The ensemble gives CMA an uncertainty measure for AI forecasts, relevant to organizations planning around weather-sensitive transport, energy and disaster response.
Fengqing Ensemble V1.0 cleared an operational review at the National Meteorological Centre. China Science Daily reported on October 16 that the China Meteorological Administration (CMA) system had met conditions for operational use. Its independent testing produced an 11.0-day effective forecast range for the ensemble mean of global 500-hPa geopotential height, one assessment measure cited in the report.
Fengqing uses an ensemble, a set of separate forecast members, to show the spread of possible atmospheric outcomes. The variation among those members gives forecasters a direct measure of uncertainty. Cailian Press reported on October 17 that Fengqing V1.0 had entered operation, and Pandaily said CMA had placed it in national operational use. China Science Daily describes the milestone as review passage and operational readiness. No direct CMA or National Meteorological Centre deployment notice appears in the reports reviewed.
The review names the organizations behind Fengqing V1.0
China Science Daily says the National Meteorological Centre led Fengqing V1.0’s development with the Xiong’an Meteorological Artificial Intelligence Innovation Institute, the National Meteorological Information Centre and Sugon. It describes the release as a new-generation AI ensemble system for global short- and medium-range weather forecasting operations. The review meeting brought together experts from the Chinese Academy of Sciences’ Institute of Atmospheric Physics, numerical forecasting and meteorological-information organizations, regional meteorological bureaus, Sugon and the National Meteorological Centre.
The system’s central design choice is to turn forecast spread into a usable output. Each member represents a possible atmospheric evolution. Their range of outcomes supports probability forecasts for severe weather, according to Pandaily’s description of the system. The resulting spread gives users an explicit confidence signal for the forecast.
Cailian Press said the system was announced at CCF HPC China 2026 and reported that Fengqing had both passed review and entered operation. The report provides a public event and a second status account. It supplies no CMA deployment record.
The 11-day result applies to a defined ensemble metric
China Science Daily says the evaluation used independent samples from January 2025 through August 2026. The reported 11.0-day range applies to the ensemble mean for global 500-hPa geopotential height. The sources assign that figure to this defined metric and provide no equivalent result covering every weather variable or forecasting task.
The newspaper also said that member continuous ranked probability score and spread-error ratio, two measures used to assess an ensemble’s probability forecasts and its internal spread, performed well. It reported that selected indicators exceeded those of the European Centre for Medium-Range Weather Forecasts’ numerical ensemble and artificial-intelligence ensemble systems. Pandaily identifies this comparison as a developers’ claim and says detailed scores have yet to be published, leaving outsiders unable to assess the full comparison from a public scorecard.
The source record also includes other Fengqing releases. A World Meteorological Centre Beijing update describes Fengqing V1.5 as a data-driven global medium-range model for forecasts beyond 10 days and says it added six-hour accumulated-precipitation forecasts. Pandaily calls the earlier Fengqing model deterministic. The evidence reported for V1.0 concerns the current ensemble release.
A direct operating notice would clarify the service transition
The sources agree that Fengqing V1.0 passed an operational review. Cailian Press and Pandaily describe operation. China Science Daily says the system has conditions for it. A direct CMA or National Meteorological Centre announcement naming Fengqing V1.0 and confirming its operating status would settle that distinction.
The reports also leave the service’s public operating details unspecified. They do not identify a public forecast product, a user interface or a commercial arrangement. The immediately verifiable advance is an AI ensemble designed to make uncertainty visible to forecasters, with a review result and an independently tested metric attached to that release.
EastFrontier’s August coverage of AI weather models during Typhoon Dolphin examined how Chinese AI systems were being used alongside conventional forecasting tools during a severe-weather event. Fengqing extends that discussion to an ensemble output designed to quantify forecast spread. The next concrete signal is a CMA or National Meteorological Centre record that identifies how that output is being used in operational forecasting.