Rahul Somvanshi
Google has launched WeatherNext 3, an AI weather forecasting model that uses real-time satellite observations alongside historical weather analysis. Compared with WeatherNext 2’s six-hour forecast increments, WeatherNext 3 generates hourly forecasts using live global geostationary satellite observations and weather-station observation data used in training.
WeatherNext 3 can represent key surface variables such as temperature and moisture at 5-kilometre resolution, which means it can pick up on weather changes affecting smaller areas. This matters for people deciding whether to carry an umbrella, farmers planning their work, and companies managing wind and solar power plants.
The system is getting integrated into Google Search, Google Maps, Gemini, and Google Cloud. Google reports that independent testing by Brightband confirms the model’s accuracy improvements, and its earlier WeatherNext Cyclones research reported an average one-day lead-time advantage for cyclone track, intensity and wind-structure forecasting, which points to better hurricane and tropical storm prediction.
