Re: [Sgems-users] MultiGaussian Kriging

[Pierre G. <goo...@te...>]

Dear Sara,

The back-transform of estimated normal scores is not trivial and simply
applying a normal
score back-transform to kriging estimates will yield biased estimates. It's
analogous
to taking the exponent of lognormal kriging estimates, it won't do the
trick... Besides,
you cannot back-transform the kriging variance that way...
I have explained the methodology to do the back-transform in a few papers,
such as

Goovaerts, P., AvRuskin, G., Meliker, J., Slotnick, M., Jacquez, G.M., and
J. Nriagu. 2005. Geostatistical modeling of the spatial variability of
arsenic in groundwater of Southeast Michigan.  Water Resources Research, *41
*(7), W07013 10.102.


   Saito, H. and P. Goovaerts. 2000. Geostatistical interpolation of
positively skewed and censored data in a dioxin contaminated site.
Environmental Science & Technology, *34*(19): 4228-4235.

Both papers can be downloaded from my webpage.

The back-transform is done automatically in the beta version of a geostat
GIS software
I am developing at BioMedware, yet it works only in 2D. For the 3D modeling
of contaminated
sediments I mentioned in a previous e-mail, I am using my own code...
The other option, easier yet more CPU-intensive, is to use sgsim and compute
the E-type
estimates of all back-transformed simulated values using postsim. The
results should
in theory be similar to multiGaussian kriging, assuming the number of
realizations is large enough
and you get the kriging variance as well.

Pierre

On Tue, Feb 17, 2009 at 5:58 AM, sara sabri <sar...@ya...> wrote:

> Dear Frinds
> Please someone help me
> I would like to know if we could do MultiGaussian Kriging (MGK) with sgems.
> If so how we could backtransform the results? I tried to do it by Trans
> optin and assuming the hard data are Gaussian with a variance of 1. But the
> variance generated are very close to each other,i.e. the variance of
> variances is very small (e.g. conditional variances are between 18 to 19)
> while for the ordinary kriging the variance is between 0.01 to 10 for
> example. I really mixup. Please someone help me.
> I really appreiate your patience and am looking forward to hearing soon.
> Thanks in advance
> Sara
>
> ------------------------------
>  New Email addresses available on Yahoo!
> <http://sg.rd.yahoo.com/aa/mail/domainchoice/mail/signature/*http://mail.promotions.yahoo.com/newdomains/aa/>
> Get the Email name you've always wanted on the new @ymail and @rocketmail.
> Hurry before someone else does!
>
>
> ------------------------------------------------------------------------------
> Open Source Business Conference (OSBC), March 24-25, 2009, San Francisco,
> CA
> -OSBC tackles the biggest issue in open source: Open Sourcing the
> Enterprise
> -Strategies to boost innovation and cut costs with open source
> participation
> -Receive a $600 discount off the registration fee with the source code:
> SFAD
> http://p.sf.net/sfu/XcvMzF8H
> _______________________________________________
> Sgems-users mailing list
> Sge...@li...
> https://lists.sourceforge.net/lists/listinfo/sgems-users
>
>


-- 
Pierre Goovaerts

Chief Scientist at BioMedware Inc.
516 North State Street
Ann Arbor, MI 48104
Voice: (734) 913-1098 (ext. 8)
Fax: (734) 913-2201

Courtesy Associate Professor, University of Florida
Geostatistician, Computer Sciences Corporation
President, PGeostat LLC
710 Ridgemont Lane
Ann Arbor, MI 48103
Voice: (734) 668-9900
Fax: (734) 668-7788

http://goovaerts.pierre.googlepages.com/