Pete Davis

IGPP, SIO, La Jolla, CA92093-0225
(619) 534-2839

REM

Trial Model Exchange Format Specifications

One of the outcomes of the recent REM workshop held in La Jolla was a resolution to exchange computer codes and preliminary Earth models as part of the REM development process. To encourage general participation and to minimize the time we all have to spend reformatting our own favorite model to test against someone else's code, it was suggested that a uniform model exchange format that was both flexible and extensible be adopted. By conforming to this format, each contributor would render a favorite model or computer code immediately accessible to colleagues.

We propose the following as a model formatting scheme to meet these needs. It is our intention to provide with this format outline, computer code in C and FORTRAN both to (1) read a model and (2) return the value of a parameter and its radial derivative given a depth and frequency as input. If the specified depth falls on a first order discontinuity, the subroutine will signal that this is the case and return values for the parameter on both the upper and lower side of the discontinuity. Initially, the input model file will be in ASCII text format only. When this format is generalized to three dimensions, the computer code will be modified to accept binary input.

For the sake of uniformity, we adopt the following conventions: (1) spatial position will be expressed as normalized radius. (2) All models must begin at the Earth's center. (3) If omitted as an explicit parameter, gravity will be calculated on the basis of density.

1. Model regions

Model regions are delimited by first order discontinuities and specified by listing the (normalized) radius which constitutes their lower bound. Listing of regions is ordered beginning at the center and proceeding outward. The definition of model regions is the first information to be listed.

2. Parameters

At this stage in the development of the REM, we do not wish to limit the number of parameters which can be accommodated. While the model input format itself is certainly extensible, the computer code to process that format may not be so. Interpolation methods are allowed to vary by parameter and by region. (See below.) Each parameter is defined for a given region by a series of lines containing information about the interpolation method or scheme to be employed and the knot points or coefficients needed as input to that scheme.

We request each contributor to provide an absorption band model for each parameter where appropriate. This Q will be treated as a distinct parameter however. In general, it will be possible to specify multiple frequency bands for any single parameter. For each band, the user must specify two corners and the slope within the pass band.

3. Interpolation methods

Unfortunately, there are almost as many interpolation schemes as there are velocity models in the literature. We intend to develop code which can handle a series of representations including the following:

Table InterpolationPolynomial Representation
constant Chebyshev
linear Legendre
cubic spline natural
b spline Bezier curves
R spline
power law

Different parameters may have different interpolation schemes.

4. Example format 

   Model regions
      r1 
      r2 
      ...
      rN 
      model region termination line

   REGION: 1    PARAMETER:parameter 1      SCHEME:scheme a
      essential scheme setup variables
      knot variables/coefficients
      region/parameter termination line
 
   REGION: 1    PARAMETER:parameter 2      SCHEME:scheme a
      essential scheme setup variables
      knot variables/coefficients
      region/parameter termination line
...
 
   REGION: 1    PARAMETER:parameter m      SCHEME:scheme e
      essential scheme setup variables
      knot variables/coefficients
      region/parameter termination line

   REGION: 2    PARAMETER:parameter 1      SCHEME:scheme b
      essential scheme setup variables
      knot variables/coefficients
      region/parameter termination line
...
These guidelines are preliminary and open for discussion. Please contact us ( gmasters@ucsd.edu) if you feel that an important issue has not been addressed.

top
Pete Davis ( pdavis@mahi.ucsd.edu)

Go back to Workshop summary
Go back to List of Participants
Go to main page