UAV-based fine-scale daily relationship characterization between 3D urban surface thermal dynamics and landscape patterns

The urban landscape has altered regional thermal circulation, induced the Urban Heat Island (UHI) effect, and exposed over 50% of the global population to heat stress. This study employed an Unmanned Aerial Vehicle (UAV) thermal imager to capture fine-scale 24-hour urban 3D surface thermal dynamics and characterize its daily fine-scale relationship to landscape patterns. The results demonstrated that over 50% of the temperature drift/variations were normalized by the Digital Number Probability Density Function Fitting and Discrete Wavelet Transform-based (DDWT) method, horizontal UHI shifted from built-up areas during high-solar-altitude periods to waters at sunset, while façade insignificantly interacted with the horizontal thermal process. The machine learning model CatBoost correlating 3D landscape metrics and active-pixel-defined temperature process explained 86.0% of temperature variance and 63.5% of temperature change rate variance. The interpretations of Shapley additive explanations (SHAP) and Accumulated Local Effects (ALE) identified albedo and distance thresholds for mitigating UHI as 5 m to trees, 20 m to waters, and 20–50 m to buildings. The framework of this study provides actionable insights for neighborhood-scale UHI mitigation strategies. • 24-h 3D surface thermal dynamics at 0.3 m scale in 2 h interval was identified. • The DDWT efficiently eliminates temperature drift and temporal variation. • CatBoost models 3D interaction of temperature processes and landscape patterns. • SHAP and ALE characterize the contribution of main landscape metrics.