According to an article in *Physics World*, a joint US-French research team has launched an AI + Rare Event Sampling (RES) climate simulation framework to address the pain points of traditional climate models, such as excessively high computational costs and difficulty in predicting rare extreme weather events. This method uses AI to pre-screen meteorological simulation scenarios prone to heat waves and extreme cyclones, performing full physical simulations only on high-risk periods, with computational consumption only 1/1000th that of traditional full-domain simulations. Results show that, based on the PlaSim simplified climate model, the method's predictions for mid-latitude heat wave frequencies are highly consistent and can be used to analyze current extreme heat events such as the European heat wave. Scholars at ETH Zurich stated that this approach fills the gap in traditional RES screening criteria and can be further expanded to predict all types of extreme weather events, including heavy rain and typhoons, adapting to high-precision climate models.
