Location precision in conflict event data
By Tellurial · Last updated 30 October 2026
Every event on a map has exact coordinates. That does not mean anyone knows exactly where it happened.
In short
- Coordinates are often a stand-in for a place, not a measurement.
- Good datasets say how precise each location is. Use that field.
- Never measure a distance more finely than the location allows.
- Keep imprecise events in your counts, and say how you treated them.
What precision means
A report says “strikes near the city” or “fighting in the north of the province”. To put that on a map, someone has to choose a point, and that point is usually the centre of the named place.
On the map it looks identical to an event located to the metre. The difference is only visible in the data, which is why precision has to be stated rather than assumed. Researchers have warned that there are few best practices for these decisions, and that they matter (Douglass and Harkness, 2018).
How UCDP states it
The UCDP Georeferenced Event Dataset uses a variable called where_prec:
| Code | Meaning |
|---|---|
| 1 | Exact location known |
| 2 | Within about 25 km of a known point |
| 3 | Only the second-order administrative division is known |
| 4 | Only the first-order administrative division is known |
| 5 | A linear feature or a fuzzy area, such as a border or a long road |
| 6 | Only the country is known |
| 7 | International waters or airspace |
For codes 3 and 4 the point is typically the centre of the division. Dates are treated the same way, in a separate variable called date_prec. ACLED documents its own approach in its user guides.
How WarStrikes states it
Each event assessment carries a location precision, and coordinates carry a precision in metres. Time has its own field: exact, hour, day or unknown.
| Level | Precision used |
|---|---|
| site | About 100 m |
| city | About 5 km |
| district | About 15 km |
| region | About 100 km |
| country | About 500 km |
| unknown | No coordinates |
Coordinates are given only when the place is identifiable. When a report does not say where, the location stays unknown rather than being filled in.
Why it matters
- False hotspots: events placed at the centre of a region pile up on one point, and a map shows a cluster that never existed.
- False proximity: an event placed at a city centre can look close to a port when the report only named the city.
- False absence: leaving imprecise events out hides activity in places that reporters can reach least.
Five rules for using it
- Look at the precision field before you look at the map.
- Analyse at a scale no finer than your weakest records allow. Region-level events belong in regional totals.
- Draw uncertainty. A circle or shaded area is more honest than a dot.
- Compare distance with precision. If precision is 100 km, a distance of 40 km says little.
- Report your treatment: how many events were imprecise, and what you did with them.
A worked example
Say an asset sits 30 km from an event’s coordinates. If the event is site-level, the asset is genuinely close. If it is city-level, it is probably close, with a few kilometres of doubt. If it is region-level, the event could be on the other side of the region, and the distance tells you very little.
The same applies to WarStrikes. Events are linked to assets within 40 km of their coordinates, so a link is only as good as the location behind it. Check the precision before you rely on one. These numbers are illustrative.