Forecasting extreme weather events, like deadly heat waves, remains challenging despite advancements. Traditional supercomputers excel at predicting such rare occurrences but require significant time and energy. Conversely, AI models are effective for day-to-day forecasts but often struggle with outlier events. Researchers from the U.S. and France have developed a hybrid method called AI+RES, combining AI efficiency with traditional models’ reliability to predict rare events more accurately and rapidly.
Testing the method on simulations for heat waves in France and the U.S. Midwest showed it produced similar results using only 1% of the simulations needed traditionally. This method could also enhance AI training datasets and be scaled for other severe weather events, ultimately aiding government planning for climate adaptation. The research emphasizes the importance of accurate forecasting for public policy.
Source link
