Traditional forecasting still beats AI for the most extreme weather

AI weather forecasting models, while faster and more precise for many tasks, fail to accurately predict extreme weather events. Research indicates that these models, including GraphCast and Pangu-Weather, often underestimate high temperatures and struggle with extreme wind and cold predictions. Their reliance on historical data limits their ability to forecast unprecedented events. In contrast, traditional physics-based models utilize complex mathematical representations of the physical world, allowing them to adapt more effectively to new conditions and extreme weather scenarios.

""They do perform well on a lot of tasks, but for very extreme events—that are the most important for society—they still struggle," says Sebastian Engelke, a statistics professor at the University of Geneva."

"For record-breaking heat, like a heat wave in Siberia in early 2020, AI predictions tend to underestimate high temperatures. They're also less accurate than older models at predicting extreme wind or record-breaking cold."

"Essentially, they are reproducing what has happened in the past. If we're looking at extreme weather, and especially record-breaking events, then this has not been observed in the past."

"Traditional physics-based forecasting uses complex mathematical models to represent the physical world instead, and can more readily adapt to new conditions."

Read at Fast Company