next up previous print clean
Next: Discussion Up: Flat earth synthetic Previous: Shot-Profile Migration

Source-Receiver Migration

Figure 4 shows analogous plots to those presented in the previous discussion. The same earth model and analysis strategy is employed for these experiments, though a source-receiver geometry for the data is employed requiring sorting the data to midpoint-offset coordinates. The same migration split-step Fourier migration kernel is used. While the panel a result employs a fully populated, regular data set, the others only used shots every 10 receiver locations. Panel b was simply sorted and migrated. Panel c used the same amount of live data traces as panel b, with zero traces replacing nine out of ten traces from the full data volume. Thus, the first and last experiments migrated the same size data cube, while second and last experiments contain the same amount of non-trivial data though they are an order of magnitude different in size.

Importantly, the second panel does not show the aliasing problems present in the second panel of Figure 3 despite the same level of shot decimation. In this experiment, the subsampling of the shot-axis is partially mapped into both of the two new coordinate axes before migration. The coordinate transform of equation (1) thus distributes the axis' lower Nyquist limit equally to the new coordinates prior to migration. Since the output image coordinates inherit this same sampling, the resampled data naturally adhere to the band-limiting criterion of equation (5).

The image location and subsurface offset variables of these migrations have the same meaning as those discussed in the shot-profile section previously Biondi (2003). There have been two important modifications however due to the initial coordinate transform. First, notice the range of wavenumbers included in the second image space is drastically limited from the panel to the left showing the migration of all available shots. The resorting has effectively band-limited the image space honoring the Nyquist requirement appropriate for the image given the shot axis subsampling. Thus, the algebraic combination of source and receiver coordinates in the numerator makes this an inherently band-limited propagation method. Second, the division by two of both new axes stretches their Fourier dual domains. Notice that the alias replications in the right panel of Figure 4 appear at a wavenumber of 20 rather than the 10 seen in the shot-profile migration example. Despite the fact that the same number and sampling interval for the shot axis was used in both experiments, the division associated with the coordinate mapping has decreased the sampling interval in the space domain and stretched energy along the Fourier domain. This has happened independently of the three modes of aliasing described above and needs undoing separately as well.

 
sg
sg
Figure 4
Source-receiver migration results. Left panel imaged with all data. Center panel imaged with every tenth shot and sorted to midpoint-offset coordinates. Right panel has zero traces inserted to fill out the decimated data set migrated in the center panel to the data set's original size.
view


next up previous print clean
Next: Discussion Up: Flat earth synthetic Previous: Shot-Profile Migration
Stanford Exploration Project
5/23/2004