Data

The data were acquired over a salt dome using three cables of 3570m length with geophones every 25m. The CMP sampling inline was 13.33m and crossline of 25m. The entire survey covered 13.5km inline and 4km crossline. Although this is not an exceptionally large 3-D survey, tomography is an iterative process so reducing the cost of the migration is an important consideration. For tomography, well implemented Kirchhoff methods which can produce a sparse set of CRP gathers are ideal. An alternate approach when dealing with marine data is Common Azimuth Migration (CAM) Biondi and Palacharla (1996). CAM requires more expensive full volume imaging, but provided three advantages to me:

• it is faster than Kirchoff when performing full volume imaging
• it can easily produce model domain angle gathers Prucha et al. (1999); Sava and Fomel (2000)
• it was already implemented in SEPlib

This dataset has been previously migrated using CAM Vaillant and Calandra (2000); Vaillant and Sava (1999) at SEP. In this paper, and the previous papers, the data volume was reduced to 10.5km inline and 4km crossline. CAM operates in the frequency-wavenumber domain so the data had to be placed on a regular grid. Data regularization was performed using Azimuth Moveout (AMO) Biondi et al. (1998). In the process of performing AMO the dataset was resampled. The resampled dataset had CMP spacing of 20m in the inline and 25m in the crossline. The offset range was resampled to 50m ranging from 200m to 3400m. This paper is simply attempt to prove that the concept works in 3-D, so for computational speed I decided to concentrate on the first 2500m in depth. Therefore I began by using only the first 3 seconds of the data. By windowing the data I was able to reduce the number of frequencies needed from 176 to 108 while still being able to handle frequencies up to 54Hz.

The initial velocity model was created using the S.M.A.R.T method Ehinger and Lailly (1995); Jacobs et al. (1992). Early migration tests showed that a better migration result could be obtained by smoothing the model Vaillant and Sava (1999) . As a result the S.M.A.R.T. model was smoothed, preserving the sharp salt boundary (Figure ). Using the velocity in Figure  the data was migrated with six reference velocities and frequency range of 5 to 60 Hz Vaillant and Calandra (2000). Figure  shows the initial migration result. Note how the reflectors generally have good coherency but die out at the top of the salt and along the salt flanks.

 

elf3d.vel0
Figure 1 Initial ELF velocity model.

 

elf3d.mig.vel0
Figure 2 Initial migration using the velocity in Figure .