In October 2025, Google DeepMind's AI model WeatherNext correctly forecast that a developing storm would become Hurricane Melissa and strike Jamaica as a Category 5 system, giving five days' warning with 80 percent confidence. This marked the first time the US National Hurricane Center predicted a Category 5 hurricane while the storm was still at Category 1 strength. The model, developed by DeepMind and Google Research, was tested on historical data before being used in real-time forecasts. According to a paper published in Nature, WeatherNext provides cyclone predictions with a full day more lead time than traditional models, making its three-day forecasts as accurate as others' two-day ones. The AI was trained on broad weather data to overcome the scarcity of cyclone-specific records, allowing it to predict both storm tracks and intensity—something previous models struggled with. Mike Brennan, director of the US National Hurricane Center, said even a few extra hours can significantly improve evacuation planning and resource deployment. Kate Musgrave of the Cooperative Institute for Research in the Atmosphere, who co-authored the study, explained that predicting intensity requires fine-scale local data that global models often miss. While earlier AI systems could forecast storm tracks, they failed at predicting rapid intensification, which WeatherNext managed in the case of Hurricane Melissa.
The AI's successful prediction of Hurricane Melissa's intensity highlights a shift in how storm forecasts are made, relying on machine learning to fill gaps in traditional modeling. This could change emergency response timelines, especially in vulnerable island nations facing stronger storms. The model's ability to use general weather data to compensate for limited cyclone records suggests a new path for climate AI development.
Editorial note: AI-assisted opinion, not established fact. Full disclaimer →
