Provenance and source checking

A layer can be technically accurate and still be the wrong evidence for the question in front of you.

Provenance is the record of where information came from, when it was created or checked, what it was intended to represent, how it has been changed and what authority it carries. In Māori GIS this matters because an official public dataset, a historical record, a local name and Māori-held knowledge can all describe the same place while having very different purposes and authority.

The common source record

Use this as the default pattern across MāoriGIS.nz projects. Not every project needs every field, but the structure should remain recognisable.

Field	Record
source_id	stable project identifier
source_name	dataset, document, plan, imagery, kōrero or local source name
source_org	organisation, repository or knowledge holder
source_ref	URL, record ID, plan number, page reference or controlled internal reference
source_date	source, capture or publication date
checked_date	date the source was last checked for this project
purpose	what the source was created to represent
authority_type	official, administrative, technical, historical, Māori-held, interpreted or model-generated
processing	important transformations, joins, digitising, georeferencing or analysis
limitations	known uncertainty, scale, completeness or interpretive limits
sensitivity	public, internal, restricted or a locally defined class where needed
sharing	relevant redistribution, publication or audience conditions

A URL by itself is not adequate provenance. Web addresses change and rarely explain what happened after the source was obtained.

Six questions before use

For any source that materially affects a decision or output, answer:

1. Who created, maintains or holds it?
2. What exactly is the source?
3. What date or version does it represent?
4. What was it created to represent?
5. What kind of authority does it have for this question?
6. What should not be inferred from it?

If those cannot be answered, treat the source as discovery material until they can.

Keep authority types visible

Do not collapse different evidence into one generic idea of an authoritative source.

A useful distinction is:

• official or statutory source, such as a Gazetteer record or cadastral parcel
• administrative source, such as a council planning layer
• scientific or technical source, such as LiDAR-derived elevation
• historical source, such as a plan, minute book or archived aerial photograph
• Māori-held or locally governed source, where authority comes through the relevant people, relationships and kaupapa
• practitioner interpretation, where an analyst has derived or classified information from other evidence
• model-generated or AI-assisted output, which remains a candidate or derived result until checked

One category does not automatically outrank another for every purpose.

Evidence and interpretation

Where a project involves research or judgement, keep these separate where practical:

Type	Example
source evidence	a plan labels a route in 1912
observation	a track is visible on 2025 aerial imagery
mapped relationship	the track intersects the cadastral road parcel
interpretation	the route may warrant further access research
unresolved question	whether a legal right of access exists

This is more useful than storing one field called truth or status and hiding how the result was reached.

Keep source and working data separate

Keep original downloads or source material unchanged where practical. Perform cleaning, joins, georeferencing and analysis on working copies.

A simple structure is:

01_source
02_working
03_project
04_outputs

This makes it easier to answer two questions later: what did we receive, and what did we change?

Derived information needs provenance too

If you georeference a historical plan, extract a place name with AI, calculate a hazard intersection, classify imagery or digitise a feature, record the derivation.

Useful derived-data fields include:

source_id
method
processing_date
responsible_role
geometry_source
confidence_note
review_status

Do not let a derived polygon lose the uncertainty that existed in its source.

For ingoa wāhi

Keep the wording in the source separate from a normalised or approved display name. An official Gazetteer name, a historical spelling and a locally held name can coexist without being treated as interchangeable.

Useful fields include name_as_source, matched_name, name_source, name_source_id, name_status and review_status.

See Place names and Check places found with local AI.

For Māori land

Keep mapped geometry, Māori Land Court record context, cadastral parcel geometry and internal whenua information distinct. Record the source and date of the boundary rather than allowing an old layer to appear current simply because it still draws neatly on the screen.

See Māori land and Māori land blocks versus parcels.

For historical research

Every mapped feature should be traceable to the document, page, plan, image or record that supports it. Keep exact source wording before normalising names or interpretations.

Use the spatial research log and Archives to map.

For AI-derived information

Keep the source passage or evidence reference. Do not replace it with the model's summary. Record which material was searched and label AI-extracted or model-generated results as candidates until reviewed.

See Local AI and Māori GIS.

Before publishing

A provenance record should survive into the output in a form suited to the audience. That may be a map source note, metadata page, appendix, pop-up field or internal project register.

The public version does not need to expose restricted source details. It does need enough information to avoid making derived or approximate information look unexplained or absolute.

Use Publishing maps and Protect, share and publish.

Related mahi

• Geospatial Data Catalogue
• Worked examples
• Historical research
• Māori Data Sovereignty
• Learning pathway

Last reviewed: 4 September 2026