WeatherNext: AI model achieves breakthrough in forecasting cyclones
| Source: Google DeepMind Blog
Tags: WeatherNext, DeepMind, Google, Hurricane Forecasting, Nature, Open Source, Climate AI
DeepMind open-sourced WeatherNext, a Nature-published AI model that unifies hurricane track and intensity prediction, delivering one extra day of average lead time — equivalent to a decade of meteorological progress — and predicting 1,000 storm scenarios per cyclone.
Details
Published in Nature on August 6, 2026, WeatherNext resolves a decades-old forecasting trade-off: predicting a cyclone's track requires global-scale atmospheric data, while predicting intensity requires fine-grained local data — previously forcing meteorologists to use two separate models that couldn't be directly combined. WeatherNext trains on both global weather patterns and cyclone-specific data simultaneously, enabling unified track and intensity prediction from lower-resolution inputs than conventional systems. The benchmark result: WeatherNext's 3-day forecasts match the accuracy of prior models' 2-day forecasts — a full extra day of lead time. During the 2025 hurricane season, the model predicted Hurricane Melissa's rapid intensification and Category 5 Jamaica landfall five days in advance at 80% confidence when other systems disagreed. The 2026 season deployment runs 1,000-member ensemble forecasts per cyclone, generating probabilistic wind-impact maps across possible scenarios. Collaborators include the US National Hurricane Center, UK Met Office, CIRA, and international weather agencies. DeepMind is open-sourcing WeatherNext 2 and WeatherNext Cyclones to empower global research and operational forecasting.