"风云"卫星俯瞰全球，筑就陆海空天一体化监测体系；人工智能预报精准至公里级，暴雨预警准确率再创新高；生活气象服务指数增至70余 种，精细化气象服务涵盖衣食住行游购娱……"十四五"时期，从守护百姓衣食住行到助力国家发展大局，从筑牢国内生产生活安全防线到赋能 产业升级，气象服务已成为满足民生期待、推动经济社会发展的重要支撑。 跟随这组海报，共同感知中国气象以硬核实力护航国计民生、守护万家平安、赋能千行百业的蓬勃力量与责任担当。 · 全球首次实现气象卫星区域250米分辨率每分钟连续成像 观测 · 风云三号G星是我国首颗、全球第三颗低倾角轨道降水测量 上 • 风云三号E星是全球唯一在晨昏时段开展观测的民用极轨 气象卫星 新华社 司刊 nlist 新华社 3 27 建成临海空 本省合量量表机机系统 · 风云卫星、北斗探空、天气雷达等探测装备、技术实现 高水平自立自强 · "风雷""风清""风顺""风宇"气象预报模型 达到国际先进水平 · 远洋气象导航、全球碳源汇监测核枝等一批创新技术成果涌现 部信息 精准频报 年47周三日可控的全球、区域数国预报技术的 · 灾害性天气监测率提升到83% 强对流天气预警时间提前13% · ...

Core Insights - The China Meteorological Administration has launched an AI meteorological service system named "Fenghe," which significantly enhances smart meteorological service capabilities, featuring data fusion, real-time human-machine interaction, and intelligent tool invocation [1][4]. Group 1: Features and Capabilities - "Fenghe" allows users to input weather-related queries and receive automated responses, providing personalized suggestions across various scenarios such as travel, health, and logistics [3][4]. - The system includes five major modules: Meteorological Knowledge Center, Model Plaza, Meteorological AI Toolbox, Intelligent Agent Factory, and Evaluation Laboratory, which collectively enhance its service capabilities [4]. Group 2: Technical Framework - The design of "Fenghe" is based on a unique "1+1+N" technical framework, integrating a foundational model with professional meteorological knowledge and a large-scale meteorological service corpus of 550 billion tokens [5]. - The system employs advanced technologies such as LoRA for fine-tuning, reinforcement learning from human feedback (RLHF) for deep reasoning, and multi-agent collaborative technology to improve understanding of user needs and adaptability to various scenarios [5].

Core Insights - The China Meteorological Administration has launched the "Fenghe" model, a billion-parameter meteorological service model that integrates AI technology with meteorological expertise, aiming to transform traditional weather services into an intelligent new phase [1][2]. Group 1: Technical Features of "Fenghe" - "Fenghe" is a generative AI system specifically designed for meteorological services, built on a large language model architecture, distinguishing it from traditional numerical weather prediction models [2]. - The model boasts a powerful core with a trillion parameters, indicating its strong learning and expressive capabilities, allowing it to capture intricate atmospheric phenomena [2]. - Through multimodal integration and generative AI technology, "Fenghe" aims to provide more accurate forecasts, achieving high resolution, efficiency, and rapid response in intelligent meteorological services [2]. Group 2: Comparison with Previous Models - Previous models like "Fengqing," "Fenglei," and "Fengshun" were primarily designed for internal meteorological forecasting systems, focusing on enhancing the core technology of large model predictions [3]. - "Fenghe" is targeted at the public and industry, addressing the limitations of existing general models in understanding meteorological service needs and generating professional content [3]. - The model aims to deliver reliable, usable, and trustworthy meteorological decision-making information, surpassing the capabilities of general models in terms of professionalism, precision, safety, and cost-effectiveness [3]. Group 3: Impact on Public Services - "Fenghe" is expected to revolutionize public travel services by providing unprecedented precision and detail in travel decision support, moving beyond traditional weather forecasts [5]. - For example, it can simulate microclimate conditions in complex areas, offering specific forecasts for different locations, such as predicting sunny conditions on one slope while rain occurs on another [5]. - The service will transition from regional forecasts to point-to-point forecasts, enabling proactive intelligent interventions for optimal travel decision-making [6].

Core Viewpoint - China has launched the "Mazu" meteorological warning system at the 2025 World Artificial Intelligence Conference in Shanghai, aimed at helping developing countries respond to natural disasters [1][2]. Group 1: System Overview - The "Mazu" system is named after a widely revered Chinese sea goddess in Southeast Asia and is designed for global coverage [2]. - The system integrates multiple disaster warnings into a single platform, emphasizing simplicity and collaboration in meteorological safety [3]. - The Chinese Meteorological Administration plans to enhance the meteorological warning system through training, providing forecasters with an open-source toolkit of AI models named "Wind Thunder," "Wind Clear," and "Wind Smooth," each tailored for different time frames from 10 minutes to 3 months [3]. Group 2: Global Context - The system is part of China's efforts to establish a network of warning partnerships with other countries to address extreme weather challenges [2]. - Advanced meteorological warning systems already exist in other countries, such as the U.S. National Weather Service and Japan Meteorological Agency, which rely heavily on local supercomputers and prioritize specific disaster types [2][3]. - The "Mazu-Urban" platform, a city-level disaster warning tool, has been donated to representatives from Djibouti and Mongolia, providing risk assessments, real-time monitoring, forecasts, automated disaster reports, and AI-generated emergency plans [3]. Group 3: Implementation and Reach - Since January, a pilot version of the "Mazu" system has been launched in 35 countries across Asia, Africa, and the Pacific [3]. - The system aims to facilitate global cooperation by making its AI models open-source [3].

Core Insights - The 2025 World Artificial Intelligence Conference showcased the rapid acceleration of AI applications across various sectors, highlighting its transformative potential in industries and scientific research [1][10]. Weather Forecasting and Disaster Warning - AI models in meteorology, such as "Rain Master" and "Fuyou," can advance disaster warnings by 15 to 45 minutes, enhancing urban disaster preparedness [2][3]. - The "Rain Master" model provides detailed insights into severe convective weather, while "Fuyou" focuses on short-term heavy rainfall and thunderstorms, improving forecast updates from hourly to every 10 minutes [3]. - The "MAZU" early warning system integrates AI technology with advanced algorithms and multi-source data for comprehensive disaster alerts [3][4]. Industrial Empowerment - AI is increasingly integrated into various industries, enhancing efficiency and quality, with examples including Siemens' industrial AI system that autonomously executes production tasks [7][8]. - Over 200 companies and 150,000 workers are collaborating with Siemens' AI system, which is expected to be implemented in China within the year [8]. - Tencent's "Magical Digital Human" improves content production efficiency by 50% and reduces live-streaming costs by up to 90%, applicable in e-commerce, education, and finance [8]. Scientific Discovery and Innovation - The term "AI4S" (Artificial Intelligence for Science) signifies the use of AI in scientific discovery, marking a shift from traditional research paradigms to innovative solutions in complex systems [11]. - AI4S has achieved breakthroughs in various fields, including quantum computing and life sciences, with significant advancements such as defect-free arrangement of quantum bits and new cancer treatment targets [11][13]. - The integration of AI into the entire research lifecycle is propelling scientific discovery towards large-scale innovation [13].