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

AI-GEOSTATS: references, indicator kriging, soft data

Expand Messages
  • lorenz.dobler@gla.bayern.de
    dear members, Is there anyone with experiences in indicator kriging with soft data using gslib ? i have a continous primary variable and a continous
    Message 1 of 1 , Apr 8 9:21 AM
    • 0 Attachment
      dear members,

      Is there anyone with experiences in indicator kriging with soft data using
      gslib ?

      i have a continous primary variable and a continous (exhaustive) secondary
      variable. I allready did "normal" indicator kriging (so i allready have a
      set of transformed hard indicator data and variograms for the corresponding
      thresholds) and the result looks rather plausible. my idea now is to
      incorporate the secondary variable within indicator kriging to improve the
      estimates in regions where sample densitiy is sparse (similar to kriging
      with external drift or simple kriging with varying local means using uncoded
      "raw" values).

      the most straightforward method seems to be simple indicator kriging using
      soft prior probabilities as described by Goovaerts (1997; pp 307) and
      Deutsch & Journel (1998; pp77, i hope they mean the same by "simple kriging
      with prior means" !).

      some questions about that method just to be sure that i am on the right way:

      1. first i have to classify my secondary (continous) soft data. how do i get
      discrete classes of soft data with a "calibration scattergram". In Deutsch &
      Journel 1998; pp92 i can not recognize how they decided to classify ! what
      is the best way to classify - the more classes the better ?

      2. i have to calculate soft prior probabilities. the calibration step is to
      calculate for each class the proportions of data that do not exceed one (or
      more) threshold(s) (i have 7 thresholds!). example: i have defined 3 (or 5)
      classes of soft data (question 1) so i have to calculate 3(5) different
      frequencies of not exceeding one threshold. the classes of soft data are
      then replaced by the calculated soft probabilities (=local prior
      probabilities) => if i have more thresholds i get 3(5) different values as
      local prior probabilities for each threshold ???

      3. the rest of the method is very similar to simple kriging with varying
      local means. the residuals i get by substracting hard indicator data [0,1]
      from local soft probabilities (0 ....1) calculated in question 3. the
      residuals are used to calculate semivariograms.

      4. how can i do the kriging step within wingslib? with ik3d (what about the
      exhaustive data-file?) or with kt3d? at least i want to use all the
      advantages of indicator kriging (maps of estimated values, probabilities,

      p.s. i am looking for some basic references on indicator kriging (using soft
      data) with gslib, especially Journel (1987): Geostatistic for the
      environmental sciences, EPA project no. cr 811893. Technical report. U.S.
      EPA Lab, Las Vegas, NV. it's hard to get in germany ...


      Lorenz Dobler
      Bayer. Geologisches Landesamt
      Heßstr. 128
      80797 München
      Tel.: 089/9214-2766
      e-mail: Lorenz.Dobler@...

      * To post a message to the list, send it to ai-geostats@...
      * As a general service to the users, please remember to post a summary of any useful responses to your questions.
      * To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
      * Support to the list is provided at http://www.ai-geostats.org
    Your message has been successfully submitted and would be delivered to recipients shortly.