A satellite image provided by the National Oceanic and Atmospheric Administration (NOAA) shows Hurricane Isaias approaching the Gulf Coast of Alabama and Florida on Friday, Oct. 9 2026.NOAA/NOAA
From NPR by Jessica Meszaros
Hurricane forecasters are increasingly using artificial intelligence to make predictions — and they say the technology is rapidly transforming the field.
When the first hurricane of the Atlantic season formed this month, an AI model beat out traditional models at predicting where it would go.
In early October, Hurricane Isaias made landfall near Destin, Fla., as a Category 2 storm.
Conventional forecasting models initially predicted it would head further west, toward southeastern Louisiana.
But Google DeepMind's artificial intelligence model called it right out of the gate: the hurricane was going to hit the western Florida Panhandle.
"Google DeepMind, in particular, has really proven itself the last few years as the gold standard model," says Michael Lowry, a longtime meteorologist and hurricane expert in Miami, who has previously advised both the Federal Emergency Management Agency and the National Hurricane Center (NHC).
"That model has been, by far and away, the best performing computer model that we've ever seen. Period."
But Google DeepMind's artificial intelligence model called it right out of the gate: the hurricane was going to hit the western Florida Panhandle.
"Google DeepMind, in particular, has really proven itself the last few years as the gold standard model," says Michael Lowry, a longtime meteorologist and hurricane expert in Miami, who has previously advised both the Federal Emergency Management Agency and the National Hurricane Center (NHC).
"That model has been, by far and away, the best performing computer model that we've ever seen. Period."
A different kind of model
The National Hurricane Center first started incorporating the Google DeepMind model into its official forecasts last year.
By the end of the 2025 Atlantic hurricane season, it was already one of the best at predicting a storm's path and strength, says Brian McNoldy, a hurricane researcher for the University of Miami.
Usually, new models take years to refine after first being introduced.
"It's something we're not used to – to have a model, the new kid on the block, very quickly becoming the best model out there," McNoldy says.
"This pretty much came out of the gate in first place, and that's a pretty impressive feat."
Traditional physics-based hurricane models use real-time atmospheric data, like temperature, humidity, pressure and wind shear, as measured by satellites and aircraft.
Traditional physics-based hurricane models use real-time atmospheric data, like temperature, humidity, pressure and wind shear, as measured by satellites and aircraft.
Then they use complicated equations to predict what path a storm might travel and how strong it could get.
An AI model operates very differently: it relies more on history.
An AI model operates very differently: it relies more on history.
It identifies patterns by analyzing thousands of past storms along with atmospheric data.
"The AI models don't really know anything about the atmosphere other than the current picture," Lowry says.
"The AI models don't really know anything about the atmosphere other than the current picture," Lowry says.
"Beyond that, it's not leaning on the governing equations of the atmosphere, the physics, to make predictions."
Most of the people designing and developing these models aren't even meteorologists, Lowry notes: "They're software engineers, and they're developing neural networks."
A "neural network" is a set of computer algorithms modeled on the human brain and trained to find patterns in massive amounts of data.
As a result, the models can seem like a "black box," Lowry says: "We don't know why they might get a forecast right. We also don't know why they might get a forecast wrong."
It can be unsettling for meteorologists to depend on a system they don't totally understand – especially during storms when millions of people might be relying on their forecasts.
"It makes us a little uncomfortable, because we like to know why you're getting the right answer," McNoldy says.
Most of the people designing and developing these models aren't even meteorologists, Lowry notes: "They're software engineers, and they're developing neural networks."
A "neural network" is a set of computer algorithms modeled on the human brain and trained to find patterns in massive amounts of data.
As a result, the models can seem like a "black box," Lowry says: "We don't know why they might get a forecast right. We also don't know why they might get a forecast wrong."
It can be unsettling for meteorologists to depend on a system they don't totally understand – especially during storms when millions of people might be relying on their forecasts.
"It makes us a little uncomfortable, because we like to know why you're getting the right answer," McNoldy says.
"And I feel like we don't – or at least I don't know why."
Not all AI models have performed well, though.
