Understanding the relationship between 2D/3D variables and land surface temperature in plain and mountainous cities: Relative importance and interaction effects

• Published in: Building and environment 2023-11, Vol.245, p.110959, Article 110959
• Main Authors: Luo, Pinyang, Yu, Bingjie, Li, Pengfei, Liang, Pengpeng, Zhang, Qiuyi, Yang, Linchuan
• Format: Article
• Language: English
• Subjects: 2D/3D variables; Extreme heat; Land surface temperature; Machine learning; Shapley additive explanation (SHAP); Street view
• ISSN: 0360-1323; 1873-684X
• DOI: 10.1016/j.buildenv.2023.110959

The escalating prevalence of extreme heat events has intensified scholarly interest in understanding the nexus between urban built environments and extreme heat. This study focuses on the central urban areas of Chengdu and Chongqing, sizable Chinese cities that share a common building climate zone, yet divergent in topography and landform (i.e., plain versus mountainous). Leveraging spatial regression, machine learning, and Shapley additive explanation techniques, we scrutinize the spatial effects, relative importance, and interactive impacts of two-dimensional and three-dimensional (2D/3D) built environments on land surface temperature (LST). The results indicate that: (1) 2D/3D variables exert significant spatial effects on LST, with the cooling effect of the 3D green view index outweighing that of the 2D normalized difference vegetation index (NDVI). (2) the overall proportion of the relative importance of 3D variables on LST is greater than that of 2D variables in both Chengdu and Chongqing. (3) The 3D variables have particularly strong impacts on LST in mountainous cities, which are 6.21 % higher than those in plain cities. (4) 2D/3D variables interactively influence LST. For instance, LST will be dropped in Chengdu when floor area ratio (FAR) is above 1.5 and NDVI is below 0.53, as well as when FAR is less than 1 and NDVI is greater than 0.53. Our findings provide support and recommendations for location-specific urban sustainable development and effective heat adaptation strategies. •Leveraging machine learning and Shapley additive explanation (SHAP).•The most powerful cooling is delivered by FAR.•The sum of the relative importance of 3D variables is greater than that of 2D.•The impact of 3D variables is higher in mountain cities compared to plain cities.•There are interactive effects between the 2D/3D built environment and LST.