Edu News

UAlbany Study Puts Google’s AI Hurricane Model to the Test

TechnologyK Puspa20 Nov 2026

ALBANY, N.Y., Nov 20: A week before Hurricane Melissa slammed into Jamaica last fall, the forecast models couldn’t agree on where it was headed or how strong it would get.

One got it right — an AI model from Google DeepMind, which called both the storm’s path and its Category 5 strength. 

A new study published by University at Albany researchers is offering one of the first independent looks at how well Google’s AI hurricane model predicts storms, with a specific focus on intensification.

The study, published in Geophysical Research Letters, evaluated intensity forecasts from Google’s WeatherNext Cyclone model for every tropical cyclone between June 2022 and May 2026. 

“We decided to examine Google’s AI hurricane model as it has quickly become one of the most talked about tools in weather forecasting,” said Brian Tang, a professor in UAlbany’s Department of Atmospheric and Environmental Sciences, who co-led the study.

“Independent checks matter because they give the public more confidence when a model’s developers claim it has improved. They also offer an objective look at both strengths and weaknesses.”

Predicting the next storm

Tang, along with recent UAlbany graduates Jacob Vile and Nour Elgalad, analyzed a total of 385 tropical cyclones worldwide.

To test Google’s AI hurricane model, the researchers compared the tool’s predicted wind speeds with what each storm actually recorded and then checked its performance against official forecasts from the National Hurricane Center and Joint Typhoon Warning Center, along with various other tools that are commonly used in forecasting tropical cyclones. 

They also compared against two tools built to flag rapid intensification (when a storm’s winds jump at least 35 mph in 24 hours).

The AI model was more accurate than NOAA’s state-of-the-art hurricane model at every point from one to five days out and performed on par with human forecasters beyond the first day. It was also about as reliable as the leading tool for predicting rapid intensification over the next 24 hours, with fewer false alarms.

“Rapid intensification is still really hard to predict, especially when and how extreme it will be, because it depends on a lot of factors lining up that we don’t fully understand or observe well,” Tang said. “AI models may be better at connecting those dots. In particular, we found Google’s model is less likely to predict rapid intensification that never happens.”

The study also flagged a few weak spots. The model tends to underestimate how strong storms will get, and that gap grows the further out it forecasts.

Its performance also varied by region (it was best in the North Atlantic and worst in the eastern North Pacific), and its range of possible outcomes was often too narrow, meaning the real storm intensity frequently landed outside the model range.

“The model was trained on fairly big-picture weather data and basic storm information, so it may be missing some of the finer details needed to predict the strongest, fastest-strengthening storms,” Tang said.

From class project to publication

The research started as a class project in Tang’s graduate course last year. Vile and Elgalad expanded the project into a full analysis and saw it through peer review.

“When I began this project almost a year ago, I had no idea it would become a published paper,” said Vile, who graduated from UAlbany in the spring. “This model kept coming up in our forecasting discussions. I proposed the project just to better understand some of its biases, but Professor Tang saw how useful the results could be to the broader scientific community.”

For Elgalad, the project was a crash course in the fast-moving world of AI weather prediction.

“The project tackled an important, ongoing topic in tropical cyclone forecasting, so it exposed me to the research happening in the field right now,” she said. “Having our work published is an incredible opportunity.”

“It’s a big deal for students to turn a class project into a published paper, especially on a topic as hot as this one,” Tang added.

What’s next

Google released a new version of its model this year (not included in the study), and other AI weather models are now in development.

Tang believes forecasters will benefit from a shared, transparent way to compare them.

 “AI is a powerful new tool, but it doesn’t replace forecasters,” Tang said. “Knowing where a model tends to fall short is exactly what helps forecasters decide when to lean on it and when to trust their own judgment.”