Not all AI models have performed well, though.
There are lots of unproven weather forecasting AI models out there, Lowry says.
The two that have so far proven to be the most accurate, despite some pitfalls, are the models developed by Google DeepMind and by the U.K.-based European Centre for Medium-Range Weather Forecasts (ECMWF).
The two that have so far proven to be the most accurate, despite some pitfalls, are the models developed by Google DeepMind and by the U.K.-based European Centre for Medium-Range Weather Forecasts (ECMWF).
People retrieve belongings from a collapsed boat house in the aftermath of Hurricane Isaias in Destin, Fla., Saturday, Oct. 10, 2026. An AI model from Google Deepmind beat out traditional physics-based models at predicting the path of the storm, though it didn't do as well in forecasting the storm's strength.
Gerald Herbert/AP
Relying on a "blend" of models
Forecasters at the National Hurricane Center use several proven models – both AI and traditional – when putting together their official forecasts.
"Every storm has different forecast challenges," says NHC Director Michael Brennan.
"That's why we take that consensus approach and use that blend of the best models."
And it's still humans who have to interpret the results of those models before deciding what to tell the public.
The NHC provides updates every six hours during active hurricanes.
And it's still humans who have to interpret the results of those models before deciding what to tell the public.
The NHC provides updates every six hours during active hurricanes.
The models by themselves can sometimes show drastic changes from one update to the next. Forecasters have to translate those shifting predictions into clear guidance that people can act on.
"That's one thing that the human forecaster provides a lot of, is continuity," Brennan says.
"That's one thing that the human forecaster provides a lot of, is continuity," Brennan says.
"We're using our expertise and our understanding to translate that hurricane forecast into the hazard space, and that's what people really need to pay attention to."
Scientists expect rapid advances
The Google DeepMind model still needs some work, Brennan says.
While it topped other models in tracking the path of Isaias, it didn't do as well as the NHC's blended forecast in determining the intensity, or strength, of the storm.
Brennan's team gives regular feedback to Google DeepMind to help improve the model.
Brennan's team gives regular feedback to Google DeepMind to help improve the model.
Google didn't respond to questions from NPR for this story.
The "future of forecasting has always been changing," Brennan says.
The "future of forecasting has always been changing," Brennan says.
"It's never just one thing that's driving those improvements."
It wasn't that long ago when forecasters couldn't predict how strong a storm would be even within a couple of days, he notes.
It wasn't that long ago when forecasters couldn't predict how strong a storm would be even within a couple of days, he notes.
That changed, because of advances in prediction models and better data collected by satellites and planes.
"Errors are about half of what they were, say, at the time of [Hurricane] Katrina, in terms of intensity forecasting," Brennan says.
The exciting thing about AI, says meteorologist Michael Lowry, is how quickly it's improving.
"Errors are about half of what they were, say, at the time of [Hurricane] Katrina, in terms of intensity forecasting," Brennan says.
The exciting thing about AI, says meteorologist Michael Lowry, is how quickly it's improving.
AI updates take only months or even weeks, unlike traditional models, which are updated annually at best.
"They are making advancements in skill and hurricane predictions over the course of a year that would have taken us 10 years, a decade or more, to make," Lowry says.
But even with AI coming in fast and strong, Lowry says he's not worried about artificial intelligence replacing people.
"For me, I don't feel some sort of existential crisis with AI," Lowry says.
"They are making advancements in skill and hurricane predictions over the course of a year that would have taken us 10 years, a decade or more, to make," Lowry says.
But even with AI coming in fast and strong, Lowry says he's not worried about artificial intelligence replacing people.
"For me, I don't feel some sort of existential crisis with AI," Lowry says.
"If anything, I think it is a boon for our business."
Links :
Links :
. Mid Mediterranean map
Where it started: the Western Indian Ocean Maritime Emergency Contact Map, Edition 1, July 2025.
Example: Tunis MRCC, with its contact details
Example of the operations layer: Frontex Themis drawn on the map, with its contact at the top of the screen.
Munitions on the seabed just off Lubeck, Germany (GEOMAR)