DeepMind Says Its AI Can Predict Hurricanes Earlier Than Everyone Else
Google/DeepMind
Google DeepMind
Google Research
Google DeepMind and Google Research have developed an AI model called WeatherNext that can predict hurricanes with unprecedented accuracy, giving forecasters an extra day of lead time. The model, detailed in a Nature paper, successfully predicted Hurricane Melissa's landfall in Jamaica five days in advance as a Category 5 storm. The researchers are open-sourcing the models to aid further research.
In October 2025, Hurricane Melissa formed over the Caribbean Sea, and the WeatherNext AI model, developed by Google's DeepMind and Google Research, predicted five days before landfall with 80 percent confidence that the storm would hit Jamaica as a Category 5 hurricane. This early warning helped communities prepare, as the hurricane caused flooding and landslides across Jamaica. A paper published in Nature shows that WeatherNext can predict cyclones with unprecedented accuracy, giving forecasters on average a day more lead time than existing models. The model uses lower-resolution atmospheric data than traditional models, yet performs better, surprising researchers who admit it is a 'black box' but provides signals for physicists. WeatherNext generates a range of potential scenarios for each storm, now up to 1,000 per storm, aiding forecasters. Mike Brennan of the US National Hurricane Center called the model a valuable tool but emphasized it is one of many and that human expertise remains critical. Google DeepMind has open-sourced the WeatherNext models to allow researchers to improve them and gain insights into cyclone behavior.
Source: Wired AI —
original
