Google DeepMind’s AI Shines in Hurricane Forecasts and Heads to the Philippines
Forecasting hurricanes has long been a headache for meteorologists, but Google’s DeepMind AI model nailed it during the 2025 Atlantic season. It correctly predicted Hurricane Melissa’s path and its Category 5 strength a week out, when other models were all over the place. James Franklin, a former National Hurricane Center branch chief, called it the best guidance of the year after reviewing performances across models.
Beating Traditional Models
DeepMind’s hurricane model stood out by handling both track and intensity predictions well. Kate Musgrave, a research scientist at Colorado State University, pointed out that earlier AI efforts were good at paths—driven by big atmospheric patterns—but weak on intensity. This one fixed that by digging into historical hurricane data. Franklin explained the difference: physics-based models like NOAA’s Global Forecast System crunch equations for wind, moisture, and heat. DeepMind’s approach skips the physics lessons and learns patterns straight from past weather records.
Engineers built it with help from the National Hurricane Center and Colorado State. Forecasters there started using it for Melissa, which hit Jamaica with 185 mph winds and caused widespread damage. Still, Franklin admitted it’s a bit of a black box—you feed it data, get a forecast, but can’t always trace the reasoning. He and Musgrave figure it’ll complement, not replace, old-school models and human judgment. Check out the full story in this NPR report.
Now Tackling Typhoons in the Philippines
The same kind of AI push is reaching the Pacific, where the Philippines gets slammed as the world’s typhoon nursery. Google teamed up with Weathernews Inc. (WNI) and the local weather service PAGASA to fold DeepMind’s flood forecasting and cyclone prediction models into their toolkit. This gives forecasters a sharper view of storm impacts for quicker warnings.
Raia Hadsell, DeepMind’s VP of Research, said they’re glad to see the partnership grow and hope it keeps helping forecasts there. Yossi Matias, Google Research VP and GM, stressed that AI plus local experts like PAGASA and WNI builds better disaster prep. Prep Palacios, Google Philippines country manager, added that better storm predictions mean safer communities during typhoon season. More details in this Mindanao Times piece.
From Atlantic hurricanes to Pacific typhoons, DeepMind’s prediction models show AI pulling real weight in weather forecasting. Expect more of this as patterns from history prove their point.
