I am trying to summarize my polygons by each's distribution of some variable.

To couch the discussion, let's say we wanted to summarize the demographics of each county (of a small state) by it's demographic makeup--say, black/white/other.

What came to mind would be to treat the fill of each county's polygon as a pie chart, and slice it accordingly.

Is there a way to do this in base QGIS (or SAGA)?

If not, what about through scripting? I was thinking one might be able to split the polygon layer into three new layers--white, black & other--with one slice of each polygon corresponding to the right wedge size.

This sounds vaguely doable to me, but I've got nothing like this GIS chops it would take. Any pointers?

Here's a rough sketch of what I've got in mind for two counties (with made up data).

The give-up alternative is to use the pie charts labeling feature. Which may be more readable in the end anyway... just wondering whether this is possible anyway.


Note that the areas of the resulting sub-polygons should correspond (in proportion to the overall area) to the data points--i.e., the B sub-polygon in Providence should be 10% of the Providence's area, and the B sub-polygon in Kent should be 50% of Kent's area.

  • 3
    If I may, I would really recommend against "slicing" the fill in this way. Pie charts are notoriously hard to interpret correctly, and here you'd be making it even harder because the slices aren't going to correspond to the correct proportions, and it feels as if you're making a suggestion about location rather than proportional distribution. What about randomly distributed points of three colours as an alternative? Jan 28, 2015 at 21:41
  • It's a very good point about the improper proportion. I've edited the question to maintain relative area. Jan 28, 2015 at 22:29
  • As noted, this may not be a good idea, but it is an interesting scripting exercise, and certainly (probably) do-able. Is an arcpy solution at all useful to you?
    – phloem
    Jan 28, 2015 at 22:29
  • As far as suggesting location, wouldn't putting actual points (even if random) suggest location even more strongly (as far as interpreting at a glance)? Jan 28, 2015 at 22:29
  • @phloem I haven't broken the seal on GIS scripting yet, but this is a good enough excuse as any. I agree it may not look as appealing in reality once it's coded, but I think it's interesting to try anyway. Jan 28, 2015 at 22:33

1 Answer 1


You can use the Diagram Overlay Plugin

enter image description here

And some more instructions:

Statistics or simple pie chart in Quantum Gis

  • 1) I'm not finding that plugin in QGIS 2.6.1; 2) As mentioned, I'm willing to put pie charts as labels as a backup, but am really hoping to replace the layer fill itself with the pie chart--see my rough sketch. Jan 28, 2015 at 21:35
  • Some may say that what you're trying to do is a 'bad idea' - especially for a map. If you simply used the chart plugin (assuming you can install the plugin) you can remove the fill from the polygons and just use the icons from the plugin...otherwise, maybe a map isn't what you're after... Jan 28, 2015 at 21:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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