Operational tropical cyclone forecasting with AI.

AI-based WeatherNext Cyclones (WN-C) improves tropical cyclone forecasts by over one day in lead time, achieving state-of-the-art accuracy with large ensembles.

• Why it matters: Accurate and timely tropical cyclone predictions are critical for saving lives and reducing damage, yet current models face limitations in forecast precision and lead time.
• What they did: WN-C was trained on global analysis data and historical cyclone records to generate 15-day ensemble forecasts of cyclone track, intensity, and size, using coarse atmospheric inputs.
• The result: In evaluations from 2023 to 2025, WN-C outperformed leading models, especially with larger ensembles, enabling more reliable forecasts and better capturing rare events for improved warning systems.