.The in-line direction is East-West,
corresponding to the approximate North-South direction
in the model as displayed in Figure 1.
Notice that the figure displays the model rotated
with respect to its ``true'' orientation;
that is,
the ``true North'' of the model points to the West of the figure.
The salt body in the model exhibits steep flanks near the crest and a rough surface on the top of the shelf. These characteristics cause severe distortions in the wavefield propagating through the salt. The reflectors below the salt area are thus poorly illuminated by data acquired with narrow-azimuth marine-like geometry. Consequently, the imaging of subsalt reflector is spotty even when using full-wave equation methods Ober and Oldfield (1999). Furthermore, deep dipping reflectors cannot be imaged because of the limited spatial extent of the data set. To reduce the computational cost of the modeling effort, the data were acquired on a dense grid only on a subset of the model. Good reference reflectors are: the bottom of the salt, the flat strong reflector at the bottom of the model (not marked in Figure 1), and the two sand lenses marked as "Lenses" in Figure 1. The bottom of the salt can be imaged pretty well in most of the areas, with the exception of the root proximities, where the interfaces are steeply dipping.
Before common-azimuth migration, the narrow-azimuth data were transformed to effective common-azimuth data by applying Azimuth Moveout Biondi et al. (1998). The regularized common-azimuth data set was binned with a 20 meters CMP spacing in both the in-line and cross-line directions, and with 100 meters sampling along the in-line offset direction. With 100 meters offset spacing, the moveouts of the shallow events are aliased. However, because the dips along the offset can be safely assumed to be always positive, aliasing by a factor of two can be easily overcome by both wave-equation migration and Kirchhoff migration Biondi (1998b). The data were muted with a ``deep'' mute because the early arrival are contaminated by all sorts of modeling noise. This mute affected the imaging of the shallow events. A more careful mute could accomplish both goals of noise removal and shallow events preservation. Kirchhoff migration is more flexible than common-azimuth migration with respect to the input-data geometry. Therefore, the original narrow-azimuth data were migrated by Kirchhoff migration.