Biodiversity, the variation within and between species and ecosystems, is essential for human well-being and the equilibrium of the planet. It is critical for the sustainable development of human society and is an important global challenge. Biodiversity research has become increasingly data-intensive and it deals with heterogeneous and distributed data made available by global and regional initiatives, such as GBIF, ILTER, LifeWatch, BODC, PANGAEA, and TERN, that apply diferent data management practices. In particular, a variety of metadata and semantic resources have been produced by these initiatives to describe biodiversity observations, introducing interoperability issues across data management systems. To address these challenges, the InteroperAble Descriptions of Observable Property Terminology WG (I-ADOPT WG) was formed by a group of international terminology providers and data center managers in 2019 with the aim to build a common approach to describe what is observed, measured, calculated, or derived. Based on an extensive analysis of existing semantic representations of variables, the WG has recently published the I-ADOPT framework ontology to facilitate interoperability between existing semantic resources and support the provision of machine-readable variable descriptions whose components are mapped to FAIR vocabulary terms. This contribution will focus on how the I-ADOPT framework can be applied to represent variables commonly used in the biodiversity domain.
The interactions and diversity of organisms within and across the Earth’s ecosystems play
critical roles in the coevolution of the biosphere and the broader Earth system. Human societies
have evolved with and are part of this complex system [
Over the last decades, the need to share data and knowledge on biodiversity has led to
numerous local, regional, and global initiatives resulting in an unprecedented biodiversity data
mobilization [
LifeWatch Italy is one of the national branches of LifeWatch ERIC, the European e-science
infrastructure for biodiversity and ecosystem research (LifeWatch ERIC1) [
The emerging European Long-Term Ecosystem, critical zone and socio-ecological Research
Infrastructure (eLTER RI5 [
The British Oceanographic Data Centre (BODC) is managed and operated by the UK’s National Oceanography Centre (NOC), an independent self-governing organization with a status of charitable company limited by guarantee. BODC is the marine component of the Natural Environment Research Council’s (NERC’s) Environmental Data Centre network. Its mission is to develop, coordinate, and provide specialist data services for marine science communities; to enable innovative use and re-use of data; to ensure long-term curation of valuable and unique marine data resources; and to champion Open Data. BODC runs the NERC Vocabulary Server (NVS)9, an internationally known and globally used infrastructure for the management of and access to controlled vocabularies related to the marine science and associated domains. Vocabularies served by the NVS underpin the SeaDataNet and EMODnet infrastructure. They have been adopted by the Ocean Biodiversity Information System (OBIS) to support data harmonization within and interoperability of the OBIS Darwin Core MeasurementOrFact Extension [12]. These vocabularies, describing and identifying sensor and platform types and models, variables measured or derived, units, methods, and many other essential metadata elements, are used by an increasing number of scientific marine data networks across the globe.
The information system PANGAEA10 is jointly managed by the Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research (AWI) and the Centre for Marine Environmental Sciences (MARUM) at the University of Bremen. PANGAEA is a certified (CoreTrustSeal), trustworthy long-term operating repository providing continued access to more than 400,000 datasets from various sub-disciplines of Environmental Sciences. A large part of PANGAEA’s dataset consists of biodiversity data, ranging from fossil records to modern faunal and floral 6https://cran.r-project.org/web/packages/Taxonstand/ 7http://www.theplantlist.org/ 8https://unece.org/integrated-monitoring 9https://www.bodc.ac.uk/resources/products/web_services/vocab/ 10https://www.pangaea.de/ observations collected through research infrastructures, research projects, and programs as well as individual researchers. The required observed variables are managed in a controlled vocabulary that is routinely undergoing an automated semantic annotation process [13] to link, e.g., taxonomic vocabularies such as WORMS or ITIS to harmonize species-related data. Access to biodiversity datasets which are routinely tagged by a DOI is enabled through support for a number of community-specific and cross-domain standards.
The Terrestrial Ecosystem Research Network (TERN)11 collects and preserves biodiversity data together with other critical terrestrial ecosystem data from the continental scale to field sites across Australia to support researchers and policymakers’ immediate and long-term needs, including state and regional governments. TERN has developed several valuable data services to support the discovery, analysis, and re-use of data. One of the core services is enabling data harmonization and integration through semantic data model12 (i.e., ontology) and machine-interpretable vocabularies. The TERN data model is developed based on the SOSA (Sensor, Observation, Sample, and Actuator)13 and comprises vocabularies for representing ecological survey data, including biodiversity, vegetation, and soil. The datasets are enriched with machine-interpretable vocabularies representing variables measured, features, units, methods, instruments, platforms, organization, and people. The vocabularies are available publicly through the TERN Linked Data Services14. They are developed either in-house based on community-endorsed specifications (e.g., Australian Soil and Land Survey Field Handbook and AusPlots Rangelands Survey Protocols Manual) or imported from external semantic resources (e.g., QUDT15, GCMD16, and NERC). The extensible data model and rich collections of vocabularies are building blocks for improving TERN data curation and discovery.
