RE: [ai-geostats] polygon changes
- Dear Mr. Boogaart,
Thanks for your response and I'll see if I can be clearer in my question.
My interest is in detecting and characterizing changes in polygons between
T1 and T2. So essentially, comparing two polygon datasets. I think it's a
little narrower than space and time in Vector GIS, as there are numerous
methods available for point data, and while one way is to use polygon
centroids with spatial information attached as attributes, I feel this does
not effectively capture the complexity of the polygon data (ie boundary
complexity, orientation, proximity to other features etc).
So with two polygon datasets which are derived from peak values in an
intensity surface (moving infestation), some polygons in T1 will intersect
polygons in T2, some will not. The polygons therefore vary in space and in
time. So my concern is, where there is intersection, can all of a polygon
in T2 be considered to have come from the intersecting polygon in T1? Is
there consistent shape complexity between both? Where there is no
intersection, but proximity (to some threshold), can a relationship between
T1 and T2 be established? More importantly, what spatial relationships
(beyond spatial intersection) can be detected? How can I detect shrinking,
expanding, moving, coalescing? Much information can be derived from
visually looking at polygons in T1 and T2, but I am hoping to find a method
to characterize / detect the change more formally.
I hope this clears up my original post. As far as my background goes, I'm
just starting a graduate thesis, and have completed a bachelors(geography)
and advanced diploma(GIS). So I have much to learn about spatiotemporal
Thanks for your interest,
From: Gerald van den Boogaart [mailto:boogaart@...]
Sent: Tuesday, November 01, 2005 4:30 AM
To: Colin Robertson
Cc: 'AI Geostats mailing list'
Subject: Re: [ai-geostats] polygon changes
Dear Mr. Robertson,
> Do any of you have experience/advice on spatiotemporal analysis ofTo me your question reads like: Did anybody of you at any time considered
> polygon data?
in vektor GIS. And clearly one can lecture hours on that, while "real world"
GIS Systems typically don't support that properly. However to give any sort
of advice one needs a more specific question:
How do your polygones vary?
What do they represent?
Do they overlap in space / in time / in spacetime?
Are they polygons by nature or just areas approximated by polygons.
What are the questions, you would like to ask your data?
Do you have a polygon process consisting of many polygons or do you observe
one individual polygon changing over time.
Do your polygons change smoothly or discontinuesly.
Are the polygons individuals or perculations?
Do they form a coverage (i.e. is everything occupied)?
Is there a map of relevant additional covariates?
Is their any sort of representatitivty in the dataset?
What is your personal background (no need of writing well known things to
Are you writing a thesis on the theory of analysing polygons or do you work
a specific problem.
Gerald v.d. Boogaart
Am Samstag, 29. Oktober 2005 02:12 schrieben Sie:
> Dear List,
> Do any of you have experience/advice on spatiotemporal analysis of
> polygon data? I've read about Sadahiro's methods
> (http://okabe.t.u-tokyo.ac.jp/okabelab/sada/docs/pdf/CSIS%20DP25.pdf )
> on the subject and his ideas look promising to me, but I have not seen
> them implemented anywhere, nor have I seen anything comparable. Has
> anyone have experience with these or other methods for analyzing changes
> in polygons over time?
> Colin Robertson
> MSc Candidate
> Dept of Geography
> University of Victoria
> (250) 885-8065