Urban HEAT Map
Hyper-resolution Environmental AI-based Temperature maps to empower local authorities to make decisions on heat mitigation interventions.
Can we create an easy-to-use hyper-resolution machine learning-based model for urban environments to empower decision makers and local authorities to make well-informed decisions regarding heat intervention, public health impacts, and mitigation methods? With urban population expected to increase further and urban heat waves projected to intensify and become more frequent (Seto et al. 2012; Zhao et al. 2021), a hyper-resolution urban heat mapping platform that considers individuals and communities’ vulnerabilities will be critical to ensure reduced mortality and promote public health and wellbeing, especially from disadvantaged communities.
Super resolution machine learning algorithms (e.g., super-resolution generative adversarial networks) will be used to take the city/district level urban environmental information and derive a block-level or neighborhood level urban heat map for city planners to identify which areas need prioritization based on heat extremes and human vulnerability. Machine learning inpainting methods (e.g., partial convolutions) will also be used for estimating meteorological variables for areas that lack sensors or weather stations. Thus far, the team has made connections with the city of Miami, which is interested in piloting this program, and additional researchers with expertise in urban heat modeling, data fusion, and remote sensing as mentors and advisors to help support the team’s progress.
A Convergence Research Project
In order to create near-term predictions and long-term climate projections of urban heat, the team is using a multi-disciplinary approach, including the disciplines of meteorology and climate science, urban informatics, machine learning, sociology, economics and environmental policy.
This research is also a multi-sector effort, including academic institutions for an established and sound foundation based on theoretical and physical principles, research laboratories for access to cutting edge and state-of-the-art machine learning, and local government for a nuanced understanding of the actionable needs and inherent limitations faced by officials and decision makers.
How has the CORE Institute helped the team?
Our team greatly benefited from the CORE bootcamp and in-person incubator activities. The CORE bootcamp provided thorough and well-paced foundational training on the principles of convergence research and also provided concrete examples of what successful convergence research looks like by sharing insights from past projects. The pacing of the virtual CORE bootcamp allowed for time to synthesize learnings at our own pace and for flexible scheduling in the case of professional scheduling conflicts. The CORE bootcamp also provided a space to engage with other CORE Fellows and for team creation based on shared research interests. The in-person incubator helped with building professional relationships among our team members that will very likely last beyond the scope of our project of interest, helping with a sense of community in convergence research. The in-person incubator also allowed us to sit down and work together, and identify potential failure points in our proposed conceptual framework and data availability.
