Next: Conclusion
Up: Vaillant & Calandra: Common-azimuth
Previous: Application to real data
Figures 3 and 4 show
comparisons between CAM and Kirchhoff migration results. The Kirchhoff
algorithm used is derived from a preserved-amplitude approach and
selects the most energetic arrival. Both CAM and Kirchhoff migration
use exactly the same velocity model.
Some of the most significant differences between both approaches are
well-known: in the CAM algorithm, finite frequency wave propagation is
modeled avoiding the asymptotic approximations necessary for Kirchhoff.
Also, for deep targets, the migration cost increases as Nz3 (number of
depth samples) for Kirchhoff, whereas CAM cost only increases as
Nz2. However, Kirchhoff methods allow target-oriented migrations
where CAM has to perform downward-continuation of the whole wavefield.
Figure 4 shows that the in-line sections around
the salt body are relatively comparable in quality. Globally, CAM
seems to give better results at imaging sediments bending against the
salt flank on the left-hand side.
The most important differences are shown by the horizontal sections
(Figure 3):
at a depth of 900m, CAM enhances complex high-frequency turbiditic patterns
in shallow layers. At the same location, the Kirchhoff
image appears at a considerably lower frequency and blurred along the
in-line direction.
There are potentially 3 factors that could explain CAM's better accuracy
compared to Kirchhoff migration:
- Wave propagation is handled completely
differently: CAM iteratively propagates the wavefield by regular depth
steps, as opposed to Green functions for Kirchhoff that are evaluated
by ray tracing, wavefront construction or other methods.
- Since Green functions are not calculated for every source but
rather pre-computed on a coarser grid, Kirchhoff algorithms usually
include critical interpolations of traveltime maps and other attributes
that may reduce accuracy for high-frequency details in the image. The
only interpolations performed with CAM are the extended split-step
scheme, which essentially addresses rapid velocity variations.
- Even if Kirchhoff algorithms can easily handle irregular
geometries, the resulting image incorporates acquisition footprints,
especially in depth slices.
On the contrary, CAM requires regular geometry and acquisition
problems are addressed during preprocessing through the AMO
operator. Thus, the whole imaging stage is done with regularized
geometry and potentially enables higher resolution in the final image.
kir-cam-zslice
Figure 3 Comparison between Kirchhoff (top)
and CAM (bottom) imaging results: depth slice at z=900m. The Kirchhoff
image has a lower frequency content, a higher noise ratio and is
blurred along the in-line direction.
Next: Conclusion
Up: Vaillant & Calandra: Common-azimuth
Previous: Application to real data
Stanford Exploration Project
4/27/2000