Ø I have also specific question about geohex. Do you know anyone who has
population-normalized
Ø geohex (or something similar) in at least national level, better if
global.
Ø Or can you suggest a method to create this kind of grid?
In the spirit of brainstorming, the areas would have to be different shapes
as well as different sizes. They could be built up by merging adjacent
tiles, be they hexagons, squares or something out of an Escher drawing.
One approach:
· Lay out the base tiles over the area of interest, with a known
population for each tile.
· Systematically walk through unused adjacent tiles, joining them
together until you reach the desired population level.
· Either hex or square tiles should work. Hex shapes merge to look
like natural boundaries. Rectangular shapes merge to look like man made
boundaries.
As far as the systematic walk goes, you could use several options. I once
did a similar project using the Peano ordering. It is a bit esoteric, but
two cells which are adjacent in the Peano ordering are also adjacent on the
grid. Thus a sequence of cells traversed in Peano order can be merged into a
single contiguous shape.
The algorithm became:
Repeat until all cells in the grid are consumed
Current shape ß empty
Scan from current cell in Peano order, until the current shape
has the right population
Add the current cell to the current shape
Output the current shape.
I have no idea if this approach would meet your needs. It doesnt attempt to
do clustering, but it does result in a set of non-overlapping shapes of
similar population which cover a large area. Im sure others can come up
with better solutions.
I have Java code which implements the algorithm. In my case, I was
generating synthetic data for a spatial benchmark, where randomly sized,
non-overlapping shapes modeled country, state and county boundaries.
Changing the code to work for population would be easy.
- John Morris
_______________________________________________
Geowanking mailing list
[email protected]
http://geowanking.org/mailman/listinfo/geowanking_geowanking.org
