Championing FAIR data management in ECMWF Data Stores
At ECMWF, we are championing the FAIR data principles across all our data stores - making data Findable, Accessible, Interoperable, and Reusable. The aim is to make it easier for our users to discover and integrate data into their workflows, helping users get more value from our products and services.
Our data stores offer open data across a wide range of sectors, serving a wide range of communities including climate science, atmospheric research, natural hazard response and weather forecasting. The FAIR principles ensure this data and the accompanying metadata are well-structured and richly described, making it accessible to humans, machines, and now AI systems alike. This empowers users across the communities to freely discover, integrate and build on our data, whilst reinforcing the transparency that open data promises. It also helps build a trustworthy and interoperable data ecosystem where AI and machine learning models can interrogate and train on the high-quality data we serve.
Implementing the FAIR principles plays an important role in repositioning the data stores to address changing data-consumption patterns, evolving user needs, and facilitating broader federation with external platforms and data spaces. This commitment increases the usefulness of our data products across a broader community, facilitating accessibility, collaboration and ultimately enhancing the value of the data stores to our users.
Understanding our FAIR score
Our data stores are regularly assessed against a FAIR maturity model, allowing us to score how fully the FAIR principles have been implemented across our data stores. We publish the latest FAIR score as a percentage on the individual live status pages allowing users to track our FAIR maturity across data stores, measure progress against our improvement targets, and ensure all new catalogue entries meet the highest FAIR standards. We calculate the FAIR score using the online F-UJI FAIR checker. F-UJI is an outcome of the FAIRsFAIR “Fostering FAIR Data Practices In Europe” project funded by the European Union’s Horizon 2020 project (H2020-INFRAEOSC-2018-2020).
Our FAIR score percentage provides a quick and understandable measure of how well each catalogue entry aligns with good data management practices - according to the FAIR principles. The assessment evaluates aggregated metadata from our dataset catalogue landing page, persistent identifier service in which the dataset catalogue is registered (Datacite), and the machine-actionable metadata services we utilize (CSW & STAC).
F-UJI uses a minimum of 17 metrics to assess catalogue maturity, each evaluated on an ordinal scale with partial credit awarded depending on the level of maturity achieved. The aggregated score determines the overall percentage shown on the dashboards. Users may input a dataset catalogue entry Digital Object Identifier number into the online F-UJI FAIR checker to obtain a more detailed breakdown of our scores. Since the F-UJI metrics are versioned and may be revised or improved, we display the metric version employed for users to compare before running their own FAIR checks. The latest F-UJI metrics are available here.
Detailed FAIR metrics assessment
The table below outlines the F-UJI metrics in the current version and describes our specific efforts we are implementing to achieve FAIR compliance. Our approach is phased and ongoing.
Principle | FAIR Metric | Description | |
Findable | F1 | Globally unique identifier (FsF-F1-01D) | Each catalogue entry has a unique HTTP URL for individual identification and enhanced discovery. |
F1 | Persistent identifier (FsF-F1-02D) | We assign Digital Object Identifiers (DOIs) via DataCite. This supports long-term stewardship and digital preservation by ensuring that the catalogue entries (metadata) remain accessible even if the associated data are deprecated or withdrawn. | |
F2 | Descriptive metadata (FsF-F2-01M) | We provide standard citation metadata (creator, title, identifier, type, date, publisher, keywords) in machine-readable formats on landing pages and via an external DataCite register. | |
F3 | Metadata links to the data (FsF-F3-01M) | Metadata includes unique URLs for API endpoints and downloads in machine-readable format. | |
F4 | Machine-readable metadata (FsF-F4-01M) | Metadata for each catalogue entry is retrievable programmatically through structured formats on the landing page. | |
Accessible | A1 | Access levels and conditions (FsF-A1-01M) | We ensure that the metadata explicitly states the access level of the dataset (e.g., open, restricted, or embargoed) and any conditions under which the data can be accessed. |
A1 | Standardized metadata protocol (FsF-A1-02M) | Metadata (HTML & JSON-LD) is accessible via HTTPS without login when the catalogue entry is queried. | |
A1 | Standardized data access (FsF-A1-03D) | Data is accessible via HTTPS through download forms and our standard API service endpoint. | |
A2 | Persistent metadata (FsF-A2-01M) | We ensure metadata for deprecated entries remains accessible even after data withdrawal, supported by our data deprecation procedures. | |
Interoperable | I1 | Formal knowledge representation (FsF-I1-01M) | We ensure that the metadata is machine readable and semantically structured in JSON-LD, RDF formats. |
I2 | Registered semantic resources (FsF-I2-01M) | We ensure that the metadata uses standardised vocabularies including Schema.org, Dublin Core and Datacite. | |
I3 | Qualified relationships (FsF-I3-01M) | Related catalogue entries are referenced with explicit, semantically qualified relationships. | |
Reusable | R1 | Metadata specifies the contents (FsF-R1-01M) | We ensure that the metadata includes multiple machine-readable elements describing dataset contents. |
R1.1 | Licence information (FsF-R1.1-01M) | We ensure that licence information determining the terms of reuse of the data are clearly stated in SPDX format. | |
R1.2 | Provenance information | Provenance metadata (publication date, creator, publisher) are included in machine-readable format. | |
R1.3 | Community standards | We follow ISO 19115 and INSPIRE (Infrastructure for Spatial Information in Europe) core metadata standards for geospatial and Earth science data. | |
R1.3 | Recommended file formats (FsF-R1.3-02D) | Data is provided in community-standard formats: GRIB (meteorological), NetCDF (multidimensional), CSV (tabular). | |
These efforts represent a phased implementation, with ongoing refinement and improvement as we progress. Current maturity levels vary across metrics and may be tracked via the online F-UJI FAIR checker.
