Srme PDF
Srme PDF
Srme PDF
Leharne Fountain
SUMMARY
The presence of free-surface-related wave phenomena is a
classic problem in marine seismic data processing. Over
the years, the industry has relied heavily on conventional
multiple suppression methods such as predictive
deconvolution and differential move-out filtering to
remove surface-related multiples from marine seismic
data. These methods are based on rather specific
assumptions about the subsurface and characteristic
differences between primaries and multiples. Since these
assumptions are often not met in the field, the
effectiveness of these methods may be limited. SurfaceRelated Multiple Elimination (SRME) is a relatively new
method that removes all surface-related multiples,
without using any additional information about the
subsurface. Application of SRME to offshore Australia
data sets results in much improved results, where
relatively weak primary reflections become more
interpretable.
Key words: Multiples, autoconvolution, SRME.
INTRODUCTION
Removal of free-surface multiples from seismic reflection
data is an essential pre-processing step in seismic imaging in
offshore Western Australia. Due to the high velocity
contrasts at the water bottom, first layer multiples tend to
decay slowly and degrade the quality of a large part of the
seismogram severely. In addition, peg legs are generated off
structurally complex 3D sedimentary bodies to create a
complicated set of reverberations that can easily obscure
primary reflections from relatively weak sedimentary
reflectors.
In complex geological environments where primary/multiple
energy ratios are generally low, it is essential to employ
multiple elimination methods that require no a priori
information, either structural or material, about the
subsurface geology, and which leave unaffected all relevant
information present in the data.
Surface Related Multiple Elimination (SRME) removes all
multiples that are introduced by a particular surface in the
Earth. In order to remove these multiples, both the geometry
and the reflection coefficients at this surface need to be
known. Since this information is readily available for the
water surface, it is possible to remove all multiples that are
generated by the water surface, without using any additional
information about the subsurface. The method is fully data
th
driven, meaning that only the data itself is used to predict the
multiples. As a result, user interaction is minimized.
In the following, we discuss the basic methodology of
SRME, and show some results of its application to a dataset
from the NW Shelf in offshore Western Australia.
SURFACE-RELATED MULTIPLE
ELIMINATION
Surface-related multiple elimination is applied in three steps
(Verschuur and Berkhout, 1997). The first step includes the
removal of all non-physical noise, regularisation of the
measured data to obtain a constant grid of sources and
receivers, the interpolation of missing near offsets and
missing intermediate offsets, and the removal of the direct
wave and its surface reflection. Since the method is datadriven, the quality of the data after multiple removal depends
heavily upon the pre-processed data.
The second step is the prediction of multiples. The
prediction is based on the observation that any surfacerelated multiple can be predicted through temporal and
spatial convolutions of the measured wavefield with itself
(Berkhout, 1982).
In the last step, the predicted multiples are subtracted from
the input data, using the minimum energy criterion, which
states that, after the subtraction of the multiples, the total
energy in the seismogram should be minimized.
For a long time, the SRME method has been considered to be
promising but too expensive and too difficult to run in
production processing. However, due to both increased
computer performance and increased understanding of the
crucial data preparation steps, the industry seems to be
moving towards a broader application of the method, and it
has even replaced more conventional methods in some
onboard processing projects.
Current acquisition configurations prohibit the application of
3D SRME. By assuming that no lateral variation occurs in
the cross-line direction, each individual streamer from a 3D
survey is assumed to pertain to a 2.5D configuration. After
an inline projection of each streamer, the 2D SRME method
can be applied. Small deviations from the 2.5D assumption
can be overcome in the adaptive subtraction process.
However, it is important to realize that most conventional
demultiple methods based on predictive deconvolution and
differential move-out filtering intrinsically assume localized
1D configurations, thereby ignoring any inline variation.
Extended Abstracts
RESULTS
SRME was applied to a 2D seismic line from the Carnarvon
Basin (NW Shelf area). Data from this area is known to be
severely contaminated with multiples that are generated by
near-surface carbonates. A very strong reflection coefficient
at the top and bottom of the carbonates results in a strong
surface multiple problem, and strong interbed multiples are
also a severe problem. Removal of multiple energy remains
the foremost obstacle to successful seismic imaging in
offshore Western Australia.
The full mechanism for
generating the multiple wavefield has never been properly
determined, nor has a satisfactory means been developed to
remove the multiples. The water bottom in the survey area is
very shallow (75 m), resulting in a strong train of short
period multiple energy (refer to Figure 1). Event amplitudes
are characteristically strongest at the near- to mid-offsets.
The following pre-processing steps were applied: Muting of
the direct arrival and its surface reflection, removal of
refracted wavefields, wavefield regularization, anti-aliasing
filtering and near offset interpolation.
Long et al.
ACKNOWLEDGEMENTS
The authors thank PGS Australia Pty. Ltd. for permission to
publish these results.
REFERENCES
Berkhout, A. J., 1982, Seismic Migration, Imaging of
acoustic energy by wavefield extrapolation, vol. 14A:
Theoretical aspects, Elsevier, Amsterdam.
Verschuur, D. J., and Berkhout, A. J., 1997, Estimation of
multiple scattering by iterative inversion, Part II: Practical
aspects and examples: Geophysics 62, 1596-1611.
CONCLUSIONS
Computational advances and increased understanding of the
crucial preparation steps are responsible for an increased
interest in the application of the SRME method to large 3D
data volumes. SRME requires no a priori information about
the subsurface, and as such it is fully data-driven. As a
result, very limited user interaction is needed.
th
Extended Abstracts
Long et al.
(a)
(b)
(c)
(d)
Figure 1. A single shot gather from a 2D dataset in the Carnarvon Basin, NW Shelf Australia. The raw shot is shown in (a).
Multiples predicted from the raw gather by SRME are shown in (b). The shot gather after the predicted multiples have been
subtracted is shown in (c). Note the improved strength of primary events (arrow). The gather depicted in (d) is the difference
between (a) and (c).
Figure 2. Autocorrelations of a single shot gather from 2D data shown in Figure 1. The autocorrelation of the raw shot is
shown on the left. The autocorrelation of the same shot after SRME is shown on the right. The ringing at near offsets evident
in the raw shot has been eliminated by SRME.
th
Extended Abstracts
Long et al.
(a)
(b)
(c)
Figure 3. Stacked sections of 2D data from the NW Shelf, offshore Western Australia. The raw stack is shown in (a) and the
stack after SRME has been applied is shown in (b). For comparison purposes, the stack after Tau-P deconvolution and radon
demultiple is shown in (c).
th
Extended Abstracts