Appearance
To represent locational uncertainty well, store it as explicit data — an uncertainty radius and a confidence level beside every coordinate — and visualise it with buffers or fuzzy symbols rather than crisp pins. The cardinal sin of historical GIS is false precision: writing six decimal places onto a place you only know to the nearest parish. Uncertainty is information; record it, do not bury it. This is a best-practice checklist for doing that consistently.
What is locational uncertainty?
It is the difference between the coordinate you assigned and where the place or event truly was. Historical sources create it constantly: "near the old mill", a destroyed village, a parish named but not pinpointed, a georeferenced map with 40 m of residual error. The goal is not to eliminate uncertainty — often impossible — but to measure and carry it through the analysis.
Precision versus accuracy: why the difference matters
These are not the same:
- Precision = how finely you state a location (number of decimals).
- Accuracy = how close that statement is to the truth.
A coordinate of 51.752341, -1.257610 for a place you know only "somewhere in the parish" is highly precise and probably inaccurate. Truncating to match what you actually know — and recording the rest as an uncertainty radius — is more honest.
How do you store uncertainty as data?
Add fields. A location is not just lat, lon; it is lat, lon, radius_m, confidence, basis:
json
{
"place": "Cropwell mill",
"lat": 52.943, "lon": -0.992,
"uncertainty_radius_m": 500,
"confidence": "low",
"basis": "parish centroid; exact site lost"
}These extra columns let later users, and your future self, judge how far to trust each point — and let analysis tools weight or filter by confidence.
How should you visualise uncertain locations?
Match the symbol to the certainty. Crisp pins shout "exactly here"; that is usually a lie. Better options:
| Uncertainty | Symbol | Reads as |
|---|---|---|
| Tight (< 50 m) | Small precise marker | "Pinned" |
| Moderate | Graded circle sized to radius | "Around here" |
| Large / vague | Faded blob or heat surface | "Somewhere in this area" |
In QGIS, a data-defined buffer makes this automatic:
text
# QGIS: marker size or buffer driven by the uncertainty field
buffer( $geometry, "uncertainty_radius_m" )Should you drop the shakiest points?
Rarely silently. Keep low-confidence locations but tag them, then run a sensitivity analysis: compute your result with all points, then with only high-confidence ones, and compare. If the conclusion holds either way, it is robust; if it flips, the uncertainty is material and must be reported. Quietly deleting awkward points hides exactly the information a reader needs.
A working uncertainty checklist
- Record
radius_m,confidenceandbasisfor every located feature. - Truncate coordinate precision to what the source supports.
- Size map symbols to the uncertainty; avoid crisp pins for vague places.
- Propagate georeferencing residuals into the uncertainty radius.
- Run analyses with and without low-confidence points.
- Document the rules you used, so others can reproduce them.
How does this connect to gazetteers?
Established gazetteer models already expect imprecise places. Pleiades and the World Historical Gazetteer let a location be an extent or carry a precision note rather than a single false point. Aligning your uncertainty fields with those conventions makes your data interoperable and reuses standards thousands of historians already read.
Key Takeaways
- Treat uncertainty as data: store a radius, a confidence level and the basis beside each coordinate.
- Beware false precision — precise is not the same as accurate.
- Truncate coordinates to the precision your source actually supports.
- Visualise with buffers, graded circles or fuzzy symbols, not crisp pins.
- Keep low-confidence points but tag them and run sensitivity tests.
- Fold georeferencing residuals into the recorded uncertainty radius.
- Align with gazetteer extent/precision conventions for interoperability.
Frequently Asked Questions
What is locational uncertainty in historical GIS?
It is the gap between the coordinate you assigned to a historical place or event and where it actually was. It arises from vague sources, lost place names, imprecise georeferencing and changing landscapes, and it should be recorded, not hidden.
How do I store uncertainty alongside a coordinate?
Add explicit fields: a radius or extent in metres, a confidence level, and the source of the estimate. A point with a 500 m uncertainty radius is far more honest than a bare coordinate implying metre precision.
How should I visualise uncertain locations on a map?
Use buffers or graded circles sized to the uncertainty, fuzzy or faded symbols, or a heat surface rather than crisp pins. Crisp markers imply a precision the evidence does not support.
What is the difference between precision and accuracy here?
Precision is how finely you state a location (decimal places); accuracy is how close it is to the truth. Many historical datasets are precise but inaccurate — six decimals on a guess — which is the trap to avoid.
Should I exclude very uncertain locations from analysis?
Not necessarily. Keep them but tag them, then run the analysis with and without low-confidence points to test how much they change the result. Excluding silently is worse than including with a flag.
Is there a standard for recording spatial uncertainty?
Conventions vary, but recording an uncertainty radius in metres plus a confidence code is widely understood, and gazetteer models like those used by Pleiades and the World Historical Gazetteer support representing imprecise or extent-based locations.