shallow water demultiple

ASEG 2010 - Sydney, Australia 1

Shallow Water Demultiple Barry Hung Kunlun Yang Joe Zhou Qing Long Xia CGGVeritas CGGVeritas CGGVeritas CNOOC Ltd Singapore Singapore Singapore Tianjin, P.R. China [email protected] [email protected] [email protected] [email protected]

INTRODUCTION

Efficient attenuation of water-layer related multiples remains one of the challenging issues in shallow water situations. For instance, conventional surface-related multiple elimination (SRME) method is known to have difficulties in attenuating shallow water multiples because the primary water-bottom reflection is not always present in shallow water seismic data due to the near offset gap (e.g., Verschuur, 2006). This kind of multiples become even problematic in revealing the subsurface structures if the water-bottom has a hard acoustic

response that generates a lot of very strong water-bottom multiples. Conventionally, predictive deconvolution in the x-t or -p domain is used in processing workflow for attenuating shallow water multiples (Ali et al., 2002). However, -p deconvolution is highly sensitive to precise amplitude consistency of the traces across a gather (Schoenberger and Houston, 1998). Wave equation approaches have been proposed to handle the issue of inaccurately recorded water-bottom data for multiple attenuation in shallow water environment (e.g. Pica et al., 2005). Whilst they have the advantage of relaxing the requirement in SRME that there must be a receiver at each shot location, they need structural model of water-bottom for predicting water-layer multiples. In very shallow water situations, the model may not be easy to obtain. The methodology of estimating primaries from multiples for coping with missing water-bottom reflections has been demonstrated to be a viable alternative for attenuating shallow water multiples (Biersteker, 2001). Using multichannel prediction operator that is estimated from the multiples, it provides a way of estimating the water-bottom reflections from the water-layer multiples (Hargreaves, 2006). Sparse inversion can also be utilised in the process for avoiding the need of adaptive subtraction that is required in SRME (van Groenestijn and Verschuur, 2009). Using an approach that is similar to the methodology, we demonstrate a workflow that also involves conventional SRME for effectively attenuating surface multiples in shallow water environment.

METHOD There are two ways to describe the process of SRME. One is the iterative approach that is described by Berkhout and Verschuur (1997): P = P(I + A P)-1 (1) where P is the acquired data, P represents the primary response and A is the surface operator that involves source properties and free-surface reflectivity. The other is the inversion method that is proposed by Biersteker (2001): P = P P F (2) where F is a multichannel prediction filter and is equivalent to a scaled version of primaries (Hargreaves, 2006). These two ways of implementation can be utilised in turn to attenuate free-surface multiples in shallow water environment by first tackling short-period water-layer multiples and then handling

SUMMARY Multiples due to shallow water are observed in seismic data acquired in various places such as the Gippsland Basin of Australia. These short period multiple reflections often pose problems to the interpretation of geological structures. They are not easily handled by conventional surface-related multiple elimination (SRME) methods because the recorded primary water-bottom reflection, which is required by SRME, is often indistinct in shallow water situations due to the near offset gap. Hence, predictive deconvolution in the x-t or -p domain is frequently used for attenuating shallow water multiples. However, besides multiples, deconvolution also attenuates primary events that have a periodicity which is close to that of the water-layer. In this paper, we present a workflow that involves first attenuating short-period water-layer related multiples (WLRMs) a process that we term shallow water demultiple (SWD); and then suppressing other long-period free surface multiples using conventional SRME. SWD is a wavefield-consistent method that first makes use of WLRMs in the data to reconstruct the missing water-bottom primary reflection and then uses the reflection for predicting shallow WLRMs. It is data driven and takes into account the spatial varying nature of subsurface structures. Since the WLRM model predicted by SWD has similar amplitude and phase as the input data, very short matching filters, which are not possible if deconvolution is used, can be utilised in the adaptive subtraction process. We demonstrate, through real-data examples, that our workflow provides an optimal multiple attenuation solution in shallow water environment in comparison with conventional methods such as -p deconvolution or SRME alone. Key words: Shallow water, multiple attenuation

