2. Find Natural Capital Data service - Approach to classifying records

Created: 05 June 2025 Updated:

Natural capital search

The NCEA ‘Find Natural Capital Data’ search service enables users to discover natural capital relevant data from third party sources, with over 8000 records currently accessible. It also provides access to an expanding collection of NCEA data, reports, and publications all from a single, centralised digital location.  

The service offers two search methods:  

• A Quick Search, allowing users to enter keywords directly.  

• A Natural Capital Search, which uses a natural capital vocabulary to guide users through relevant options.  

Search results can be viewed as a list or displayed on a map, with additional filters available to refine results. This functionality supports the use of natural capital terminology and significantly improves the ease and efficiency of locating relevant natural capital data.

Natural Capital Vocabulary

The natural capital vocabulary used in the NCEA Natural Capital Search was developed by a Task and Finish group of natural capital experts from across all NCEA partners. This is a new and experimental vocabulary that will undergo further refinement based on user feedback and evaluation.

Terms and definitions in the vocabulary have been adapted from established standards where possible including:  

• UK Biodiversity Action Plan (UK BAP) broad habitats 

• European nature information system (EUNIS) 

• Common International Classification of Ecosystem Services (CICES) 

• Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) 

• Marine Evidence based Sensitivity Assessment (MarESA) 

• NCEA’s Conceptual Ecological Model (CEM)

The full list of terms and definitions can be found in the Natural Capital Vocabulary table.

The Natural Capital hierarchy diagram provides a cut of the vocabulary table to demonstrate how the vocabulary is structured across the three tiers of the natural capital search hierarchy with the colours used to link categories and sub-categories by theme. The arrows show an example classification of a record tagged as ‘Natural asset’ (L1), ‘Terrestrial and freshwater habitats’ (L2) and ‘Broadleaved, mixed and yew woodland’ and ‘Coniferous woodland’ (L3).

Large language model approach

There is a large body of existing data that could be relevant when taking a natural capital approach to evidence gathering and decision making. However, metadata records for these datasets do not use consistent natural capital language, making them difficult to find and access. Enriching the metadata with natural capital terms from a controlled vocabulary, allows users to search for data using the Natural Capital Search and widens access to natural capital relevant data. 

Manually classifying all natural capital relevant legacy datasets would take a large amount of time and resource. The NCEA is trialling an innovative approach to classify metadata records with natural capital terms using a large language model. This saves time that would be spent manually classifying records and means a larger number of datasets can be tagged and made accessible through the Find 

Natural Capital Data search service. The classification process can also be applied to new data sources and efficiently repeated as the vocabulary of natural capital terms is refined. 

The large language model approach is being used experimentally and will continue to be improved. If you see results that appear to have been classified incorrectly, we encourage you to notify the team via the feedback form so that we can improve the service.