Google WeatherNext 3 Uses Live Satellite Data for Sharper AI Forecasts

Google Research and DeepMind have released WeatherNext 3, an AI weather model that learns directly from real-time satellite data rather than relying on traditional physics simulations. Previous AI systems, including WeatherNext 2, were trained on numerical weather prediction data that Google says can carry a six-hour delay. WeatherNext 3 instead generates a fresh forecast every hour, which is especially relevant for fast-changing conditions such as rainfall, temperature shifts, and developing storms.

The model produces forecasts at multiple resolutions, including five kilometers for temperature and humidity, ten kilometers for other surface variables, and 25 kilometers for atmospheric values like wind speed. That is about five times sharper than WeatherNext 2’s 25-kilometer grid with six-hour intervals. Google says WeatherNext 3’s finer grid and weather-station training help capture coastlines, valleys, mountains, and other local terrain details, with regions in Latin America, Africa, and the Asia-Pacific expected to benefit most where high-cost regional models have been less available.

WeatherNext 3 was trained for precipitation using NASA’s satellite-based IMERG dataset and Google’s own global precipitation analysis built on satellite radar. The article reports improvements in Continuous Ranked Probability Score for medium-range global forecasts, including up to 60 percent over IMERG, 30 percent over MRMS, and ten percent over rain gauges at short lead times. The model also predicts wind speeds at 100 meters and provides cloud cover and solar irradiance values, giving wind farms, solar installations, and grid operators more targeted data for planning supply and demand.

WeatherNext 3 now powers weather features in Google Search, the Gemini app, Google Maps, the Google Maps Platform Weather API, and Google Earth Engine. Researchers, developers, and businesses can query the data through BigQuery and Earth Engine or download it in bulk from Google Cloud Storage. Google says users planning a day or more ahead should see up to 50 percent more accurate precipitation forecasts, while still pointing people to national weather services for official forecasts, severe weather warnings, and safety advisories.