WeatherNext
Our most advanced weather forecasting AI technology
WeatherNext, a new family of AI models from Google DeepMind and Google Research, produces state-of-the-art weather forecasts.
The challenge of global weather forecasting
The weather dictates how we dress in the morning, helps us generate green energy – and creates storms that threaten communities. In a world of increasingly extreme weather, fast and accurate forecasts are crucial.
But predicting the weather remains challenging. Weather systems are chaotic, involving massive amounts of data, and we can’t account for - or measure - every variable. Even small inaccuracies in measurement are quickly magnified.
AI models can help us make predictions that are faster and more reliable than those created with traditional methods. These AI models are already transforming how we forecast the weather, helping us make better decisions, and improving disaster response, grid reliability, and global food security.
We're sharing our models with scientists and forecasters, to accelerate their work and ultimately benefit billions of people around the world.
AI is bringing faster, more accurate weather predictions to the world.
WeatherNext models are faster and more efficient than traditional physics-based weather models and yield superior forecast reliability. The gains in forecast performance could enable better preparation to help save lives in the face of extreme weather events, and enhance the reliability of sustainable energy and supply chains.
WeatherNext model family
WeatherNext, our state-of-the-art AI-based weather forecasting technology, from Google DeepMind and Google Research.
WeatherNext Graph
WeatherNext Graph represents a significant step forward for AI weather prediction. It offers more accurate and efficient deterministic forecasts (i.e. single prediction), compared to the best deterministic systems in use today.
Why use this model?
For applications where efficiency and temporal resolution are paramount, and deterministic forecasts are easier to work with.
Forecast per time and location
1 single weather forecast
Temporal resolution
6 hour
Lead time
10 day
WeatherNext Gen
WeatherNext Gen accurately generates an ensemble forecast (i.e. a range of likely future weather scenarios), better than current ensemble models most widely used today. This helps decision-makers better understand weather uncertainties and risks of extreme conditions.
Why use this model?
Our latest model, with a longer lead time. Probabilistic forecasts capture a wider range of possible weather trajectories, ideal for aiding applications like cyclone tracking.
Forecast per time and location
Ensemble of up to 50 weather forecasts
Temporal resolution
12 hour
Lead time
15 day
Access WeatherNext on Google Cloud
Live forecasts four times per day and historical data are accessible for research, analysis, and to inform decision-making.
