We gratefully thank the developers and contributors to the following software used in this work: R v. 4.5.3 and the following R packages: arrow v. 23.0.1.2, cpp11 v. 0.5.3, crsuggest v. 0.4, glue v. 1.8.0, grateful v. 0.3.0, gt v. 1.3.0, ISOcodes v. 2026.3.28, knitr v. 1.51, pointblank v. 0.12.3, quarto v. 1.5.1, rmarkdown v. 2.31, testthat v. 3.3.2, tidyselect v. 1.2.1, tidyverse v. 2.0.0, zip v. 2.3.3.
The eData Data Reporting Format is a format for the standardisation of data on pollutant occurrence in the environment. By converting this data to a single, consistent format, it makes it easier to validate, compare, analyse, and reuse, in line with the FAIR principles.
Why standardise?
These data are typically published as concentrations (in water) or mass ratios (in soil, sludge and biota). Fully interpreting these measurements requires considerable metadata, relating to the time, place, conditions, methodology, and context of sampling. However, no single standard exists for these metadata, and study authors and sampling teams tend to use a variety of inconsistent definitions, units, and schema depending on their scientific domain and the study’s individual needs.
This is not, in itself a problem. However, as environmental risk assessment becomes more data-based and larger scale, the need to use and re-use large quantities of data increases. When these datasets are consistent, validated, and interoperable, much time can be saved. When they are not, researchers much go through the tedious process of understanding, validating and remapping by hand; an often non-reproducible process. Thus a harmonised format - and associated validation and formatting functions - can represent a significant timesaver for anyone in such a position.
Why aquatic toxicology?
Starting with a field we know well allows us to test the effectiveness of the underlying design philosophy and technology.
Why an R package?
Although an R package is not a conventional form for a data structure, as our extraction, analysis, workflow is heavily R-based it was a pragmatic choice. By using the existing package architecture, it was possible to integrate the format closely with downstream tools, and use existing package dependencies and versioning to keep them updated as the format changes. In future, we plan to extend the format to other forms.
Who supported this work?
We gratefully acknowledge the assistance and input of Kim Leirvik, Li Xi, Camden Karon Klefbom, Pierre Blévin, Anna Maria Hatland, Sophie Mentzel, Viviane Giradin, Walter Zobl, and the members of the PARC Project P7.2.2 team in the creation of this format.
This work is supported by the Research Council of Norway (RCN-315969 EXPECT), EU-HEU-101057014 Project “PARC”, and NIVA’s Computational Toxicology Program (NCTP, RCN-160016).
Declaration of Generative AI Use
The Claude family of Large Language Models was used in the creation of the package for the generation of boilerplate and some R code.