Figure depicts the migrated image obtained using our benchmark velocity model. We use this image to relate all our velocity analysis results. Since this velocity model is not perfect, the migrated image is not perfect, either. Various gathers show substantial moveout, particularly inside the ``bowl'' (around depth 2500-3000 m).
Since our benchmark velocity model is not perfect anyway, and given the intrinsic non-linearity of migration velocity analysis, we decided to back away from this model and use a heavily smoothed version of it as our background model. Figure shows the smoothed model and the corresponding migrated image. The angle-gathers clearly indicate slowness inaccuracies which we try to correct using our WEMVA with differential image perturbations method.
Following the strategy used for our synthetic model, we run residual migration for a wide range of velocity ratios, and then pick at every location the value which corresponds to the flattest gathers. We compute the image perturbation using the differential equation (12), and scale it with the picked residual migration ratio. We use the image perturbation in Figure to invert for the slowness perturbation. We run 1 non-linear iteration and 9 linear iterations to obtain the slowness perturbation presented in the right panel of Figure .
Finally, we take the inverted image perturbation in Figure and update the background slowness in Figure to obtain the slowness model in Figure . This figure also shows the image obtained by migrating the data using this updated slowness model.
Comparing the background and improved slowness models in Figures and , we observe improved flatness in the upper part of the image (around depth 1500-2500 m). We also observe better definition of the ``bowl'' (around depth 2500-3000 m). However, the bottom-right corner of the image degrades slightly after inversion, possibly as a result of boundary effects or of poor picking during the residual migration step.
Comparing our improved slowness model and the benchmark model in Figures and , we also observe a few interesting differences. Again, the ``bowl'' (around depth 2500-3000 m) is better defined using our improved slowness model, although the upper part of the model is still flatter in the image obtained with the benchmark model.