1. WAVEFIELD PREDICTION OF WATER-LAYER-MULTIPLES
Ruiqing He
University of Utah
Feb. 2004

2. Outline Theory Synthetic experiments Application to Unocal data
Introduction
Theory
Synthetic experiments
Application to Unocal data
Conclusion

3. Introduction accurate wavefield prediction of multiples. Other works:
Primary-preserving multiple removal demands
accurate wavefield prediction of multiples.
Other works:
- Delft
- Amundsen, Ikelle, Weglein, etc.
Multiple prediction is the prerequisite for multiple subtraction. If the multiple prediction is more accurate, then the subtraction is more effective.
Water layer multiples

4. Outline Theory Synthetic experiments Application to Unocal data
Introduction
Theory
Synthetic experiments
Application to Unocal data
Conclusion

5. Berryhill and Wiggins’s Methods
Kirchhoff Summation
Forward extrapolate traces
down to water bottom
Off-shore
seismic data
Kirchhoff Summation
Forward extrapolate bottom
traces up to receivers
 Emulated Muliples
Multiple
attenuation
Filtered Subtraction
The problem lies in the receivers recorder neither up-coming waves nor down-going waves’ pressure, the receivers recorder their superposition waves’ pressure.
Water surface
Receiver line
Water bottom

6. The proposed method FD: Finite Difference
Decomposition of
receiver-side
ghosts
Wave forward
extrapolation
to the water bottom
Off-shore
seismic data FD FD FD
Multiples
with last
round-trip
in water layer
Primary-
preserving
multiple
removal
Wave forward
extrapolation
to the receivers DS
filtering
FD: Finite Difference
DS: Direct (simple) Subtraction
Other multiple
attenuation

7. Why Finite Difference? - speed - convenience
Advantage
- speed
- convenience
- capability: heterogeneous medium
Because Kirchhoff method is based on homogenous assumption, it has difficulty to handle heterogeneous media.
Disadvantage
- dispersion?: reality, high-order FD

8. Types of Water-Layer-Multiples
LWLM: Multiples that have the last round-trip in the water layer.
Other WLM: other water-layer-multiples except LWLM.
In very strict situations, any other WLM can have an identical LWLM that has the same traveltime and waveform in seismic traces. In most cases, LWLM and other WLM are similar (e.g. 102 and 201 in the previous talk).

9. Wavefield Extrapolation of RSG
Mirror image of
the Receiver line
Water surface
Receiver line U
RSG

10. Decomposition of RSG RSG + DATA f U Mirror image of the Receiver line
Water surface
Here f is an inexplicit function that can be implemented by many means. f
Receiver line U

11. Outline Theory Synthetic experiments Application to Unocal data
Introduction
Theory
Synthetic experiments
Application to Unocal data
Conclusion

12. Synthetic Model water BSR Depth (m) Salt dome Sandstone 1500 3250
water
BSR
Depth
(m)
BSR: Boundary Simulating Reflectors.
Salt dome
Sandstone
1500
Offset (m)
3250

13. Synthetic seismic data
400
Time
(ms)
The water-bottom multiples are so strong that they blur the reflection from the salt dome.
2500
Offset (m)
3250

14. Decomposed RSG
400
Time
(ms)
2500
Offset (m)
3250

15. Predicted LWLM
400
Time
(ms)
2500
Offset (m)
3250

16. Waveform Comparison between Data & RSG
Amplitude
2400
600
Time (ms)

17. Waveform Comparison between Data & LWLM
Amplitude
2400
600
Time (ms)

18. Waveform Comparison between Data & RSG+LWLM
Amplitude
2400
600
Time (ms)

19. Elimination of RSG & LWLM
400
Time
(ms)
This result is primary-preserving, but not all water-layer related peglegs are eliminated. As mentioned before, other WLM are not predicted, and remain there, but there are similar to the predicted LWLM, so they can be attenuated by additional subtraction algorithms.
2500
Offset (m)
3250

20. Further Multiple Attenuation
400
Time
(ms)
2500
Offset (m)
3250

21. Outline Theory Synthetic experiments Application to Unocal data
Introduction
Theory
Synthetic experiments
Application to Unocal data
Conclusion

22. Unocal field data
600
Time
(ms)
2400
Offset (m)
3175

23. Inadequate RSG Decomposition
600
Time
(ms)
Because the real seismic waves are propagation in 3D, one single receiver line is not enough for exact RSG decomposition. I leave this challenging topic for future research.
2400
Offset (m)
3175

24. Emulated LWLM 600 Time (ms) 2400 3175 Offset (m)
Emulate LWLM as Berryhill and Wiggins did, but with FD rather than Kirchhoff methods.
2400
Offset (m)
3175

25. Waveform comparison between Data & Emulated LWLM
Amplitude
1400
Time (ms)
2400

26. Attenuation of WLM
600
Time
(ms)
2400
Offset (m)
3175

27. Attenuation of WLM

28. Attenuation of WLM
600
Time
(ms)
2400
Offset (m)
3175

29. Attenuation of WLM

30. Subtracted WLM
600
Time
(ms)
2400
Offset (m)
3175

31. Outline Theory Synthetic experiments Application to Unocal data
Introduction
Theory
Synthetic experiments
Application to Unocal data
Conclusion

32. Conclusion Theoretically revives Berryhill and Wiggins
methods for primary-preserving removal of one
kind of water-layer-multiples.
Requirements are practically obtainable,
and can be derived from seismic data.
Applicable to field data with approximations.
The differences between this research and Berryhill & Wiggins’s methods are:
1: It uses FD rather than Kirchhoff methods.
2: It decomposes Receiver-Side-Ghosts in synthetic data.
3: It makes distinctions among water-layer related multiples.
4. It results in primary-preserving multiple removal.
Overcomes the Delft method by alleviating
acquisition requirements and the need to know
the source signature.

33. Future Work Ghost decomposition for field data.
3D to 2D seismic data conversion.
Multiple subtraction.

34. Reference Berryhill J.R. and Kim Y.C., 1986, Deep-water pegleg and
multiples: emulation and suppression: Geophysics Vol. 51,
Wang Y., 1998, Comparison of multiple attenuation methods
with least-squares migration filtering: UTAM 1998 annual
report,
Wiggins J.W., 1988, Attenuation of complex water-multiples
by wave-equation-based prediction and subtraction:
Geophysics Vol.53 No.12,

35. Thanks
2003 members of UTAM for financial support.