The system has a global effective forecast time of 11 days, marking that China’s AI weather forecast has reached the world’s advanced level.
AI-generated summary
"Fengqing" is a new generation AI collective forecasting system led by the National Meteorological Center, which relies on the core nodes of the national supercomputing Internet for training.
China News Service, Beijing, October 19 (Reporter Zhang Su) Test results based on independent samples from January 2025 to August this year show that the "Fengqing Ensemble V1.0" global ensemble average forecast of 500 hPa geopotential height has an effective forecast time limit of 11.0 days, and key indicators such as member forecast continuous classification probability score (CRPS) and dispersion-error ratio performed well...
The reporter learned from an interview on the 19th that the "Fengqing" artificial intelligence global short- and medium-term collective weather forecast system V1.0 successfully passed the operational review and was put into operation. Many experts from the Institute of Atmospheric Physics of the Chinese Academy of Sciences, the Earth System Numerical Prediction Center of the China Meteorological Administration and other institutions participated in the review.
"Fengqing Collection V1.0" is a new generation of artificial intelligence (AI) ensemble forecasting system led by the National Meteorological Center and jointly developed by the Xiongan Meteorological Artificial Intelligence Innovation Institute and the National Supercomputing Internet.
According to the R&D personnel, the system relies on the 100,000-ka AI4S computing cluster at the core node of the National Supercomputing Internet to complete training and has become a business application meteorological AI ensemble forecasting model that achieves full autonomy in data, models, and computing power. Its global effective forecast timeliness reaches 11 days, and some indicators are better than the European Center for Medium-Range Weather Forecasts numerical ensemble forecast system and artificial intelligence ensemble forecast system, marking that China's AI weather forecast has reached the world's advanced level.
It is reported that "Fengqing" has been applied and effective in real-time business. In September 2024, during the actual combat test in response to Typhoon "Capricorn", "Fengqing" locked the landing path of the typhoon more than 6 days before it made landfall, grabbing a valuable window period for Guangdong, Guangxi, and Hainan to make disaster prevention decisions. In August last year, "Fengqing" predicted scattered heavy rains in North China 8.5 days in advance.
In addition, "Fengqing" is still being upgraded. Its V1.5 version has achieved new improvements in various performance indicators. It can generate global weather forecast products for the next 0 to 15 days, updated every 6 hours, and with a resolution of 25 kilometers within 3 minutes. The number of effective forecast days exceeds 10.5 days.
The R&D team stated that in the future, it will continue to promote model optimization, focusing on improving temporal and spatial resolution, deepening the application of multi-source data fusion, and further exploring the potential of data processing and pattern analysis, so that "Fengqing" can play a greater role in extreme weather warning, regional refined forecasting and other fields.
AI outlook — possibilities, not facts
Continue to promote model optimization and improve spatiotemporal resolution
Very likely · Within months
2026 will celebrate the 90th anniversary of the discovery of the Liangzhu ruins. The fourth "Liangzhu Forum" was held recently. Scholars from many countries discussed the uniqueness of Liangzhu civilization in rice farming, collective collaboration and spiritual beliefs, emphasizing the diversity of the origin paths of human civilization and the value of civilization diversity.

Chinese researchers have discovered that dendrites in neurons act as minicomputers, integrating learned rules with sensory data to enable flexible decision-making in mice. This finding offers new potential for developing more adaptable AI systems.

On October 20, the 4th "Liangzhu Forum" released six Liangzhu archaeological research results, including the "Complete Collection of Liangzhu Jade Artifacts". Experts said that these results demonstrate the latest progress in Liangzhu archeology and emphasize the important role of international cooperation in the study of Liangzhu ruins.

By analyzing 25 years of global ionospheric observation data, a team from Chuo University and Kyoto University confirmed for the first time that the ionosphere has a "saturation effect" on solar activity. This subverts the previous intuitive understanding that "the stronger the sun, the more active the ionosphere is" and helps improve the reliability of GPS positioning and satellite communications.
During the 4th "Liangzhu Forum", Wang Wei, the principle director of the Chinese Archaeological Society, accepted an exclusive interview to analyze the unique growth method of Liangzhu civilization that relied on water due to rice cultivation and gathered due to water control, and discussed the local growth and foreign exchange and mutual learning of Chinese civilization.
The International Research Center for Sustainable Development Big Data released "Global Sustainable Development Data Products (2025)" in Beijing on October 19. This series includes 9 sets of data products, covering the land surface, offshore and atmosphere, with a longest span of 38 years and a maximum resolution of 10 meters. It aims to provide scientific support for the assessment of global sustainable development goals.