Dow

nloa

ded

02/0

6/14

to 1

15.1

13.1

28.4

5. R

edist

ribut

ion

subje

ct to

SEG

licen

se or

copy

right;

see T

erms o

f Use

at htt

p://lib

rary.s

eg.or

g/

Shallow Water Demultiple Hung, Yang, Zhou and Xia


remaining long-period surface multiples generated from other subsurfaces. For the first step, since water-bottom reflections are not available and hence the iterative method is not possible for predicting water-layer multiples, an approach that is similar to the inversion method is used for estimating Fw - the multichannel prediction filter associated with the water-bottom. In this case, the design window for Fw should include either simple or peg-leg (or both) water-layer multiples. Deterministic information of bathymetry that is normally available from the navigation data can be used for designing gaps in the inversion process for obtaining Fw. In deriving Fw, the spatial noncausality property of the operator is taken into account in our implementation (Hung and Notfors, 2003). By convolving the resultant operator with the input data, the water-layer multiple model can then be generated. Since the estimation process of Fw has already included the effect of the surface operator, the multiple model has the correct amplitudes and phase for the water-layer multiples. Hence, in the subtraction process, direct subtraction or adaptive subtraction using very short matching filters is adequate. This minimises the risks of changing other events significantly, especially those primary events that are close to the multiples. This first step constitutes the process that we call shallow water demultiple (SWD). In the second step, the iterative approach (conventional SRME) is used for handling free-surface multiples that have longer period. Since the water-layer multiples have already been handled, the data corresponding to the multiple generator, i.e. the water-bottom in this case, needs to be first muted off to form the input for SRME. In practice, the length of the mute time is associated with the operator length of Fw. With this data preconditioning, relatively simple extrapolation methods can be utilised for SRME. Moreover, targeting long-period multiples allows more flexible control of adaptive subtraction parameters.

REAL DATA EXAMPLES Figure 1 shows an input stack of a line that was acquired offshore Australia. The water depth in this area is in the range of 50 to 90 m. Since the nearest recorded offset was 182 m, the recorded water-bottom reflection is in the post-critical range. Hence, as indicated by the arrow in Figure 1b, the water-bottom reflection is not well defined and it poses problems to conventional SRME. By deriving Fw from the data, the estimated water-bottom reflection at near offsets can then be reconstructed. Figure 1c displays the derived Fw at selected shot locations. With the estimated prediction operators, SWD was then applied to the data and the result is depicted in Figure 2. It can be observed that most of the water-layer multiples have been effectively attenuated. Its effectiveness can be compared with that of -p deconvolution which is used routinely in processing workflow for removing short-period multiples. Figure 3a displays a common offset (magnified) section of the result obtained by -p deconvolution. Comparing with the SWD result shown in Figure 3b, it can be seen that -p deconvolution is less effective in attenuating the water-layer multiples and has the tendency of damaging the primaries.

Using the result of SWD as an input to SRME, other free-surface multiples that are generated by deeper subsurfaces can be attenuated. Figure 4a depicts the result obtained by the combination of SWD and SRME and Figure 4b shows difference stack. To illustrate the importance of including SWD in the processing workflow for shallow water, conventional SRME alone was applied to the data and the corresponding results are displayed in Figure 5. A comparison between the two results clearly demonstrates that the lack of well-defined water-bottom reflection in the data degrades the performance of SRME. This can further be illustrated by the magnified sections in Figure 6. Another example is a survey acquired offshore Western China where the water depth is much shallower (25 to 30 m), resulting in an indistinct water-bottom reflection as shown from a stacked line in Figure 7. Consequently, conventional SRME has difficulties in suppressing the water-layer multiples, as illustrated in Figure 8. However, including the process of SWD in the workflow, it can be observed from Figure 9 and 10 that the free-surface multiples are significantly attenuated. Therefore, a combination of SWD and SRME provides an effective workflow for attenuating surface multiples in shallow water situations.

