INFLUENCE FACTORS OF LAND SURFACE TEMPERATURE INVERSION USING THERMAL INFRARED HYPERSPECTRAL REMOTE SENSING SATELLITES DATA

Land surface temperature (LST) is widely used in research fields such as numerical forecasting, global circulation models, and regional climate models. For the remote sensing data from satellites with thermal infrared detection capability, the land surface temperature (LST), land surface emissivity (LSE), and atmospheric influence are mixed together. Using different assumptions and approximations for the radiative transfer equations and surface emissivity, various LST algorithms have been proposed. Among these algorithms, the split-window (SW) algorithm is currently most widely used. Besides, with the rapid development of machine learning, new ideas have been emerged for quantitative remote sensing inversion. For a hyperspectral remote sensing satellite with over 20 thermal infrared channels, machine learning methods such as random forest and artificial neural network can be selected to build an integrated separation and inversion algorithm for LST and LSE. In this paper, the influencing factors of LST inversion using thermal infrared hyperspectral satellites data is discussed, taking the SW algorithm and integrated machine learning algorithm as examples, and the contribution of these factors to the LST inversion error is analysed. We hope this paper could provide valuable reference for the design, index analysis and error calculation for remote sensing satellites with thermal infrared hyperspectral detection capability.