Can AI Help Our Cities Beat the Heat? Inside the University of Michigan’s AI for Urban Heat Resilience Hackathon

This article describes the University of Michigan's AI for Urban Heat Resilience Hackathon, co-sponsored by AWS, which explored using machine learning to generate thermal imagery from standard RGB photographs to address urban heat mapping challenges.

• Extreme heat is the leading weather-related cause of death in the US, claiming over 2,300 lives in 2023
• Urban heat disproportionately affects low-income communities, older adults, and outdoor workers
• Current satellite thermal imagery lacks sufficient resolution and frequency for street-level urban planning
• Hackathon teams developed three approaches: Conditional U-Net, Pix2Pix GANs, and cross-sensor alignment
• Key challenge: preserving spatial structure in predicted thermal maps, especially in homogeneous scenes
• Performance improved with strong material contrast between impervious and vegetated surfaces
• AWS provided SageMaker and Bedrock infrastructure to remove computational barriers for researchers
• University of Michigan plans to expand datasets across additional cities and climates

The hackathon demonstrates how university-industry collaboration can advance AI solutions for public health challenges while maintaining equitable design principles.