source: Google DeepMind: Introducing WeatherNext 3, our most advanced and accurate global weather AI model
level: technical
google deepmind and google research have introduced weathernext 3, a new ai weather forecasting model. it uses real-time satellite data instead of traditional physics simulations. the model produces hourly forecasts at up to 5-kilometer resolution, five times sharper than the previous version. it is now integrated into google search, gemini, maps, and cloud. the goal is to provide more accurate and localized weather predictions for users worldwide.
the model trains directly on live geostationary satellite mosaics and sparse weather station data. this reduces the six-hour lag common in numerical weather prediction models. for precipitation, weathernext 3 shows up to 60 percent improvement in continuous ranked probability score against imerg satellite data. it also forecasts 100-meter wind speeds and solar radiation for renewable energy planning. independent evaluations by brightband rank it as the most accurate global model to date.
weathernext 3 is available through google cloud, bigquery, and earth engine for researchers and developers. the update improves longer-term precipitation forecasts by up to 50 percent in some regions. it particularly benefits areas in latin america, africa, and asia-pacific that lack high-resolution forecasting due to supercomputing costs. the model also predicts cyclone tracks and station-level variables natively. this makes it useful for agriculture, supply chains, and emergency response.
why it matters: hourly, high-resolution forecasts can improve planning for renewable energy, agriculture, and disaster response, especially in underserved regions.
source: Google DeepMind: Introducing WeatherNext 3, our most advanced and accurate global weather AI model