Great Yellow’s Mission: From Extractive to Regenerative, and Why We Exist
Here at Great Yellow, our core mission is to shift the economy from extractive to regenerative. In this, we share a common purpose with many in the nature space: to make landscape recovery projects successful and help them proliferate. We are in the midst of a biodiversity and climate crisis; solving this requires collaborative teamwork effort across multiple domains and stakeholders to create investable outcomes.
This is a solvable problem, and doesn’t require huge leaps in technology, or even policy. What it does require is getting the basics right together.
The Challenge: The Complexity of Landscape Recovery
Landscape recovery projects sit at the intersection of many domains: ecosystem services, ecology and habitats, agriculture and land use, document and project management, environmental modelling, finance, geospatial and temporal boundaries, scientific measurements, stakeholder management and regulation.
Across projects we see similar hurdles. Fragmented data and processes that are hard to scale, including ineffective data organisation and retrieval across a huge variety of documents and datasets (“heterogeneous data”). Multiple stakeholders with different interpretations of terminology. The challenges of collecting and labelling data from multiple third parties and stakeholders and managing multiple disconnected tools, and of responding to multiple stakeholder requests for reports in different formats and with different data requirements.
Users don’t reliably tag documents, leading to poor search, filtering, and discoverability. Constructing efficient and effective folder structure is difficult and error prone. We need scalable ways to automatically classify documents and populate metadata to enable programmatic access, search, and downstream processing.
There is a clear need for concise, consistent terminology across these domains to facilitate interoperability of systems and effective data retrieval.
The Opportunity: A Chance to Build Stronger Foundations
There are a plethora of different methodologies that can assist here, including rule-based tagging, supervised machine learning, LLM-based classification, unsupervised clustering and auto-discovery. But one such approach seems particularly appealing: ontologies and knowledge-graph classification.
Ontologies let us declare structure and apply semantic meaning to documents and measurements, and support semantic search: queries that understand meaning, not just keywords. For example: “Find all bird surveys in wetland habitats in Norfolk since 2022”.
Why Ontologies Matter: Making Sense of Complexity: What is an Ontology?
Having a shared language of concepts and the relationships between them is crucial, not only for interoperability of data across systems, but for semantic search, reasoning, systems integration and even grounded natural language applications.
So what actually is an ontology? Well, firstly, it builds on something called a “taxonomy”: A hierarchical classification system that organises concepts into parent–child relationships in a tree-like structure.
Example: “Project Data” → “Environmental Data” → “Habitat Surveys” → “Bird Count Reports”. But taxonomies are not enough to model the relationships.
An ontology can be described as a machine-readable model of a domain, defining not only concepts but also the relationships between them, supporting reasoning, interoperability, and semantic search.
Example: A graph showing how specific interventions relate to species, habitats, funding sources, and monitoring results.
Pre-existing ontologies are freely available for several domains, and when they are designed according to certain standards (e.g. Basic Formal Ontology (BFO)) they can be consistently combined. Ontologies are critical in the development of graph networks and the application for grounded, traceable high-performance retrieval-augmented generation.
The Current Landscape: What’s Out There Today (and What’s Missing)
Multiple ontologies are open-sourced and freely available, often consistent and interoperable with top-level frameworks like BFO. Whilst there are some ontologies on the boundaries of ecology and landscapes, none specifically cover the emerging domain of landscape recovery. Some ontologies of particular interest for landscape recovery include
- BFO for top-level ontologies
- SOSA/SSN for observations/samples/procedures
- GeoSPARQL 1.1 for features, geometries, and WKT/GeoJSON literals
- CRS for coordinate reference systems
- QUDT for units and quantities (including money)
- OWL-Time for project timelines
- FOAF / ORG for people & organisations
- DOAP / ORG for project and stakeholder management
- DCTERMS / DCAT / SKOS for docs, datasets, and controlled code lists
- AGROVOC / Agronomy / AGRO for Agriculture & Land Use
- SWEET for Earth and Environmental Terminology
- FIBO for finance and ESG
- ENVO / ESM / ESCO / EUNIS for ecology & habitat classifications
Our Work So Far: Experimenting, Learning, Building
We have been experimenting with compiling a combined ontology spanning the required domains and testing with sample data. This included manually tagging sample documents and testing effective retrieval via SPARQL (a common and standard query language for graph databases). We also developed the ontology using tooling such as Protégé and freely available graph applications. We built sample data and competency questions to test information retrieval.
