Appearance
To record coordinate uncertainty for a place, store a single uncertainty radius in metres alongside your latitude and longitude, and document what that radius is based on. The widely understood convention is Darwin Core's coordinateUncertaintyInMeters — one number that says "the true location lies within this radius of the point I recorded". This guide takes you step by step from choosing a radius to storing it in GeoJSON, with practical defaults and the pitfalls that quietly corrupt spatial analysis.
What does coordinate uncertainty actually capture?
Three different things roll into one radius:
- Georeferencing error — how precisely you pinned the location (a surveyed point versus a guess from a description).
- Source vagueness — "near Exeter" is inherently fuzzier than a street address.
- Feature extent — a region's coordinate is a centroid; the place is kilometres wide.
Your recorded radius should reflect the largest of these, because that is the limit of what your coordinate can honestly claim.
Step 1 — Add an uncertainty field
Start with the simplest portable model: latitude, longitude, and a radius.
text
place_id,lat,lon,uncertainty_m,uncertainty_basis
P0101,50.7184,-3.5339,30,GPS-grade point on church
P0102,50.9,-3.5,5000,"region centroid, named area only"
P0103,51.0632,-1.3080,400,"parish centroid"The uncertainty_basis column is not optional. A radius without a stated basis cannot be checked or revised later, and reviewers will not know whether 400 means "measured" or "felt about right".
Step 2 — How big should the radius be?
Size it to the worst contributing factor, and when torn, round up. Understating uncertainty is the more dangerous error because it makes downstream analysis falsely confident.
| Source quality | Typical radius | Example |
|---|---|---|
| Surveyed / GPS point | 5–30 m | A church door |
| Address geocode | 50–200 m | "12 High Street" |
| Parish or settlement centroid | 200–1000 m | A village name |
| Named region only | 2–20 km | "the Fens" |
| Vague historical description | 5 km+ | "a day's ride north of York" |
These are starting defaults, not rules — adjust to your sources, but never quietly record a 30 m radius for a place you only know by region.
Step 3 — Point with radius, or polygon?
For most data a point plus radius wins on simplicity and portability: one coordinate, one number, every tool understands it. Switch to a polygon only when the uncertainty is strongly non-circular — for instance a settlement known only to lie somewhere along a river, where a circle would wrongly include the hills on either side. A useful rule: if a circle would mislead more than it helps, draw the polygon.
How do I record uncertainty in GeoJSON?
GeoJSON has no native uncertainty field, so put the point in geometry and the uncertainty in properties under a documented key.
json
{
"type": "Feature",
"geometry": { "type": "Point", "coordinates": [-3.5339, 50.7184] },
"properties": {
"name": "St Petrock parish",
"uncertainty_m": 400,
"uncertainty_basis": "parish centroid, GBHGIS polygon"
}
}Document the property names in a README so consumers know uncertainty_m is a radius in metres, not a diameter or a different unit. An undocumented convention is barely better than no convention.
Why should I never just drop uncertain coordinates?
Because a blank is read as "we have no idea", when the truth is usually "we have a good idea, give or take". Leaving the field empty discards real information and often pushes a place off the map entirely. Recording the estimate and its radius together lets a careful user filter — "show me only places certain to within 500 m" — which is impossible if you simply omitted the shaky ones.
Key Takeaways
- Record one radius in metres alongside lat/lon, following
coordinateUncertaintyInMeters. - The radius captures georeferencing error, source vagueness and feature extent — size it to the worst.
- Always store an
uncertainty_basisnote; a bare number cannot be audited. - Use a point plus radius by default; reach for a polygon only when uncertainty is non-circular.
- In GeoJSON, keep the point in geometry and the radius in documented properties.
- Never drop uncertain coordinates — an honest radius beats a misleading blank.
Frequently Asked Questions
What is coordinate uncertainty for a place?
It is a quantified statement of how far the true location could be from the coordinate you recorded, usually expressed as a radius in metres around a point. It captures georeferencing error, source vagueness and the size of the feature itself.
What is the simplest way to record uncertainty?
Add an uncertainty_radius_m field next to your latitude and longitude, following the Darwin Core coordinateUncertaintyInMeters convention. A single radius is enough for most projects and is widely understood.
How big should the uncertainty radius be?
Size it to the worst contributing factor: a few metres for a surveyed point, hundreds of metres for a parish centroid, kilometres for a named region. When in doubt, round up rather than understate.
Should I use a point or a polygon for an uncertain place?
Use a point with a radius for most data because it is simple and portable. Switch to a polygon when the real uncertainty is strongly non-circular, such as a place known only to lie along a particular river.
How do I record uncertainty in GeoJSON?
Keep the coordinate in the geometry and put the radius and its basis in the feature's properties, for example an uncertainty_m number and an uncertainty_basis string. GeoJSON has no native uncertainty field, so a documented convention is essential.
Why not just leave uncertain coordinates out?
Because a recorded coordinate with an honest radius is far more useful than a blank, and a missing value is often misread as 'unknown location' rather than 'approximate location'. Record the estimate and its uncertainty together.