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

GEOSTATS: Sensitivity analysis & geostatistics

Expand Messages
  • Gregoire Dubois
    Dear all, happy new year to the 1290 subscribers of ai-geostats ! 2000 has started in a very silent way. I hope this will stimulate a bit the discussions on
    Message 1 of 3 , Jun 16, 1905
    • 0 Attachment
      Dear all,

      happy new year to the 1290 subscribers of ai-geostats !

      2000 has started in a very silent way. I hope this will stimulate a bit
      the discussions on the list. For what concerns the management of the
      web pages, I won't be able to make any major updates and changes
      before March because of technical reasons (unless someone is willing
      to pay my phone bills to maintain the web pages via modem).

      I have few questions on the mapping of geostatistical data. I'm working
      on the mapping of radioactivity in the environment after a Chernobyl type
      accident. Two approaches to the mapping are clearly needed: the first one
      should be as fast as possible for immediate counter-measures. The second
      one allows more time to be taken and is needed for the long-term
      management of contaminated territories.
      While the long-management allows the organisation of new sampling
      campaigns, it is not the case for the short term management. The last will
      be based on information provided by automated sampling networks.
      Geostatistics have proven to be adequate methods for the mapping of a
      variable which is presenting large scale structures (due to the plume
      of contamination) as well as a strong local variability due to the role
      played by daily rainfall which is the main factor in the deposition of
      the radioactivity. Rainfall can occur over very small areas, rendering
      the description of the deposition more difficult.
      Geostatistics, however, require many assumptions to be verified and
      the modelling of the spatial correlation can be problematic. There are
      many different methods that are solving different problems.
      So which one should be chosen by default ?
      Few errors can appear in the data and have a great impact on the
      structure of the semivariogram. Therefore, the automatic modelling of
      the spatial correlation should not be an acceptable solution.
      Neural networks are promising techniques for the automatic mapping
      of environmental data but they still require to be combined
      with geostatistical techniques to be really efficient. As far as I know,
      there are only very few publications where a fully automated technique
      provides better estimates than ordinary kriging.
      GIS are now providing more advanced tools for spatial statistics. A
      possible drawback is the misuse of the functions when used by non
      experts. This again underlines the need for robust and automated
      methods. Intensive cross-validation techniques could be a solution
      for the definition of optimal parameters but, because of the many
      parameters that have to be tested, takes too much time. Also remains
      the problem of the risk to apply cross-validation techniques on a dataset
      that would involve few wrong measurement that can be identified
      only by expert-knowledge.

      I believe many answers would be found in works on sensitivity analysis
      (impact of data, parameters and models on the estimates) of the main
      spatial interpolation techniques. It would also provide many answers to
      the debate "what is the best interpolation method".

      Any references, papers or comments are more than welcome.

      Best regards

      Gregoire



      Gregoire Dubois
      Section of Earth Sciences
      Institute of Mineralogy and Petrography
      University of Lausanne
      Switzerland

      Currently detached in Italy

      http://curie.ei.jrc.it/ai-geostats.htm

      ____________________________________________________________________
      Get free email and a permanent address at http://www.netaddress.com/?N=1
      --
      *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!
    • Bill Thayer
      Gregoire and the rest of the list, I am interested in using sensitivity analysis and value of information (VOI) analysis to design sampling plans for human and
      Message 2 of 3 , Jan 13, 2000
      • 0 Attachment
        Gregoire and the rest of the list,

        I am interested in using sensitivity analysis and value of information
        (VOI) analysis to design sampling plans for human and ecological risk
        assessment. I would be interested in references or thoughts on the subject
        of VOI and spatial statistics.

        VOI is useful when one is more interested in determining the variables that
        have the most effect on the decision to be made (e.g., whether remedial
        action is required; choosing from amongst remedial alternatives) in
        addition to identifying the variables that have the largest effect on the
        uncertainty in a predicted quantity (e.g., soil concentration).













