Blue-Green space seasonal influence on land surface temperatures across different urban functional zones: Integrating Random Forest and geographically weighted regression.

Abstract

As climate change and urbanization progress, the urban heat island issue will affect more people. Urban blue-green spaces can effectively mitigate the urban heat island effect, and their structure and morphology significantly impact the degree of mitigation. To identify the most effective blue-green space distribution for mitigating the heat island effect across different urban function zones (UFZ), we selected 14 landscape metrics of blue-green spaces in the main urban area of Nanjing. Using the Random Forest model, we identified the four metrics with the most significant contribution, and then applied the Geographically Weighted Regression (GWR) model to obtain explicit spatial-related implications. We found that GWR model outperforms others (R²_Mar. = 0.574, R²_Aug. = 0.482, R²_Sept. = 0.618, R²_Oct. = 0.567, R²_Dec. = 0.460). in March the Landscape Shape Index of green spaces has the greatest impact (Feature Importance (FI) = 0.350), in August, the average size of green spaces is most influential (FI = 0.206). Patch Density of green spaces plays the most significant role in September (FI = 0.251) and October (FI = 0.253). Industrial areas are most impacted by green space structure (coef_LSI3 = 0.49, coef_AREA-MN8 = -0.53, coef_PD9 = 0.45). The influence of water bodies on Land Surface Temperature (LST) is weaker in winter, with minimal differences across different functional zones. This study introduced an effective method for reducing the number of independent variables in linear modeling, and elucidated that the optimization of blue-green space design should be flexibly adjusted according to urban functional zones.