Why the National Hurricane Center bet big on AI to forecast Isaias

The National Hurricane Center placed a significant bet on AI by shaping its forecasts for Hurricane Isaias around Google DeepMind's hurricane model rather than traditional physics-based models like the European and GFS. This decision reflects the improving reliability of AI weather models and a remarkably fast shift toward routine use by human forecasters. The NHC forecast Isaias would become a Category 2 storm from its first advisory, a prediction that contradicted many physics-based models showing a weaker storm on a different track. If accurate, this will be a major victory for Google and give residents more time to prepare. The NHC's confidence stems from experience—last season it correctly anticipated Hurricane Melissa's rapid intensification nearly three days in advance based largely on DeepMind projections—and from its partnership with Google to train the model. NHC director Michael Brennan noted that AI models have shown more consistency between runs, and that DeepMind performed extremely well last season. With climate change enabling more rapid intensification, the NHC has shifted from a stair-step approach to making more aggressive early intensity forecasts, as seen with Melissa, Helene, and Milton. Brennan emphasized that AI is another tool in the toolbox, not a replacement, and that training on accurate hurricane-specific data has improved performance.

(Source：CNN)