FitzRoy forecasts rain with high accuracy and realism
Skillful and realistic precipitation forecasting with reinforced diffusion
Machine Learning
Summary
Predicting rain accurately is hard because traditional methods either miss details or make blurry guesses that underestimate storms. The authors introduce FitzRoy, a new system combining advanced AI techniques to make rain forecasts that are both precise and realistic. It works well for up to 24 hours ahead and even spots extreme weather events better than current models. This helps provide timely, reliable warnings for floods.
What this means in practice
- •For weather forecasters: Generate more accurate and realistic short-term precipitation forecasts for regions across the United States, including extreme rainfall events.
- •For emergency response teams: Use improved precipitation predictions to anticipate and prepare for flood emergencies, supporting better decision-making during disasters.
Authors
Ozan Sener, Anand Avati, David Hafner, Jierui Lin, Laura Manduchi, Xiaoying Pang, Langwen Huang, Torsten Hoefler, Russ Schumacher, Vladlen Koltun
Abstract
Accurate precipitation forecasting remains elusive as physics-based models struggle to resolve mesoscale convection and machine learning often yields physically inconsistent, blurred results that consistently underestimate extreme events. We present FitzRoy, the first system that simultaneously achieves state-of-the-art predictive skill and physical realism over the continental United States. Through a reinforced diffusion framework, where a conditional discrete diffusion model is refined via reinforcement learning, FitzRoy effectively resolves the long-standing trade-off between pointwise accuracy and structural fidelity. Across lead times up to 24 hours, FitzRoy consistently matches or exceeds state-of-the-art models while preserving structural realism and capturing extreme events. Its low-latency inference further enables deployment in real-time operational settings. In case studies such as the 2025 Texas Hill Country floods and the 2026 Missouri floods, FitzRoy signals record-breaking accumulations that existing models largely missed. By bridging the gap between realism and skill, FitzRoy establishes a new frontier for high-fidelity precipitation forecasting.