WeatherNext: AI Revolution in Cyclone Forecasting
AI model provides an extra day of early warning for cyclone predictions.
1A New Era in Cyclone Forecasting
WeatherNext has developed an AI model that predicts a cyclone's track, intensity, and wind structure. This model provides an extra day of predictive accuracy compared to previous models. For example, three-day forecasts are as accurate as older models' two-day forecasts. This advancement is equivalent to roughly a decade of progress in meteorology.
2Developed Through Global Collaboration
The WeatherNext model was developed in collaboration with AI researchers at Google DeepMind and Google Research, along with experts from the National Hurricane Center (NHC), the Cooperative Institute for Research in the Atmosphere (CIRA), and the UK Met Office. This team ensured the model's global applicability.
3Real-World Impact
During the 2025 hurricane season, the WeatherNext model enabled the NHC to make a historic forecast for Hurricane Melissa by predicting its rapid intensification and landfall in Jamaica. This provided critical preparation time for local teams. The model now predicts 1,000 different scenarios for each cyclone to support decision-makers.
4Open-Source Model
WeatherNext has open-sourced its WeatherNext 2 and WeatherNext Cyclones models. This move aims to empower the research community and amplify AI's impact. The models assist local forecasters in preparing for natural disasters, supporting renewable energy growth, and anticipating extreme weather events.
5Technological Advancement
WeatherNext Cyclones predicts global weather patterns and cyclone tracks up to 15 days in advance, starting from global atmospheric conditions during Hurricane Milton (October 2024). Running a 1,000-member ensemble generates localized probability maps from tropical storms to hurricane-force winds.
6Innovative Model Approach
The WeatherNext model predicts a cyclone's track and intensity using a single AI model. This is made possible through a unique approach to low-resolution inputs. The model was trained on global weather dynamics and expert-curated historical cyclone observations.
7Tested on Historical Data
WeatherNext Cyclones was tested on historical cyclones from 2023 to 2024. The model gained more than 24 hours of lead time advantage in predicting cyclone tracks, intensity, and wind structure compared to other top weather models. This improvement is equivalent to a decade of progress based on trends over the last 20 years.
8Future Implications
The WeatherNext model has the potential to reduce the impact of natural disasters by improving cyclone forecast accuracy. Its open-source nature will help the research community enhance this technology and build more resilient communities. This is a significant step in areas like renewable energy and extreme weather preparedness.
Original article: DeepMind — WeatherNext: Siklon Tahminlerinde Yapay Zeka Devrimi