National Hurricane Center tests AI forecasts from Google DeepMind

In short: The National Hurricane Center has been using AI-based weather forecasts, including Google DeepMind’s WeatherNext, as extra guidance when predicting hurricane paths.

What's going on

Reports say forecasters at the National Hurricane Center are looking at AI guidance to help judge where a hurricane might travel. One example mentioned is Hurricane Isaias, where AI forecasts were used alongside the usual tools.

Traditional weather models work like a big physics calculator. They simulate the atmosphere by crunching equations about air, heat, and moisture. WeatherNext works differently, it learns patterns from past weather data and applies them to what is happening now (like learning from many old storms to make a best guess about a new one).

The AI system can also create many possible forecasts quickly. You can think of it like generating several different “what if” routes for the storm, so forecasters can compare where it might go and how strong it might get. In some cases, reports say the AI pointed to a different track than traditional models, although the forecasts moved closer together as the storm got nearer.

What to watch

The key point from forecasters is that AI is not replacing meteorologists. Humans still review satellite images, aircraft measurements, and standard models, then choose and explain the official forecast. Watch for how often AI forecasts beat or match older models, and whether the public forecast becomes clearer about uncertainty, meaning how wide the range of possible storm paths could be.

Source: NYTimes