Using GeoAI to Reveal the Contribution of Urban Park Green Space Features to Mitigate the Heat Island Effect

Abstract

Although urban parks have a mitigating effect on the urban heat island phenomenon, it remains unclear how different types of green spaces contribute differentially to this effect. The problem can be addressed through the integration of geographic information and AI, forming GeoAI technologies and methods. This study takes the green spaces within the Outer Ring of Shanghai as an example to explore the interaction between various features of green parks and cooling effects. We extracted information on 2,118 major urban park green spaces through GIS, and obtained LST and NDVI data by inverting Landsat 8-OLI remote sensing images. Then, we further used SVM to label and identify underlying features of ground, water surfaces, grasslands, trees etc. to establish an urban park database. In UHI analysis, we defined the buffer zone using the ERM method, and the temperature difference obtained was used as the dependent variable for cooling intensity. In terms of data analysis, scikit-learn and Shapley Additive Explanation revealed the contributions of various indicators of park green spaces to alleviating the UHI. Using GeoAI, we combine geographic spatial science and Explainable AI. we established a thermal comfort evaluation model for park green spaces. We found that the area, NDVI, and surrounding FAR of green spaces have a significant impact on cooling intensity, and other indicators such as area, NDVI, TP, and LP also have a certain effective range. This mechanism of influence will provide important directional recommendations for decision-makers in the initial planning and design of future urban park green spaces.