Retrieval of high spatial resolution mountainous land surface temperature considering topographic and adjacency effects

Land surface temperature (LST) is a key parameter reflecting the interaction between land and atmosphere. Currently, thermal infrared (TIR) quantitative remote sensing technology is the only means to obtain large-scale, high spatial resolution LST. Accurately retrieving high spatial resolution mountainous LST (MLST) plays an important role in the study of mountain climate change. The complex terrain and strong spatial heterogeneity in mountainous areas change the geometric relationship between the surface and satellite sensors, affecting the radiation received by the sensors, and rendering the assumption of planar parallelism invalid. In this study, considering the influence of complex terrain in mountainous areas on atmospheric downward radiation and the thermal radiation contribution of adjacent pixels, a mountainous TIR radiative transfer model based on the sky view factor was developed. Combining with the atmospheric radiative transfer model MODTRAN 5.2, a nonlinear generalized split-window algorithm suitable for high spatial resolution MLST retrieval was constructed and applied to Landsat-9 TIRS-2 satellite TIR remote sensing data. The analysis results indicate that neglecting the topographic and adjacency effects would lead to significant discrepancies in LST retrieval, with simulated data showing LST differences of up to 2.5 K. Furthermore, due to the lack of measured MLST in the field, the MLST accuracy obtained by this retrieval method was indirectly validated using the currently recognized highest-accuracy forward 3D radiative transfer model DART. The MLST and emissivity were input into the DART model to simulate the brightness temperature at the top of the atmosphere (TOA) of Landsat-9 band 10, and compared with the brightness temperature at TOA of Landsat-9 band 10. The RMSE (Root Mean Square Error) for the two subregions was 0.50 and 0.61 K, respectively, indicating that the method proposed can retrieve high-precision MLST.