Google DeepMind AI Improves Hurricane Prediction Accuracy

Google DeepMind open-sourced an AI model, WeatherNext, capable of earlier and more accurate hurricane predictions.

Google DeepMind has open-sourced an AI model named WeatherNext, which is designed to provide earlier and more accurate hurricane warnings. The model demonstrates a significant advancement in forecasting cyclones, offering predictions days earlier than existing systems. Researchers from Google DeepMind and Google Research developed WeatherNext, which can predict a storm's track and intensity using lower-resolution weather data.

The WeatherNext model has shown a substantial leap forward in predicting cyclones. Its predictions three days out are as accurate as previous models' predictions two days out, effectively providing an extra day of lead time. This additional time can be crucial for organizing evacuations, staging supplies, and moving resources in response to hurricane risks. The model's performance was consistent even when adopted into forecasters' operations.

Despite its effectiveness, researchers do not yet fully understand how WeatherNext achieves such accurate predictions with lower-resolution data. The model was trained to be proficient in both general weather and cyclone forecasting. This development is particularly significant as predicting both the track and intensity of hurricanes is critical, and earlier AI models struggled with intensity predictions.

Find the right AI tool for your business

Chat with Insta and get matched to the right tool in seconds.