1

I have just compared the random point generator functions of QGIS and ArcGIS and had confusing results. My setup was: Create random points within an existing extent with a minimum dist between points of 5m - max number of points was set to 40000 (which is far beyond the max amount possible with 5m dist).

The results are :

ArcGIS: overall point count 3699; mean dist to next pt: 7.53m; std dist to next pt: 2.04

QGIS: overall count 13993; mean dist to next pt: 5.31m; std dist to next pt: 0.32

parameters used:

ArcGIS: Create Random Points: Constraining Extent: an existing polygon dataset. number of points: 40000, min allowed distance / linear unit: 5m, create multipart: no, environments/random numbers / random number generator / seed: experimented with different values and different generator types

QGIS / Random Points In Extent, extent from the same polygon dataset). number of points: 40000, min dist: 5m, CRS: 31256;

I've been searching for any kind of reason and found ArcGIS Random Point generator producing same points every iteration. But changing the seed-value in ArcGIS did neither change the number of points nor the mean distance value significantly. Esri doesn't answer my question ... and I have to use ArcGIS for that task.

So, does anybody have an idea what could be responsible for this huge difference and how to solve the problem.

7
  • 2
    What do you mean by "Esri doesn't answer my question"? Do you have access to Tech Support? This doesn't seem like a GIS SE issue, since we are not Esri support.
    – Vince
    Commented Oct 24, 2019 at 11:20
  • 1
    So do you know what QGIS is using as the random number generator? ArcMap has 3 different algorithms (which you have played around with) but you have not provided any information about how QGIS is generating numbers? My gut feeling is that this is the source of the issue how the different software implement their random number generators.
    – Hornbydd
    Commented Oct 24, 2019 at 13:49
  • 1
    If I had to guess (which we do, because we can't see the source of the ESRI tool) I would say that the qgis algorithm is using a much larger number of iterations to try and place each random point. The ArcGIS version is "giving up" much earlier and accordingly placing less points.
    – ndawson
    Commented Oct 24, 2019 at 23:19
  • 1
    Out of curiosity, I replicated your experiment (using my own data, of course) and came up with similar results: QGIS created ~3.5x more points than Arc, which resulted in the output points being more densely "packed in". I also noticed that QGIS took far longer to run than Arc (several minutes for QGIS vs a couple of seconds for Arc), which lends credence to @ndawson and the idea that Arc is giving up sooner - although the time difference could be algorithm efficiency. Further, although Arc said that it had exhausted the 40k input point limit, only ~8k were output. I trust QGIS...
    – Stu Smith
    Commented Oct 26, 2019 at 18:01
  • 1
    I conducted a second experiment comparing QGIS and Arc. In this experiment, only 100 points were randomly dispersed over 0.9 hectare, which assured that all points were utilized (as opposed to my previous experiment with 40k points, of which only a subset were placed). In this second experiment QGIS and Arc generated results that were very similar: QGIS: mean = 50.6, std dev = 25.7. Arc: mean = 52.8, std dev = 33.4.
    – Stu Smith
    Commented Oct 27, 2019 at 17:33

0

Your Answer

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

Browse other questions tagged or ask your own question.