Appearance
Work with historical parish boundaries when your evidence is parish-shaped — registers, tithe apportionments, the 1851 ecclesiastical census, poor-law and overseers' accounts — and you need to aggregate, map or compare those records spatially. Do not reach for a parish layer when your data already aggregates to registration districts, hundreds, counties or modern wards: forcing it into parishes manufactures false precision and wastes weeks of georeferencing. The deciding question is never "are parishes interesting?" but "is the unit of observation in my sources actually the parish?"
Why does the unit of observation decide everything?
A GIS join is only as honest as the match between your table's key and your polygon's key. If a baptism register names "St Mary, Whitchurch" then a parish polygon is the natural container and your counts are meaningful. If a census abstract reports a registration sub-district that spans four parishes, you cannot disaggregate it without inventing data. Before drawing a single polygon, write down the smallest spatial unit named in every source you intend to join. That column is your real resolution; a finer geometry only lies to you.
When are parish boundaries the right tool?
Reach for them when several of these hold:
- Your records carry a parish name or a parish-level code (Chapman code, Kain & Oliver ID).
- You are studying something that was administered by the parish: poor relief, settlement, tithe, churchwardens' rates, vestry decisions.
- You need to normalise counts by area or population within the parish system.
- You are comparing the same parishes across two or more dates and changes matter.
If two or more of these are true, the curation cost pays off.
When should you avoid parish polygons?
Skip them when the unit mismatch is large, the period predates reliable cartography (much of England before the 1840s tithe surveys), or your question is genuinely point-based — individual addresses, monuments, find-spots. In those cases geocode the points directly and let any parish attribution be a derived field, not the spine of your model. A common failure is spending a month reconstructing 1700 boundaries when a gazetteer point and a valid-from date would have answered the question.
What does the setup actually cost?
| Task | Typical effort | Main risk |
|---|---|---|
| Acquire base layer (Kain & Oliver / GBHGIS) | 0.5 day | Licence and attribution terms |
| Reproject to EPSG:27700 + clean topology | 0.5–1 day | Slivers, overlaps, gaps |
Add valid_from / valid_to / parish_type | 1–2 days | Undated boundary changes |
| Reconcile names to your source spellings | 2–5 days | Spelling drift, dedications |
| QA the table-to-polygon join | 1 day | Many-to-one mismatches |
Budget roughly a working week for a county before you map anything substantive. That estimate is the trade-off you are accepting.
How do you model parishes that change over time?
Treat the boundary as a versioned feature. Each polygon gets explicit temporal validity and a stable identifier so a split becomes new births, not an in-place edit:
sql
-- PostGIS: snapshot of parishes valid in a target year
SELECT p.parish_id, p.name, p.parish_type, p.geom
FROM parishes_hist p
WHERE p.valid_from <= DATE '1851-01-01'
AND (p.valid_to > DATE '1851-01-01' OR p.valid_to IS NULL);To attach records to the right version, join on both name/id and date range. A quick spatial check for the topology errors that wreck area calculations:
bash
# find overlaps and gaps after editing in QGIS
ogr2ogr -f GPKG checked.gpkg parishes.gpkg \
-dialect SQLite \
-sql "SELECT a.parish_id, b.parish_id, ST_Area(ST_Intersection(a.geom,b.geom)) AS ov
FROM parishes a, parishes b
WHERE a.parish_id < b.parish_id AND ST_Overlaps(a.geom,b.geom)"Which signals tell you the approach is failing?
Stop and reconsider if you see these red flags: more than ~10% of source rows fail to match a polygon name; you keep "fixing" matches by editing geometry to fit the data; your valid_to column is mostly empty because you cannot date the changes; or every interesting result depends on a single interpolated boundary. Any of these means the parish layer is carrying more uncertainty than signal, and a coarser, better-documented unit will tell a truer story.
Key Takeaways
- Use parish boundaries only when the parish is the actual unit of observation in your sources.
- The smallest unit named in your records — not the prettiest geometry — sets your real resolution.
- Budget about a week per county for acquisition, reprojection, dating and name reconciliation.
- Distinguish ecclesiastical from civil parishes after 1866/1894 with an explicit attribute.
- Model change with
valid_from/valid_toand stable IDs; never edit a polygon in place across a boundary change. - Store data in EPSG:27700 (British National Grid) and keep a WGS84 copy for the web.
- Walk away if match rates collapse, dating is impossible, or results hinge on one interpolated edge.
Frequently Asked Questions
When should I use parish boundaries instead of modern administrative units?
Use parishes when your sources are themselves parish-organised — baptism registers, the 1851 religious census, tithe apportionments, poor-law returns. If your data is organised by registration district, county or modern ward, a parish layer adds error rather than precision.
Where can I get ready-made historical parish boundary data for England and Wales?
The most complete free dataset is the Historic Parishes of England and Wales (Kain & Oliver) hosted by the UK Data Service, plus GBHGIS parish polygons. For Scotland use the NRS/GROS civil parish layers; Ireland is best served by Townlands.ie and the OSi historic boundary sets.
Do ecclesiastical and civil parishes have the same boundaries?
Often, but not always. After the 1866 Poor Law Amendment and the 1894 Local Government Act, civil parishes diverged from ecclesiastical ones. Always record which type each polygon represents in an attribute field.
How do I handle parishes that split or merged over time?
Model time explicitly: give each polygon valid-from and valid-to date fields and treat a split or merge as the death of one feature and the birth of others. Never edit a polygon in place across a boundary change.
What coordinate system should historical parish polygons use?
For Britain use British National Grid (EPSG:27700); store a WGS84 (EPSG:4326) copy for web maps. Reproject once, on import, and keep the projection definition in your metadata.
How accurate are reconstructed parish boundaries?
Pre-1840 boundaries are often interpolated from tithe maps, perambulations and 19th-century surveys, so positional accuracy can be tens to hundreds of metres. Treat them as approximate and document the source for every edge.