These are just a few regional implementations demonstrating a highly diverse landscape of data management practices, even when dealing with the same type of resources. This makes it clear that there are still some dificult hurdles to overcome in order to achieve interoperability and convergence between infrastructures in the same domain, and even more challenging across domains. Initial progress has been made in providing machine-readable descriptions of sensors and their observation types through the OGC’s Sensor Web Enablement SensorML, Observations and Measurements (O&M), the Sensor Observation Service (SOS) as well as the W3C’s/OGC’s Semantic Sensor Network (SSN) ontology 17. This work has been further refined within the OGC SensorThings API specification, as well as the upcoming version 3 of O&M. However, “deep metadata” that further contextualizes observations (e.g., methodology or observable properties) is typically represented as coarsely qualified classes. What exactly falls into these classes is currently unconstrained and could be anything ranging from unstandardized free-text to standardized descriptions accessible via fully resolvable URIs.
To address the challenge to properly model and describe observable properties (also called variables), representative members of the above initiatives decided to collaborate under the 11https://www.tern.org.au/ 12https://ternaustralia.github.io/ontology_tern/ 13https://www.w3.org/TR/vocab-ssn/ 14https://linkeddata.tern.org.au/ 15https://www.qudt.org/ 16https://earthdata.nasa.gov/earth-observation-data/find-data/idn/gcmd-keywords 17https://www.ogc.org/docs/is umbrella of the Research Data Alliance. The Working Group InteroperAble Descriptions of Observable Property Terminology (RDA I-ADOPT WG)18 was born in 2019 and is now in its ifnal phase. This group of international terminology providers and data center managers set itself the objective to produce an Interoperability Framework paving the way for seamless terminology alignment for variable descriptions.
This I-ADOPT WG has a strong focus on variables observed in environmental research because it leverages existing eforts to accurately encode what was measured, observed, derived, or computed in relation to Earth’s systems. However, many of the principles it leans on are relevant to or connected with other domains. The construction of the framework has been informed by a review of current practices used in the community. The working group is also iteratively testing and refining the framework through a set of in-depth use cases. Much like a generic blueprint, the refined conceptual framework will be a basis upon which terminology developers can formulate or refine their local design patterns to more easily align with others and to avoid insurmountable inconsistencies. By utilizing the recommendations of the I-ADOPT framework, terminology developers may expand the applicability and interconnectivity of their local resources in a collective attempt to uniformly represent complex properties observed across the environmental sciences, from marine, atmospheric, and terrestrial Earth sciences to biodiversity.
1. Collect user stories 3a. Transform into use cases 3b. Identify functional requirements 2. Collect conceptual models and terminologies 4. Identify information requirements 5b. Test the framework 5a. Define the I-ADOPT Framework 6. Develop design patterns
The I-ADOPT WG has been working for about two years to develop the common framework for agreed-upon components to represent diferent aspects of observable properties. The timeline followed over the last few years involved completing six diferent tasks (see Figure 1): 1. Collect user stories from the community about needs related to semantic representation of observable properties involving nitrogen-related observable properties, following a submission template19. 18https://www.rd-alliance.org/groups/interoperable-descriptions-observable-property-terminology-wg-i-adopt 19https://github.com/i-adopt/users_stories 2. Collect semantic representations from the community, including terminologies, standard vocabularies, ontologies, and conceptual models20. 3. Derive framework requirements by transforming user stories into concrete use cases, categorizing use cases, and defining functional needs within each category 21. 4. Analyze cataloged semantic artefacts to derive the framework’s information requirements – a set of conceptual classes and relationships common to the varying representation mechanisms, whether explicit or implicit. 5. Develop the I-ADOPT Framework, including core variable components, their relationships and definitions, and iterate on feedback from the community. 6. (ongoing) Use the I-ADOPT Framework to develop aligned design patterns across several participating terminology resources for selected use cases.
During the framework development, many potential framework components were considered, discussed, and iterated upon, but only a small subset of those are included in the current recommendation. Discussions of the diferent possible components that were in the end not considered essential to the framework, such as statistical measures, denominations of objects of interest, and property dimensions and units, are documented through the group’s GitHub page and supplemental documents. These eforts resulted in an ontology for providing a machine-readable representation of the framework.
