Predicting regional gray swans via translocation: AI weather models and Dubai's unprecedented 2024 rainfall.

AI weather models accurately forecast the unprecedented 2024 Dubai storm, with GraphCast and AIFS predicting rainfall up to 8 days in advance, highlighting their potential for extreme event prediction.

• Why it matters: Predicting rare and extreme weather events remains a challenge, yet these "gray swans" can have devastating impacts, making it crucial to improve forecasting capabilities for such anomalies.
• What they did: Researchers analyzed forecasts from GraphCast, AIFS, and FuXi, focusing on their ability to predict the record rainfall, and investigated the mechanisms behind their success, including "translocation" learning from similar events.
• The result: Findings reveal that models succeed mainly through "translocation"—learning from stronger, comparable events elsewhere—offering insights to enhance AI models for better regional extreme weather prediction.