Air pollution is a leading cause of premature deaths, and detailed tracking of its distribution at a local level is a significant priority in terms of policy and long-term reductions. While low-cost sensors, satellite products, and other data sources are able to track certain gases, they often differ in spatial and temporal coverage.
This study will develop a Bayesian data integration approach with Gaussian Processes using data from low-cost sensors and other globally available datasets. Our focus is to integrate new remote sensing data products from the TEMPO satellite and explore the use of an atmospheric chemistry & transport model (WRF-Chem) to add physical constraints in forecasting. The resulting integrated map and emulator will enable rapid predictions and policy recommendations.

In high school, I enjoyed delving into atmospheric modeling using open-access data. However, I was limited by my computer’s computational power and (lack of) access to advanced physics-based models. I am now excited to work on a project involving these tools, as well as learn from others doing research on pollution at MIT.