AI Models Change Hurricane Forecasting for the Better

Meteorologists are watching a wave of artificial intelligence that's changing hurricane preparations. NPR's Jessica Meszaros reports that AI models are rapidly reshaping forecasting, with Google DeepMind's system already outperforming many long-trusted, physics-based models at predicting storm tracks, including this month's Hurricane Isaias, which it correctly pegged to hit Florida's western Panhandle while traditional models pointed toward Louisiana. Experts like Miami hurricane specialist Michael Lowry now call DeepMind the "gold standard" for track prediction, and the National Hurricane Center has begun folding it into official forecasts.

The shift isn't seamless. AI systems function as "black boxes," relying on patterns from thousands of past storms rather than atmospheric equations, which leaves meteorologists uneasy about why a forecast is right—or wrong. AI has also lagged on intensity predictions, and not all experimental models perform well. As NHC Director Michael Brennan explains, forecasters still depend on a blend of human judgment, traditional models, and vetted AI. The Hurricane Center's forecasts pointed out the differences between the models as Isaias approached, per the New York Times. For the full picture on how this technology is changing hurricane season, read the original piece at NPR.