; finally, the accuracy of the proposed algorithm is evaluated from simulation data. Simulation results show that the root mean squared errors (RMSEs) of the urban canopy BTs are about 0.2 K and 1.2 K using the XGBoost algorithm and split window (SW) algorithm, respectively. The urban LST and LSE RMSEs using the proposed algorithm are approximately 0.36 K and 0.022, respectively. Applied to Beijing and Wuhan, China, the algorithm yields slightly lower urban LST compared to ASTER level-2 products, with LST/LSE RMSEs of 0.72 K and 0.017, respectively.