Haneet Wason, Felix Oghenekohwo, and Felix J. Herrmann … · 2019. 9. 6. · Haneet Wason, Felix...

1
Haneet Wason, Felix Oghenekohwo, and Felix J. Herrmann Compressed sensing in 4-D marine – recovery of dense time-lapse data from subsampled data without repetition We P3 12 Conclusions Acknowledgements We would like to thank all the sponsors of our SINBAD project for their support. References Abma, R., Ford, A., Rose-Innes, N., Mannaerts-Drew, H. and Kommedal, J. [2013] Continued development of simultaneous source acquisition for ocean bottom surveys. 75th EAGE Conference and Exhibition, doi: 10.3997/2214-4609.20130081. Baron, D., Duarte, M.F., Wakin, M.B., Sarvotham, S. and Baraniuk, R.G. [2009] Distributed compressive sensing. CoRR, abs/0901.3403. Beasley, C.J. [2008] A new look at marine simultaneous sources. The Leading Edge , 27(7), 914–917, doi: 10.1190/1.2954033. Berkhout, A.J. [2008] Changing the mindset in seismic data acquisition. The Leading Edge , 27(7), 924–938, doi: 10.1190/1.2954035. Donoho, D.L. [2006] Compressed sensing. Information Theory, IEEE Transactions on, 52(4), 1289–1306. Hampson, G., Stefani, J. and Herkenhoff, F. [2008] Acquisition using simultaneous sources. The Leading Edge, 27(7), 918–923, doi:10.1190/1.2954034. Hennenfent, G., Fenelon, L. and Herrmann, F.J. [2010] Nonequispaced curvelet transform for seismic data reconstruction: a sparsity-promoting approach. Geophysics, 75(6), WB203–WB210. Lumley, D. and Behrens, R. [1998] Practical issues of 4d seismic reservoir monitoring: What an engineer needs to know. SPE Reservoir Evaluation & Engineering, 1(6), 528–538. Oghenekohwo, F., Esser, E. and Herrmann, F.J. [2014] Time- lapse seismic without repetition: reaping the benefits from randomized sampling and joint recovery. 76th EAGE Conference and Exhibition. Wason, H. and Herrmann, F.J. [2013] Time-jittered ocean bottom seismic acquisition. SEG Technical Program Expanded Abstracts, 1–6, doi:10.1190/segam2013-1391.1. Processing time-lapse data jointly leads to improved recovery of both vintages (baseline and monitor) and time-lapse signal Given the context of randomized subsampling, calibration errors ! deteriorate recovery of the time-lapse signal ! improve recovery of the vintages Since calibration errors are inevitable in the real world, the insistence on repetition in time-lapse surveys can be relaxed Possible to recover dense time-lapse data from randomized subsampling and joint processing Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 no error [12.2 dB] error 1.0 m [8.5 dB] 4-D recovery (Joint recovery method) error 2.8 m [3.8 dB] 0% overlap [2.0 dB] 50% overlap in acquisition matrices As the calibration error increases, the 4-D signal recovery decreases significantly (b,c). Moreover, the recovery becomes comparable to the recovery with 0% overlap between the surveys, i.e., randomized acquisition without repetition (d). Monitor recovery (Joint recovery method) Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Residual no error [13.9 dB] error 1.0 m [14.5 dB] error 2.8 m [15.5 dB] 0% overlap [18.3 dB] (a), (b), (c), (e), (f), (g) 50% overlap in acquisition matrices (a) (b) (c) (d) (e) (f) (g) (h) 0% overlap Independent vs. Joint recovery method Randomized sampling in (4-D) marine increasing error increasing SNR Time-lapse seismic Current acquisition paradigm: ! repeat expensive dense acquisitions & independent processing ! compute differences between baseline & monitor survey(s) ! hampered by practical challenges to ensure repetition Compressive sampling paradigm: ! cheap subsampled acquisition, e.g., via time-jittered marine subsampling ! may offer possibility to relax insistence on repeatability ! exploits insights from distributed compressed sensing Distributed compressed sensing ! joint recovery model (JRM) ! different vintages share common information A z }| { A 1 A 1 0 A 2 0 A 2 z z }| { 2 4 z 0 z 1 z 2 3 5 = b z }| { b 1 b 2 x 2 = z 0 + z 2 x 1 = z 0 + z 1 baseline monitor differences vintages common component no repetition => 0% overlap in baseline and monitor acquisitions (shot positions) Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 Source position (m) Time (s) 0 250 500 750 1000 0.5 1 1.5 2 independent [11.6 dB]* 4-D signal Monitor recovery and residual independent residual joint residual joint [13.9 dB] joint [12.2 dB] independent [-10.6 dB] Processing data jointly (by exploiting common features of surveys) leads to improved recovery of both vintages and time-lapse signal. Note: all results for 50% overlap in acquisition matrices with no calibration errors. * [SNR dB] : signal-to-noise ratio Aim: recovery of both vintages & time-lapse signal from incomplete data Source position (m) 200 400 600 800 1000 1200 Recording time (s) 50 100 150 200 250 300 350 400 450 Baseline Monitor Monitor w/ error error 1.0 m error 1.4 m * Baseline o Monitor x Monitor with error in repeated shot positions Time-jittered marine acquisition (for 50% overlap in baseline & monitor acquisitions) (a) (a) (b) (c) (d) (e) (f) (c) (d) (b) subsampled and blended data (50 seconds of total recording time) Receiver position (m) Recording time (s) 0 500 1000 70 80 90 100 110 120 Released to public domain under Creative Commons license type BY (https://creativecommons.org/licenses/by/4.0). Copyright (c) 2015 SLIM group @ The University of British Columbia.