CONCLUSIONS By targeting the short-period and long-period multiples separately using multichannel prediction filter and conventional SRME respectively, we have demonstrated that the workflow helps to overcome the difficulty faced by SRME alone in attenuating free-surface multiples due to shallow water. The advantage that both methods have the same data arrangement such as interpolation/extrapolation and data ordering means that the workflow provides an effective and yet efficient way of shallow water demultiple.

ACKNOWLEDGMENTS The authors would like to thank Nexus Energy Ltd for permission to present the data and results and CGGVeritas for permission to publish this work.

REFERENCES Ali R., Verschuur D. J., Drummond J., Morris S. and Haughey G., 2002, Shallow water multiple prediction and attenuation, case study on data from the Arabian Gulf: 72nd Meeting, SEG, Expanded Abstracts, 2229-2232. Berkhout A. J. and Verschuur D. J., 1997, Estimation of multiple scattering by iterative inversion, Part I: Theoretical considerations: Geophysics, 62, 1586-1595. Biersteker J., 2001, MAGIC: Shells surface multiple attenuation technique: 71st Meeting, SEG, Expanded Abstracts, 1301-1304. Hargreaves N., 2006, Surface multiple attenuation in shallow water and the construction of primaries from multiples: 76th Meeting, SEG, Expanded Abstracts, 2689-2693.

Dow

nloa

ded

02/0

6/14

to 1

15.1

13.1

28.4

5. R

edist

ribut

ion

subje

ct to

SEG

licen

se or

copy

right;

see T

erms o

f Use

at htt

p://lib

rary.s

eg.or

g/



0 125m

Hung B. and Notfors C., 2003, Seismic Trace Interpolation Using Noncausal Spatial Filters in the F-X-Y Domain: 65th Meeting, EAGE, D021. Pica A., Poulain G., David B., Magesan M., Baldock S., Weisser T., Hugonnet P. and Herrmann P., 2005, 3D surface-related multiple modeling, principles and results: 75th Meeting, SEG, Expanded Abstracts, 2080-2083.

Schoenberger M. and Houston L. M., 1998, Stationarity transformation of multiples to improve the performance of predictive deconvolution: Geophysics, 63, 723-737. van Groenestijn G. J. A. and Verschuur D. J., 2009, Estimating primaries by sparse inversion and application to near-offset data reconstruction: Geophysics, 74, A23-A28. Verschuur D. J., 2006, Seismic multiple removal techniques past, present and future: EAGE Publications.

Figure 1. (a) Input stack. (b) A magnified section to illustrate the ill-defined water-bottom reflection indicated by the arrow. (c) Derived prediction operators at selected shot locations.

Figure 4. (a) Stack after SWD and SRME. (b) Difference stack.

(a) (c)

(b)

Figure 3. (a) Common offset section after -p deconvolution. (b) Common offset section after SWD.

(a)

(b)

Figure 2. Stack after SWD.

(a) (b)

Dow

nloa

ded

02/0

6/14

to 1

15.1

13.1

28.4

5. R

edist

ribut

ion

subje

ct to

SEG

licen

se or

copy

right;

see T

erms o

f Use

at htt

p://lib

rary.s

eg.or

g/



Figure 5. (a) Stack after SRME alone. (b) Difference stack.

Figure 6. Magnified sections of the area indicated by the green box in Fig. 5b. (a) Input. (b) Difference stack for

SWD+SRME. (c) Difference stack for SRME alone.

(b) (a)

(c) (b)(a)

Figure 7. Stack of input data for example 2. Figure 8. Stack of the result after SRME.

Figure 9. Stack of the result after SWD+SRME. Figure 10. Difference stacks for (a) SRME alone. (b) SWD+SRME.

(b)

(a)

Dow

nloa

ded

02/0

6/14

to 1

15.1

13.1

28.4

5. R

edist

ribut

ion

subje

ct to

SEG

licen

se or

copy

right;

see T

erms o

f Use

at htt

p://lib

rary.s

eg.or

g/

shallow water demultiple

Documents