TERN/BDR implementation evidence for alignments, examples, imports

[leskneebone]

TERN and BDR provide useful implementation evidence for ecological observation and sampling workflows. Some TERN terms appear to duplicate or closely specialise SOSA/SSN terms, some look like profile or SHACL material, and some may point to modelling patterns that deserve clearer SOSA/SSN guidance.

The working alignment artifact is:

ssn/rdf/ontology/alignments/tern-ssn-mapping.ttl

related: #272

Summary

TERN and the Australian Biodiversity Data Repository provide implementation evidence for ecological observation and sampling workflows. An exploratory alignment suggests that many core TERN observation/sampling terms map cleanly to SOSA/SSN, while several TERN patterns may identify areas where SOSA/SSN guidance or alignment material could be improved.

In particular, ecological sites and transects appear to exercise the sosa:SpatialSample pattern; surveys and site visits appear to exercise sosa:ExecutionCollection; TERN sub-sampling exercises sample derivation chains; and TERN value wrappers, units, uncertainty, and calibration expose practical modelling questions around results and result quality.

TERN/BDR may be useful implementation evidence for guidance, examples, profile material, or alignment notes. In some cases, TERN Ontology itself may contain classes and properties that SOSA/SSN could import, mirror or customize for implementation in the SOSA/SSN namespace.

The main proposed direction is to treat TERN as implementation evidence for ecological field workflows rather than as a vocabulary to be collapsed into SOSA/SSN.

The draft alignment follows the style of existing SOSA/SSN alignment modules:

• direct owl:equivalentClass, rdfs:subClassOf, and rdfs:subPropertyOf axioms;
• explanatory comments for non-obvious modelling choices;
• no reified mapping records;
• profile/domain vocabulary listed in prose where no direct SOSA/SSN mapping is proposed.
• selective skos:editorialNote annotations following conventions (such as in sosa-sdo.ttl), used to flag
  terms that may be candidates for SOSA/SSN examples, guidance, classes, or properties.

Several TERN patterns appear especially relevant to current SOSA/SSN work:

• ecological sites and transects as spatial samples;
• surveys and site visits as execution collections;
• material sample derivation and sub-sampling;
• result values, value wrappers, units, and uncertainty;
• calibration as an execution pattern;
• profile terms for ecological methods, feature types, observation themes, and local metadata.

