3

I am using a point cluster tool in QGIS to consolidate data points based on their longitude and latitude. I note that in the Layer Styling panel that Distance can be set to metres at scale, map units, etc. For distance, is this a measure of radius or diameter of the circle around each cluster's centroid? My base is the OpenStreetMap, so the distance is measured in metres.

In other words, if I set distance to 100,000 metres, is this a 100km radius or 100km diameter for the points of data pulled in?

Improve this question

2 Answers 2

Reset to default
4

A quick empirical test reveals it is the radius.

The answer is neither. It looks like a cluster of points is first created, with its own rules. It looks like all points within the given distance of a seed point are clustered, then a weighted centroid is computed. (Depending on how this seed point is computed, we can expect two similar looking sets to have a different cluster)

If there are more than 2 points, the centroid can be further away than distance/2 from a point, so the distance is not a radius. At the same time, other points could be closer than distance/2 from the centroid while still be excluded from the cluster, so it is not a diameter either.

enter image description here

Improve this answer
3
  • Your screenshot shows the contrary. OP asks about "radius or diameter of the circle around each cluster's centroid". Points at the bottom: centroid is 51 m (102/2) m away form both points. A circle around the centroid with radius of 100 m (thus: diameter of 200 m) would clearly contain both points, thus they would be clustered - on your screenshot, they are not. If measurment is diameter, 100 m diameter means 50 m for radius. In this case, indeed the two points are not clustered as the centroid is 51 m (102/2) m away form both points, so the circle does not cover the points.
    – Babel
    Commented Nov 22, 2022 at 21:00
  • 1
    @Babel Interesting observations! It may be even more complex than that (see edits).
    – JGH
    Commented Nov 22, 2022 at 22:08
  • indeed very interesting! I revised my answer to match your findings.
    – Babel
    Commented Nov 22, 2022 at 22:21
3

Short answer

See @JGH's answer for the distance. In addition, be aware: distances are in cartesian measurements of the project CRS, so the projection used heavily influences the result.

Details

The projection (project CRS) also plays a role. Changing the project CRS also changes the points that are clustered (empirically tested, see screenshots below).

To make things even more complex: using OpenStreetMap, you probably work in EPSG:3857, WebMercator. This projection heavily distorts lengths. So you will not get any meaningful "real world distances" with this projection, even when using meters at scale. Just to be aware of this.

Points 1&3 (distance 86.1m) are clustered, points 1&2 (distance 108 m) are not (layer and project in EPSG:3857): enter image description here

The same settings, with only project CRS changed to EPSG:4326: as you see, all three points are clustered...: enter image description here

Improve this answer

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.