From Urban Heat to Human Impact: A Machine Learning Framework for Equity-Sensitive Climate Adaptation

Climate change is intensifying heat extremes in cities, where dense built environments amplify thermal exposure and increase risks to human health, productivity, and well-being. Urban Heat Island (UHI) effects, driven by impervious surfaces, reduced vegetation, and altered surface energy balances disproportionately burden urban populations, particularly vulnerable groups such as the elderly, low-income households, and those with limited access to cooling or green space. Nature-based solutions (NBS), including urban greening and de-sealing, offer substantial potential to mitigate urban heat by restoring local cooling processes. However, effective heat adaptation requires more than identifying locations with the largest temperature reductions: policymakers must also consider how many people are affected, which population groups are exposed, and where elevated heat coincides with social vulnerability and limited adaptive capacity.

This contribution presents a prototype machine learning-based framework for modeling hyperlocal land surface temperature (LST) as a function of land cover and urban form, and for linking heat exposure to human-centered impacts, using Leipzig, Germany as a case study. At the core of the framework is a machine-learning model that predicts pixel-level LST from satellite-derived land-cover information, spectral indices, and selected indicators of urban morphology. By learning the relationship between local land-cover configurations and surface temperature under city-wide climatic conditions, the model generates high-resolution heat maps that reveal fine-grained spatial variation in thermal exposure and its underlying drivers.

To move beyond purely physical measures of urban heat, predicted LST patterns are integrated with socio-economic and demographic indicators, including population density, land use, and proxies for vulnerable population groups. This coupling enables heat exposure to be assessed in relation to the distribution of people and social characteristics across the urban landscape, highlighting where high temperatures intersect with heightened risks to health and well-being. The framework thus supports an explicitly impact-oriented perspective on urban heat.

Finally, the framework provides a basis for evaluating the cooling potential and distributional implications of nature-based solutions, such as targeted greening or de-sealing interventions. By linking land-cover changes to both thermal effects and population exposure, the approach can inform spatially targeted, socially aware heat mitigation strategies and support urban climate adaptation planning that prioritizes both effectiveness and equity.