Research on the coupling mechanism between green space morphology and the urban heat island effect based on machine learning: a case study of Dali City, China.

Green space configurations represent a critical pathway for mitigating the urban heat island (UHI) effect. However, the underlying mechanistic link between green space morphology and UHI dynamics remains underexplored. This study investigates the relationship between green space morphological attributes and UHI in Dali City through morphological spatial pattern analysis and machine learning approaches. Key findings include: (1) UHI exhibited significant negative correlations with altitude, Normalized Difference Vegetation Index, core, average building height, islet, bridge, branch, and edge, while showing positive associations with Normalized Difference Built-up Index (NDBI), population density, road network density, perforation, and loop. (2) Core, branch, islet, and edge demonstrated greater explanatory power than average building height and NDBI. Enhancing connectivity among green space patches was found to significantly improve cooling efficiency. (3) Random forest model integrating green space morphological factors (proposed model) outperformed a benchmark random forest model excluding morphological parameters in terms of fitting accuracy. (4) The random forest model surpassed Ordinary Least Squares, Spatial Error Model, Spatial Lag Model, and Geographically Weighted Regression models in predictive performance. These results establish a methodological framework for evaluating the landscape morphology-UHI relationship and offer empirical guidance for urban planning strategies to mitigate heat island effects.