Diurnal land surface temperature responses to urban morphology during heatwaves: insights from SDGSAT-1 and explainable machine learning

Understanding how urban morphological structure regulates land surface temperature (LST) is critical for mitigating the urban heat island (UHI) effect under extreme heatwave conditions. However, the interactive and threshold effects of urban morphology on diurnal LST variation remain insufficiently understood. In this study, diurnal LST was retrieved from high-resolution, radiometrically consistent thermal infrared data acquired by SDGSAT-1. A two-dimensional (2D)–three-dimensional (3D) urban morphology indicator system was constructed using multi-source remote sensing datasets describing buildings, vegetation, and anthropogenic activity. XGBoost models combined with SHapley Additive exPlanations (SHAP) were applied to quantify nonlinear and interactive effects on daytime and nighttime LST. The results indicate that: (1) vegetation and building environments dominate daytime LST variations, whereas nighttime LST is jointly influenced by vegetation, buildings, and anthropogenic factors; (2) tree height (TH) and building height (BH) exhibit nonlinear response relationships revealed by SHAP, with enhanced daytime cooling but intensified nighttime warming when TH exceeds ~15 m and BH surpasses ~10 m; and (3) interaction effects show that high TH and BH combined with low building density (BD) produce optimal daytime cooling, while elevated vertical morphology is associated with higher nighttime LST.