Project Grant 2530596

This Project Grant award of $500,000.00 from the National Science Foundation's Geosciences Program (CFDA 47.050) is funding the development of a transformative AI-based framework for generating high-fidelity, physically consistent, and uncertainty-calibrated geoscience data. The goal is to overcome limitations in observational infrastructure and computational cost to produce enhanced datasets that can improve decision-making for disaster preparedness, emergency response, and infrastructure planning. Key innovations include physics-guided loss functions, text-prompted guidance from expert knowledge, and conformal prediction techniques to provide rigorous confidence intervals. The project will validate the approach using diverse, high-impact datasets such as satellite precipitation observations and radar data, with specific applications focused on improving tropical cyclone prediction and wildfire risk detection. By integrating explainable AI with physical principles, the research aims to improve public trust in scientific models and provide actionable insights for meteorologists, policymakers, and emergency managers. The award period is from October 1, 2025 to September 30, 2028.