Google's AI-Driven Weather Forecast Upgrade Boosts Accuracy

Google has upgraded its weather forecasting system using AI to improve accuracy by up to 50% in predicting rain, according to internal benchmarks. The new model, called WeatherNext 3, is already integrated into Google Maps and Search, replacing the previous system.

The key change lies in how the model receives data: it ingests live geostationary satellite imagery every hour, bypassing the six-hour lag of traditional weather forecasting systems. This allows for more accurate predictions at a resolution of 5 kilometers, compared to the previous 25-kilometer grid.

WeatherNext 3 outperforms not only Google's previous model but also traditional numerical weather prediction systems operated by the U.S. National Weather Service and European meteorological agencies, according to internal testing. The system also provides more detailed forecasts for wind speeds, cloud cover, and solar radiation levels.

While Google is not alone in developing AI-based weather models, Microsoft, Nvidia, and European research institutions have published competing approaches, the company's announcement did not name specific rivals or provide external validation of WeatherNext 3's accuracy gains.