CAREER: Mechanistically Informed Modeling of 3D Urban Morphology and Real-Time Exposure to PM2.5

Outdoor air pollution, especially particulate matter with a diameter of 2.5 microns or less, is a major threat to public health. Pollution levels in cities often vary among neighborhoods. Building design, street layout, traffic, and weather affect how pollution spreads and accumulates. However, current air pollution models often cannot fully explain why some places become pollution hot spots or how city design could reduce exposure. This project will study how the shape of cities, such as building heights and street layouts, affects air pollution levels. The research will combine air quality measurements, geospatial and traffic data, and artificial intelligence (AI) models to understand where and when pollution accumulates. The project will produce accessible tools to help city planners and engineers to identify risks and explore mitigation strategies. Students will help deploy air sensors and analyze the data while workshops will train planners and public health officials to use these tools to support healthier cities. This project will develop a mechanistically-informed modeling framework that integrates 3D urban morphology data, real-time environmental observations, and interpretable AI to quantify how the built environment influences PM2.5 exposure. It will include two interconnected objectives. First, annual air quality models will be developed for six major U.S. cities by combining standardized morphological indicators from the Local Climate Zone (LCZ) framework (e.g., sky view factor, building surface fraction) with regulatory monitoring and low-cost sensor data to train interpretable machine learning models. These models will link urban structural features to pollution concentrations and use Shapley Additive explanations (SHAP) to interpret the nonlinear relationships. Second, high-resolution hourly exposure models will be developed in Lexington, Kentucky using a dense sensor network, mobile monitoring, and time-series traffic and meteorological data to capture dynamic pollution patterns. Model performance will be validated through cross-validation and additional field campaigns with undergraduate and graduate students participating in sensor deployment and data analysis through hands-on STEM training. The resulting pollution maps and morphological risk indicators will be interpreted into an interactive geospatial dashboard for scenario analysis. Overall, this project will combine interpretable AI, advanced monitoring, and urban morphological analysis to advance scalable exposure modeling methods and provide transferable insights into how urban form modulates pollution dispersion. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2540973 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Tianjun Lu | Institution: University of Kentucky Research Foundation, LEXINGTON, KY | Award Amount: $585,497 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2540973 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2540973.html