Seasonal synergistic management of urban heat island effect and PM₂.₅ pollution: Insights from interpretable LightGBM-SHAP machine learning model

Abstract

Urban heat island (UHI) effect and fine particulate matter (PM_2.5) have emerged as two major challenges in urban environmental management, with their interactions further exacerbating threats to global public health. However, strategies for the synergistic management of these two challenges remain limited. To address this gap, we proposed a new synergistic management framework incorporating temporal and spatial dimensions and empirically apply it to Guangdong-Hong Kong-Macao Greater Bay Area (GBA). We analyzed the seasonal spatial distribution of UHI effect and corresponding PM_2.5 concentrations, identifying their spatial synergy using bivariate spatial analysis. Furthermore, an interpretable LightGBM-SHAP machine learning model was applied to explore the key driving factors and underlying mechanisms jointly influencing UHI and PM_2.5. The results showed that spatially synergistic regions of UHI and PM_2.5 were mainly concentrated in the central and northwestern parts of the GBA, particularly in Guangzhou, Foshan, and Zhaoqing. The spatial synergy peaked in summer with a coverage of 73.73 %, highlighting the need for prioritized intervention during this season, while it declined markedly to 46.95 % in winter. Across all seasons, building density (BD) and building shape index (BSI) were identified as key drivers with positive synergistic effects, whereas green space ratio (GSR) and mean annual precipitation (MAP) exhibited negative synergistic impacts. Additionally, factors such as nighttime light (NL), blue space ratio (BSR), edge density (ED), and built-up land ratio (BLR) showed synergistic influence in specific seasons. This study can provide support for the development of more targeted and seasonally adaptive strategies for the synergistic management of urban thermal environments and air quality.