[postgis-users] Cluster / Grouping

[trevorw]

I have an unusual problem and I'm wondering if I may be able to do this
in PostGIS.

I have large database of bird surveys conducted over a large area, which
has been divided into a regular grid for survey purposes. I would like
to look at reporting rate to assess the relative abundance across the
landscape. By this I mean bird x was seen y out of z surveys. The
problem lies in that the number of surveys done in different spatial
units is not consistent. Some areas were surveyed many times and other
areas were not surveyed at all due to access restrictions.

I would like to create a program that would step across the grid and
merge polygons until minimum number of surveys was conducted in the
aggregated spatial unit. The end result would be an irregular layer of
areas but the number of surveys conducted within them would be almost
identical thus making reporting rate a more accurate (although far from
perfect) measure of relative abundance. It seems like a neighbourhood
type of question and I've tried to implement a solution in GRASS using
raster functions and converting back to vectors but I'm not entirely
satisfied with the results.

As I'm aware that the answer will change depending on starting location
it would be nice to be able to get it to start for a number of different
positions and compare the results.

I also considered using R for this purpose, but in examining the cluster
routines, they really were structured to solve other problems and didn't
pursue that too far.

Any suggestions on how this problem might be tackled in PostGIS would be
much appreciated.

Thanks

T
-- 
Trevor Wiens