AnswersPublished15 sources
What China’s Fengqing AI Weather Ensemble Does—and What Its 11-Day Result Means
Fengqing Ensemble V1.0 produces 51 member global probability forecasts out to 360 hours; China’s Meteorological Administration says formal operations are scheduled to start November 10, 2026. The ensemble passed operational review at the National Meteorological Centre.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Fengqing Ensemble V1.0 produces 51 member global probability forecasts out to 360 hours; China’s Meteorological Administration says formal operations are scheduled to start November 10, 2026.
- The ensemble passed operational review at the National Meteorological Centre. Its multiple forecast members represent a range of possible atmospheric outcomes, unlike a single deterministic forecast.[17][19]
- The published material provides headline verification results but not detailed scores or a fully documented, independent comparison with other systems.[18][19]
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What is the China Meteorological Administration’s Fengqing AI global short- and medium-range ensemble forecasting system (CMA-AIM-GFS-Fengqi. Article summary: CMA-AIM-GFS-Fengqing-ENS-V1.0, or “Fengqing Ensemble V1.0,” is CMA’s AI system for global short- and medium-range probability weather forecasting, developed under the National Meteorological Centre’s leadership.[9] Octob. Topic tags: general, general web, academic, education. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts wi
Fengqing Ensemble V1.0 is a China Meteorological Administration (CMA) system for global short- and medium-range probability forecasting. It passed operational review at the National Meteorological Centre, but an official CMA notice says formal operational service is scheduled to begin on November 10, 2026. The clearest reported performance result is an 11-day effective forecast range for one atmospheric measure: ensemble-mean 500 hPa geopotential height.17
19
What the system forecasts
CMA says Fengqing Ensemble V1.0 will issue global forecasts twice a day, covering 0–360 hours in six-hour steps at 0.25° horizontal resolution. Each forecast has 51 ensemble members and includes upper-air and surface weather variables, such as geopotential height, humidity, wind, temperature and sea-level pressure.19
An ensemble provides several possible forecast evolutions rather than just one. The spread among its members can help describe forecast uncertainty and support probability forecasts. That is the key distinction from a deterministic system, which gives a single predicted evolution.
Review passed; formal service is scheduled for November 10
The operational review took place at the National Meteorological Centre. The system passed after expert review, which established that it was ready for operational use.17 News reports on October 17 described it as already in operation, but CMA’s later official notice gives a more specific timetable: formal operational service is scheduled to start November 10, 2026.
19
CMA identifies the National Meteorological Centre as the lead developer, working with the Xiong’an Meteorological AI Innovation Research Institute and the National Meteorological Information Centre. The review report also names Sugon among the development partners.17
19
How it relates to the earlier Fengqing model
The ensemble system shares the Fengqing name with an earlier CMA weather AI model developed with Tsinghua University. CMA and Tsinghua describe that earlier model as integrating atmospheric physics into its design and training, including physical conservation properties.33 Those details help explain the earlier model’s approach, but they do not establish that the ensemble uses the same architecture or constraints.
Some reports describe Fengqing forecasts as reaching 15 days at 25-kilometre resolution in about three minutes, but those specifications refer to the earlier model’s later V1.5 release—not verified specifications for Ensemble V1.0.38
19 CMA’s ensemble notice instead specifies its own 51 members, 0.25° resolution and 0–360-hour forecast window.
19 The official information provided does not identify the ensemble’s precise neural-network architecture or forecast-generation runtime.
What the 11-day result tells us—and what it doesn’t
CMA says tests using independent samples from January 2025 through August 2026 found an 11-day effective forecast range for the ensemble-mean global 500 hPa geopotential-height forecast. It also reports favourable continuous ranked probability score (CRPS) and spread–error ratio results, two measures used to assess ensemble forecasts.19
That 11-day figure is specific to the stated variable and verification result. It should not be read as evidence that every kind of weather—including heavy rain, typhoons or severe convection—can be forecast accurately 11 days ahead. The published notice does not provide the detailed CRPS and spread–error scores needed to judge the size of the reported performance.19
A China News Service report says some Fengqing Ensemble indicators were better than those of ECMWF numerical and AI ensemble systems.18 The supplied reporting does not give detailed, matched scores across forecast variables, regions and lead times, so it is not enough to establish broad superiority.
The same report says the system was trained using a 100,000-card AI4S computing cluster at a National Supercomputing Internet core node, and describes its data, model and computing resources as domestically controlled.18 This is a reported development claim; it does not provide an independent audit of every component. Nor does the training-cluster figure indicate how much computing is needed to generate each operational forecast.
The practical takeaway
Fengqing Ensemble V1.0 adds a CMA operational system for global probability forecasting, with 51 forecast members and output extending to 360 hours. Its review status and scheduled start date are distinct: it passed review, while CMA says formal operations are due to begin November 10, 2026. The reported 11-day result is promising but narrow; more detailed verification is needed to assess performance across weather variables and hazards.17
19
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "What China’s Fengqing AI Weather Ensemble Does—and What Its 11-Day Result Means"?
Fengqing Ensemble V1.0 produces 51 member global probability forecasts out to 360 hours; China’s Meteorological Administration says formal operations are scheduled to start November 10, 2026.
What are the key points to validate first?
Fengqing Ensemble V1.0 produces 51 member global probability forecasts out to 360 hours; China’s Meteorological Administration says formal operations are scheduled to start November 10, 2026. The ensemble passed operational review at the National Meteorological Centre. Its multiple forecast members represent a range of possible atmospheric outcomes, unlike a single deterministic forecast.[17][19]
What should I do next in practice?
The published material provides headline verification results but not detailed scores or a fully documented, independent comparison with other systems.[18][19]