56GEOSTATS: Re: SISIMPDF
- Jan 31, 1997thank you clayton for an excellent response. for others, the original
questions are listed below. I was in error and there is no problem
witht the gslib code. Clayton makes a note about the MAXCT[X,Y,Z]
parameters in the include file. The stock version of sisimpdf.for sets
the following values:
One needs to be careful to reset the maxctz value to a value somewhat
greater than 1 if running 3D simulations. This was not my problem, but
it did take me a while to figure this out.
I have two questions regarding indicator simulation:
1. What role does the sill play when simulating categorical data (e.g.
sisimpdf.for, from gslib)? As nodes are created the probability that
each node will fall into a category is a function of the kriging weights
and the post. prob (eq. V.33 gslib). Page 170 of the gslib manual says:
"There is no need to standardize the parameters to a sill of one since
only the relative shape affects the kriging weights." Therefore, for a
given range and nugget would various sill values alter the simulation?
2. After reading some of the previous email listings from this
newsgroup, I noticed someone asked about an errata to the gslib manual.
I found a possible error in the calculation of the rotation matrix in
sisimpdf.f routine. Using the parameters ang1=45, ang2=-45
ang3=0,anis1=1.0, anis2=0.1 for a 3d realization, I was unable to
produce dips of 45 degrees. The stikes (ang1) were produced correctly,
but the dips were vertical and any change to ang2 had little or no
effect on the dips. I recoded the calculation of the rotation matrix
according to Gendzwill and Stauffer, 1981, table 1 (see gslib ref 57)
and the dips were then modeled correctly. Is there a problem in the way
I am specifying the rotation angles in the original version or is there
an error in the code?
Clayton Deutsch wrote:
> I think the confusion relates to the angle conventions in the GSLIB
> programs. There are no known problems in the rotation matrices of
> GSLIB (even after my testing with your cases). Nevertheless, don't
> read further if you have recoded the rotation matrix and are getting
> results you like -- you will only get more confused.
> In answer to your two questions to GEOSTATS:
> 1. As explained in the GSLIB book, the variogram sill plays no role in
> indicator simulation; the estimated distributions depend only on the
> kriging weights and not the kriging variance. For example, a
> variogram with a nugget (c0) of 0.1 and an isotropic spherical model
> with a variance contribution (c1) of 0.9 and some arbitrary range
> would give the same kriging weights if c0=0.001 and c1=0.009 (the
> sill reduced by a factor of 100). Of course, the range and relative
> shape must remain unchanged.
> 2. The results of using parameters ang1=45, ang2=-45, ang3=0.0,
> anis1=1.0, and anis2=0.1 give correct results with the sisimpdf
> program!?! An XY horizontal slice shows continuity in the NW-SE
> direction (because the principal direction of continuity has been
> rotated 45 degrees from North (to NE) and then down 45 degrees).
> Other slices also show the expected anisotropy. I observed no
> vertical dips. Changing anis2 will have little affect on any of
> the orthogonal XY, XZ, or YZ slices (main effect visible on a
> vertical section from SW to NE).
> It is critical that you have the MAXCT[X,Y,Z] parameters large
> enough to reproduce your variogram range in any particular
> Regarding your message on Tuesday:
> 1. ang1=-15 is ONLY the same as ang1=165 when ang2/ang3 are both 0.0;
> otherwise, the results are quite different.
> 2. ang1=-15, ang2=-110, ang3=0.0, anis1=1.0, and anis2=0.1 will not
> give features with a strike of N15W and a dip of 70 --> the
> features will have a strike of N75E and a dip of 70 degrees to
> the SW.
> Use ang1=105, ang2=-70, ang3=0.0, anis1=1.0, and anis2=0.1 for
> strike of N15W and dip of 70 to SW.
> Use ang1=-75, ang2=-70, ang3=0.0, anis1=1.0, and anis2=0.1 for
> strike of N15W and dip of 70 to NE.
> Best Regards, Clayton V. Deutsch
> Phone: 415 723-4142 Fax: 725-2099 Dept. of Petroleum Engineering
> Email: clayton@... Stanford, CA 94305-2220
Desert Research Institute
Water Resources Center
P.O. Box 60220
Reno, NV 89506
Voice: (702) 674-7523
FAX: (702) 673-7397
*To post a message to the list, send it to ai-geostats@....
*As a general service to list users, please remember to post a summary
of any useful responses to your questions.