WeatherNext – [An] AI Model Achieves Breakthrough In Forecasting Cyclones
--
https://deepmind.google/blog/weathernext-ai-model-achieves-breakthrough-in-forecasting-cyclones/ <-- shared technical Google DeepMind blog post
--
https://doi.org/10.1038/s41586-026-10953-2 <-- shared paper
--
https://deepmind.google/science/weatherlab/ <-- shared data
--
https://github.com/google-deepmind/weathernext <-- shared GitHub repository
--
H/T @Juliet Rothenberg | Product Director of Earth & Resilience AI at Google
[this post should not be considered an endorsement of a particular organisation or their approach]
“[The Google WeatherNext AI team] are showing how the WeatherNext AI model from Google DeepMind and Google Research has achieved state-of-the-art accuracy in predicting a cyclone’s track, intensity, and wind structure. On average, the WeatherNext Cyclones model gives forecasters an extra day’s worth of predictive accuracy- delivering an advance equivalent to roughly a decade of historical meteorological progress 🌀
Here is how WeatherNext is transforming cyclone forecasting:
Gaining an Extra Day of Advanced Warning: WN 3-day forecasts are as good as what prior models were able to provide for 2-day forecasts, giving critical time for emergency response.
Overcoming Traditional Trade-offs: WN bridges the gap between massive global atmospheric currents (which steer a cyclone’s path) and fine-grained thermodynamic processes around its core (which drive its intensity) into a single AI model.
Unprecedented Ensemble Scale: Using Functional Generative Networks (FGNs), WN now generates 1,000-member ensembles in less than a minute on a TPU to capture rare, consequential tail-risks like sudden rapid intensification – which means forecasters can see a broader range of possible scenarios.
Real-World Impact: During the 2025 Atlantic hurricane season, the WN model helped the National Hurricane Center (NHC) make a historic forecast for Hurricane Melissa by predicting rapid intensification and landfall five days in advance.
[The] teams are open sourcing the operationalized models (WeatherNext Cyclones and WeatherNext 2), alongside a compact version (WeatherNext 2-mini) that can run on a single TPU in a free public Colab notebook – all with a goal of empowering local organizations worldwide.
Weather affects everyone. By combining advanced AI with the real-world expertise of human forecasters, we can build a collaborative ecosystem that saves lives and helps communities adapt to a changing climate…”
--
“Tropical cyclones are among the most dangerous and costly weather phenomena, yet forecasting them remains a profound scientific challenge. Here, we introduce WeatherNext Cyclones (WN-C), an AI operational weather model producing state-of-the-art ensemble forecasts for track, intensity, and size of tropical cyclones worldwide. Trained on a combination of global analysis data1 and a global database of historical tropical cyclones2,3, WN-C generates large ensembles of possible global weather and cyclone scenarios extending 15 days into the future. Evaluated on tropical cyclones from 2023–2025, the track, intensity and wind radii predictions from WN-C offer an average of a day or more of lead time advantage over leading operational models, an improvement in accuracy comparable to the progress seen over the last decade of operational development. [The researchers] achieved these results using inputs orders of magnitude coarser than regional models, suggesting that high resolution is not a strict prerequisite for state-of-the-art intensity forecasting and that this coarser atmospheric data contains more intensity signal than previously recognised. Including predictions from WN-C in a weighted-average consensus ensemble substantially improves its skill. The scalability of WN-C enables up to 1,000-member ensembles which better capture rare events over conventional 50-member ensembles. By providing state-of-the-art operational ensemble guidance to human forecasters, this work represents a step-change towards more reliable and timely forecasts and warnings that can help protect lives and mitigate the devastating impacts of tropical cyclones…”
#Google #DeepMind #GoogleResearch #AI #ensembles #FunctionalGenerativeNetworks #WeatherNext #cyclone #operationalised #model #modeling #forecasting #spatialanalyis #spatiotemporal #track #intensity #windstructure #hurricane #weather #climate #metrology #cyclonetrack #risk #hazard #emergencyresponse #planning #tool #earlywarning #scale #magnitude #path #track #thermodynamic #scenarios #opensource #impacts #tropicalcyclones #WeatherNextCyclones #weathermodel #atmospheric #predictions #mitigation #warning #robust #publicsafety #infrastructure
@Google | @WeatherNext