Traditional weather forecast models are physics-based and incorporate data such as upper-air measurements, weather station information and satellite observations. This data is processed by supercomputers.
AI models use historical weather data to learn patterns in the atmosphere.
"The saying is, 'All models are wrong; some are useful,' which is dead on," notes Alabama Weather Network's James Spann.
By comparison, he says these AI models, especially Google DeepMind, have proved to be "darn good" with tropical weather.
"They're on to something, and obviously, it's outperforming the physics-based models," he tells AFN.
With Isaias, the Google model very consistently showed landfall would be around Pensacola and Navarre in the Florida Panhandle. The physics-based models, meanwhile, predicted landfall would occur about 70 miles west, near the Mississippi-Alabama border.
"With the hurricane landfall, the point of landfall is important because the big impact will be, generally, in terms of storm surge, along and east of the center," Spann explains. "That's the onshore flow."
In simpler terms, if a hurricane makes landfall near Pensacola rather than farther west near the Mississippi-Alabama border, as Isaias did, the areas exposed to the most dangerous storm surge can be different.
Spann also commends the Google AI model for predicting the hurricane's intensity, even though the damage was greater on the west side of the storm in this case.
"This was complicated. This was not your standard tropical system," he says. "This thing was rapidly becoming cold core right as it moved inland. It was a mess … and that's one of the reasons we had the severe damage in Pensacola and up into South Alabama on the west side of the circulation center."
South Alabama was still in bad shape due to the storm.
