Project Grant 2615743

The National Science Foundation Office of Integrative Activities awarded the University of California San Diego's Scripps Institution of Oceanography $516,390 on August 15, 2026, under the Geosciences program (CFDA 47.050) to develop mathematical and computational methods for calibrating atmospheric models used in weather forecasting and climate simulation.

The project addresses the challenge of tuning dozens of adjustable parameters that control small-scale atmospheric processes such as cloud formation. Current calibration approaches are slow, expensive, and partly manual. The work leverages a new differentiable atmospheric general circulation model with automatic differentiation capability and integrates machine learning techniques to improve model accuracy, stability, and interpretability. The central research questions focus on determining the spatiotemporal scales over which atmospheric gradients provide useful calibration information, optimizing hybrid models that combine physical and machine-learned components, and identifying optimal sub-grid representations.

The project spans three years, with performance at La Jolla, California, and runs through July 31, 2029. Beyond atmospheric science applications, the resulting methods have transfer potential to other fields involving chaotic nonlinear systems, including plasma physics, turbulence, and geophysics. The work includes graduate student training at the interface of artificial intelligence and Earth science, building workforce capacity in both domains.