DESC0024737

The urban heat island e?ect is a critical health concern that promises to intensify the negative impacts of climate change during the 21st century. Within urban centers, the heat landscape varies rapidly over short distances due to factors including local vegetation, built environment, albedo, and anthropogenic heat sources. Recognizing the disproportionate vulnerability of certain individuals and communities to heat-related illnesses and fatalities, it is important to understand the spatial distribution of heat. To address this challenge, our solution will use machine learning merge heterogeneous satellite data sources. The model will deliver urban surface temperature with an unprecedented combination of high spatial and temporal resolution. The resulting 70-meter temperature dataset will be made accessible through a user-friendly web visualization tool. The data will updated every 10 minutes, allowing for real-time monitoring. Additionally, users will have the option to overlay this temperature data with maps of socioeconomic, demographic, and other factors. During Phase I we will demonstrate the feasibility of modeling urban surface temperature by adapting state-of-the-art di?usion models to the problem of statistical downscaling. We will build and deploy the model on commercial cloud infrastructure. As a case study, we will share the visualization interface with the City of Boston and urban researchers in the area. Our product benefits municipal planners, weather warning and alert systems, and energy provision, whether supply-side, market intermediaries or customers. More granular temperature data enables precise energy optimization in buildings, reducing energy consumption and emissions, and aids in city planning for sustainable, comfortable, and equitable urban environments.