The theoretical result is based on the assumption that the numerical algorithm used to propagate the wavefield is exactly the same. Unfortunately, SEP still lacks a downward-continuation shot-profile migration capable of handling severe velocity variations, though Brad Artman is close to succeeding in getting one up and running 2002. Therefore, for the moment I had to run an ``imperfect'' test and I compare the results of migrating synthetic data by source-receiver downward continuation and reverse-time shot-profile migration Biondi and Shan (2002).
Figure 1 shows the shot with source location at .5 kilometers used for the test. Figure 2 shows the zero offset (stack) image produced by both migrations methods. The panel on the left shows the image produced by shot-profile migration, and the panel on the right shows the image produced by source-receiver migration. The two images are similar, except for a small difference in frequency content caused by the fact that I did not enter a perfect impulsive source in the shot-profile migration to avoid dispersion. Not only the flat reflector is imaged similarly in the two images, but also the strong ``ghost'' reflectors caused by the triplication of the wavepath Valenciano and Biondi (2002), visible between the surface locations of 0 and 1 km, are almost identical.
Figure 3 shows the subsurface offset-domain common image gathers at the surface location of 300 meters: panel a) shot-profile migration, panel b) source-receiver migration). Again the images are similar for both the ``true'' reflector and the ``ghost'' reflectors. Figure 4 shows the angle-domain common image gathers obtained from the offset-domain gathers shown in Figure 3 after a slant stack transformation Sava et al. (2001). Notice that the ``true'' reflector gets imaged at both positive and negative aperture angle because of the wavepath triplications. The ``ghost'' reflectors get imaged in the aperture-angle gap between the two branches of the true reflector.
Figure 5 demonstrates that the artifacts disappear if the whole data set (400 shots) is imaged. It shows the zero offset image [panel a)] and the angle-domain common image gather [panel b)] obtained by source-receiver migration when all the shot records are included in the data. I did not migrate all the shots by reverse time migration, because it would have taken considerable computer resources.
Shot-trip
Figure 1 Shot profile used for the tests. |
Cig-trip-100-both
Figure 3 Offset-domain common image gathers obtained by slicing the migrated cubes at the surface location of 100 meters: a) shot-profile reverse-time image, b) source-receiver downward-continuation image. |
Ang-Cig-trip-100-both
Figure 4 Angle-domain common image gathers obtained by slicing the migrated cubes at the surface location of 100 meters: a) shot-profile reverse-time image, b) source-receiver downward-continuation image. |
Mig-Ang-Cig-trip-300-sr-full
Figure 5 Stacked image (panel a)) and angle-domain common image gather at surface location of 100 meters (panel b)) obtained by source-receiver downward-continuation migration of the whole data set (400 shots). |
I have proven theoretically that source-receiver migration is exactly equivalent to downward-continuation shot-profile migration.
The results of the migration tests that I show strongly support this theoretical result, though they are not the ultimate proof, since I was limited to run a reverse-time shot-profile migration instead of a downward-continuation shot-profile migration.
I would like to thank Bill Symes of Rice University for kindly providing the synthetic data set that was used for the tests shown in the paper.