It is sometimes thought that prestack migration footprint is a 3-D problem that is due to the poor spatial sampling inherent in prestack 3-D data. Footprint is often visible on 3-D time slices as a hatched interference pattern, as in Figure 1. However, it is not always recognized that prestack migration can generate a characteristic footprint on 2-D data as well as 3-D data. It is worthwhile to work through a simple example of how footprint is generated on 2-D data so that the 3-D problem becomes more understandable.

Fig. 01
Figure 1. An example of footprint in the form of a cross-hatched interference pattern on a 3-D timeslice.

A decimation test with an extremely simple model is sufficient to show how footprint is generated by 2-D prestack time migration once the source interval exceeds the receiver interval. We put a single flat reflector at 400 m depth in a constant velocity medium of 1000 m/s. The receiver interval is 10 m, and the full dataset (before decimation) has a source interval of 10 m. We generate 61 end-on shots, each with 96 channels, using a Kirchhoff modelling algorithm. Figure 2 shows one of the unmigrated shot gathers. The synthetic seismograms have been generated with a 30 Hz Ricker wavelet.

Fig. 02
Figure 2. An unmigrated shot gather from the test dataset with a 30 Hz Ricker wavelet.

The advantage of using such a simple model is that it is very simple to generate the correct migrated image. In fact, NMO-stack produces exactly the right result for a flat reflector, so we will be most happy if prestack migration of each shot gather could produce as good a result as the NMO-corrected source gather in Figure 3. Notice that NMO-stretch reduces the resolution of the reflector on the right-hand side. If we prestack migrate the same source gather with a Kirchhoff algorithm, we get the result in Figure 4. Notice that the flat reflector is positioned correctly, along with the NMO-stretch, but there are prominent migration smiles that look as though they have been generated by the left and right ends of the shot gather.

Fig. 03
Figure 3. A single NMO-corrected shot gather.

One way of understanding how these migration artifacts have been generated is by taking a quick look at how the integrals in Kirchhoff migration are evaluated. The migration of each source gather is given by an integral of the form:

Equation 01

We are obviously not worrying about the exact details of the integral here. The factors within the square brackets […], if they were included, would include functions that determine exactly how the weighted summation of the shot gather at location xs along hyperbolic trajectories is done across the input shot gather p(t, xs, xr), as a function of time, t, and receiver position, xr.

Although the integral over receiver position has limits from minus infinity to plus infinity, we only have receivers that extend from the beginning to the end of the receiver spread. Therefore we truncate the integral at the beginning and end of the receivers by ignoring the first and third terms in this equation:

Equation 02

The origin of the artifacts in the migrated image of Figure 4 is due to the fact that we are only evaluating the middle term in this equation, and ignoring the other two.

Fig. 04
Figure 4. Migrated image from one prestack time migrated shot gather.

The prestack migration algorithm includes an integral over sources, as well as receivers:

Equation 03

In general, the integral over shots is simply converted into a summation, regardless of the number of sources, or the interval between source locations. Once again we must truncate an integral because we only have sources from the beginning to the end of the seismic line:

Equation 04

We evaluate only the middle term in this equation and neglect the first and third term, which is what generates the familiar edge effects at the beginning and end of the seismic line that we encounter even with poststack migration.

Fig. 05
Figure 5. Migrated image from two prestack time migrated shot gathers.

Figure 5 shows the image with two migrated shots. The flat reflector is relatively well imaged, but the edge effects from each shot are obviously not cancelling each other out at all. Figure 6 shows the results with 6 shots (shot interval = 10 x receiver interval). The pattern of edge effects now looks quite regular, like something we would call a footprint. What shot interval needs to be used for the edge effect from one shot to merge smoothly with the edge effect from the next shot, so that the edges disappear?

Fig. 06
Figure 6. Migrated image (shot interval = 10 x receiver interval).

Figures 7 to 11 show the migrated images as the shot interval is reduced from 5 times the receiver interval to a shot interval that is equal to the receiver interval. The interference pattern of edge effects (the migration footprint) does not fully disappear until the shot interval equals the receiver interval. Figure 12 shows the image from the NMO stack of all 61 shot gathers. The best prestack migrated result (Fig. 11) is not quite a good as the NMO stack because it has edge artifacts at the beginning and end of the seismic line.

Fig. 07
Figure 7. Migrated image (shot interval = 5 x receiver interval).
Fig. 08
Figure 8. Migrated image (shot interval = 4 x receiver interval).
Fig. 09
Figure 9. Migrated image (shot interval = 3 x receiver interval).
Fig. 10
Figure 10. Migrated image (shot interval = 2 x receiver interval).
Fig. 11
Figure 11. Migrated image (shot interval = receiver interval).

The purpose of this short article is to illustrate that migration footprint is a problem with 2-D data as well as with 3-D data. The simple example used in this article has also shown that, without any additional compensation, migration footprint arises in 2-D prestack migration once the shot interval is greater than the receiver interval.

Fig. 12
Figure 12. Sum of all 61 NMO-corrected shot gathers.

End

     

About the Author(s)

Peter Cary has B.Sc. and M.Sc. degrees in physics, and a B.A. degree in philosophy from the University of Toronto, and a Ph.D. in geophysics (1987) from Cambridge University, England. He worked for Chevron both in Calgary and in La Habra, California from 1982 to 1984 and was Manager of Geophysical Research with Pulsonic Geophysical Ltd. from 1988 to 1996. He is presently with Sensor Geophysical Ltd. in Calgary. He has p resented and published many papers on seismic processing, and served as technical program chairman of the SEG 2000 Annual Meeting and of the 1993 CSEG Annual Meeting. He is the past-president of the CSEG, after having served as president in 2004-05. He was also 2nd V.P. of the CSEG in 1996-97. He was an associate editor (seismic processing) of Geophysics from 1998-2001. One of his specialties is processing and writing software for multicomponent seismic data.

References

Appendices

Join the Conversation

Interested in starting, or contributing to a conversation about an article or issue of the RECORDER? Join our CSEG LinkedIn Group.

Share This Article