Page 412 - Acquisition and Processing of Marine Seismic Data

7.8 QC IN MULTIPLE SUPPRESSION 403

be reconstructed at missing offsets for near off- complex subsurface areas, however, the NMO
set traces to obtain a regular distribution of approximation fails and NMO velocities are
traces for all existing offsets using data regular- generally inaccurate, which results in unflatten-
ization and interpolation. However, interpo- ing of the primaries after NMO. If the spatial
lated data is not as good as real data, sampling is poor, then the method may not pro-
especially for a complex 3D subsurface. duce satisfactory results due to the severe alias-
Although new innovations exist for 3D SRME ing problems in the τ-p domain. Since the
algorithms, a regularization process for the method relies on residual moveout differences,
requirement of regular shot and receiver distri- a careful and accurate velocity picking is neces-
bution for all offsets is the main issue for 3D sary. It is effective on multiples at moderate to
implementation of SRME algorithms and this deep waters depending on the streamer length,
makes it difficult and expensive to implement. since the efficiency of the Radon velocity filter-
This geometry regularization and its iterative ing increases with increasing offsets where the
nature can make it a time-consuming, and hence residual moveouts between primary and multi-
expensive, process depending on the input ples become more prominent. However, it may
data volume. Although it has unique advan- fail when the moveout difference decreases,
tages, it predicts only surface-related multiples such as in the case for peg-leg multiples or mul-
and interbed multiples are not handled. In addi- tiple energy in near offset traces. The remaining
tion, applications indicate that SRME works multiples in near offsets, however, can be
better for near offsets as compared to the far removed by a subsequent inner mute, if neces-
offset range. sary. A careful picking of the mute zone in the
Wave equation multiple rejection (WEMR) τ-p domain is necessary to make the primaries
predicts multiples by extrapolation with wave unaffected, which strongly makes the method
equation modeling and estimation of the sea- operator dependent. In general, amplitudes of
floor reflectivity (Fig. 7.35G). Its major advan- the primaries are also reduced to some extent
tage is that it works fine when the seabed is after Radon filtering, as is the case in
irregular, by means of dip, curvature and vari- Fig. 7.35H between 150 and 400 ms.
able reflectivity. It requires remarkable compu- Table 7.1 summarizes the advantages and dis-
tational time for wave field extrapolation. In advantages of the multiple suppression
addition, spatial aliasing, if it exists, and lack methods commonly used in the marine seismic
of near offset traces may cause issues in multiple industry today. In general, the most effective
prediction, and in ideal conditions, the WEMR method in multiple suppression is CDP stack-
technique requires data from zero to as large off- ing, which removes most of the coherent noise
sets as possible. from the data. The methods that do not require
Radon velocity filtering has a similar theoret- subsurface information, especially those inde-
ical basis as f-k filtering, which exploits the resid- pendent from the subsurface velocity distribu-
ual moveout differences between the primaries tion and that do not require an accurate 2D/
and multiples in the CDPs after NMO correc- 3D velocity field, such as predictive deconvolu-
tion. This time, however, the velocities of pri- tion or SRME, are preferable since they are less
mary reflections are used in NMO correction sensitive to operator experience. To determine
and discrimination is done in the τ-p domain the most suitable technique for a specific dataset,
after a parabolic Radon transform. Radon several tests and trials using more than one
demultiple is successful in attenuating multiples demultiple method on a small test dataset is
if the subsurface is not complex (Fig. 7.35H). In strongly recommended.