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

Re: GEOSTATS: intro to splines

Expand Messages
  • soroko@netcom.ca
    Hi, The classic reference, as M.Kay pointed out is: Briggs, I. 1974. Machine contouring using minimum curvature Geophysics, 39 And an improvement to the
    Message 1 of 4 , Oct 4, 1997
      Hi,

      The "classic" reference, as M.Kay pointed out is:

      Briggs, I. 1974. "Machine contouring using minimum curvature" Geophysics,
      39

      And an improvement to the method (gridding with tension to reduce the erroneous inflections) is
      in:

      Smith, W.H.F. and Wessel, P. 1990. "Gridding with continuous curvature splines in tension".
      Geophysics. Vol.55, No.3, p.293-305.

      Fortran Code is offered by Swain in:

      Swain, C.J. (1976) "A Fortran IV Program for Interpolating Irregularly Spaced Data using the
      Difference Equations for Minimun Curvature". Computers and Geoscience. Vol.1, p.321-240.

      And I haven't studied it intensively, but a more recent paper which uses directionalized tension
      as a way to incorporate simple anistropy.

      Mitasova, H. and Mitas, L. (1993). "Interpolation by Regularized Spline with Tension: I. Theory
      and Implementation". Mathematical Geology. Vol.25, No.6.


      I hope this helps!

      Jason Soroko
      University of Ottawa
      soroko@...



      > After checking out kriging I'm interested in what seems to be an
      > alternative: splines, so I'm looking for a simple introduction to splines,
      > thin plate smoothing splines, for the analysis of meteorological data
      > (average data) (simple as for example "An Introduction to Applied
      > Geostatistics" by Isaaks and Srivastava for kriging). Can anybody suggest
      > some references?!

      Matthew Kay responded:

      Dear Beatrice

      I have used a form of splines called minimum curvature (MC)

      A "classic" reference is:

      Briggs, I. 1974. "Machine contouring using minimum curvature" Geophysics,
      39

      An example is presented in:

      Mendonca, C. & Silva, J. (1974). "Interpolation of potential field
      data by equivalent layer and minimum curvature", Geophysics (60)

      However, I prefer kriging over splines/MC, as kriging can incorporate
      anisotropies present in your data.

      I may be wrong, but as far as i know, the only way anisotropies can be
      dealt with in MC, is by varying the search ellipse (an inferior solution).

      It has been shown by Matheron (1980) that a formal equivalence exists
      between kriging and splines, with splines being equivalent to kriging with
      a particular covariance model. Thus MC may give poorer results than
      kriging when your data's covariance (model) is different to the fixed one
      used by MC when there are few data in the area being interpolated.

      In zones where you have good spatial sampling then kriging and MC perform
      equally well.

      I hope this helps

      cheers

      matthew


      --
      *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!
    • Syed Abdul Rahman
      A couple of caveats: Before embarking on work using splines, or neural networks, etc, might be worth noting that all work on the same premise, i.e. an
      Message 2 of 4 , Oct 5, 1997
        A couple of caveats:

        Before embarking on work using splines, or
        neural networks, etc, might be worth noting that
        all work on the same premise, i.e. an underlying
        spatial correlation structure, and an interpolation
        component.

        Thin plate splines can be equivalent to same
        covariance models. Likewise radial basis function
        neural networks, which can be equivalent to
        kriging using a particular covariance model,
        depending on the radial basis function used.

        Unfortunately, splines and RBF networks are
        mostly packaged as black boxes (with some
        particular covariance model). Or, if "customization"
        is allowed, the spatial correlation structure can
        sometimes be modified, within certain limits,
        but it is still not as intuitive as graphical variogram
        modeling. The danger here is that the spatial
        analysis part -- the most important component
        in an interpolation exercise -- is relegated to the
        computer.

        Re: anisotropies:

        1) One can use the earlier suggestion, i.e. elliptical
        or ellipsoidal search neighborhoods when using
        splines or RBF networks, or

        2) Go back to the code, and modify the distance
        measures to include the anisotropy ratio, i.e.
        exagerating distances along directions of minimum
        continuity and compressing them along directions of
        maximum continuity.

        Note that a normal variography phase is still recommended
        to derive the anisotropy ratios in the first place.

        Regards,

        Syed

        --
        *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.