Google’s Hurricane Model Performed Well This Season

An anonymous reader cites a report from Ars Technica highlighting the impressive performance of Google DeepMind’s Weather Lab in cyclone track forecasting, which began in June. Preliminary analysis by Brian McNoldy reveals that DeepMind’s model significantly outperformed the US National Weather Service’s GFS model, achieving a five-day forecast error of 165 nautical miles compared to 360 nautical miles for GFS. Additionally, Google’s AI model often surpassed the official forecasts from the National Hurricane Center and other consensus models. This performance indicates a potential shift in reliance towards AI-based forecasting methods.

Want More Context? 🔎