Project Grant 2616237

The National Science Foundation Office of Integrative Activities awarded The Washington University $996,716 on August 1, 2026, under the Geosciences program (CFDA 47.050) to develop physics-informed generative artificial intelligence models that quantify aerosol effects on marine low clouds.

The research team will create three AI models trained on satellite observations and field measurements to isolate and quantify how aerosols—including wildfire smoke—influence ocean clouds. Using North American wildfire plumes perturbing low clouds over the North Atlantic as natural experiments, the first two models will predict counterfactual aerosol concentrations and cloud properties absent wildfire influences, while the third will quantify probability distributions during wildfire periods. The models integrate transformer-based architectures with a Lagrangian approach that explicitly encodes emissions and air mass history to distinguish aerosol effects from meteorological influences. The project will also develop human-in-the-loop, agentic systems for broader application.

Understanding aerosol effects on marine low clouds—which cool Earth by reflecting sunlight—improves environmental prediction capability, helping communities prepare for extreme weather, protect public health, and guide infrastructure investment. The research strengthens the nation's AI-ready scientific workforce by training students and researchers in cutting-edge AI methods.

Performance occurs at Washington University in University City, Missouri, with a period of performance from August 1, 2026, through July 31, 2029.