Assessing the Relationship Between Seasonal Urban Heat Island Effects and Forest Structure in Hangzhou City Using the XGBoost Model

As a critical component of urban ecological infrastructure, urban forests play a pivotal role in regulating regional climate and mitigating the urban heat island (UHI) effect. However, existing studies have predominantly focused on single temporal snapshots or aggregate spatial scales, with limited attention to the seasonal dynamics of urban forest landscape patterns and a lack of systematic quantification of their nonlinear regulatory mechanisms. Empirical evidence from subtropical cities remains particularly scarce. In this study, Hangzhou was selected as the study area. Land Surface Temperature (LST) was retrieved using the Google Earth Engine (GEE) platform, and the Thermal Field Variance Index was employed to classify UHI intensity. Six representative forest landscape indices were selected to construct an evaluation framework. Pearson correlation analysis and the XGBoost model were further applied to quantify the relationships between landscape patterns and seasonal LST variations. The results reveal that: (1) LST in Hangzhou exhibits pronounced seasonal variability, following the order of summer > spring > autumn > winter. Areas without UHI effects dominate in spring, summer, and autumn, whereas the extent of strong UHI zones increases markedly in winter. (2) All landscape indices are significantly correlated with seasonal LST; forest ratio and forest largest patch index show negative correlations, while forest patch density, forest landscape shape index, number of patches, and landscape division index (DIVISION) are positively correlated. (3) The XGBoost model indicates that DIVISION consistently exhibits high contribution across all seasons, identifying it as a key determinant of LST variation. These findings provide a scientific basis for optimizing urban forest landscape configuration and developing effective UHI mitigation strategies.