Page 411 - Acquisition and Processing of Marine Seismic Data

402 7. SUPPRESSION OF MULTIPLE REFLECTIONS

Its advantage is that it does not require correct predictive deconvolution for shorter period
subsurface velocity information obtained from multiples and cannot be handled for long
primaries, but needs the velocity of the multi- periods. Parameter determination is an impor-
ples, which can be assumed to be between tant point and autocorrelograms of the input
1480 and 1520 m/s in most cases. traces are used to determine the correct predic-
Although f-k filtering has a simple theory and tion lag and operator length values for an effi-
is quite straightforward to apply, practical cient multiple suppression. This situation
applications indicate that the selection of the means that the success of the method depends
proper mute zone is important in discrimination on the experience of the processor. In general,
of primary and multiple amplitudes in the f-k several tests and trials are required to determine
domain. It generally fails in the case of strong these parameters accurately. Practical applica-
multiple energy due to hard seafloor conditions tions indicate that the method also works well
and leaves a significant amount of residual mul- in the τ-p domain where the multiples are more
tiple amplitudes in the output. The most impor- periodic for all constant p values (Fig. 7.35E).
tant disadvantage is that the method also The SRME method is based on work done at
removes amplitudes of primaries in near-offset Delft University (Verschuur et al., 1992), and
traces, because the primaries also remain flat- therefore also is known as the Delft approxima-
tened in near offsets after NMO correction. tion. It is the most common and one of the most
Therefore, the performance of f-k filtering in successful methods used in the seismic proces-
multiple suppression is poor for small offset sing industry today. Results of SRME are gener-
data of short streamers, even if the subsurface ally superior as compared to the results of the
geology is not complex. This situation is also other commonly used demultiple methods. It
valid for the seafloor reflection: since its velocity is the only technique that almost completely
is close to the multiple velocity, the seabed removes surface-related multiples as well as
reflection hyperbola is also flattened after peg-legs (Fig. 7.35F). Like other demultiple
NMO correction, which results in a significant methods, SRME has its own advantages and dis-
removal of the amplitudes of seabed and shal- advantages. The most important advantage is
low reflections. These disadvantages generally that the conventional SRME technique does
make f-k filtering an unfavorable tool as com- not require any a priori information about the
pared to other techniques for multiple elimina- subsurface, such as seafloor reflectivity, or 2D/
tion. Fig. 7.35C shows the output of f-k 3D velocity distribution. However, it requires
filtering to remove the multiples. Although it a regular shooting geometry and equal shot
works fine for first-order seabed multiples, and receiver interval, so that a shot location
peg-legs still exist in the output. should exist at each receiver location, since
One of the well-known multiple suppression every receiver location is also used as a source
methods, successfully used for over 40 years, is location. This prerequisite, however, is not the
predictive or gapped deconvolution, based on case for most 3D acquisition geometries. The
the periodicity of the multiples in the time quality of SRME output is controlled by several
domain to design an operator that identifies factors, such as wide shot spacing resulting in
and removes the multiples (Fig. 7.35D). It is aliasing, narrow spread for narrow azimuth
especially effective on short period multiples data, feathering, large crossline sampling due
and fails if the multiple period is longer than to the wide cable spacing in 3D acquisition, or
approximately 200 ms, since the multiples are missing near and/or far offsets. During the com-
not perfectly periodic at far offsets on the shot putations, all receiver locations from zero to
records. This situation is partially tolerated by maximum offset are used. The input data must