Google launches WeatherNext 3, an AI weather model with five times sharper resolution than its predecessor
The updated model incorporates real-time satellite data to produce hourly forecasts at up to 5-kilometer resolution, promising up to 50 percent more accurate precipitation predictions at least a day in advance.
What happened
Google announced WeatherNext 3, an updated AI weather forecasting model that produces global forecasts at five times higher resolution than its previous model, WeatherNext 2. The new model generates forecasts every hour based on the most recent satellite observations and can visualize temperature, moisture, and other variables at up to 5-kilometer resolution, compared to WeatherNext 2's 25-kilometer grid updated every 6 hours. Google reports precipitation forecasts are up to 50 percent more accurate when looking at least a day in advance. WeatherNext 3 is designed to handle fast-moving weather systems and fills gaps in regions with fewer ground-based rain gauges, particularly outside the US and Europe. The model also forecasts renewable energy generation, including wind speed predictions at 100 meters altitude. WeatherNext 3 has been integrated into Google Search, Maps, Gemini, and other Google products, and Google has collaborated with the US National Hurricane Center and agencies in Asia.
Context
AI weather models like WeatherNext 3 represent a departure from traditional physics-based supercomputer simulations, which involve solving complex atmospheric equations but incur time delays. By recognizing patterns in historical and live observational data, AI models can generate faster predictions. The incorporation of real-time satellite data in WeatherNext 3 addresses a key limitation of previous AI models and fills forecasting gaps in regions with limited ground infrastructure. The model's ability to predict renewable energy generation reflects the increasing energy demands of AI systems themselves; Google framed this capability as important given rising data center energy consumption. However, AI models are expected to complement rather than replace physics-based forecasting, with weather agencies typically consulting multiple prediction sources before issuing warnings.