        >Dear all,
        >
        >happy new year to the 1290 subscribers of ai-geostats !
        >
        >2000 has started in a very silent way. I hope this will stimulate a bit
        >the discussions on the list. For what concerns the management of the
        >web pages, I won't be able to make any major updates and changes
        >before March because of technical reasons (unless someone is willing
        >to pay my phone bills to maintain the web pages via modem).
        >
        >I have few questions on the mapping of geostatistical data. I'm working
        >on the mapping of radioactivity in the environment after a Chernobyl type
        >accident. Two approaches to the mapping are clearly needed: the first one
        >should be as fast as possible for immediate counter-measures. The second
        >one allows more time to be taken and is needed for the long-term
        >management of contaminated territories.
        >While the long-management allows the organisation of new sampling
        >campaigns, it is not the case for the short term management. The last will
        >be based on information provided by automated sampling networks.
        >Geostatistics have proven to be adequate methods for the mapping of a
        >variable which is presenting large scale structures (due to the plume
        >of contamination) as well as a strong local variability due to the role
        >played by daily rainfall which is the main factor in the deposition of
        >the radioactivity. Rainfall can occur over very small areas, rendering
        >the description of the deposition more difficult.
        >Geostatistics, however, require many assumptions to be verified and
        >the modelling of the spatial correlation can be problematic. There are
        >many different methods that are solving different problems.
        >So which one should be chosen by default ?
        >Few errors can appear in the data and have a great impact on the
        >structure of the semivariogram. Therefore, the automatic modelling of
        >the spatial correlation should not be an acceptable solution.
        >Neural networks are promising techniques for the automatic mapping
        >of environmental data but they still require to be combined
        >with geostatistical techniques to be really efficient. As far as I know,
        >there are only very few publications where a fully automated technique
        >provides better estimates than ordinary kriging.
        >GIS are now providing more advanced tools for spatial statistics. A
        >possible drawback is the misuse of the functions when used by non
        >experts. This again underlines the need for robust and automated
        >methods. Intensive cross-validation techniques could be a solution
        >for the definition of optimal parameters but, because of the many
        >parameters that have to be tested, takes too much time. Also remains
        >the problem of the risk to apply cross-validation techniques on a dataset
        >that would involve few wrong measurement that can be identified
        >only by expert-knowledge.
        >
        >I believe many answers would be found in works on sensitivity analysis
        >(impact of data, parameters and models on the estimates) of the main
        >spatial interpolation techniques. It would also provide many answers to
        >the debate "what is the best interpolation method".
        >
        >Any references, papers or comments are more than welcome.
        >
        >Best regards
        >
        >Gregoire
        >
        >
        >
        >Gregoire Dubois
        >Section of Earth Sciences
        >Institute of Mineralogy and Petrography
        >University of Lausanne
        >Switzerland
        >
        >Currently detached in Italy
        >
        >http://curie.ei.jrc.it/ai-geostats.htm
        >
        >____________________________________________________________________
        >Get free email and a permanent address at http://www.netaddress.com/?N=1
        >--
        >*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!

        **************************************************
        William C. Thayer

        Environmental Science Center
        Syracuse Research Corporation
        6225 Running Ridge Road
        North Syracuse, NY 13212-2510
        phone: (315) 452-8424
        fax: (315) 452-8090
        email: thayer@...
        **************************************************

        --
        *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!
      • Gerard Heuvelink
        ... Dear Bill, There are lots of references that I could give you, but let me give you a few recent book/proceedings titles that should provide a useful
        Message 3 of 3 , Jan 17, 2000
        • 0 Attachment
          > I am interested in using sensitivity analysis and value of information
          > (VOI) analysis to design sampling plans for human and ecological risk
          > assessment. I would be interested in references or thoughts on the subject
          > of VOI and spatial statistics.
          >
          > VOI is useful when one is more interested in determining the variables that
          > have the most effect on the decision to be made (e.g., whether remedial
          > action is required; choosing from amongst remedial alternatives) in
          > addition to identifying the variables that have the largest effect on the
          > uncertainty in a predicted quantity (e.g., soil concentration).

          Dear Bill,

          There are lots of references that I could give you, but let me give you a few
          recent book/proceedings titles that should provide a useful 'introduction', with lots
          of references. Please note that these are references on (spatial)
          sensitivity/uncertainty analysis, not on VOI (with which I am not familiar).

          Chan, K., S. Tarantola and F. Campolongo, 1998, (Proceedings of the) Second
          International Symposium on Sensitivity Analysis of Model Output. Please contact
          Stefano Tarantola (stefano.tarantola@...) who I think will be most willing to
          send you a copy of the book for free.

          Heuvelink, G.B.M., 1998, Error Propagation in Environmental Modelling with GIS.
          London: Taylor & Francis. (Please forgive me for promoting my own work.)

          Lowell, K. and A. Jaton, 1999, Proceedings of the Third Symposium on Spatial
          Accuracy Assessment in Natural Resources and Environmental Sciences. Ann
          Arbor: Ann Arbor Press.

          Please let me also draw your attention to the FOURTH Symposium on Spatial
          Accuracy Assessment in Natural Resources and Environmental Sciences, to be
          held in Amsterdam, 12-14 July 2000 (http://www.gis.wau.nl/Accuracy2000/).

          Finally, it might be interesting for you to get in touch with Daniel Griffith
          (dgriffith@...), a spatial statistician that has published a lot on
          spatial undertainty analysis, or with Steve Stehman
          (svstehma@...), a statistican who is specialised in sampling
          theory, applied to forestry/remote sensing. Both are from Syracuse, so that
          should be convenient.

          I hope this is of use to you, best regards, Gerard Heuvelink

          --------------------------------------------------------------
          Gerard B.M. Heuvelink
          Institute for Biodiversity and Ecosystem Dynamics
          University of Amsterdam
          Nieuwe Prinsengracht 130
          1018 VZ Amsterdam
          The Netherlands

          Tel: +31(0)20-525 7448 (Secretary 7451)
          Fax: +31(0)20-525 7431
          email: g.b.m.heuvelink@...
          http://www.frw.uva.nl/soil/Welcome.html

          Have you recently visited http://www.gis.wau.nl/Accuracy2000 ?
          --
          *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!
        Your message has been successfully submitted and would be delivered to recipients shortly.