Enhanced Downscaling of Urban Land Surface Temperatures Using a Land Cover–Enhanced Nonlinear Model with Landsat-8/9 and Sentinel-2 Imagery

Abstract

Urban temperatures are rising due to climate change and rapid urbanization, leading to the urban heat island (UHI) effect, significantly affecting local climates. Satellite-derived land surface temperature (LST) is crucial for understanding urban thermal dynamics. Existing satellite thermal infrared sensors have a coarse spatial resolution that fails to accurately capture the complex thermal variations within cities. This limitation affects the assessment of UHI effects and hinders effective mitigation strategies. To address these challenges, we developed a land cover-enhanced nonlinear model named high-resolution urban thermal sharpener per land cover (HUTS-LC), which builds on the high-resolution urban thermal sharpener (HUTS) algorithm. The proposed method uses high spatial resolution visible and near-infrared data from Sentinel-2 to enhance the LST derived from Landsat-8/9 data. Our model was tested in Perth, Australia. Validated by ground measurements, HUTS-LC demonstrated a significant improvement in accuracy, yielding a Pearson’s correlation coefficient of 0.85, a root mean square error (RMSE) below 3 °C, a mean absolute error less than 2.5 °C, and a normalized RMSE under 7 percent. The results were compared with the original HUTS and linear regression methods, exhibiting an outperformance of HUTS-LC and making it a valuable tool for urban thermal studies involving high-resolution LST data.