Robot-ready by construction: an ontology-driven, SHACL-validated submission pipeline for national ecological monitoring data
Andrew Tokmakoff · AI-Readiness Metrics and Metadata for Biodiversity Data
The short versionDriving both field data capture and RDF export from one semantically annotated specification makes monitoring data standards-compliant and machine-ready from the start.
What this was about
Andrew Tokmakoff described how TERN Ecosystem Surveillance moved from a decade-old Android app and Postgres database to EMSA (Ecological Monitoring System Australia), built for the federal government, where a single data model specification drives both the offline-capable field app (a PWA) and the exporter that pushes RDF to Australia's national Biodiversity Data Repository. Protocol steps and attributes are mapped to IRIs, an OpenAPI 3.0 specification generates UI elements (e.g. lookup-table references) and validates JSON on the way in, and the exporter maps flat JSON into the graph-based ABIS and TERN ontologies, resolving vocabularies live via SPARQL and validating output with SHACL shapes. Data therefore arrive 'robot-ready by construction' rather than by later remediation.
Why it matters. It demonstrates a working alternative to retrofitting standards onto legacy data: specification-driven collection with validation at both boundaries, feeding a national biodiversity knowledge graph.
In the room
- TERN is Australia's national terrestrial ecosystem research infrastructure (funded through NCRIS); Ecosystem Surveillance, based in Adelaide since 2011, runs a long-term standardised plot network, mainly in the rangelands.
- Legacy system: AusScribe Android app, Postgres and CouchDB; new system: EMSA protocols and software, with a PWA field app and a Strapi headless CMS back end ('Core') on Postgres.
- Two routes to standards-compliant data: collect first and map later, or, as with EMSA, start from protocol definitions that drive both collection and RDF creation.
- Data are structured surveys, not just occurrences; protocol steps and procedures are mapped to IRIs, making data self-describing.
- An OpenAPI 3.0 specification (close to JSON Schema) is a machine-readable contract: the app builds its UI from it (e.g. an x-lutRef veg growth form lookup) and the exporter reads the same key to map codes into the graph.
- The exporter maps flat JSON to graph models (ABIS national profile, TERN ontologies, EMSA protocol-specific SHACL shapes), resolving EMSA vocabularies via SPARQL against GraphDB with caching, so it adapts as protocols evolve.
- Quality gating at both boundaries: JSON Schema validation in the app and on submission; SHACL validation of RDF on export.
Notable moments
Transcript
Automatically generated captions can contain mistakes, especially in names and technical terms. Times are relative to the room recording.
Read the transcript ↓
Related talks.
Reliable LLM-assisted curation of ecological survey data: a deployed agent bridging field collection and data curation
An LLM agent that proposes findings for curator approval, with confirmed patterns turned into deterministic rules, catches plausibility errors that schema validation misses.
The WildLIVE Portal: Towards Scalable Wildlife Monitoring through the Integration of Citizen Science and Human-in-the-Loop AI Workflows
WildLIVE combines a model registry, provenance-tracked human verification and FAIR publication to move camera trap data from raw images to publishable data.
Delivering AI for biodiversity data in the public sector: data handling, security and humans-in-the-loop when small teams build at pace
Small public-sector teams can use AI tools productively if they set clear boundaries between prototype, dev, test and production and keep expert humans in the loop.