Comparing the Impacts of Urban Form on Urban-Rural Temperature and Human-Perceived Temperature Differences Using Interpretable Machine Learning Models

Hangying Su, Renlu Qiao,
Comparing the Impacts of Urban Form on Urban-Rural Temperature and Human-Perceived Temperature Differences Using Interpretable Machine Learning Models,
Sustainable Cities and Society,
2025,
106932,
ISSN 2210-6707,
https://doi.org/10.1016/j.scs.2025.106932.
(https://www.sciencedirect.com/science/article/pii/S2210670725008042)
Abstract: Previous research has primarily focused on temperature differences between urban and rural areas to understand more severe heat challenges faced by urban residents. However, human actual perceived heat is also shaped by humidity and wind speed. Given that urban form affects not only temperature directly but also urban ventilation, we examine how urban form influences differences in human-perceived temperature between urban and rural areas (ΔHPT), comparing this relationship to the more commonly studied urban-rural temperature differences (ΔT). Utilizing data from 238 Chinese cities, this study employs a Monte Carlo simulation-based interpretable machine learning framework to analyze these relationships. The results show that findings regarding the impact of urban form on ΔT cannot be directly extended to ΔHPT. Specifically, the contribution of urban form group to ΔHPT (37.6%) is significantly greater than its contribution to ΔT (29%), even exceeding the contribution of the natural factor group. Furthermore, certain urban form characteristics exhibit different relationships between ΔHPT and ΔT. For example, in highly compact cities, while further increases in urban compactness reduce ΔT, it paradoxically exacerbates ΔHPT. Our findings challenge the effectiveness of current heat mitigation strategies based solely on ΔT and provide valuable insights for developing urban form optimization strategies that not only reduce temperatures but also alleviate the actual heat stress experienced by urban residents.
Keywords: Urban heat island; Human-perceived temperature; Urban form; Heat stress; Interpretable machine learning model