AI driven urban heat equity atlas integrating machine learning for climate risk assessment

Urban heat island (UHI) exposure is unevenly distributed and often co-locates with social vulnerability, motivating tools that integrate physical heat burden with environmental justice considerations. We developed an Intelligent Urban Heat and Equity Atlas for 55,871 U.S. urban census tracts, integrating remotely sensed heat indicators, urban form, meteorology, and tract-level sociodemographic vulnerability measures. Machine-learning models were trained to predict tract-level annual daytime UHI intensity (UHI_annual_day) using spatially grouped cross-validation, and outputs were summarized into two GIS-ready equity products: (i) the Vulnerability-Adjusted Urban Heat Index (VA-UHI), a normalized composite prioritization index combining heat exposure and vulnerability, and (ii) a complementary Heat Exposure Equity Index (HEEI) for equity-focused comparison across vulnerability strata. Across the national tract sample, higher VA-UHI values systematically coincide with higher vulnerability, indicating that the most heat-burdened neighborhoods frequently overlap with areas facing structural socioeconomic disadvantage. Category-based comparisons further show that “Extreme” VA-UHI tracts exhibit higher vulnerability index scores than “Low” VA-UHI tracts (e.g., 0.412 vs. 0.242; + 0.170 index units), reflecting differences in normalized index magnitude rather than calibrated increases in health risk. These atlas outputs support interpretable, tract-scale prioritization for urban heat mitigation and equity-informed planning.