AI-driven assessment of uncertainty and quality terminology in Copernicus data to enable policy uptake
Application to the Copernicus Land Monitoring System
Copernicus data needs to be fit for legal applications to enhance its use in policymaking. Quality must be quantified and traceable to international standards. While various pixel-level quality indicators exist, only uncertainties provide the reliability required for legal context. This study aims to evaluate the uncertainty characterization and the consistency of quality nomenclature across Copernicus. We developed an AI-driven methodology to identify and classify quality layers—defined as information on observation count, quality flags, data spread, classification probability, errors, and uncertainty. We applied the methodology to the Copernicus Land Monitoring Service (CLMS), analysing 94 datasets and 757 layers. The main barrier was the lack of programmatic metadata access, prompting recommendations to enable similar studies across Copernicus.
We identified 189 quality layers across 61 datasets; 14 datasets reported uncertainties, but only 8 followed metrological guidelines. These could support use cases demonstrating how propagating uncertainties to policy indicators can enhance the uptake of Copernicus data in policymaking. We found inconsistent quality nomenclature in CLMS, with terms such as uncertainty, error, and standard deviation used interchangeably, and generic labels like flag or confidence varying in meaning across datasets. Copernicus services should standardize both the content and terminology of quality layers. A better definition of the uncertainty budgets and coverage factors is also needed.
PIERRE Gaia;
URRACA VALLE Ruben;
2025-09-24
Publications Office of the European Union
JRC143507
978-92-68-31853-9 (online),
1831-9424 (online),
EUR 40458,
OP KJ-01-25-481-EN-N (online),
https://publications.jrc.ec.europa.eu/repository/handle/JRC143507,
10.2760/3375300 (online),
Additional supporting files
| File name | Description | File type | |
|---|
