Google DeepMind’s AI model called Hurricane Isaias’s Florida Panhandle landfall while physics-based models pointed toward Louisiana

Hurricane Isaias hit Destin, Florida, as a Category 2 storm on October 9. Google DeepMind’s AI model, which the National Hurricane Center has used in its official forecasts since 2025, pointed to the western Florida Panhandle from the start, while conventional models first sent the storm toward southeastern Louisiana. A preliminary HurriLab analysis found it timed landfall within a few hours, three days out.

In early October, Hurricane Isaias was crossing the Gulf of Mexico, and the forecast models disagreed about where it would come ashore. The traditional physics-based models leaned west, toward southeastern Louisiana or coastal Mississippi, with a slower storm. On October 9, Isaias made landfall near Destin, on the Florida Panhandle, as a Category 2 hurricane.

The clever bit

Google DeepMind’s model skips the physics equations that traditional models solve. It found patterns in thousands of past storms and the atmospheric data around them, then takes the current picture and projects the storm forward. Forecasters at the National Hurricane Center (NHC) never lean on a single model, so it sits alongside the physics models. The NHC started folding it into official forecasts last year, as part of a blend of the best performers.

From its first runs on Isaias, the DeepMind model pointed at the western Florida Panhandle. A preliminary verification by HurriLab, published by hurricane specialist Michael Lowry, found it was closest to the real landfall spot at Destin and timed landfall within a few hours, three days out. The physics-based European and American ensembles put landfall too far west and a full day or more too late. Over the 2025 season, the NHC’s own verification report found the DeepMind ensemble slightly beat its official Atlantic track forecasts from 12 to 72 hours.

Google DeepMind, in particular, has really proven itself the last few years as the gold standard model.

— Michael Lowry, hurricane specialist, WPLG-TV Miami, via npr.org

Hooray

A model that learned from old storms pointed at the right stretch of coast from day one, and hurricane forecasters now have it in the blend they use to warn people when to get ready and when to leave.

What comes next

NHC Director Michael Brennan’s team gives Google DeepMind regular feedback to improve the model. AI models can be updated in weeks or months, while traditional ones get updated once a year at best, and Lowry says the gains in a single year would have taken a decade or more. The NHC sticks with a consensus of the best models, because, as Brennan puts it, every storm has different forecast challenges.

How solid is it?

In use: Being used outside a research-only setting. This describes deployment, not proof of effectiveness.

Real-world forecast checked against a real landfall. NPR (October 26, 2026) reports the model’s Isaias call and the National Hurricane Center’s use of it, with named comments from NHC Director Michael Brennan and hurricane experts Michael Lowry and Brian McNoldy. A preliminary HurriLab verification, published by Lowry on October 16, compares every major model on Isaias’s track and landfall. The NHC’s own 2025 verification report found the Google DeepMind ensemble mean slightly beat the official Atlantic track forecast at 12 to 72 hours.

See for yourself.

Selected for Oct 29, 2026. Reviewed Oct 29, 2026, 5:52 AM MT. Source dates may be earlier than this edition.