Google Unveils Advanced Weather Forecasting AI Model with Hourly Updates
Alphabet Inc., the parent company of Google, has launched WeatherNext 3, its most advanced global weather forecasting AI model. This new model integrates real-time satellite data and provides hourly forecast updates at higher resolution than its predecessor.
The WeatherNext 3 model generates forecasts at multiple spatial resolutions, visualizing key surface variables at 5-kilometer resolution, other surface variables at 10 kilometers, and atmospheric variables at 25 kilometers. This is roughly five times sharper than the previous WeatherNext 2 model, which produced forecasts on a 25-kilometer grid in 6-hour increments.
The new model ingests live global geostationary satellite data and trains directly on sparse weather station observation data, allowing it to generate a new forecast every hour. Independent evaluations by Brightband show precipitation forecasting improvements of up to 60% against NASA's IMERG baseline, 30% for MRMS, and 10% against rain gauge measurements for early lead times.
WeatherNext 3 also introduces predictions for renewable energy production, including 100-meter wind speeds for wind-energy output and high-resolution cloud cover and sun radiation levels for solar farms. Google is integrating WeatherNext 3 across various platforms, including Google Search, Gemini app, Google Maps, Google Maps Platform Weather API, and Google Earth Engine.