Achieving FAIR Data Principles at the Environmental Data Initiative, the US-LTER Data Repository

Abstract

The Environmental Data Initiative (EDI) is a continuation and expansion of the original United Stated Long-Term Ecological Research Program (US-LTER) data repository which went into production in 2013. Building on decades of data management experience in LTER, EDI is addressing the challenge of publishing a diverse corpus of research data (Servilla et al. 2016). EDI’s accomplishments span all aspects of the data curation and publication lifecycle, including repository cyberinfrastructure, outreach and training, and enhancements to data documentation methodologies used by the environmental and ecological research communities. EDI is managing almost 43,000 unique data packages and their revisions from a community of nearly 2,300 individual data authors, most of which are contributed by LTER sites, and are openly accessible and documented with rich science metadata in the Ecological Metadata Language (EML) standard. Here we will present how EDI achieves FAIR data principles (Wilkinson et al. 2016, Stall et al. 2017), and report data use metrics as a measure of success. The FAIR principles serve as benchmarks for EDI’s operation and management: the data we curate are Findable because they reside in an open repository, with unique and persistent digital object identifiers (DOIs) and standard metadata indexed as a searchable resource; they are Accessible through industry standard protocols and are, in most cases, under an open-access license (access control is available if required); Interoperability is achieved by archiving data in commonly used file formats, and both metadata and data are machine readable and accessible; rich, high quality science metadata, with automated congruence and completeness checking, render data fit for Reuse in multiple contexts and environments, along with easily generated data provenance to document their lineage. The success of this approach is proven by the number and spatial and temporal extent of recent re-analyses and synthesis efforts of these data. Although formal data citations are not yet common practice, a Google Scholar search reveals over 400 journal articles crediting data re-use through an EDI DOI. However, despite improved data availability, researchers still report that the largest time investment in synthesis projects is discovering, cleaning and combining primary datasets until all data are completely understood and converted to a similar format. Starting with long-term biodiversity observation data EDI is addressing this issue by implementing a pre-harmonization of thematically similar data sets. Positioned between the data author’s specific data format and larger biodiversity data stores or synthesis projects, this approach allows uniform access without the loss of ancillary information. This pre-harmonization step may be accomplished by data managers because the dataset still contains all original information without any aggregation or science question specific decisions for data omission or cleaning. The data are still distributed into distinct datasets allowing for asynchronous updating of long-term observations. The addition of specific and standardized metadata makes them easily discoverable.

Earth and Space Science Informatics Perspectives on Integrated, Coordinated, Open, Networked (ICON) Science

“…This hinders data from being Open and/or Findable, Accessible, Interoperable, and Reusable (FAIR; Wilkinson et al, 2016). Further, specific criteria to implement the FAIR Guiding Principles (Gries et al, 2019;Jones et al, 2019) inevitably vary across disciplines and data types as inconsistencies in interpretations of the principles have grown (e.g., Kinkade & Shepherd, 2021;Mons et al, 2017;Stall et al, 2019). Importantly, FAIR does not mean Open; data can be Open without being FAIR, and vice versa (see What is the difference between "FAIR data" and "Open data" if there is one?).…”

Beyond ecosystem modeling: A roadmap to community cyberinfrastructure for ecological data‐model integration

“…Scientists, managers, and policymakers increasingly rely on models to understand the impact of decisions on ecological processes (Arneth et al, 2014;Bonan & Doney, 2018;Smith et al, 2019).…”

“…Data play a critical role in modeling activities; however, due to their sheer volume and diversity, they can be difficult to locate and obtain as sifting through deluge of data manually is impractical (Reichstein et al, 2019;Waide et al, 2017 Gries et al, 2019;Open Science Framework, Sullivan et al, 2019). When those repositories are part of jointly searchable networks (e.g., DataONE-www.datao ne.org), it could further allow developers to leverage one set of tools for many sources.…”

“…[R10] To make data FAIR, we recommend data producers use consistent naming structures (e.g., Assistance for Land‐surface Modelling activities convention, also please see Appendix A for more details) and open file formats (e.g., comma‐separated values, netCDF; Hart et al., 2016). [R11] Next, data should be stored in data repositories where datasets are versioned, data citations are provided, and that support [R12] standard, searchable metadata, and machine‐readable Application Programming Interfaces (e.g., the Oak Ridge National Laboratory Distributed Active Archive Center, Cook et al., 2016; Environmental Data Initiative, Gries et al., 2019; Open Science Framework, Sullivan et al., 2019). When those repositories are part of jointly searchable networks (e.g., DataONE—http://www.dataone.org), it could further allow developers to leverage one set of tools for many sources.…”

DBgen: A Python Library for Defining Scalable, Maintainable, Accessible, Reconfigurable, Transparent (SMART) Data Pipelines

“…DBgen is designed to reduce the barrier to creating databases for complicated data sources in accordance with the FAIR data principle [1]. The FAIR principle, which states that data should be Findable, Accessible, Interoperable, and Reusable, is widely accepted and has shown utility in a variety of fields of scientific research including medicine [2,3,4], meteorology & oceanography [5,6,7,8], oral speech studies [9], botany [10], mass spectrometry [11], and many others. Although many agree that it is important to make data FAIR, a great deal of scientific data remains stored in a way that does not meet these principles.…”