Hurricane Isaias: the US Hurricane Centre aligns its forecasts with Google DeepMind’s IA model

From the very first bulletin, the National Hurricane Centre based its forecasts on WeatherNext 3 rather than on physical models, and announced a Category 2 storm when the system did not even have a name.

Since the formation of Isaiah, the forecasters of the National Hurricane Center (NHC) in Miami made a new choice: to set their official forecasts on Google DeepMind's AI model, the latest version of which is called WeatherNext 3, rather than on conventional physical models such as the European model or the US GFS. From the first bulletin, when the storm was not yet named, the NHC announced a category 2, when physical models hesitated and predicted a weaker system on another trajectory. So far, this bet is true, CNN reports.

Trust won with Melissa

This confidence comes from experience: last season, the NHC announced that Hurricane Melissa would reach category 5 when she was only in category 1, almost three days before her devastating arrival in Jamaica, relying mainly on DeepMind. The centre's director, Michael Brennan, highlights the greater coherence of IA models from one execution to the next, and the fact that the NHC helped Google train the model on its own trajectory analyses, making it better on intensity than the first IA models, especially good for trajectory.

Brennan insists, however, that AI remains a tool among others, combined with physical models and human expertise. Isaias, the first hurricane in an Atlantic season held back by a super El Niño, was heading for the northeast coast of the Gulf of Mexico.