Exploring the non-linear impacts of urban features on land surface temperature using explainable artificial intelligence

• DOI: 10.1016/j.uclim.2024.102045
• Publication Date: 2024-06-28T01:14:12Z
• Authors: Fei Feng; Yaxue Ren; Chengyang Xu; Baoquan Jia; Shengbiao Wu; Raffaele Lafortezza
• Keywords: Impervious surface; Urban Heat Island; Urban ecosystem; Land Cover; Elevation (ballistics)

High land surface temperatures (LST) have emerged as crucial threats to urban ecosystems and sustainable development. To better understand mitigate their impacts, it is essential analyze the contributing features. Against this background, we developed a random forest model enhanced by Explainable Artificial Intelligence (XAI) impact features of LST in Beijing, China. By applying XAI method, our results suggest that major Beijing are elevation (44.19%), compactness impervious (17.27%), Normalized Difference Vegetation Index (11.12%), proportion area (8.04%), tree height (3.83%). Compactness exhibited an overall cooling effect, which became weaker at high values. increased with building height, trend reached 5 m. The most important impacting inner city buildings, whereas outer these surfaces. study applies explain non-linear interactions between features, offering innovative insights policy-makers develop planning strategies. Our findings increasing green spaces water bodies well controlling density can effectively heat dense areas enhance effects.