Provenance (Origin) of Research Data

Minimum Description of Provenance

Since the research data are very diverse (i.e. obtained experimentally, by observation, derived or compiled on the basis of other data sets, obtained through simulations or from reference or official databases), they also require a tailored level of detail in the description of their provenance. In a review article titled Data Provenance Standards and Recommendations for FAIR Data, Jauer and Deserno (2020) compared the requirements of four metadata systems, namely the general DataCite schema, the ECRIN recommendations for clinical research, the Working Group on Data Citation (Research Data Alliance) recommendations for experimental research and the Data Quality Collaborative schema. According to their findings, each metadata record should contain at least the six basic pieces of information listed in the table below.

Some repositories prescribe a specific metadata schema for the description of provenance, whereas others leave the choice of the metadata format to the authors. It is recommended that you provide minimal metadata, but even better, you should follow one of the schemas presented below. If a structured metadata schema is not prescribed, you can record the metadata in the form of a ReadMe file or data paper.

For projects within Horizon Europe, the EU Grants - AGA: Annotated Model Grant Agreement specifies in Annex 5: Communication, dissemination, open science and visibility that a minimum description of provenance must contain the following metadata:

• information about the scientific publication (author(s), title, date of publication, publication venue - journal or publishing house);
• funding source (Horizon Europe or Euratom), including project name, acronym and number;
• licensing terms;
• persistent identifiers for the publication, authors, and, if possible, for their organisations and the grant (see also the article on persistent identifiers of non-digital objects).

If possible, the metadata must include persistent identifiers for any research output or any other tools and instruments needed to validate the conclusions of the publication.

General Metadata Schemas

A good example of basic information that should accompany research data is the DataCite metadata schema, which is suitable for various research fields. Information is divided into mandatory, recommended and optional. It is used by repositories such as Zenodo, Dryad and Figshare.

Similar to the DataCite schema is the Dublin Core metadata schema, which contains all of the DataCite categories (some of which are named differently), as well as more specific ones, such as repository entry date, license and audience. Zenodo, Dryad and Figshare are compatible both with Dublin Core as well as with DataCite. Dryad is also compatible with the RDF Data Cube Vocabulary and OAI-ORE – Open Archives Initiative Object Reuse and Exchange schemas.

Domain-specific Description of Provenance

For many research fields, especially in the field of natural sciences and engineering, where many people participate in research projects and the acquisition of data and their processing are very complex, simple metadata schemas for describing the provenance of data are not sufficient. The UK Digital Curation Centre and Research Data Alliance maintain a list of over 40 domain-specific metadata schemas, covering all major research fields. Of these, the following are worth mentioning due to their wide adoption, for example:

• Data Documentation Initiative (DDI) for the description of data from the fields of social sciences, humanities and economics;
• Crystallographic Information Framework (CIF), which is used to report crystal structures in the Acta Crystallographica journal and related publications;
• NeXus for storage and exchange of data on experiments with X-rays, neutrons and muons;
• Minimum Information for Biological and Biomedical Investigations (MIBBI), which brings together more than 40 minimum standards for various research subfields within biology and biomedicine;
• Darwin Core for preservation and exchange of biodiversity data;
• Genome Meatdata, which defines 61 criteria for describing genomic data;
• Geographic Information Metadata (ISO 19115-1:2014) for geographic data and services,
• etc.

In addition to standardized metadata schemas, there is also a series of minimum requirements for reporting experiments, which were created at the initiative of the scientific community within a particular research field and are currently still under development. Most of them are available for life sciences. Some of these are, for example:

• Minimum information about a microarray experiment (MIAME),
• Minimum information about a proteomics experiment (MIAPE),
• Minimum information about a spinal cord injury experiment (MIASCI),
• Minimum information about a simulation experiment (MIASE),
• Standard for exchange of nonclinical data (SEND),
• Standardized battery reporting guidelines,
• CONSORT (for randomized controlled clinical trials),
• QUOROM (for meta-analyses of controlled clinical trials),
• STARD (for reporting diagnostic tests),
• etc.

In most cases, domain-specific repositories prescribe an appropriate, field-specific metadata schema that is most suitable for the type of data these repositories store. A good example is the metadata schema of the Environmental Information Data Centre, the UK's national repository for land and water sciences, which includes all aspects of information about data, data collection process and quality control.

If a suitable domain-specific repository does not yet exist for your research field and you will deposit the data in a general or institutional repository, CTK UL advises you not to settle for the basic metadata schema used by these repositories. Instead, we recommend that you follow the domain-specific metadata schema or minimum research reporting requirements that apply to your field. If neither of these two options exist, you can use the Core Scientific Metadata Model (CSMD) generic metadata schema for experimental data.

Example of Provenance Description (PCCS Standard)

As a representative example of a detailed provenance description for complex research, we will look at the Provenance and Context Content Standard (PCCS). This standard has been developed since 2011 by the Earth Science Information Partners (ESIP), which was founded by NASA in 1998 to improve methods for preserving, discovering, accessing and reusing Earth data. PCCS provides technical standards for identifying, capturing and tracking all relevant details that ensure the validity and repeatability of Earth observations. PCCS has also been outlined by NASA as a mandatory reporting format for new missions. The source table can be viewed on the ESIP website.

For the purpose of presentation on this website, we have summarized only part of the table.