Unraveling the demographic disparity in urban heat risk through explainable machine learning: A case study of Chongqing, China

Increasing urban heat risks, with cascading adverse effects, disproportionately affect vulnerable populations. While previous studies have examined urban heat exposure and its influencing factors, the relationship between heat exposure, heat mitigation resources, and demographic disparities remains underexplored. To investigate these demographic disparities in urban heat risk, this study proposes an explainable machine learning framework that integrates heat exposure data with accessibility to multi-dimensional mitigation resources and quantifies their relationships with various demographic factors. Using Chongqing, China, as a case study, we apply the Urban Heat Risk Index (UHRI) to quantify heat exposure and the G2SFCA method to measure accessibility to mitigation resources. Leveraging XGBoost-SHAP models, we further analyze the role of demographic factors in shaping these disparities. The results reveal that vulnerable populations, such as the elderly and low-income groups, face higher heat exposure and lower access to mitigation resources, with age factors contributing to 32–42 % of the heat risk disparity. Notably, blocks with more than 35 % middle-aged residents exhibit greater heat resilience, whereas high school-educated populations exceeding 63 % experience a significant decline in accessibility of mitigation resources. These findings highlight the urgent need for targeted interventions to address heat risk disparity and promote equitable climate adaptation planning. • We explored urban heat risk disparity via explainable machine learning method. • Vulnerable social groups face higher heat exposure and fewer mitigation accessibility. • Age factors had a greater relative importance on both exposure and mitigation. • Urban core benefits from better accessibility of heat mitigation resources. • Heat-vulnerable blocks are primarily in suburban industrial areas.