Urban Thermal Archetype Classification in the Context of Urban Development Transformation Using Machine Learning Techniques
Open Access
- 30 July 2025
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
- Vol. 18 (19391404) , 20035-20053
Abstract
In pursuit of urban climate resilience, it is crucial to characterize the spatial heterogeneity of urban thermal environments for spatially targeted mitigation, with the Local Climate Zone (LCZ) framework emerging as a prevailing and powerful tool. However, the inherent complexity and highly dynamic nature of urban-atmosphere interactions necessitate complementary approaches to observe and interpret subtle (e.g., inter-LCZ) built-environment alterations and associated thermal responses, particularly in contexts like urban renewal. This study collected 21 indicators of urban composition, urban form, urban function, and socio-economy to provide a more complete and detailed portrayal of real-world built-environments in Wuhan. Then, four machine learning approaches, including k-means, Agglomerative Hierarchical Clustering (AHC), Gaussian Mixture Model (GMM), Density-Based Spatial Clustering of Applications with Noise (DBSCAN), and Autoencoder combined with k-means (AE+k-means) were used to recognize urban thermal archetypes in large volumes of geospatial data. The clustering results were evaluated using both internal and external metrics, with urban thermal performance presented using reliable and representative summer land surface temperature (LST) data reconstructed through rigorous data selection and spatio-temporal image fusion. Results revealed the robust k-means to be optimal. The derived urban thermal archetypes demonstrated improved spatial coherence and enhanced interpretation of LST variations, with an increase in R² from 0.31 to 0.48 as compared to LCZ. Sankey diagram further certified the superiority of our approach in capturing LST heterogeneity, particularly in high-temperature areas. Then, geographically weighted regression (GWR) was applied to reveal the spatially varying relationships between urban indicators and LST. This integrated analysis provided valuable insights into localized drivers of urban heat, supporting the development of targeted and practical climate-adaptive planning strategies tailored to specific urban contexts, particularly in the context of urban renewal.Keywords
Funding Information
- National Natural Science Foundation of China (52308079)
- Natural Science Foundation of Hubei Province (2023AFB080)
- Key Laboratory of Spatial Intelligent Planning Technology (20240302)
This publication has 97 references indexed in Scilit:
- Exploring the combined cooling effect of street canyon geometry and the surrounding built environmentEnvironmental Science and Pollution Research, 2024
- Integrating planar and vertical environmental features for modelling land surface temperature based on street view images and land cover dataBuilding and Environment, 2023
- Identifying urban morphological archetypes for microclimate studies using a clustering approachBuilding and Environment, 2022
- Effects of block morphology on the surface thermal environment and the corresponding planning strategy using the geographically weighted regression modelBuilding and Environment, 2022
- ATLAS: Software for analysing the relationship between urban microclimate and urban morphology in a tropical cityBuilding and Environment, 2022
- Future “local climate zone” spatial change simulation in Greater Bay Area under the shared socioeconomic pathways and ecological control lineBuilding and Environment, 2021
- Improved machine-learning mapping of local climate zones in metropolitan areas using composite Earth observation data in Google Earth EngineBuilding and Environment, 2021
- Identification of representative buildings and building groups in urban datasets using a novel pre-processing, classification, clustering and predictive modelling approachBuilding and Environment, 2018
- GeoDa: An Introduction to Spatial Data Analysis2009
- Silhouettes: A graphical aid to the interpretation and validation of cluster analysisJournal of Computational and Applied Mathematics, 1987