I have a point dataset of solar farm locations. The points are grouped according to which substation they feed into (see below). I would like to test the hypothesis that electrical output variability within each group is related to the pattern the points form. That is, 10 points within a 10x10km area may resemble a straight line, wavy line, perimeter of a circle, semi-circle, triangle, dumbbell etc. However, I’m struggling to find any algorithms which can help to categorise the groups of points or even the correct terminology to describe the possible patterns.

Most spatial analysis techniques simply describe data as “random”, “uniform” or “clustered” but I want to classify the types of cluster. UK Solar Farms

I’ve found I can identify lines in point patterns with PAST (http://folk.uio.no/ohammer/past). Hierarchical clustering in R looks suitable for dumbbell i.e. two groups forming a larger group. But I can’t identify any techniques for the other possible patterns.

  • How deterministic is the shape hypothesis/ building process within the 10 by 10 km pattern space and how do you hande the scale parameter? If you look at different scales at a circle can be a line, a arc or a circle for example.
    – huckfinn
    Feb 24, 2016 at 0:04

1 Answer 1


I’d say being less scientific is feasible approach in this case: I’ve made few shapes you mentioned and generated 10 random points around them:

enter image description here

Next I’ve computed Euclidean minimum spanning tree connecting them:

enter image description here

I guess by simply going through resulting shapes you can pick the best fit from limited number of shapes (straight, L-shape, circle, etc)

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.