第12号台风"红霞"于7月26日凌晨在广东惠州市惠东县平海镇沿海登陆[1]。这次台风的预报过程再次引发公众对台风预报技术的关注。
国家气象中心台风与海洋气象预报中心主任赵伟指出，"台风的路径或移动速度的突然变化很难预判"[1]，这正是台风预报工作中最大的难点所在。为了应对这一挑战，气象部门不断探索和改进预报方法。其中，AI气象大模型在3至5天台风路径预报上已表现出优于传统物理模式的准确性[1]。在台风"摩羯"登陆前，"风清"AI模型提前6天准确锁定了其登陆路径[1]。针对台风降水预报，气象部门采用集合预报方法，通过融合多个模式和多个初始场进行综合分析[1]，以提升预报的可靠性。
Typhoon "Hongxia," the 12th typhoon of the season, made landfall along the coast of Pinghai Town, Huidong County, Huizhou, Guangdong Province in the early morning of July 26 [1]. The event has drawn renewed attention to the complexities and uncertainties inherent in typhoon forecasting, even as meteorological agencies deploy increasingly sophisticated technologies to improve prediction accuracy.
Forecasting the precise behavior of typhoons remains a significant scientific challenge. According to Zhao Wei, director of the Typhoon and Marine Meteorological Forecast Center at the National Meteorological Center, "sudden changes in a typhoon's path or movement speed are particularly difficult to predict" [1]. To address these limitations, meteorological agencies employ ensemble forecasting methods for precipitation prediction, integrating analysis from multiple models and initial conditions [1].
Recent advances in artificial intelligence have shown promise in enhancing forecast reliability. AI meteorological models have demonstrated superior accuracy compared to traditional physical models in typhoon path forecasting over three to five-day periods [1]. Notably, the "Fengqing" AI model successfully locked in the landing path of Typhoon "Mojiake" six days before its actual landfall [1], exemplifying the potential of machine learning to extend the predictability window for these high-impact weather events.