Back to news
AI & Machine Learning
5d ago

DeepMind's WeatherNext Model Enhances Cyclone Forecasting Accuracy

Aug 8, 2026
AI Summary

DeepMind's WeatherNext model has achieved significant advancements in cyclone forecasting, providing an extra day of predictive accuracy compared to previous models. The technology, which is now open-sourced, aims to improve preparedness for severe weather events and support meteorological research worldwide.

  • The WeatherNext model enables accurate cyclone forecasts, offering an additional day of warning time.
  • Cyclones are highly destructive, causing over 700,000 deaths and $1.4 trillion in economic losses globally over the past 50 years.
  • The model was developed by Google DeepMind and Google Research in collaboration with the National Hurricane Center and other meteorological agencies.
  • WeatherNext achieved state-of-the-art accuracy in predicting cyclone track, intensity, and wind structure, equating to a decade's worth of meteorological progress.
  • It was tested on historical cyclone data and demonstrated a lead time advantage of more than 24 hours over existing models.
  • The model uses a combination of global weather dynamics and historical cyclone observations, trained on nearly 20 terabytes of data.
  • WeatherNext Cyclones can generate a 15-day forecast in under a minute, capturing a wide range of potential cyclone scenarios.
  • The model operates effectively at a lower spatial resolution than traditional models, surprising researchers with its accuracy.
  • The code and model weights for WeatherNext are being open-sourced to facilitate further research and operational forecasting.
  • The Weather Lab platform has been updated to visualize cyclone forecasts and global weather predictions using WeatherNext technology.
  • The initiative aims to foster collaboration among researchers and meteorological agencies to enhance weather forecasting capabilities and community resilience.
deepmindweather forecastingcyclonesai modelbreakthrough