[19046] in s-news-athena
[S] Summary: Block kriging in S+.
daemon@ATHENA.MIT.EDU (Iiris-Tiina Annikki Sivori)
Thu Aug 26 04:45:10 1999
Date: Thu, 26 Aug 1999 10:39:46 +0200 (MET DST)
From: Iiris-Tiina Annikki Sivori <i.t.a.sivori@geofysikk.uio.no>
To: s-news@wubios.wustl.edu, ai-geostats@gis.psu.edu
Message-Id: <Pine.OSF.4.05.9908260946130.16002-100000@gauss.uio.no>
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Dear list members,
Many thanks to all those who answered my question! You have been of great
help (and inspiration).
It looks like many people are working on this issue, there are routines
being developed right now, that will appear soon. There are also
existing functions, as you see when you read on.
Here comes a summary of the answers I received to my request for
block kriging routines in S+:
(Answers are listed up in the order they arrived.)
------------------------------------------------------------------
Prof. Brian Ripley writes:
"...Block kriging is just kriging applied to blocks. All you have to do is
to calculate the covariance between blocks (by integration): I know of no
pre-packaged code to do this."
-------------------------------------------------------------------
Richard Duin forwarded a message from Christian Lajaunie,
who has also been working on the subject; He has written a piece of
S+code to do block kriging.
Christian writes:
..."it involves packing data into blocks on a regular grid, covariance
calculation at block support, and block kriging or random kriging (a
version of kriging in which the data locations are assumed random
uniform within blocks. This makes calculations much simpler when the
data
density is high and irregular). Block-block covariances are calculated
in
C functions attached by dyn.load(). I can see potential problems which
makes
me hesitating to propose the code:
- The code is not very up to date and was not written with the
aim to be distributed.
- The documentation is in French and is sparse (for the same reason)
- The variogram models have a structure not compatible with
spatial stats. This structure has up to now
some limitations: all the elementary models are not covered (the list
cover
the exponential, the spherical, the |h|^alpha (alpha<2), the gaussian,
and
the cubic model), and no support of geometric anisotropy is ensured .
Conversion is certainly quite easy, but requires some effort.
I think that the best, if you are interested, would be to send you just
the part concerned with the between-block covariance calculation, which
you
could easily include in your kriging program. This program is in 2-D and
assumes
rectangular blocks with edges parallel to the coordinate axis."
-----------------------------------------------------------------------
Doug McIver suggested using a freeware package called geoeas which has
block kriging. It's availible from:
http://www.sph.umich.edu/~aelon/geoeas/
or
ftp://math.arizona.edu/incoming/unix.geoeas/
with more info at:
http://curie.ei.jrc.it/software/geoeas.htm
----------------------------------------------------------------------
Matthew Pringle quided me to
http://www.usyd.edu.au/su/agric/acpa/
where in few weeks time there will appear a rigorous and flexible
FORTRAN-based program that will do block kriging. It was not finished yet
when he wrote.
- Follow the link to "Freeware Download". Unless there is a last minute
name change, the program will be called 'KRAIG'.
Further, Matthew has written some S+ code that estimates the within-block
variance given the variogram model, its parameters and a specified block
size.
He writes:" It would not be too hard to modify this to include the
data for interpolation and return the block kriged estimate, however I am
afraid it would become quite cumbersome...". He recommends to wait until
'KRAIG' appears.
-----------------------------------------------------------------------
Med Bennett
recommends the GSLIB programs - He writes:"...the code is free but if you
spend $50 US you can get the book and CD-Rom with it."
Program may be downloaded from
http://www.ualberta.ca/~smpe/people/clayton/gslib.html
--------------------------------------------------------------------------
Andrew Lister
told about Robin Reich at Colorado State University, who has written free
modules for s+ that will conduct block kriging.
R. Reich has developed several functions for doing kriging and cokriging
that he uses in his class in spatial statistics. The Splus code is
available on his page:
http://www.cnr.colostate.edu/~robin
Additional information may be received by
mailing him at
robin@picea.cnr.colostate.edu
---------------------------------------------------------------------------
Again, thank you all very much for your answers!
Best regards,
Tiina Sivori
*******************************************************************************
tiina@geofysikk.uio.no
Tiina Sivori
Institutt for Geofysikk, UiO
tlf: +47 - 22 85 52 06
fax: +47 - 22 85 52 69
tlf private: +47 - 23 00 97 16
+47 - 93 42 93 72 (mob)
*******************************************************************************
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
