Post mortem

The update to the interval slowness contains rapid lateral variations. These variations have a wavelength approximately the same as the cable length (2 km). The slowness model had a node spacing of 1 km horizontally and .25 km vertically. While some of these variations may be actual features of the true interval slowness model, it is more likely that they are artifacts. Gaps in reflector point coverage and variations in quality of velocity information are possible causes (Fowler, 1988). The shape of the multiples in Figure 2 do indicate some short wavelength features (less than a cable length), and eventually these features can be estimated; but for an initial update after a migration with v(z), it is best to try for a laterally smooth update.

 

deltaw.smoothed
Figure 12 A smoothed version of the update to the model in the previous figure. The smoothing was over 3 nodes laterally and vertically. This removes the features that resonate with the cable leaving only features with lateral wavelengths longer than the cable.

For the example shown, the inversion was unconstrained and no model covariance matrix was used. Smoothing the result show in Figure 11 gives the slowness model update shown in Figure 12. The low wavenumber change to the background model is more evident. It would have been more correct get a smoother update by including a model covariance matrix in the inversion. I will do this in the future.

It is encouraging that the algorithm recognized the slowness decrease (velocity increase) we would expect over and around the dome where deeper, faster rocks are thrust up. The next step is to remigrate the data from scratch with the new velocity model and see if the stack is better and if the structural image improves.