Aurora: AI Breakthrough Enhances Global Extreme Weather Forecasting

As climate-related disasters grow increasingly frequent, an international research team led by the University of Amsterdam has developed a powerful AI model named Aurora. Trained on over one million hours of Earth system data, Aurora is designed to efficiently predict air quality, ocean waves, and extreme weather events, with the potential to significantly improve global disaster preparedness and response.

Unlike traditional forecasting methods that rely on complex numerical models and costly supercomputing resources, Aurora uses machine learning to dramatically reduce computational demands. This makes high-quality forecasts more accessible, especially in regions with limited infrastructure. Developed collaboratively with multiple institutions, the model boasts a 1.3-billion-parameter architecture and has demonstrated exceptional performance across various benchmarks: outperforming traditional models in 74% of air quality predictions, achieving 86% accuracy over numerical simulations in wave forecasts, and surpassing seven leading forecasting agencies in all tropical cyclone tests. In 92% of scenarios, Aurora also led in high-resolution weather predictions, particularly excelling in extreme events.

Researchers emphasize that Aurora can shorten development cycles and enable resource-limited nations and organizations to quickly access critical climate data. The model is openly available, allowing users to fine-tune it for specific tasks with their own data—no retraining from scratch required. Future applications may include flood and wildfire prediction, agricultural yield forecasting, and renewable energy management.