Synergising Machine Learning and Remote Sensing for Urban Heat Island Dynamics: A Comprehensive Modelling Approach

This study evaluates the effectiveness of sustainable urban regeneration projects in mitigating Urban Heat Island (UHI) effects through a place-based approach. Geographic Information Sys-tems (GIS) and satellite imagery were integrated with machine learning (ML) models to analyse the urban environment, human activities, and climate data in Turin, Italy. A detailed analysis of the ex-industrial Teksid area revealed a significant reduction in Surface Urban Heat Island In-tensity (SUHII), with decreases of -0.94 in summer and -0.54 in winter following regeneration interventions. Using 17 variables in the Random Forest model, key determinants influencing SUHII were identified, including building density, vegetation cover, and surface albedo. This study quantitatively highlights the impact of increasing green spaces and enhancing surface materials to improve solar reflectivity, with findings showing a 19.46% increase in vegetation and a 3.09% rise in albedo after mitigation efforts. Furthermore, the results demonstrate that inte-grating Local Climate Zones (LCZs) into urban planning, alongside interventions targeting these key variables, can further optimise UHI mitigation and assess changes. This comprehensive ap-proach provides policymakers with a robust tool to enhance urban resilience and guide sus-tainable planning strategies in response to climate change.