Researchers develop hybrid AI model to predict rare extreme weather events

An international team of researchers from the United States and France has developed a new hybrid forecasting method designed to accurately predict rare, high-impact weather events, such as once-in-a-millennium heatwaves. The study, led by members of the Climate and Data Theory Group at the University of Chicago, was published in the journal Physical Review Letters.

Traditional weather forecasting relies on supercomputers that are reliable but resource-intensive, while newer AI-based models often struggle with extreme phenomena not present in their training data. The new hybrid approach bridges this gap by combining the speed and efficiency of artificial intelligence with the physical reliability of traditional models. According to Pedram Hassanzadeh, an associate professor of geophysical sciences, this method is specifically designed to simulate extreme events that pose the greatest societal risks while consuming significantly fewer computational resources.

This development aims to improve long-term disaster preparedness by enabling faster and more accurate assessments of rare climate probabilities. The research provides a scalable solution for meteorologists and climate scientists looking to modernize forecasting frameworks without sacrificing the rigor of traditional physics-based simulations.