WeatherNext 3 is also supposed to be much better at predicting precipitation.
Google DeepMind has released the WeatherNext 3, an updated version of its last AI-powered weather model that it claims offers even more detailed forecasts across key variables like temperature, moisture and wind speed. DeepMind notes the new model should be particularly useful to companies working in renewable energy production, because of its ability to forecast “100-meter wind speeds (roughly at turbine-height) for precise wind-energy output, alongside high-resolution cloud cover and sun radiation levels to help solar farms.”
WeatherNext 3’s ability to offer hourly and higher-resolution forecasts (in a 5-kilometer square grid rather than the 25-kilometer square grid of WeatherNext 2) is all due to the new data it’s trained on, DeepMind says. Rather than primarily relying on data from numerical weather prediction models, which include a six-hour lag, WeatherNext 3 incorporates live satellite data, which gives the model “a rich, continuously updating view of the atmosphere.” The model’s ingestion of sparse weather station data means it can also offer more detailed forecasts of variables like humidity, which can usually change dramatically from one kilometer to the next.
Combining precipitation data from NASA and Google’s own satellite analysis also means WeatherNext 3 is better at predicting things like rain and snow. DeepMind says the model produces “up to 50 percent more accurate precipitation forecasts — with the greatest improvements in regions where forecasts have historically been less reliable.”
WeatherNext 3 will power weather experiences in Google Search, the Gemini app and Google Maps, along with the Google Maps Weather API and the Google Earth Engine. If you’re curious about experimenting with the model yourself, you can give it a spin right now in the Google Weather Lab. If you want to take the next step and build something with Google’s model, the original WeatherNext model is also open-
