Page 409 - Acquisition and Processing of Marine Seismic Data

400 7. SUPPRESSION OF MULTIPLE REFLECTIONS

7.8 QC IN MULTIPLE SUPPRESSION successful demultiple application, and it is
recommended that a second-pass velocity anal-
The expected output of a multiple attenuation ysis be performed after multiple suppression.
algorithm is the seismic data with all types CDP stacking is the earliest method to sup-
of multiples removed, with the primary reflec- press multiples. It exploits the residual moveout
tions unaffected. Existing multiple suppression differences between the primaries and multiples
methods try to accomplish this by considering in CDPs after NMO correction. Stacking can
different characteristics of the multiples, suppress a significant amount of coherent and
which enable us to discriminate the multiples incoherent noise and can improve the S/N ratio
from primaries in different domains, such as by up to 20 dB, depending on the number
time-distance, frequency, τ-p, etc. These charac- of traces involved in stacking. Its efficiency
teristics include their moveout difference, perio- increases as the moveout differences between
dicity, velocity or dip discrimination. The primaries and multiples increase, and therefore
efficiency and success of each approach strongly long offsets are crucial in the stacking process for
depend on the criteria used to attack the multi- an effective multiple removal. In addition, veloc-
ples via the characteristic features of the multi- ity distribution of the subsurface is required for
ples. Therefore, theoretical and practical NMO correction, which also controls the accu-
factors of each approach constituting their limi- racy of the residual moveouts between the pri-
tations should be properly evaluated before the maries and multiples, especially in shallow
application. This section discusses the advan- parts where even small changes in NMO veloc-
tages and mostly the limitations of the demulti- ity may produce a significant amount of residual
ple methods, to provide more insight into how moveout differences at far offsets. Therefore, a
to determine the suitable method for a particular careful velocity analysis is crucial in the effective
multiple type in seismic data. Fig. 7.35 compares removal of multiples by stacking. Picking of the
outputs of different demultiple methods on the velocities corresponding to the enclosures of the
same seismic data from a shallow continental multiple amplitudes in semblance plots, which
shelf. Comparisons can be made between the appear throughout the low-velocity zones (e.g.,
output of a particular method and a stack close to 1500 m/s of water velocity axis in rela-
and/or brute stack section. The brute stack in tively deeper parts) should be strongly avoided
Fig. 7.35A shows that the section is highly con- (Chapter 9). In Fig. 7.35B, the efficiency of stack-
taminated by surface-related multiples as well ing on the elimination of different types of noise,
as peg-legs. including multiples, is illustrated. Although
The most effective solution can be deter- first-order seabed multiples and peg-legs deeper
mined by considering the principles behind than approximately 700 ms remain intact in the
these techniques for solving a specific multiple stack section, most of the higher-order long
issue. Sometimes a number of different demulti- period multiples are completely suppressed
ple methods must be tried on the test data in after stacking.
order to cope with different multiple types. Dif- The f-k domain muting is a moderately effec-
ference sections between input and output may tive approach in multiple suppression. It works
indicate what is removed and what is left in the on the residual moveout differences between the
data after multiple removal. Autocorrelation primaries and multiples in the CDPs after NMO
sections may help in analyzing the effectiveness correction. Its efficiency increases as the residual
of a particular demultiple method. In general, moveout difference increases, which requires
velocity analysis is fairly improved after a long offsets, as in the case for CDP stacking.