Skip to main content

Find historical ingoa wāhi in documents

Research question: Which places are repeatedly mentioned in a historical document collection, and what geographic evidence is strong enough to map them?

Useful for: iwi/hapū researchers, Treaty and historical researchers, cultural heritage teams, place-name researchers and GIS practitioners supporting archival work.

What you will produce​

  • a source register
  • a place-research table containing mapped and unmapped candidates
  • a name-variants table
  • a source-linked GIS layer for locations supported by evidence
  • a map showing confirmed, approximate and unresolved research states
  • a list of next research questions

The key rule​

A name in a document is evidence that the wording appears in that source. It is not automatically evidence of a modern coordinate.

The workflow keeps these stages separate:

Conceptual flow

Working structure​

Use three connected research tables rather than a synthetic download pack:

TableUse
place_researchTrack documentary references, geographic evidence and research status.
source_registerKeep source details and provenance consistent across the project.
name_variantsRecord different source forms without silently replacing historical spelling.

A place mentioned in a source can remain unlocated. Do not create coordinates simply because the name appears in a document. The spatial research log provides a reusable source and evidence structure.

Source and checking​

For every mapped or candidate place, retain the source document, page or plan, exact wording, checked date, geographic evidence and location status. Use the common Provenance and source checking pattern rather than recording only a URL.

Keep documentary evidence, Gazetteer matching and geographic interpretation in separate fields so a later researcher can see where each conclusion came from.

Step 1: preserve the documentary reference​

For each hit record:

  • exact wording as printed
  • document title
  • author/organisation
  • date
  • page
  • short context paraphrase
  • stable URL or archive reference

Do not modernise spelling in the source-text field.

If OCR produced the candidate, keep the OCR form as well. Macrons, ligatures, damaged type and historical orthography can all affect search results.

Step 2: use Local AI for candidate finding, not adjudication​

Use the Local AI workflow to search large PDF collections. A useful extraction request asks for candidate place wording plus document and page.

Do not ask the model to invent coordinates or decide that two historical names are the same place.

A strong table includes:

record_id
source_name
source_page
source_wording
candidate_name
context_note
checked_status

Step 3: check the New Zealand Gazetteer​

Search the New Zealand Gazetteer. Each Gazetteer name has a stable Name ID. Record that ID rather than storing only copied coordinates.

The Gazetteer distinguishes official, recorded, collected and other name statuses. That status describes the Gazetteer's record; it does not determine every historical or local use of a name.

The Gazetteer also provides downloadable CSV and GIS layers through LINZ Data Service, which is useful when checking many candidates.

Step 4: search historical geography​

For unresolved names use multiple evidence types:

  • historical maps and survey plans
  • Māori Land Court block names and records
  • LINZ Gazetteer history
  • NZ Gazette notices
  • Papers Past and other archives
  • historic aerial imagery where the period allows it
  • rivers, ridges, settlements and other contextual features described in the source

Record the evidence in separate rows or fields rather than replacing the original source wording.

Step 5: georeference a map when it materially helps​

If a historic plan or map contains the relevant name, use Georeferencing historical maps.

Record:

  • map title/reference
  • map date
  • control points
  • transformation used
  • residual/error information
  • whether the name position was printed as a point, area label or general landscape label

A label printed across a valley should not become a centimetre-precise point merely because GIS requires geometry.

Step 6: assign geometry only when the evidence supports it​

Use a location_type field such as:

  • point
  • line
  • area
  • approximate_area
  • unlocated

Use a separate confidence or location_status field to describe the research state, for example candidate, supported, conflicting, unresolved.

Avoid turning uncertainty into fake precision. An unresolved record can remain in the research table with no geometry.

Step 7: create the GIS research layer​

For located candidates, create a GeoPackage feature layer and keep a non-spatial evidence table alongside it. Link them using record_id or place_id.

Useful GIS fields include:

place_id
display_name
source_wording
gazetteer_id
location_type
location_status
source_id
source_page
geometry_source
notes

This allows a map feature to remain connected to the actual evidence.

Step 8: map patterns carefully​

Once enough references are checked, GIS can show:

  • repeated mentions
  • name variants
  • journeys or sequences
  • relationship to awa, ridges or land blocks
  • names that move or change across source periods
  • areas where evidence conflicts

Map uncertainty visibly. Use different symbols or line styles rather than hiding unresolved records.

Quality checks​

  1. Every mapped feature has at least one geographic evidence source.
  2. Source wording is preserved exactly enough to return to the document.
  3. Gazetteer status has not been treated as the only possible historical naming evidence.
  4. Coordinates have not been invented by AI.
  5. Approximate areas are displayed as approximate.
  6. Georeferenced map distortion is recorded.
  7. Macrons and Unicode survive CSV → QGIS → output.
  8. The map can be traced back to document/page references.

What the result means​

The result is a research index connecting documentary references to geographic evidence. It can support deeper historical, whenua and Treaty research.

It does not prove whakapapa, customary rights, the exact extent of a historical place, or that every source used the name in the same way.

Continue with Historical research, Place names, Archives to map, Māori Land Court research and Local AI.

Last reviewed: 4 September 2026

Last updated on