... inversion^3.1

The least squares inversion algorithm used by Jive3D uses no damping, which substantially reduces the dependence of the inversion result on the starting model. In tests using real and synthetic data, Jive3D has consistently recovered the same final model from several different starting models.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... administrator^4.1

However, if you plan to produce graphical output using the scripts provided or run the utility calcerr on large problems, GMT and LAPACK must be installed on your system, which will require administrator-level access.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... line.^5.1

This is due to that fact that some implementations of Fortran 77 cannot read unformatted data from a character array.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... targets^5.2

This is slightly awkward to set up, as the absolute hydrophone positions must be specified for each shot, rather than their positions relative to the shot location. Jive3D does not run at its greatest efficiency under this configuration.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... run^6.1

Fortran 77 does not support dynamic memory allocation. Since the array sizes required for different inversion problems vary by several orders of magnitude, the correct sizes for each problem must be specified at compilation time.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... listed^6.2

Unless forward-modelling only has been requested.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... monotonic^6.3

It is, of course, possible that a triplication exists that has not been mapped due to the initial fan density being too low. This will lead to failures in the second stage two-point ray-tracing. It is important to attain a balanced fan density that maps out the essential characteristics of the function but that is not unnecessarily large.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... misfit^6.4

This is equivalent to an un-normalised $\chi^2$ misfit estimator.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... iteration^6.5

`Iteration' is used here in the context of the inversion `steps' and `iterations' described in section 3.4. This should not be confused with the iterations of the conjugate gradient solution to the optimisation problem discussed here.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... compatible^6.6

The Bourne shell was chosen because it is a `lowest common denominator' shell that is present on most systems. The scripts described here are all compatible with the Bourne again shell (bash) and alternative scripts designed to work with the C-shell are also provided (these have a .csh extension).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...gv^6.7

The utility gv is available in the public domain and may be downloaded from many software archives such as the Sunsite archivehttp://sunsite.org.uk/.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... bytes^7.1

For large inversion problems, double precision (which uses 8-byte or 64-bit numbers) is required to maintain numerical stability during the matrix inversion.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... files^7.2

Note that if these files have been produced from a different iteration of the inversion, the results from the uncertainty analysis will be wrong.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... longer^7.3

The inversion method used by jive3d requires neither the full, explicit calculation of the Hessian matrix nor does it require this matrix to be inverted.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... substituted^7.4

This is to ensure that there is no confusion between parameter values and parameter uncertainties, which could be caused if they were placed together in the same file.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... phase^8.1

In practice, it is not always easy to identify the phase of a horizon with complete confidence. Often a trial-and-error approach is necessary, where data are modelled under several possible phase interpretations which are then selected according to the feasibility of the models generated.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... used^8.2

If a bathymetry model is available that provides sufficient resolution, it can of course be used but it is important to check that the model is compatible with the reflections in the data.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.