Mapping Table

TERN term or pattern	Current draft mapping	Comment	Issues for investigation
tern:Deployment	owl:equivalentClass sosa:Deployment	Current SOSA includes sosa:Deployment; the previous draft used ssn:Deployment.	Check whether TERN deployment semantics include only arrangement/use of assets, or also project/campaign metadata that should stay in a profile.
tern:FeatureOfInterest	owl:equivalentClass sosa:FeatureOfInterest	Appears to be a close conceptual match.	Confirm TERN does not impose ecological-domain constraints that make this narrower than generic SOSA.
tern:Observation	owl:equivalentClass sosa:Observation	Direct duplicate or close alignment.	Confirm current TERN shapes around values and attributes do not imply a materially different result model.
tern:ObservationCollection	owl:equivalentClass sosa:ObservationCollection	Directly useful for ecological observation bundles.	Check whether TERN collections are homogeneous in the same sense as current SOSA collection guidance.
tern:Platform	owl:equivalentClass sosa:Platform	Close fit where the thing hosts systems or procedures.	Some TERN platform subclasses may be regional/domain categories rather than generic alignment terms.
tern:Procedure	owl:equivalentClass sosa:Procedure	Close fit for reusable methods, protocols, and workflows.	TERN tern:Method is currently modelled as skos:Concept; clarify relationship between TERN Method concepts and SOSA Procedure individuals.
tern:Sample	owl:equivalentClass sosa:Sample	Good generic alignment.	Confirm whether all TERN samples are intended as SOSA samples, especially spatial/site-like samples.
tern:Sampler	owl:equivalentClass sosa:Sampler	Close fit; current SOSA notes a Sampler need not be a physical device.	Check whether TERN sampler usage includes people, teams, or software, and whether this needs guidance.
tern:Sampling	owl:equivalentClass sosa:Sampling	Close fit for sampling activities.	Confirm whether TERN sampleCollectionTime should map to resultTime, phenomenonTime, startTime, or endTime in different cases.
tern:Sensor	owl:equivalentClass sosa:Sensor	Close fit.	Check whether TERN distinguishes instrument, observer, and software in ways that should be profiled rather than asserted as equivalence.
tern:System	owl:equivalentClass sosa:System	Current SOSA has sosa:System; old draft used ssn:System.	Confirm compatibility with existing SSN system/capability modules.
tern:Input	rdfs:subClassOf ssn:Input	Retained as compatibility mapping to older SSN input pattern.	Investigate whether current SOSA should expose a non-deprecated/common input class, or whether this should remain legacy only.
tern:ObservableProperty	rdfs:subClassOf sosa:Property	sosa:ObservableProperty is deprecated; generic sosa:Property is safer.	Decide whether a close-match note to deprecated sosa:ObservableProperty is helpful or distracting.
tern:Result	rdfs:subClassOf sosa:Result	sosa:Result is deprecated but still present.	Current SOSA prefers hasResult / hasSimpleResult; consider whether TERN value wrappers should be documented as a profile pattern instead of class alignment.
tern:Attribute	rdfs:subClassOf sosa:Property	TERN attributes behave like properties/qualities of entities.	Review whether sosa:Property or ssn:Property is preferred in current modules.
tern:Intervention	rdfs:subClassOf sosa:Actuation	TERN intervention is a domain-specific actuation-like activity.	Check whether all TERN interventions intentionally change the state of the world, as SOSA Actuation requires.
tern:MaterialSample	rdfs:subClassOf sosa:MaterialSample, dwc:MaterialSample	Current SOSA has a precise sosa:MaterialSample class; this is stronger than the old generic sosa:Sample mapping.	Investigate whether owl:equivalentClass is too strong because TERN may also include Darwin Core-specific specimen expectations.
tern:Site	rdfs:subClassOf sosa:Platform, sosa:SpatialSample; skos:editorialNote candidate example	Sites may host systems/procedures and also represent spatial sampling features.	Potentially contentious: being both platform and spatial sample may be valid in TERN, but SOSA guidance may need examples.
tern:EcosystemProcessesSite	rdfs:subClassOf sosa:Platform, sosa:SpatialSample; skos:editorialNote candidate example	Similar to tern:Site, but more domain-specific.	Decide whether this belongs in a generic alignment or only a TERN/BDR profile.
tern:Transect	rdfs:subClassOf sosa:SpatialSample; skos:editorialNote candidate class	SOSA SpatialSample examples include traverse-like spatial sampling features.	Check whether transect endpoints and direction should map to GeoSPARQL geometry or remain profile properties.
TERN platform subclasses	rdfs:subClassOf sosa:Platform	FixedPlatform, MobilePlatform, CosmOzStation, EarthObservationSatellite, and FluxTower fit as platform specialisations.	Most are probably regional/domain vocabulary; do not propose as SOSA core terms.
tern:Survey	rdfs:subClassOf sosa:ExecutionCollection; skos:editorialNote candidate guidance	Survey can bundle observations, samplings, and related activities.	Confirm whether all surveys are execution collections, or whether some are plans/campaigns better modelled as procedures or deployments.
tern:SiteVisit	rdfs:subClassOf sosa:ExecutionCollection; skos:editorialNote candidate guidance	Site visits look like field execution bundles.	Need examples: a site visit may be a single visit event, a collection of executions, or both.
tern:Calibration	rdfs:subClassOf sosa:Execution; skos:editorialNote candidate guidance	Calibration is an execution-like activity, often involving systems and procedures.	Current SOSA has no explicit calibration class; this may be useful implementation evidence for guidance or extension material.
tern:hasAttribute / tern:isAttributeOf	rdfs:subPropertyOf sosa:hasProperty / sosa:isPropertyOf	Good generic property relation mapping.	Confirm whether TERN attributes are always properties of features, or also metadata attributes of datasets/resources.
tern:hasSubSample / tern:isSubSampleOf	rdfs:subPropertyOf sosa:hasSample / sosa:isSampleOf; skos:editorialNote candidate property	TERN defines this as a child sample derived from a material sample.	Current SOSA also has hasOriginalSample, isOriginalSampleOf, and ultimate-sample relations; these may be more precise for sub-sampling chains.
tern:hasObservation, tern:hasSampling, tern:hasSubActivity	rdfs:subPropertyOf sosa:hasMember	Useful for collections such as survey/site visit.	Need domain/range review. If the subject is not a collection, sosa:hasMember may be too broad or wrong.
tern:hasSurvey	rdfs:subPropertyOf sosa:isMemberOf	Links an activity to the survey under which it was conducted.	Check whether this should instead be a deployment/campaign relation or a provenance relation.
tern:hasSite / tern:isSiteOf	rdfs:subPropertyOf sosa:hasFeatureOfInterest / inverse	Links activities or features to site-like spatial samples.	Domain varies in TERN; mapping may be correct for site visits/samplings but not for every use.
tern:hasSiteVisit / tern:isSiteVisitOf	rdfs:subPropertyOf sosa:isFeatureOfInterestOf / inverse	Intended to connect a site to visits concerning it.	Direction and semantics need checking carefully; may be better as profile relation rather than generic SOSA property mapping.
tern:hasSamplingPoint / tern:isSamplingPointOf	rdfs:subPropertyOf sosa:hasFeatureOfInterest / inverse	Sampling points are likely spatial samples/features of interest.	Need to clarify whether a sampling point is a sample, a site, a feature, or a geometry marker in TERN.
tern:resultDateTime	rdfs:subPropertyOf sosa:resultTime	TERN notes historical datatype precision concerns with older SOSA.	Check current SOSA resultTime datatype flexibility; this may remove need for a TERN-specific property.
tern:sampleCollectionTime	rdfs:subPropertyOf sosa:resultTime	Represents time at which a sample was collected during sampling.	Could sometimes be phenomenonTime, startTime, or endTime; needs examples before fixing.
tern:hasValue	rdfs:subPropertyOf sosa:hasResult	TERN value wrapper can be viewed as a result node.	Risk: sosa:hasResult is functional in current SOSA; review multi-value or attribute-value patterns in TERN.
tern:hasSimpleValue	owl:equivalentProperty rdf:value	TERN definition says simple value is equivalent to rdf:value of a value node.	Do not equate with sosa:hasSimpleResult without care; domains differ.
tern:unit	rdfs:subPropertyOf qudt:unit	TERN asks users to use QUDT units.	Current SOSA examples mention units but do not impose a unit property; QUDT relation seems better than schema:unitCode.
tern:uncertainty	No direct axiom; profile/domain vocabulary	Related to result quality, but it is a datatype property on TERN value nodes.	Investigate relation to sosa:resultQuality, DQV, QUDT uncertainty quantity kinds, and value-wrapper design.
tern:hasCalibration, calibration result properties	No direct axiom; profile/domain vocabulary	Calibration pattern is important implementation evidence.	Could motivate SOSA/SSN guidance for calibration executions, inputs, procedures, systems, and result quality.
tern:undertakenAfter, tern:undertakenBefore, related method sequencing	No direct axiom; profile/domain vocabulary	These appear to sequence methods/procedures.	Investigate whether PROV, OWL-Time, P-Plan, or SOSA procedure composition guidance is the right home.
Feature types, observation themes, parameters, methods	Mostly SKOS/profile mappings	TERN uses controlled vocabulary links such as hasFeatureType, hasObservationTheme, hasParameter, hasMethod.	Likely profile guidance rather than SOSA core; may be useful for BDR/TERN SHACL examples.

