Adaptive Urban Heat Mitigation Through Ensemble Learning: Socio-Spatial Modeling and Intervention Analysis

Urban Heat Islands (UHIs) are intensifying under climate change, exacerbating thermal exposure risks for socially vulnerable populations. While the role of urban environmental features in shaping UHI patterns is well recognized, their differential impacts on diverse social groups remain underexplored—limiting the development of equitable, context-sensitive mitigation strategies. To address this challenge, we employ an interpretable ensemble machine learning framework to quantify how vegetation, water proximity, and built form influence UHI exposure across social strata and simulate the outcomes of alternative urban interventions. Drawing on data from 1660Dissemination Areas in Vancouver, we model UHI across seasonal and diurnal contexts, integrating environmental variables with socio-demographic indicators to evaluate both thermal and equity outcomes. Our ensemble AutoML framework demonstrates strong predictive accuracy across these contexts (R2 up to 0.79), providing reliable estimates of UHI dynamics. Results reveal that increasing vegetation cover consistently delivers the strongest cooling benefits (up to 2.95 °C) while advancing social equity, though fairness improvements become consistent only when vegetation intensity exceeds 1.3 times the baseline level. Water-related features yield additional cooling of approximately 1.15–1.5 °C, whereas built-form interventions yield trade-offs between cooling efficacy and fairness. Notably, modest reductions in building coverage or road density can meaningfully enhance distributional justice with limited thermal compromise. These findings underscore the importance of tailoring mitigation strategies not only for climatic impact but also for social equity. Our study offers a scalable analytical approach for designing just and effective urban climate adaptations, advancing both environmental sustainability and inclusive urban resilience in the face of intensifying heat risks.