Google has announced the rollout of its updated AI weather model, WeatherNext 3, which aims to enhance the accuracy of rain and snowfall predictions. This new model is designed to provide forecasts with what the company describes as “unprecedented resolution,” producing a global picture that is five times sharper than its predecessor, WeatherNext 2.

Advancements in Forecasting Technology

According to Google, the WeatherNext 3 model leverages real-time weather observations to improve its predictive capabilities. Samier Merchant, a research engineer at Google Research, noted, “One of the main developments is for [WeatherNext 3] to go beyond what data most global AI models train on. We’re able to leverage fresher and richer observational data sets.” This approach allows the model to make faster predictions by recognizing patterns in historical weather data, contrasting with traditional forecasting methods that rely on supercomputers simulating atmospheric physics.

Enhanced Resolution and Speed

The new model incorporates live satellite data, enabling it to generate forecasts every hour based on the latest observations. This is a significant improvement over WeatherNext 2, which produced forecasts every six hours on a 25-kilometer grid. WeatherNext 3 can visualize certain variables, such as temperature and moisture, at a resolution of up to 5 kilometers. This increased speed and resolution are particularly beneficial for predicting rain and snow from fast-moving weather systems, especially in regions with fewer ground-based rain gauges.

Google claims that users can expect precipitation forecasts to be up to 50 percent more accurate when looking at least a day in advance, particularly in areas outside of the US and Europe where traditional forecasting methods may fall short.

Integration with Renewable Energy Forecasting

In addition to weather predictions, WeatherNext 3 is designed to support renewable energy generation forecasts, including wind speed predictions at heights relevant to turbine operation. Ferran Alet, a research scientist at Google DeepMind, emphasized the importance of making renewable energy appealing as global energy needs continue to rise.

The WeatherNext 3 model is now integrated into various Google products, including Search, Maps, and Gemini. Google has collaborated with the US National Hurricane Center and other agencies in Asia to enhance weather forecasting using its AI models. While AI models like WeatherNext 3 are becoming increasingly accurate, they are expected to complement traditional physics-based models, as weather agencies typically consider a range of predictions to issue warnings.

For further details, you can read the full article on The Verge.