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

AI-GEOSTATS: answers to the simulate in block support question

Expand Messages
  • Adrian Martinez Vargas
    Dear mail list there are the answers to the simulate in block support question: You have to simulate in an SMU scale and after that migrate to your block
    Message 1 of 1 , Oct 25, 2003
    • 0 Attachment
      Dear mail list there are the answers to the simulate in block support question:



      You have to simulate in an SMU scale and after that migrate to your block support.



      For example:



      If you want to simulate a 10x10 block, you can use a 1x1 simulation grid and after that migrate this 1x1 simulation to your 10x10 block.

      This migration is a simple moving average but you need to create the grids with the same size, exaclty in order to 100 blocks in SMU scale corresponds to 1 block in estimation scale.





      Alessandro Henrique Medeiros Silva

      Geologist - Technical Services

      AHSilva@...



      *********************************************************************



      Hi,



      good question. If you only need the uncertainty of the blocks then regularise to the new support and simulate. This is less processor intensive. If the information inside the block other than value is of importance, i.e. connectivity of high values, patterns etc, then simulate to a fine grid. The second option is better if you need to pass the values to a post-processing algorithm.



      regards



      Benjamin Warr



      Dr Benjamin Warr
      Research Fellow

      INSEAD - Center for the Management of Environmental Resources (CMER)
      Fontainebleau, France

      email: Benjamin.Warr@...
      http://soil.society.futures.free.fr

      http://terra2000.free.fr/

      http://benjamin.warr.free.fr



      **************************************************



      Simulate on a dense grid, then average over blocks.

      gstat, http://www.gstat.org/ can do it directly.
      --
      Edzer



      ****************************************************



      Adrian

      I am not totally familiar with gslib but if it doing
      proper block kriging, you should be able to simulate
      directly from those results.

      Much simpler (and more stable distribution wise) is to
      simulate a dense grid and average into blocks. Around
      100 points per block should give you better than 0.5%
      accuracy - as per my 1975 paper ;-)

      Isobel
      http://geoecosse.bizland.com/whatsnew.htm



      ***************************************************



      >>
      >It can be do, transforming the initial histogram of point samples to block support, or simulating in a very dense grid?
      >


      both alternatives lead to what you want, However the second followed by
      re-blocking the simulated dense grid to the block size you need is more
      rigourously correct


      regards



      [Non-text portions of this message have been removed]
    Your message has been successfully submitted and would be delivered to recipients shortly.