How the Theia/OZCAR information system make data from the French OZCAR critical observatory network more FAIR?

The Theia/OZCAR Information System (IS) aims to facilitate the discovery and reuse of in-situ data documenting the continental surfaces collected by French research organizations and their foreign partners, historically managed and disseminated in different databases and portals built independently by different communities. This includes data from the 22 observatories of the French Critical Zone Research Infrastructure, OZCAR-RI, which document the different environmental compartments of the critical zone. Their data were historically documented and distributed by information systems that used their own vocabulary and format for sharing data. The challenge for the Theia/OZCAR IS (Braud et al. 2020) is to federate this heterogeneous data (more than 300 variables), managed by different communities, into a common system and make it FAIR (Findable, Accessible, Interoperable, Reusable ; Wilkinson et al. 2016) for both humans and machines. This effort that started in 2017 is a contribution to national and European ecosystems for sharing and analysing multidisciplinary Earth system data that are being set up. At national level, the Theia/OZCAR IS is part of the French Earth System Data Terra Research Infrastructure (RI), which aims to provide access and analysis services for data from the entire Earth system and to strengthen interdisciplinary research. It also plays a role in the national research and innovation program OneWater and its OneWater Data Platform, which aims to improve access and interoperability of water data from French research and operational services. At the European level, Theia/OZCAR IS contributes to eLTER-RI data catalogue. The Theia/OZCAR IS was co-constructed with the scientific community, data producers and IT teams involved in data management. The needs expressed by scientific users were to be able to discover data by variable name and to download data from different producers in the same archive, with harmonised formats. Data producers also expressed the need for statistical reports on data use and/or a monitoring dashboard interface. The system has been designed taking into account that the FAIR principles apply to both metadata and data. A common data model, called ”pivot model”, has been defined to harmonise and standardise the description of data between the different data producers, from the granularity of dataset down to observation (time series of a variable at one location) and to set up information fluxes between the observatories’ information systems and the Theia/OZCAR IS (Fig. 1). It is based on several metadata standards (ISO 19115, O&M, DataCite). The Theia/OZCAR thesaurus, a controlled vocabulary for variable names and objects of interest, has been developed to support efficient data discovery and interoperability services. Its objectives are: to enable a user-friendly data discovery service thanks to simplified variable names, to offer precise descriptions of observations thanks to detailed variable names, to enable the implementation of standardised data exchange services that requires concepts describing the act of observation (observed variable and characteristics involved in sampling). The Theia/OZCAR thesaurus is published on the Web according to the FAIR principles: It is formally described using knowledge representation language : it implements the SKOS and I-ADOPT framework ontologies (Coussot et al. 2024), which allows variable names to be broken down into atomic elements (with at least one Property being measured and one Entity being observed) and to make it easier to establish semantic alignments with terms from other international disciplinary thesauri (alignments with EnvThes and GEMET are carried out); It is indexed in a searchable resource with access for both human (skosmos interface) and machines (SPARQL endpoint); The vocabulary and each of its terms have unique persistent web identifiers; Metadata for the vocabulary (license, provenance) and the terms (definition, synonyms) are provided and are sufficiently precise to enable users to understand what each term means. It is formally described using knowledge representation language : it implements the SKOS and I-ADOPT framework ontologies (Coussot et al. 2024), which allows variable names to be broken down into atomic elements (with at least one Property being measured and one Entity being observed) and to make it easier to establish semantic alignments with terms from other international disciplinary thesauri (alignments with EnvThes and GEMET are carried out); It is indexed in a searchable resource with access for both human (skosmos interface) and machines (SPARQL endpoint); The vocabulary and each of its terms have unique persistent web identifiers; Metadata for the vocabulary (license, provenance) and the terms (definition, synonyms) are provided and are sufficiently precise to enable users to understand what each term means. The Theia/OZCAR IS consists of several parts (Fig. 2): a data ingestion module, and data access and discovery services including a terminology service. The data ingestion module enriches the observation information supplied by producers with an additional harmonised vocabulary. For data discovery and access, the Theia/OZCAR IS includes a data portal with faceted search. Data visualisation, and APIs for data download and usage statistics services are currently been developed. The download service will make it possible to select several observations from different producers and to download the data at once in two open, self-describing formats (CSV and NetCDF) from a object-based S3 storage repository. To ensure interoperability of data for machines, a standardised data catalogue service (OGC CSW) is already implemented and used by Data Terra RI’s data sharing systems. A standardised data exchange service based on SensorThings API is under construction. The observations and datasets are indexed in a searchable resource with human (data portal) and machine (OGC data catalogue service, SensorThings API) access and are described by rich metadata with a plurality of precise attributes. As the data and metadata refer to terms from a vocabulary that follows the FAIR principles, this contributes to the semantic interoperability of the data. Data license and data provenance are provided with the data. Improving the fairness of data involves not only implementing semantic and technical features, but also supporting communities to adopt FAIR practices. Seminars on Data Management Plans and open licenses have been organised to encourage data producers to describe their data management and documentation practices and to associate data with a clear license. Supporting communities as they move towards FAIR data, and providing them with the human resources they need to make progress, remains one of the major challenges of the project.