Loading ...
Sorry, an error occurred while loading the content.

2506RE: [ai-geostats] isatis expertise?

Expand Messages
  • Mark Sweeney
    Apr 20, 2006

      As a first pass you should try looking for duplicate values. From my
      experience Isatis falls over when kriging with duplicate data.

      Isatis has a useful function to remove duplicate data, it can be found
      in the "TOOLS" menu system.


      Mark Sweeney



      Address: Level 5, 80 Albert Street, Brisbane, QLD, 4022
      PO Box 15224, City East, Brisbane
      Phone: Work: +61 (0)7 3012 8499
      Mobile: +61 (0)407 357 877
      E-mail: Work: Mark@...


      -----Original Message-----
      From: acullma@... [mailto:acullma@...]
      Sent: Thursday, 20 April 2006 10:24 PM
      To: ai-geostats@...
      Subject: [ai-geostats] isatis expertise?

      Dear list,
      I am trying to compare kriging results from different geostats software.
      The sample size is smaller than 300, I use real data as well as
      values. The variograms are omnidirectional and not nested (a spherical
      an exponential one). I started with ordinary kriging. While GSLIB and
      SAS give similar results (the differences are probably be due to
      differences between solaris/fortran and windows/??), ISATIS keeps
      producing errors of the form

      Error during Kriging
      Rank of the point (in selection) = 331
      -> Inversion problem at rank = 3

      The error message changes with the kriging neighbourhood but seems to be
      indifferent towards changes in the sample size and semivariogram model.
      SAS and GSLIB don't even show matrix inversion problems when performing
      kriging using a global neighbourhood.
      I've been through the parameter files over and over but I can't find any
      mistake. But it's my first time with ISATIS, so I'm a little lost here.

      Anybody able to give me a hint on where I commit the mistake?

      Uh: Version is 4.0.6 for windows. And of course ISATIS works well for a
      different data set...

    • Show all 2 messages in this topic