Suggested Review Questions

1. Should ecological sites, stations, plots, quadrats, and transects be documented more clearly as sosa:SpatialSample examples?
2. Is sosa:ExecutionCollection the right general pattern for field surveys and site visits that bundle observations and samplings?
3. Should SOSA/SSN provide clearer guidance for sample derivation chains, especially hasSample, hasOriginalSample, and isSampleOfUltimateFOI?
4. How should value-wrapper models align with current sosa:hasResult and sosa:hasSimpleResult, given that sosa:hasResult is functional?
5. Should calibration be handled only as a generic sosa:Execution, or does it deserve explicit guidance or an extension pattern?
6. How should result uncertainty be expressed: as sosa:resultQuality, DQV quality measurements, QUDT quantity/value structures, or profile-level properties?
7. Which TERN terms are best left as TERN/BDR profile vocabulary rather than candidates for SOSA/SSN terms?

Notes

The current alignment draft at alignments/tern-ssn-mapping.ttl uses SOSA/SSN terms where they are already available and avoids asserting mappings for TERN terms that appear to be ecological profile vocabulary, local metadata, governance, identifiers, or controlled-vocabulary links.

The most useful WG contribution may be a small issue around one or two patterns, not the full mapping. Possible first candidates are:

• sosa:SpatialSample examples for ecological sites/transects;
• sosa:ExecutionCollection examples for surveys/site visits;
• result-value and uncertainty guidance using TERN/BDR as evidence.