Transcript of Haneet Wason, Felix Oghenekohwo, and Felix J. Herrmann … · 2019. 9. 6. · Haneet Wason, Felix...

Page 1: Haneet Wason, Felix Oghenekohwo, and Felix J. Herrmann … · 2019. 9. 6. · Haneet Wason, Felix Oghenekohwo, and Felix J. Herrmann Compressed sensing in 4-D marine – recovery

Haneet Wason, Felix Oghenekohwo, and Felix J. Herrmann Compressed sensing in 4-D marine – recovery of dense

time-lapse data from subsampled data without repetition We P3 12

Conclusions

Acknowledgements

We would like to thank all the sponsors of our SINBAD project for their support.

References

Abma, R., Ford, A., Rose-Innes, N., Mannaerts-Drew, H. and Kommedal, J. [2013] Continued development of simultaneous source acquisition for ocean bottom surveys. 75th EAGE Conference and Exhibition, doi: 10.3997/2214-4609.20130081.

Baron, D., Duarte, M.F., Wakin, M.B., Sarvotham, S. and Baraniuk, R.G. [2009] Distributed compressive sensing. CoRR, abs/0901.3403.

Beasley, C.J. [2008] A new look at marine simultaneous sources. The Leading Edge , 27(7), 914–917, doi: 10.1190/1.2954033.

Berkhout, A.J. [2008] Changing the mindset in seismic data acquisition. The Leading Edge, 27(7), 924–938, doi:10.1190/1.2954035.

Donoho, D.L. [2006] Compressed sensing. Information Theory, IEEE Transactions on, 52(4), 1289–1306.

Hampson, G., Stefani, J. and Herkenhoff, F. [2008] Acquisition using simultaneous sources. The Leading Edge, 27(7), 918–923, doi:10.1190/1.2954034.

Hennenfent, G., Fenelon, L. and Herrmann, F.J. [2010] Nonequispaced curvelet transform for seismic data reconstruction: a sparsity-promoting approach. Geophysics, 75(6), WB203–WB210.

Lumley, D. and Behrens, R. [1998] Practical issues of 4d seismic reservoir monitoring: What an engineer needs to know. SPE Reservoir Evaluation & Engineering, 1(6), 528–538.

