Nature子刊近日发表了上海科学智能研究院（上智院）、复旦大学和中国气象局国家气候中心联合研发的一款名为「伏羲」的次季节气候预测AI大模型，该模型在次季节气候预测领域取得突破性进展，有效提升了全球降水预测能力，特别是在我国长江中下游地区。
「伏羲」模型在技术上实现了两个关键性创新：一是引入了海气相互作用过程，特别是将热带大气季节内振荡（MJO）纳入模型，有效提升了对MJO的预测技巧，将MJO的预测技巧从30天提升至36天，大幅超过了欧洲中期天气预报中心（ECMWF）的S2S模式的预测时长。二是创新性设计了智能扰动生成模块，能够在当前气候系统状态下刻画未来一个预报时次气候系统演变的概率特征，有效抓住气候系统的物理不确定性。
「伏羲」模型能够在7秒内完成长达42天的全球日平均预报，对于农业规划、资源管理、灾害准备等具有重要意义。该模型不仅能够有效预测极端天气事件，如热浪干旱、寒潮洪水等，还能通过构建显著图识别导致极端事件发生的潜在信息，为应对极端天气事件提供有效的工具和策略。
国务院印发的《气象高质量发展纲要（2022—2035 年）》明确提出要「提前一个月预报重大天气过程」，而次季节气候预测技术是实现这一目标的关键。「伏羲」模型的研发，不仅填补了次季节气候预测「可预报性沙漠」的空白，也为全球气候预测领域提供了新的解决方案。
英语如下：
News Title: “Fuxi AI: New Breakthrough in China’s Sub-Seasonal Climate Forecasting”
Keywords: Fuxi Model, Sub-Seasonal Forecasting, Agricultural Support
News Content: A major artificial intelligence (AI) model for sub-seasonal climate forecasting named “Fuxi” has been published in a Nature-affiliated journal by a joint research team from the Shanghai Institute for Advanced Studies (SIS), Fudan University, and the National Climate Center of the China Meteorological Administration. This model has made a breakthrough in the field of sub-seasonal climate forecasting, significantly enhancing global precipitation prediction capabilities, particularly in the middle and lower reaches of the Yangtze River in China.
The Fuxi model has achieved two key technological innovations: first, it incorporates the interaction between the ocean and the atmosphere, particularly incorporating the Madden-Julian Oscillation (MJO) into the model, effectively improving its prediction skills for the MJO, increasing the prediction period from 30 days to 36 days, significantly surpassing the forecast duration of the S2S (Subseasonal to Seasonal) model of the European Centre for Medium-Range Weather Forecasts (ECMWF). Second, the model has an innovative design for an intelligent perturbation generation module, which can depict the probabilistic characteristics of the future climate system evolution within a forecast time step under the current climate system state, effectively capturing the physical uncertainties of the climate system.
The Fuxi model can complete a 42-day global daily average forecast in just 7 seconds, which is of great significance for agricultural planning, resource management, and disaster preparedness. The model not only effectively predicts extreme weather events such as heatwaves, droughts, cold waves, and floods but also identifies potential information leading to these extreme events through the construction of significant maps, providing effective tools and strategies for coping with extreme weather events.
The State Council’s issued “Outline for the High-Quality Development of Meteorology (2022-2035)” clearly proposes to “forecast major weather processes one month in advance.” Sub-seasonal climate forecasting technology is a key to achieving this goal. The development of the Fuxi model not only fills a blank in the field of sub-seasonal climate forecasting, known as the “desert of predictability,” but also provides new solutions for the global climate prediction sector.
【来源】https://www.jiqizhixin.com/articles/2024-08-19-5
Views: 1