科技科学

Operational Tropical Cyclone Forecasting with AI

Nature作者：Ferran Alet2026年8月6日正文已收录本站
  • Article
  • Published:
  • Tom R. Andersson  ORCID: orcid.org/0000-0002-1556-99321 na1,
  • Ilan Price  ORCID: orcid.org/0000-0003-4765-27031 na1,
  • Stratis Markou  ORCID: orcid.org/0009-0008-0434-65231 na1,
  • Andrew El-Kadi  ORCID: orcid.org/0009-0009-6510-08311 na1,
  • Dominic Masters  ORCID: orcid.org/0009-0000-6945-83251 na1,
  • Amy Li2,
  • Samier Merchant3,
  • Natalie Williams  ORCID: orcid.org/0009-0008-9281-13163,
  • Gregory Thornton1,
  • Ken MacKay1,
  • Olivia Graham3,
  • Akib Uddin3,
  • Ben Gaiarin1,
  • Devaja Shah1,
  • Elinor Kruse1,
  • Wallace Hogsett4,
  • David Zelinsky4,
  • John Cangialosi4,
  • Jonathan Martinez  ORCID: orcid.org/0000-0002-0510-59824,5,
  • James Franklin5,
  • Mark DeMaria5,
  • Kate Musgrave  ORCID: orcid.org/0000-0002-5394-39135,
  • Caroline L. Bain  ORCID: orcid.org/0000-0002-2993-93086,
  • Helen Titley  ORCID: orcid.org/0000-0003-1654-98266,
  • Jacklynn Stott  ORCID: orcid.org/0009-0004-7859-33841,
  • Remi Lam  ORCID: orcid.org/0000-0003-4222-53581,
  • Aaron Bell3,
  • Paul Komarek1,
  • Matthew Willson  ORCID: orcid.org/0000-0002-8730-19271,
  • Alvaro Sanchez-Gonzalez1 &
  • …
  • Peter Battaglia  ORCID: orcid.org/0000-0003-3622-71111 

Nature (2026) Cite this article

  • 594 Accesses

  • 11 Altmetric

  • Metrics details

We are providing an unedited version of this manuscript to give early access to its findings. Before final publication, the manuscript will undergo further editing. Please note there may be errors present which affect the content, and all legal disclaimers apply.

Abstract

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. We 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.

Similar content being viewed by others

Subjects

Author information

Author notes

  1. These authors contributed equally: Ferran Alet, Tom R. Andersson, Ilan Price, Stratis Markou, Andrew El-Kadi, Dominic Masters

Authors and Affiliations

  1. Google DeepMind, London, UK

    Ferran Alet, Tom R. Andersson, Ilan Price, Stratis Markou, Andrew El-Kadi, Dominic Masters, Gregory Thornton, Ken MacKay, Ben Gaiarin, Devaja Shah, Elinor Kruse, Jacklynn Stott, Remi Lam, Paul Komarek, Matthew Willson, Alvaro Sanchez-Gonzalez & Peter Battaglia

  2. University of Waterloo, Waterloo, ON, Canada

    Amy Li

  3. Google Research, Mountain View, CA, USA

    Samier Merchant, Natalie Williams, Olivia Graham, Akib Uddin & Aaron Bell

  4. NOAA/NWS/NCEP National Hurricane Center, Miami, FL, USA

    Wallace Hogsett, David Zelinsky, John Cangialosi & Jonathan Martinez

  5. Cooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, CO, USA

    Jonathan Martinez, James Franklin, Mark DeMaria & Kate Musgrave

  6. UK Met Office, Exeter, UK

    Caroline L. Bain & Helen Titley

Authors

  1. Ferran Alet
  2. Tom R. Andersson
  3. Ilan Price
  4. Stratis Markou
  5. Andrew El-Kadi
  6. Dominic Masters
  7. Amy Li
  8. Samier Merchant
  9. Natalie Williams
  10. Gregory Thornton
  11. Ken MacKay
  12. Olivia Graham
  13. Akib Uddin
  14. Ben Gaiarin
  15. Devaja Shah
  16. Elinor Kruse
  17. Wallace Hogsett
  18. David Zelinsky
  19. John Cangialosi
  20. Jonathan Martinez
  21. James Franklin
  22. Mark DeMaria
  23. Kate Musgrave
  24. Caroline L. Bain
  25. Helen Titley
  26. Jacklynn Stott
  27. Remi Lam
  28. Aaron Bell
  29. Paul Komarek
  30. Matthew Willson
  31. Alvaro Sanchez-Gonzalez
  32. Peter Battaglia

Corresponding authors

Correspondence to Ferran Alet, Tom R. Andersson, Ilan Price or Stratis Markou.

Supplementary information

Supplementary Information (download PDF )

This file includes: Supplementary Text, Supplementary References, and Supplementary Display Items. Supplementary Text, Section 1: Extended Methods (Architecture and training details, Computational requirements, Verification metrics, Data description, Evaluation protocol). Section 2: Extended Results (2025 deterministic evaluation, Extended global weather forecasting, Extended baseline comparisons, Tracker/Co-training ablations, Wind speed probabilities, Significance tests, Historical progress analysis, Geospatial skill decomposition, WeatherNext-C Mini evaluation). Supplementary Display Items: 19 Supplementary Figures and 2 Supplementary Tables. These items collectively illustrate deterministic and probabilistic evaluation metrics, baseline performance comparisons, historical rate of progress, model ablations, geospatial error decompositions, and evaluations of the lightweight WN-C Mini model.

Peer Review File (download PDF )

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Alet, F., Andersson, T.R., Price, I. et al. Operational Tropical Cyclone Forecasting with AI. Nature (2026). https://doi.org/10.1038/s41586-026-10953-2

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1038/s41586-026-10953-2