Oghenekohwo, F., Esser, E. and Herrmann, F.J. [2014] Time-lapse seismic without repetition: reaping the benefits from randomized sampling and joint recovery. 76th EAGE Conference and Exhibition.

Wason, H. and Herrmann, F.J. [2013] Time-jittered ocean bottom seismic acquisition. SEG Technical Program Expanded Abstracts, 1–6, doi:10.1190/segam2013-1391.1.

Processing time-lapse data jointly leads to improved recovery of both vintages (baseline and monitor) and time-lapse signal Given the context of randomized subsampling, calibration errors !  deteriorate recovery of the time-lapse signal !  improve recovery of the vintages

Since calibration errors are inevitable in the real world, the insistence on repetition in time-lapse surveys can be relaxed Possible to recover dense time-lapse data from randomized subsampling and joint processing

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

no error [12.2 dB]

error ≈ 1.0 m [8.5 dB]

4-D recovery (Joint recovery method) error ≈ 2.8 m

[3.8 dB] 0% overlap

[2.0 dB]

50% overlap in acquisition matrices

As the calibration error increases, the 4-D signal recovery decreases significantly (b,c). Moreover, the recovery becomes comparable to the recovery with 0% overlap between the surveys, i.e., randomized acquisition without repetition (d).

Monitor recovery (Joint recovery method)

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Residual

no error [13.9 dB]

error ≈ 1.0 m [14.5 dB]

error ≈ 2.8 m [15.5 dB]

0% overlap [18.3 dB]

(a), (b), (c), (e), (f), (g) 50% overlap in acquisition matrices

(a) (b) (c) (d) (e) (f) (g) (h)

0% overlap

Independent vs. Joint recovery method

Randomized sampling in (4-D) marine increasing error increasing SNR

Time-lapse seismic Current acquisition paradigm:

!  repeat expensive dense acquisitions & independent processing

!  compute differences between baseline & monitor survey(s) !  hampered by practical challenges to ensure repetition Compressive sampling paradigm:

!  cheap subsampled acquisition, e.g., via time-jittered marine subsampling

!  may offer possibility to relax insistence on repeatability !  exploits insights from distributed compressed sensing

Distributed compressed sensing

!  joint recovery model (JRM) !  different vintages share common information

Az }| {A1 A1 0A2 0 A2

�zz }| {2

4z0z1z2

3

5=

bz }| {b1

b2

x2 = z0 + z2

x1 = z0 + z1

baseline

monitor

differences vintages

common component

no repetition => 0% overlap in baseline and monitor acquisitions (shot positions)

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

Source position (m)

Tim

e (s

)

0 250 500 750 1000

0.5

1

1.5

2

independent [11.6 dB]*

4-D signal Monitor recovery and residual

independent residual

joint residual

joint [13.9 dB]

joint [12.2 dB]

independent [-10.6 dB] Processing data jointly (by

exploiting common features of surveys) leads to improved recovery of both vintages and time-lapse signal. Note: all results for 50% overlap in acquisition matrices with no calibration errors. * [SNR dB] : signal-to-noise ratio

Aim: recovery of both vintages & time-lapse signal from incomplete data

Source position (m)200 400 600 800 1000 1200

Reco

rding

time

(s)

50

100

150

200

250

300

350

400

450

BaselineMonitorMonitor w/ error

error ≈ 1.0 m

error ≈ 1.4 m

* Baseline o Monitor x Monitor with error in repeated shot positions

Time-jittered marine acquisition

(for 50% overlap in baseline & monitor acquisitions)

(a)

(a) (b) (c) (d) (e) (f)

(c) (d) (b)

subsampled and blended data (50 seconds of total recording time)

Receiver position (m)

Rec

ordi

ng ti

me

(s)

0 500 1000

70

80

90

100

110

120

Released to public domain under Creative Commons license type BY (https://creativecommons.org/licenses/by/4.0).Copyright (c) 2015 SLIM group @ The University of British Columbia.