Wavefield Prediction of Waterlayer Multiples

1
Wavefield Prediction of Water-layer Multiples
Ruiqing He University of Utah Oct.
2004
2
Outline

• Introduction
• Theory
• Synthetic experiments
• Application to real data
• Conclusion

3
Introduction

• Multiple classification.
• Free-surface multiples (FSM).
• - Delft, multiple series theories, etc.
• Water-layer multiples (WLM).
• - Berryhill, Wiggins, et al.

4
Berryhills Approach

• The prediction of WLM is obtained by propagating
  the received data once within the water layer.
• - Kirchhoff integral, Finite-Difference,
• Gaussian beams, Phase-shift, etc.
• The prediction is emulation.
• - Part of WLM.
• - Half is exact the other half is not exact.
• Multiple subtraction.

5
Outline

• Introduction
• Theory
• Synthetic experiments
• Application to real data
• Conclusion

6
Seismic Wave Representation
gS Ghost-source. s
Twin-source. f visit of subsurface once. g
Receiver-side ghosting.
7
Berryhills Emulation
8
FSM Prediction
Subscript g Receiver-side ghosts
(RSG). Subscript u Upcoming data that generate
RSG.
9
Multiple Classification

• Level 1
• Water-Layer Multiple (WLM).
• Non-WLM multiples (NWLM).
• Level 2 (WLM)
• Last reverberation WLM (LWLM).
• First reverberation WLM (FWLM).
• Middle reverberation WLM (MWLM).
• Definition priority.
• Water-Bottom-Multiple (WBM).

10
Types of Water-Layer Multiples
FWLM
MWLM
LWLM
Water surface
Water bottom
Subsurface reflector
11
Seismic Data Classification
Note Converted waves are not considered,
and direct waves have been removed.
12
LWLM Prediction
Data (W)

Upcoming waves (U)
f
Downgoing ghosts (D)
g
LWLM
-
For synthetic data, the operator g, f can be
exactly known. By this design, LWLM can be
exactly predicted.
13
Outline

• Introduction
• Theory
• Synthetic experiments
• Application to real data
• Conclusion

14
Synthetic Model
0
water
Hydrate
Depth (m)
Salt dome
Sandstone
1500
0
3250
Offset (m)
15
Synthetic Data
400
Time (ms)
2500
0
3250
Offset (m)
16
Predicted LWLM
400
Time (ms)
2500
0
3250
Offset (m)
17
Waveform Comparisonbetween Data RSGLWLM
Data RSG LWLM
Amplitude
2400
Time (ms)
600
18
Elimination of RSG LWLMby Direct Subtraction
400
Time (ms)
2500
0
3250
Offset (m)
19
Further Multiple Attenuationby Deconvolutions
400
Time (ms)
2500
0
3250
Offset (m)
20
Outline

• Introduction
• Theory
• Synthetic experiments
• Application to real data
• Conclusion

21
A Mobil data
22
Predicted LWLM
23
Waveform Comparison
24
WLM Attenuationwith Multi-Channel Deconvolution
25
Migration before demultiple
Migration after demultiple
26
A Unocal Data
27
Predicted LWLM
28
Waveform Comparison
At a geophone above non-flat water bottom
At a geophone above flat water bottom
29
WLM Attenuationwith Multi-channel Deconvolution
30
Migration before demultiple
Migration after demultiple
31
Outline

• Introduction
• Theory
• Synthetic experiments
• Application to real data
• Conclusion

32
Conclusion

• Berryhills approach does not need to know the
  source signature, and can be performed in a
  single shot gather, but the prediction is
  emulation.
• This method improves Berryhills approach by
  making clear classification among WLM, and using
  receiver-side ghosts to predict LWLM.
• This method exactly eliminates LWLM for
  synthetic data, and successfully suppresses WLM
  by multi-channel de-convolutions for field data .

33
Thanks

• This research is benefited from the discussions
  with Dr. Yue Wang and Dr. Tamas Nemeth of
  ChevronTexaco Co..
• I am also thankful to 2004 members of UTAM for
  financial support.