Maria, Thanks for your suggestions! Unfortunately, the cross-covariogram is a random scatter around zero all the way out. It was a good idea, though...
The UK/KT (Universal Kriging/Kriging with a trend - it seems every textbook uses different terminology) is an approach I hadn't thought much about before. Every program I've seen (GSLIB included) uses a polynomial function of some order to fit the trend, but it would be relatively easy to recode the GSLIB programs to use any function I could come up with. I'm not sure exactly how to derive the function I'd need, but it's definitely worth looking into. Even if it doesn't work directly, looking at the data residuals after some trend has been removed might give me more insight into how to solve the problem. At any rate, thanks again! -Joe ---- Original message ---- >Date: Sun, 29 Jan 2006 17:56:33 -0200 >From: "Maria N. Morales Boezio" <[EMAIL PROTECTED]> >Subject: Re: [ai-geostats] Multivariate kriging using terrain fabric >To: Joe Kington <[EMAIL PROTECTED]> > >Joe, >Did you try to compute the cross variogram of these variables? Sometimes no >correlation between data (for a zero separation vector) does not mean no >correlation for greater separation distances. If such exists, you can >perform cokrigng. >Maria > > > ---- Original message ---- >Date: Sun, 29 Jan 2006 17:48:56 -0200 >From: "Maria N. Morales Boezio" <[EMAIL PROTECTED]> >Subject: Re: [ai-geostats] Multivariate kriging using terrain fabric >To: Joe Kington <[EMAIL PROTECTED]> > > Joe, > I do not know if this can help: > In Goovaerts, P., 1997 "Geostatistics for Natural > Resources Research", Oxford University Press there > is a a technique described as Kriging with a Trend > Model (KT). It is similar to Kriging with an > external drift, but does not considers the trend as > a linear function of a secondary variable. In this > case, the trend has an unknown shape that can be > suggested by the physics of the problem (p.141 > Modeling the trend). It can be expressed as a > function of the coordinates. Your bathymetry can be > used to indicate the shape of the trend as a > function of the coordinates. > This technique can be implemented through > Fortran KT3D routine (program) from GSLIB > geostatistical library (Deutsch, C. V., and Journel, > A. G., 1998, "Gslib: Geostatistical Software Library > and User‘s Guide. Oxford University Press, New > York."). > Hope you get another helpful answers so you can > evaluate an appropiate methodology for facing your > problem. > Maria >
* By using the ai-geostats mailing list you agree to follow its rules ( see http://www.ai-geostats.org/help_ai-geostats.htm ) * To unsubscribe to ai-geostats, send the following in the subject or in the body (plain text format) of an email message to [EMAIL PROTECTED] Signoff ai-geostats
