AI Model WeatherNext Cyclones Sets New Standard for Tropical Storm Forecasts

A team of scientists introduced WeatherNext Cyclones (WN-C), an artificial-intelligence weather model that produces state-of-the-art ensemble predictions for tropical cyclone track, intensity and wind radii worldwide. By training on global atmospheric analyses and a comprehensive historical cyclone record, the system can simulate thousands of possible weather outcomes extending 15 days into the future. Evaluations covering cyclones from 2023 through 2025 demonstrate that WN-C delivers an average lead-time advantage of a day or more over the leading operational models, matching the progress seen over the past ten years of development.

Remarkably, the model achieves this using input data that are orders of magnitude coarser than those required by regional high-resolution models, indicating that fine resolution is not essential for top-level intensity forecasting. The platform supports ensembles of up to 1,000 members, better capturing low-probability, high-impact events than the typical 50-member ensembles. When WN-C forecasts are blended into a weighted-average consensus, overall predictive skill improves noticeably, offering forecasters more reliable and timely warnings to protect lives and reduce cyclone damage.