Stacking velocity information (interval velocity derived using Dix's formula from the root-mean-square [RMS] velocity) is the prime seismic velocity information obtainable from a multichannel seismic reflection (MCS) survey. However, although stacking velocities are useful in data processing (they allow correct stack and enable data migration), their use as interval velocities to predict the subbottom depth of a target reflector is, in general, not recommended, especially if the available offset is short compared to the depth of the target. In addition, semblance-velocity analyses and constant-velocity-stack analyses imply that a gross spatial average (the acquisition spread) is applied to the velocity estimation and that a very simple earth model (i.e., homogeneous horizontal layers) is assumed.
In order to obtain additional and independent velocity information and to compare it to the stacking velocities, we have adopted reflection tomography as described in detail by Carrion et al. (1993a, 1993b). We briefly recall its basic concepts below.
The space is discretized by long pixels, which are zones where the propagation velocity of seismic waves is assumed to be constant. The pixel shape may be irregular, and the lateral boundaries are vertical straight lines; the upper and lower boundaries coincide with the interpreted reflecting interfaces (see Böhm et al., 1999). The pixel shape is dictated by the available resolution with a surface acquisition geometry (Böhm et al., 2000) that allows good estimation of the lateral gradients but is poor for the vertical variations between two reflectors. A vertically averaged velocity is estimated between the upper and lower reflector.
Traveltimes are picked manually or in semiautomatic mode (Tinivella, 1998) along selected horizons on common-offset and common-shot gathers. An iterative inversion procedure is started with any initial model, even very far from the true solution. Rays are traced according to Fermat's principle (Vesnaver 1994, 1996a, 1996b) to simulate the seismic wave propagation in that velocity field and reflector position. Data inversion is completed according to the simultaneous iterative reconstruction technique (SIRT) (Van der Sluis and Van der Vorst, 1987; Stewart, 1991). At each step, the velocity distribution is updated first, obtaining an estimate of the depth location of the reflection point for each source-receiver position and for each considered reflector. Then, observing the pattern and the dispersion of these reflection points, a new reflector geometry and local velocity structure are introduced, while also considering lateral velocity gradients. The final solution is obtained by alternately iterating these two steps until a minimum dispersion is obtained for all reflectors. An example of a final solution is presented in Figure F3 (from analysis near the location of Site 1103).