Google Unveils WeatherNext 3 — AI Weather Forecasting with Higher Accuracy and Hourly Updates

Alphabet's subsidiary Google has introduced WeatherNext 3 — its most advanced artificial intelligence (AI) model for global weather forecasting to date. The new system integrates real-time satellite data and updates weather forecasts every hour, offering significantly higher resolution compared to its predecessor, WeatherNext 2.

WeatherNext 3 generates forecasts at several levels of spatial scale. Key surface indicators are visualized at 5-kilometer resolution, other surface parameters at 10 kilometers, and atmospheric variables at 25 kilometers.

As a result, the new model creates a picture of global weather with roughly five times greater detail than WeatherNext 2, which processed forecasts on a 25-kilometer grid once every six hours.

WeatherNext 3 uses real-time satellite data

WeatherNext 3 receives direct data from geostationary satellites around the world and uses relatively sparse observational data from weather stations directly for training the model. This allows the system to generate a new forecast every hour.

According to independent evaluations conducted by Brightband, the new model has achieved significant improvements in precipitation forecasting. In early forecast periods, WeatherNext 3's results improved by up to 60% compared to NASA's IMERG baseline model, by 30% compared to MRMS, and by 10% compared to rain gauge data.

The AI model also forecasts renewable energy production

One of the notable new features of WeatherNext 3 is its ability to forecast renewable energy production.

The model can forecast wind speed at a height of 100 meters, which helps estimate energy output at wind power plants. It can also forecast cloud cover and solar radiation at high resolution, which is important for planning the operation of solar power plants.

According to Google, WeatherNext 3 will also improve forecasting capabilities for Latin America, Africa, and the Asia-Pacific region, which have historically had less access to high-resolution weather forecasting services.

WeatherNext 3 is being integrated into Google's products

Google is beginning to use the new model today across several of its products and platforms, including:

• Google Search;
• the Gemini app;
• Google Maps;
• Google Maps Platform's Weather API;
• Google Earth Engine.

The company is also making global weather forecasts available in BigQuery and Earth Engine, and is offering users the ability to download large volumes of data via Google Cloud Storage.

The rollout of WeatherNext 3 marks another significant step in Google's use of AI — this time in weather forecasting, where more frequent updates and higher resolution are especially important both for everyday users and for industries such as energy, agriculture, and others.