I agree with both Marcel and Don that the first question, before any
sampling strategy can be chosen, should be what the data is going to be
used for, i.e. what is the sampling objective. Of course, Marcel was
talking from a mining perspective, I am talking from a soil science
perspective. In my sampling optimization software, I tried to include as
many different optimization criteria as possible. There are at least three
fundamentallydifferent objectives for spatial studies that I have come across:
1) to describe spatial variability. Sometimes finding certain variogram
parameters can be an aim by itself (e.g. to detect periodicity or
anisotropy). In my opinion, this might be one of the most difficult
optimization criteria to formulate (although some people definitely tried,
Don among others in a 1987 paper).
2) to optimize spatial interpolation. In my case, this would be important
in precision agriculture, in order to produce high quality maps of
soil/crop parameters and use those for remedial action. My previous e-mail
was mainly focussed on this - minimizing the kriging variance is one of the
optimization criteria you could try for this case. I gave this a shot in my
Geoderma papers that I referred to earlier.
3) to detect hot-spots or low-spots. In my case, this is very important in
soil pollution studies, where your very precisely want to delineate
polluted areas (because remediation costs money, and there are health risks
involved), but you are not very interested in the areas that are well below
the environmental threshold. I suspect that this is quite often the case in
minin studies. I tried to tackle this sort of optimization criterion in my
environmetrics paper.
Of course, one cannot always go without the other. In order to optimize
spatial interpolation, you probably need at least an indication of the
nature of the spatial variability, and preferably a variogram. I agree with
Don that a phased approach is probably best for such cases. However, I
don't think I would go for a purely random approach. In my case, I would
probably in the first stage lay out a coarse grid over the whole area, and
include some short-distance observations (either randomly selected, or
somehow clustered). This should give me an idea about the nature of the
spatial variability, which I could use to optimize my second stage,
additional sampling scheme for minimal kriging variance. Also, the spatial
simulated annealing algorithm would allow me to make full use of the first
stage samples.
Hope this helps,
JW.
**
Jan Willem van Groenigen
University of California - Davis
Dept. of Agronomy and Range Science
1 Shield Avenue
Davis, CA 95616 - 8515, U.S.A.
--
e-mail: [EMAIL PROTECTED]
http://agronomy.ucdavis.edu/groenigen
tel. (530) 752-3457
fax. (530) 752-4361
*
--
* To post a message to the list, send it to [EMAIL PROTECTED]
* As a general service to the users, please remember to post a summary of any useful
responses to your questions.
* To unsubscribe, send an email to [EMAIL PROTECTED] with no subject and "unsubscribe
ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND
Subscribe/Unsubscribe requests to the list
* Support to the list is provided at http://www.ai-geostats.org