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

1341GEOSTATS: Re: GIS, Geostats & Simulations

Expand Messages
  • Steve Zoraster
    Aug 19 8:34 AM
    • 0 Attachment
      >
      > Steve Zoraster made a good point by asking me why I left
      > conditional simulations out of the typical work sequence of a
      > geostatistician.

      <snip>

      >
      > My posting on GIS & Geostatistics was clearly biased because of
      > my background in environmental sciences. I would say I was not
      > personally attracted (yet) by the use of conditional simulations
      > because
      >
      > 1) of the heavy computing load of the calculations,

      I think that should be less and less of a problem as
      computers get faster and faster....

      > 2) I was looking more for exact interpolators,

      Certainly any environmental work dealing with ground water
      pollution or aquifer flow, etc., is likely to be subject to
      measurement error. Heck, even porosity and permiability
      measurements done in a laboratory on cores extracted from
      well bores are recognized to be subject to error. (Send neighboring
      cores from the same rock unit in the same well to two different
      laboratories....)

      > 3) I prefer to work on the real data and would like as much as possible
      > to avoid the required transformations for the Gaussian constraint,

      There are ways around the transformation to Gaussian path. Using
      Indicator Kriging is one way. However, none are easy or intuitive.

      > 4) it seems to be that simulations are much more sensitive to the
      > verification of the hypothesis of stationnarity which is probably
      > more frequently verified with the variables analysed in the mining
      > industry and in the petroleum industry, while in the environmental
      > sciences, the variables present often several populations and higher
      > local variabilities,

      Hmmm....

      > 5) the recent developments in robust variography (see Genton's work,
      > in the combined use of NN and geostats (Kanevsy) provide very good
      > results with not too much effort. Simulations are "simulations" while more
      > traditional geostatistics allow to be somehow closer to an intuitive and
      > natural approach to the estimation problem,

      I wonder. My guess is that in the petroleum industry more people are
      miss-using geostatistics, even relatively simple geostatistics than
      are using it correctly. I can't prove this, but I do know a lot
      of people who are making money as geostatistical consultants.

      And I know a number of people who are using the algorithm described in
      that paper of mine Geogorie cited, rather than try to do co-Kriging, or
      external drift type work as required, for example, in SIC97. Why? Because
      it's intuitive while allowing the imposition of interpretations on data.
      Also - and I realize I'm pushing my own stuff here - my algorithm doesn't
      dependon stationarity. (But it is also patented, so outside of research
      environments....)

      > 6) of the few published cases studies in environmental sciences that are
      > making use of simulations. Only one person on 40 participants has used
      > simulations for the Spatial Interpolation Comparison 97 (SIC97) for daily
      > rainfall measurements. Unfortunately the author didn't finalise his work and
      > his contribution was therefore not published.
      >

      Hmmm....

      >

      Actually, even I don't know exactly what those "Hmmm..."s mean. :-)

      Steven Zoraster
      --
      *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.
      *To unsubscribe, send email to majordomo@... with no subject and
      "unsubscribe ai-geostats" in the message body.
      DO NOT SEND Subscribe/Unsubscribe requests to the list!