GEOSTATS: variogram procedure of GSLIB

Expand Messages
• Dear List, I ve recently started using GSLIB and have been very pleased with my results. I am, however, sad that I don t understand exactly what s going on in
Message 1 of 1 , Dec 13, 1999
Dear List,

I've recently started using GSLIB and have been very pleased with my
results. I am, however, sad that I don't understand exactly what's going on
in the program (being fortran illiterate). The book (the first and second
editions) don't seem to explain as much as I'd hoped. My impression from
the example at the beginning of the book "A straightforward 2-d example" is
that what should be done is to model the variogram in the directions of
major and minor continuity (longest and shortest ranges of the variogram)
and use this information in the parameter files. That is fairly clear; what
I'm not clear about is the case where 1. the sills are not the same for the
two variograms(zonal anisotropy I believe; the example in GSLIB doesn't seem
very different from the geometric example), 2. where the variogram map is
not exactly oval (the major and minor axes are not perpendicular to one
another, which seems to be often assumed), and 3. where it seems that the
"best fit" model does not have the same combination of structures (e.g.,
spherical and exponential in the major direction, spherical and spherical in
the minor).

Furthermore, I don't quite see how programs like Sage and Surfer calculate
anisotropy ratios (major/minor ranges of anisotropy); when using their
procedure, you fit your model to all directions at once. It seems to me
that to get a range parameter in a given direction, you would need to
statistically fit a model in that direction. But from what I can tell, Sage
and Surfer 7 somehow estimate these ranges based on the one
"omni-directional" best-fitting model. This approach seems different from
what is implied in the example in the GSLIB books.

Finally, the anisotropy ratio returned in Surfer does not seem to be the
ratio of the major range/minor range when I examine the directional
variograms. The angles Surfer return seem correct, but the ratio of the
ranges seem somewhat different. I really like the variogram procedure in
Surfer 7, however....

Lastly, I did a little experiment using my data. I fit a good-looking
variogram, and then a really absurd one, with the same general shape, and
found that the results of kriging (in Surfer) were strikingly
similar....Perhaps I'm going overboard worrying about the precise
form/anisotropy ratios? (probably a case by case situation)

I feel that these are probably rather banal questions, but I searched
through the old mail archives and didn't find any answers. Thanks for any