The I-ADOPT Framework ontology22 is a concise ontology (see Figure 2) made of four classes (Variable, Property, Entity, Constraint) and six object properties (hasProperty, hasObjectOfInterest, hasContextObject, hasMatrix, hasConstraint, constrains). It is designed to facilitate interoperability between existing variable description schemes (including domain-specific ontologies, semantic models, and structured controlled vocabularies) and to support further development of machine-readable variable descriptions that re-use components mapped to FAIR vocabulary terms. This first oficial version of the ontology has been developed by a core group of terminology experts and users from the I-ADOPT WG.
The framework is variable-centric and can be used to describe result values of any type of data acquisition events, be it a human-based observation, a sensor-based measurement, a calculation, or a simulation. The Variable describes WHAT has been observed, measured, simulated, or calculated independently of WHERE (site description, geographical coordinates), HOW (procedure, protocol) and WHEN (measurement time, time resolution) the data acquisition has taken place, although in some cases the HOW might be important to include in the variable description, as it might pose certain constraints on it. This makes the application of this concept reusable in diferent settings giving meaning to the determined value. Variables exist as concepts in many vocabularies (see CF Standard Names23 or CSDMS Standard Names24) to 20https://github.com/i-adopt/terminologies 21https://github.com/i-adopt/usecase_analysis 22https://w3id.org/iadopt/ont/ 23https://cfconventions.org/Data/cf-standard-names/47/build/cf-standard-name-table.html 24https://csdms.colorado.edu/wiki/CSDMS_Standard_Names 0..n hasContextObject 0..n
1..1 Property provide metadata for the values provided in databases. They are also known as observable properties (see Observable Property Vocabulary25) or parameters (see BODC Parameter Usage Vocabulary (P01)26 and EnvThes27). This variety of approaches and the lack of a common representation strategy lead to a situation where variable concepts from diferent terminologies are still not directly comparable. Consequently, the data that is annotated with these concepts is not interoperable and requires substantial manual eforts to integrate.
In the following paragraphs, we explain the implementation of the framework by the use of an example28, describing a complex variable that requires all of the components of the framework. The variable description is “concentration of endosulfan sulfate in wet flesh of ostrea edulis” (see Figure 3) and refers to the quantitative result (i.e., requiring a magnitude and unit) of a measurement.
The I-ADOPT ontology, inspired by the atomisation approach of the Complex Property Model [14] and The Scientific Variables Ontology 29 [15, 16], conceives the Variable as a compound concept consisting of at least one entity (the ObjectOfInterest) and its Property, but very often includes further entities contextualizing the target object of observation. Splitting the Variable into constituent concepts enables the reuse and the mapping of these components in the context of other variables. The ObjectOfInterest (endosulfan sulfate) can be identified independently of an observable characteristic (concentration). By concatenating the constituent elements with prepositions the preferred label of a variable can be automatically created (compare with the upper side of the table in Figure 3). However, the I-ADOPT Framework does not provide any recommendation how the label of the variable should be composed or which order to follow as its main contribution targets machine-readability of the concept. Not included in the variable description are units, statistical measures (unlike in the Complex Property Model), measurement methods, and time-related or geographical location information. Units are essential information for describing measures, but a quantitative variable might be expressed in diferent units which requires units be modeled independently of variables. An I-ADOPT variable associated with a 25http://registry.it.csiro.au/def/environment/_property 26https://www.bodc.ac.uk/resources/vocabularies/parameter_codes/ 27http://vocabs.lter-europe.net/EnvThes/ 28The following prefixes will be used throughout the examples: obo:<http://purl.obolibrary.org/obo/> ; lifewatch:<http://thesauri.lifewatchitaly.eu/PhytoTraits/index.php?> ; nerc:<http://vocab.nerc.ac.uk/collection/> ; worms:<http://marinespecies.org/aphia.php?p=taxdetails&id=> 29http://www.scientificvariablesontology.org/ concentration of endosulfane sulfate in wet flesh of ostrea edulis The wet weight concentration of the specified analyte in the specified organism or part thereof.
nerc:P01/current/IC000344 unit can, however, be interpreted as a more specific variable. By isolating the Variable concept from these supplementary information elements, it is possible to use the I-ADOPT framework in a broader context interoperating with other representations of observations. In this way the Variable with its explicit constituents can be conceived as an extended and machine-readable conceptualization of the respective WHAT concept of these models (e.g. ObservableProperty in SOSA30, Characteristic in OBOE31).
