Google has announced the release of its AI weather forecasting model, "WeatherNext 3." The most significant feature of this update is the ability to directly incorporate real-time satellite data in addition to traditional analysis data. This reduces the time lag between current weather conditions and forecast generation, enabling predictions at one-hour intervals.
Model ReleasesGoogleWeatherNext 3
Google Announces WeatherNext 3: Weather Model with Improved Accuracy via Satellite Data Integration
This article is a translation. Read the Japanese original
WeatherNext 3 offers a substantial improvement in spatial resolution compared to previous models. While the predecessor, WeatherNext 2, provided updates every six hours at 25km intervals, WeatherNext 3 delivers forecasts with a resolution of up to 5km. Furthermore, physical information such as terrain has been partially integrated to calculate surface temperatures and dew point temperatures for specific locations, achieving more detailed predictions.
According to Google's announcement, the forecast accuracy for upper-atmosphere conditions has improved by approximately 5% compared to the previous model, which is equivalent to extending the accurate forecast lead time by about six hours. Google also stated that the forecast accuracy for surface temperatures at specific points has improved by up to 30%. The model is integrated into Google services including Google Search, Gemini, and Maps, and is expected to be utilized in fields such as agriculture and renewable energy.
Sources
- Update to Google’s AI weather model improves forecast accuracy （Ars Technica AI, 2026-09-09）
- Google Official Blog