August 10, 2026 [ report
](https://phys.org/editorials/)
Paul Arnold
Author
Gaby Clark
Scientific Editor
Robert Egan
Senior Editor
Tropical cyclones are among the most destructive natural forces on the planet, but predicting where they will go and how strong they will become is beset by challenges. Current methods are often good at tracking storm paths but struggle to anticipate changes in intensity.
Enter artificial intelligence. Researchers from Google DeepMind and Google Research have developed WeatherNext Cyclones, or WN-C, an AI system that outperformed leading systems in storm tracking, intensity forecasting and wind extent. Their findings are published in the journal Nature.
A new AI approach #
To build their new AI system, the teams trained it on decades of global weather analyses and a database of nearly 5,000 historical tropical cyclones. This allowed the system to learn how large-scale atmospheric patterns interact with the physics of an individual hurricane.
Traditional models either look at the whole planet to see where the wind is pushing a storm or zero in on a small area to see how strong the storm is getting. Doing both at once takes too much computing power.
But the new AI system can do both simultaneously. It studies the global picture to see where a storm is headed while also tracking specific details like wind speed and overall size. The AI can use large-scale weather patterns to help predict these details about the storm.
When forecasters run the model, it can produce 50 or more different samples. This gives a clearer picture of the uncertainty surrounding a storm's future path and intensity.
Extra warning time #
The team tested WN-C against historical storm data from 2023 to 2025. They also put the model to work on active storms in real time, publishing predictions on Google Weather Lab and operating a version of the model in the North Atlantic and East Pacific during 2025.
When looking five days into the future, the AI was dramatically more accurate at pinpointing a storm's location than Europe's primary global forecasting system. That would give communities more than 30 additional hours of warning time at the same level of accuracy.
When it came to predicting a storm's peak wind speeds, its three-day intensity forecasts were noticeably sharper than the United States' top specialized hurricane model. "WN-C's improvements in track and intensity accuracy represent a gain of one day or more in forecast lead time advantage," the study authors wrote in their paper.
WN-C also caught sudden, dangerous surges in wind speed when a storm rapidly intensifies within 24 hours. The AI improved the success rate for spotting these events while also reducing false alarms.
"This work shows the potential for AI to provide forecasters at weather agencies with more accurate and reliable guidance, ultimately leading to better-informed decisions that can save lives and protect communities."
Written for you by our author Paul Arnold, edited by Gaby Clark, and fact-checked and reviewed by Robert Egan—this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive.
If this reporting matters to you, please consider a donation (especially monthly). You'll get an ad-free account as a thank-you.
Publication details
Ferran Alet et al, Operational Tropical Cyclone Forecasting with AI, Nature (2026). DOI: 10.1038/s41586-026-10953-2 Journal information:
[Nature](https://phys.org/journals/nature/)
[
](http://www.nature.com/nature/index.html)
© 2026 Science X Network
Citation: AI cyclone forecasts could add 30 hours of warning time (2026, August 10) retrieved 11 August 2026 from https://phys.org/news/2026-08-ai-cyclone-hours.html