Cooling the Concrete Jungle: An AI-Vision framework for Mapping and Decoding Urban Surface Temperature at the Microscale

Urban Heat Islands (UHIs) transform cities into sweltering hotspots, significantly altering microclimates, and affecting human comfort, health, and urban sustainability. Traditional methods of studying UHIs are often limited by coarse spatial resolution or expensive data requirements, hindering effective and equitable mitigation strategies at the microscale. This study introduces a novel approach, that leverages high-resolution satellite visible range imagery, LANDSAT data, and meteorological data to develop deep learning model to predict or project future land surface temperatures (LST) at a fine scale (30m) across diverse urban environments. We use data from New York City Metro area as a baseline to accurately map microscale thermal variations, revealing critical drivers of urban heat. We then focus on the City of Newark, New Jersey, for more detailed analyses to provide actionable insights for urban planning and public policy aimed at improving urban life, addressing environmental justice and advancing climate equity. We further extend the model’s applicability through finetuning, demonstrating its robustness across global cities presenting different urban landscapes and climatic conditions. Our results enhance understanding of UHI dynamics and pave the way for city-specific, fine-scale climate interventions such as targeted green infrastructures and reflective surfaces in critical areas. This research stands at the intersection of climate change adaptation and urban sustainability, offering a scalable tool for cities worldwide to anticipate and mitigate the adverse effects of UHIs.