Combining measurements across platforms
Air pollution varies over distances and timescales that no single observing system captures well. Fixed monitors provide long records at a few locations; mobile platforms sample many locations intermittently; aircraft and satellites add different spatial and vertical information.
For my dissertation, I combined Google Street View mobile monitoring with fixed-site sensor networks to estimate pollutant concentrations at 30 m and 15 min resolution. The method uses compressive sensing to infer spatiotemporal structure from the observations. The resulting fields resolve gradients along freight corridors and near small industrial sources that are poorly represented by sparse monitoring networks or coarse atmospheric models. I used these fields for monitor siting and for evaluating localized emission controls.
Cross-platform data fusion. Physics-constrained statistical models for combining satellite, aircraft, mobile, and fixed-site observations while retaining measurement-specific uncertainty.
Measurement design. Bayesian experimental design for choosing observations based on the question they are meant to answer, such as wildfire-smoke exposure or near-road gradients.