Validation Procedure
Without repeated validations on different computer systems, it is impossible
to distinguish Hardware and Software Errors. The magnitude of Type IV (model errors) can be
controlled by variation of the depth discretization of the model.
Therefore the procedure has to be as follows:
- A continuously defined velocity model (like IASP91) has to be used to
avoid errors due to interpolation between predefined depth samples.
- From this continuous model, several differently discretized versions have
to be used. MKCLR2MODEL allows to define a discretization interval and
to use two different discretization methods: "spherical" which produces
depths samples that are equidistant in spherical coordinates, and "flat", which
produces depth samples that are equidistant in flat earth coordinates.
- By variation of the depth sampling interval and the sampling mode of MKCLR2MODEL it
can be determined at which discretization the Type IV Errors may be neglected: this
is the case when a further improvement of discretization does not result in
further reduction of the total error.
- Type V errors are made sufficiently small by applying an optimization process to the
seismic ray's take off angle.
- The remaining differences between reference travel times and computed travel times
are the sum of type I, type II and type III errors. This sum should be below the level
defined by the reference's roundoff error, but it is not desireable to make it zero.
eof.