How Google DeepMind’s AI predicted Hurricane Isaias better than traditional models
Google DeepMind's AI weather model correctly pointed to Hurricane Isaias's Florida landfall days ahead, while traditional models pointed toward Louisiana. Here's how AI is changing hurricane forecasts, and its limits.
Artificial intelligence is quietly changing how we forecast hurricanes. During Hurricane Isaias this October, Google DeepMind’s AI weather model correctly pointed to where the storm would hit, days before some of the most trusted traditional models.
What happened with Hurricane Isaias?
- Isaias peaked at about 120 mph winds and made landfall near Destin, Florida, as a Category 2 hurricane (about 105 mph) in early October 2026.
- Several days before landfall, Google DeepMind’s model often showed a Florida landfall.
- Traditional “physics-based” models, such as the American GFS, at first pointed further west, toward Mississippi or Louisiana.
- The National Hurricane Center (NHC) increasingly leaned toward the AI and other corrected guidance, and the storm hit Florida.
On October 7, the NHC noted the split itself: AI models on the east side, physics-based models on the west side.
How does an AI weather model work?
Traditional weather models use the laws of physics. They divide the atmosphere into a giant 3D grid and calculate how air, heat and moisture move. This needs huge supercomputers.
AI models work differently. They learn patterns from decades of past weather data, then predict what’s likely to happen next. Once trained, they can produce forecasts much faster, and run many versions (called an “ensemble”) to show a range of possible paths.
Physics models calculate the weather. AI models recognize it, based on everything they’ve seen before.
How good is it?
- The NHC started using DeepMind’s model in its official forecasting process in 2025, and by the end of that season it ranked among the top models for both path and strength.
- Hurricane expert Michael Lowry called it “the best performing computer model that we’ve ever seen.”
- For Isaias, it led on track (where the storm went).
The limits
- Strength: for Isaias, the AI trailed the NHC’s official blended forecast on how strong the storm would be at landfall.
- Small details: it missed a band of very strong winds on the storm’s western side near landfall. Some experimental physics models handled this better.
- The “why” question: scientists don’t fully understand why the AI gets some forecasts right. University of Miami researcher Brian McNoldy said, “we don’t know why they might get a forecast right.”
That’s why NHC Director Michael Brennan says forecasters use a blend of the best models plus human judgment, not any single model.
Why it matters
Knowing where a hurricane will hit even a day earlier can help with evacuation plans, emergency preparation and saving lives. AI won’t replace human forecasters, but it’s becoming one of their most useful tools.
Safety tip: always follow official forecasts and instructions from your national weather service and local authorities, not individual model maps shared on social media.
Frequently asked questions
Can AI predict hurricanes?
Yes. AI models like Google DeepMind’s now help forecast hurricane paths and strength, and are used alongside traditional models by the US National Hurricane Center.
Is AI better than traditional weather models?
On hurricane tracks, Google DeepMind’s model has performed among the best. On storm strength and small details, traditional and blended forecasts can still do better.
Where did Hurricane Isaias make landfall?
Near Destin, Florida, on the Florida Panhandle, as a Category 2 hurricane in early October 2026.
Does the National Hurricane Center use AI?
Yes. The NHC began using Google DeepMind’s model in its forecasting process in 2025, combined with other models and human expertise.
