Researchers at Google DeepMind have presented an analysis showing that their artificial intelligence model provides more accurate hurricane forecasts than traditional approaches. According to an article in Nature, published in August 2026, the AI outperforms other widely used models by a day or more.
The analysis covered hundreds of hurricane forecasts from 2023 to 2025. The DeepMind model demonstrates a significant advantage in accuracy, especially in predicting storm track and intensity. This is also confirmed by the example of Hurricane Melissa, which struck Jamaica in October 2025.
The National Hurricane Center in Miami has already integrated AI into its work in 2025. Thanks to this, specialists were able to more accurately predict Melissa's rapid intensification and issue warnings several days in advance. The success was not a fluke, DeepMind asserts.
"This is a big deal," notes Amy McGovern, professor of meteorology at the University of Oklahoma. "If you can warn people about a major hurricane a day earlier, they have more time to evacuate and prepare."
Kerry Emanuel, professor at the Massachusetts Institute of Technology, welcomes the progress but emphasizes that final decisions must still be made by humans. AI will become another tool alongside satellite data and hurricane hunter aircraft.
The DeepMind model is trained on extensive datasets and is capable of accounting for complex factors that traditional physical models struggle with. The question is how quickly meteorological services worldwide will begin to actively use such developments in their daily work.
The publication in Nature underscores that the benefits of AI are already visible in practice. This paves the way for more reliable warnings that could help save lives and property in vulnerable regions.