This work is available in the open-source ontology repository: https://github.com/GreatYellow-ltd/landscape-recovery-ontology
How might you benefit as a user: By providing a common structure for project, habitat, and finance data, it makes machine-to-machine sharing over Application Programming Interfaces (APIs) far more seamless, reducing the friction of integrating systems or exchanging datasets with partners. It also opens the door to retrieval-augmented generation (RAG), where large language models can pull grounded, ontology-linked facts rather than relying on generic or unverified text. This means reports, dashboards, or even conversational interfaces can be built on traceable, interoperable data, strengthening both trust and scalability.
Why We’re Open-Sourcing: Collaboration Over Competition
Ideas around sustainable development have been around since the 1970’s, and concept of ecosystem services gained traction in the 1990’s. Nature tech is arguably in its infancy, but it’s undergoing rapid growth. All the players I’ve spoken to are united by the goal of resilient and healthy ecosystems, and it feels right to be “walking our talk” and invite collaboration.
This ontology will improve with collaborative teamwork. Our hope is that the community can pull together and benefit collectively.
We would love to hear from others operating in this space, and welcome feedback, additions and contributions. Feel free to open pull requests in the repository and get involved, share use cases, and let us know if this is helpful.
Acknowledgements: Standing on the Shoulders of Giants
This wouldn’t be possible without the prior work of others. I’d specifically like to acknowledge the ontologies used, and the following authors:
- Noy, N. F., & McGuinness, D. L. (2001). Ontology Development 101: A Guide to Creating Your First Ontology. Stanford KSL/Protégé. PDF. https://protege.stanford.edu/publications/ontology_development/ontology101.pdf
- Ceccaroni, L., & Oliva, L. (2012). Ontologies for the Design of Ecosystems. In Universal Ontology of Geographic Space: Semantic Enrichment for Spatial Data (IGI Global). Chapter PDF/TOC: https://igiprodst.blob.core.windows.net/ancillary-files/9781466603271.pdf (overview page: https://www.igi-global.com/viewtitle.aspx?TitleId=64001). Also available at: https://www.researchgate.net/publication/311903669_Ontologies_for_the_design_of_ecosystems
- BFO-2020 (Basic Formal Ontology) — GitHub repository for artifacts conformant with ISO/IEC 21838-2:2020: https://github.com/BFO-ontology/BFO-2020
- Drakou, E. G., Lemmens, R. L. G., & Ayuninshih, F. (2019). Designing an Ecosystem Services Ontology within GEOBON. Biodiversity Information Science and Standards 3: e36338. DOI: 10.3897/biss.3.36338. (UTwente record: https://research.utwente.nl/en/publications/designing-an-ecosystem-services-ontology-within-geobon)
- Bennett, B. (2010). Foundations for an Ontology of Environment and Habitat. In FOIS 2010: Formal Ontology in Information Systems (IOS Press), pp. 31–44. (ACM/IOS references and metadata) https://dl.acm.org/doi/proceedings/10.5555/1804715 . Also available at: https://www.researchgate.net/publication/221234950_Foundations_for_an_Ontology_of_Environment_and_Habitat
- Buttigieg, P. L., et al. (2016). The environment ontology in 2016: bridging domains with increased scope, semantic density, and interoperation. Journal of Biomedical Semantics. https://jbiomedsem.biomedcentral.com/articles/10.1186/s13326-016-0097-6
- Affinito, F., Holzer, J. M., Fortin, M.-J., & Gonzalez, A. (2025). Towards a unified ontology for monitoring ecosystem services. Ecosystem Services. ScienceDirect article page: https://www.sciencedirect.com/science/article/pii/S2212041625000300
- Lepczyk, C. A., Lortie, C. J., & Anderson, L. J. (2008). An ontology for landscapes. Ecological Complexity, 5(3), 272–279. https://doi.org/10.1016/j.ecocom.2008.04.001
Next Steps: Let’s Build This Together
This is an ongoing experiment. We’re inviting feedback, asking for collaboration, and shared ownership of the ontology’s evolution. Let’s do this!
