Gaps and recommendations

This study analyses how uncertainty and quality information are reported acrossCopernicus ClimateChange Service (C3S)andCopernicus Atmosphere Monitoring Service (CAMS)portfolio, extending previouswork on theCopernicus Land Monitoring Service (CLMS)with a strengthened multi-modelArtificial Intelli-gence (AI)workflow [39]. Over 6,700 dataset layers, defined here as individual variables or data subcom-ponents, were screened using fourLarge Language Model (LLM)s, confirming 996 quality-related layersinC3Sbut only 2 inCAMS. These results reveal a strong imbalance in how quality information is providedacross services. While manyC3Sdatasets include quality-related information, particularly uncertainty-related variables, their interpretation is often limited by inconsistent terminology and incomplete documentation. By contrast, althoughCAMSproduct quality is assessed through dedicated Evaluation andQuality Assurance activities and reports, its datasets largely lack explicit quality indicators at the layerlevel. The use of multipleLLMs improves robustness and reduces false detections, but results remain de-pendent on the quality of metadata and documentation. Strengthening the clarity, consistency and trace-ability of uncertainty information is essential to enable the reliable use of Copernicus data in policy, regu-lation andAI-driven applications.

BRATIC Gorica;  GOBRON Nadine;  CAPPUCCI Fabrizio; 

2026-09-22

Publications Office of the European Union

JRC147495

978-92-68-43161-0 (online),   

1831-9424 (online),   

EUR 40844,    OP KJ-01-26-378-EN-N (online),   

https://publications.jrc.ec.europa.eu/repository/handle/JRC147495,   

10.2760/2341032 (online),