Google WeatherNext 3 is the company’s updated AI weather model, designed to deliver more accurate forecasts, particularly when predicting rain and snowfall. The rollout was announced today, with Google stating that the new system can produce forecasts with what it describes as unprecedented resolution.
According to the company, WeatherNext 3 generates a global picture that is five times sharper than its previous model. It achieves this by learning from real-time weather observations, allowing it to refine its predictions across the globe.
How the New Model Works
The key development behind WeatherNext 3 is its ability to move beyond the data that most global AI models typically rely on for training. Samier Merchant, a research engineer at Google Research, noted that the system is able to draw on a wider range of information than earlier approaches, expanding the scope of what the model can learn from.
By incorporating live observations rather than depending solely on historical datasets, the model can build a more detailed view of current conditions. This approach underpins the improvements Google says it has achieved in forecast resolution and accuracy.
Improved Rain and Snow Predictions
The improvements are said to be most noticeable when forecasting precipitation, including both rain and snowfall. Predicting these conditions has historically been challenging for weather models, and Google positions the sharper resolution of WeatherNext 3 as a step towards addressing that difficulty.
The updated model represents Google’s continued effort to apply machine learning techniques to weather forecasting, with the aim of producing clearer and more reliable global forecasts.
Source
Image: theverge.com