The Property (concentration) is a type of characteristic of the ObjectOfInterest. The cardinality of the object property hasProperty is 1..1, meaning that a Variable has to have exactly one Property. The Property is in most cases a more generalized concept than the variable itself, as the variable adds more detail to the property. This has the advantage that the property term can be reused in other variable descriptions and further on allows to identify similar measurements across diferent variables (e.g., concentration regardless of the substance observed).
The Entity is any object (endosulfan sulfate) or process (e.g., erosion) that has a role in an observation. An Entity may play one of the following roles: ObjectOfInterest, ContextObject, or Matrix. The roles of the entities in the variable description are implemented as object properties in the I-ADOPT ontology. Whether the involvement of a particular entity is meaningful enough to include in the variable description, depends on the specific context.
• hasObjectOfInterest: The ObjectOfInterest (endosulfan sulfate) is the Entity whose property is observed. A Variable requires exactly one ObjectOfInterest (cardinality: 1..1). • hasContextObject: The ContextObject (Ostrea edulis) is the Entity that provides additional background information regarding the ObjectOfInterest. A Variable can have more than one ContextObject (cardinality: 0..). • hasMatrix: The Matrix (flesh ) is the Entity in which the ObjectOfInterest is contained.
A Variable might only have one Matrix (cardinality: 0..1).
The Constraint (wet) limits the scope of the observation and confines the context to a particular state. It describes properties of the involved entities (flesh ) that are relevant to the 30https://www.w3.org/TR/vocab-ssn/ 31https://github.com/NCEAS/oboe/ particular observation. Oftentimes Constraints take the form of observations themselves, fixing other properties that influence the main observation considered. It may constrain ( constrains) multiple entities playing the role of ObjectOfInterest, Matrix, or ContextObject. Constraints are optional, so the cardinality of the object property hasConstraint is 0...
The I-ADOPT ontology provides a framework to adequately address the data interoperability
requirements emphasized by the FAIR Guiding Principles [
Many variables in the biodiversity domain are human-based observations with a qualitative result value selected from a controlled list. Some elements for the Occurrence, Organism, and Taxon metadata components of the Darwin Core standard require qualitative values for which a few recommended SKOS vocabularies were created by the TDWG Darwin Core Maintenance Group32. LifeWatch Italy vocabularies provide specific terminology for functional traits of several groups of aquatic organism and alien species[17]. An example of how the I-ADOPT framework can be applied to qualitative results of human-based observations is provided in Figure 4 using a variable from the LifeWatch Phytotraits thesaurus. The variable is called shape implicitly targeting cells, which is explained in the definition of this specific variable. It is implemented as a SKOS concept and has as more specific concepts qualitative values such as sphere, cube, or ellipsoid. In dataset spreadsheets these values would normally populate the cells of the variable column, but we recommend to use instead the IRIs of these narrower concepts as manifestations of the observed value. However, a more convenient way to store this type of information would be to use RDF for the entire dataset. The constituent components shape (of cell) The approximate 3 dimensional shape of a cell, described as one or more geometric solids.
lifewatch:tema=23&/shape
Example: sphere ( lifewatch:tema=41&/sphere ) of the variable shape (of cell) are the more general concept of shape (the Property) and the ObjectOfInterest the concept cell. It is not always possible to find more general properties which could be reused in the context of other variables. For example, the Darwin Core metadata element degreeOfEstablishment (degree to which an Organism survives, reproduces, and expands its range at the given place and time) would be represented by the variable with the IRI 33 and composed by the Property degreeOfEstablishment with the same IRI and the ObjectOfInterest organism34.
The RDA WG I-ADOPT will in its final phase produce guidelines for the community on how to apply the framework. These will include design patterns to generalize variable descriptions and their modelling for their reuse in diferent settings (e.g., concentration of substance in biota). The authors of this paper will evaluate the adoption of the I-ADOPT approach in their own infrastructures. One first proof-of-concept will be the SKOS representation of the eLTER Standard Observation variables based on the I-ADOPT Framework in EnvThes. An important task will also be to work on mappings to existing observation models (e.g., OBOE) and to foster the ongoing collaboration with initiatives like OGC Observations and measurements Standard Working Group35 and DDI-CDI36 to evaluate how I-ADOPT might increase semantic interoperability of the variable concept within those approaches.
33https://dwc.tdwg.org/doe/ 34http://purl.obolibrary.org/obo/OBI_0100026 35https://github.com/opengeospatial/om-swg 36https://ddi-alliance.atlassian.net/wiki/spaces/DDI4/pages/860815393/DDI+Cross+Domain+Integration+ DDI-CDI+Review
The authors like to thank the Research Data Alliance, eLTERplus, ENVRI-FAIR (grant agreements No 871128, No 824068) and Alison Pamment (NCAS, UK) for their support.