Efficient full-waveform inversion with normalized plane-wave data

Taekhyun Kwon, Soon J. Seol, Joongmoo Byun

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

Full-waveform inversion (FWI) provides a high-resolution velocity model, but carries a high computational cost. Additionally, modern seismic acquisition, with dense sources and receivers, generates massive data, resulting in an even greater computational cost. To reduce the computational burden of FWI, we have developed an FWI algorithm using plane-wave data. Using this approach, plane-wave gathers transformed from shot gathers are used as input data in inversion. Because the number of plane-wave gathers is generally far smaller than that of common shot gathers for the same data set, we can significantly reduce the computational cost and efficiently handle a massive data set. 2-D numerical testing showed that the developed FWI algorithm was more efficient than conventional FWI using common shot gathers. However, estimation of the source wavelet is essential for successful FWI in most cases. The developed FWI algorithm here excludes the source effect by using the normalized wave field instead of estimating the source wavelet through inversion processing. When the FWI algorithm was applied to data generated by slightly different source wavelets, a good velocity model was reconstructed without any artefacts from the effects of the different source wavelets. Furthermore, it presented a stable inversion result with only small artefacts, even though we used random noise-added data. Finally, in a numerical experiment with data from the SEG/EAGE 3-D overthrust model, which includes complex structures and thin layers, the developed FWI algorithm yielded a reasonable reconstructed velocity model.

Original languageEnglish
Pages (from-to)53-60
Number of pages8
JournalGeophysical Journal International
Volume201
Issue number1
DOIs
StatePublished - 2015 Jan 1

Fingerprint

waveforms
plane waves
inversions
wavelet
Costs
shot
costs
artifact
inversion
artifacts
cost
Testing
Processing
wave field
random noise
Experiments
acquisition
estimating
receivers
high resolution

Keywords

  • Acoustic properties
  • Computational seismology
  • Inverse theory
  • Seismic tomography
  • Wave propagation
  • Wave scattering and diffraction

Cite this

@article{010ad15df7d74043bf562439b6dfb7c8,
title = "Efficient full-waveform inversion with normalized plane-wave data",
abstract = "Full-waveform inversion (FWI) provides a high-resolution velocity model, but carries a high computational cost. Additionally, modern seismic acquisition, with dense sources and receivers, generates massive data, resulting in an even greater computational cost. To reduce the computational burden of FWI, we have developed an FWI algorithm using plane-wave data. Using this approach, plane-wave gathers transformed from shot gathers are used as input data in inversion. Because the number of plane-wave gathers is generally far smaller than that of common shot gathers for the same data set, we can significantly reduce the computational cost and efficiently handle a massive data set. 2-D numerical testing showed that the developed FWI algorithm was more efficient than conventional FWI using common shot gathers. However, estimation of the source wavelet is essential for successful FWI in most cases. The developed FWI algorithm here excludes the source effect by using the normalized wave field instead of estimating the source wavelet through inversion processing. When the FWI algorithm was applied to data generated by slightly different source wavelets, a good velocity model was reconstructed without any artefacts from the effects of the different source wavelets. Furthermore, it presented a stable inversion result with only small artefacts, even though we used random noise-added data. Finally, in a numerical experiment with data from the SEG/EAGE 3-D overthrust model, which includes complex structures and thin layers, the developed FWI algorithm yielded a reasonable reconstructed velocity model.",
keywords = "Acoustic properties, Computational seismology, Inverse theory, Seismic tomography, Wave propagation, Wave scattering and diffraction",
author = "Taekhyun Kwon and Seol, {Soon J.} and Joongmoo Byun",
year = "2015",
month = "1",
day = "1",
doi = "10.1093/gji/ggu498",
language = "English",
volume = "201",
pages = "53--60",
journal = "Geophysical Journal International",
issn = "0956-540X",
number = "1",

}

Efficient full-waveform inversion with normalized plane-wave data. / Kwon, Taekhyun; Seol, Soon J.; Byun, Joongmoo.

In: Geophysical Journal International, Vol. 201, No. 1, 01.01.2015, p. 53-60.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Efficient full-waveform inversion with normalized plane-wave data

AU - Kwon, Taekhyun

AU - Seol, Soon J.

AU - Byun, Joongmoo

PY - 2015/1/1

Y1 - 2015/1/1

N2 - Full-waveform inversion (FWI) provides a high-resolution velocity model, but carries a high computational cost. Additionally, modern seismic acquisition, with dense sources and receivers, generates massive data, resulting in an even greater computational cost. To reduce the computational burden of FWI, we have developed an FWI algorithm using plane-wave data. Using this approach, plane-wave gathers transformed from shot gathers are used as input data in inversion. Because the number of plane-wave gathers is generally far smaller than that of common shot gathers for the same data set, we can significantly reduce the computational cost and efficiently handle a massive data set. 2-D numerical testing showed that the developed FWI algorithm was more efficient than conventional FWI using common shot gathers. However, estimation of the source wavelet is essential for successful FWI in most cases. The developed FWI algorithm here excludes the source effect by using the normalized wave field instead of estimating the source wavelet through inversion processing. When the FWI algorithm was applied to data generated by slightly different source wavelets, a good velocity model was reconstructed without any artefacts from the effects of the different source wavelets. Furthermore, it presented a stable inversion result with only small artefacts, even though we used random noise-added data. Finally, in a numerical experiment with data from the SEG/EAGE 3-D overthrust model, which includes complex structures and thin layers, the developed FWI algorithm yielded a reasonable reconstructed velocity model.

AB - Full-waveform inversion (FWI) provides a high-resolution velocity model, but carries a high computational cost. Additionally, modern seismic acquisition, with dense sources and receivers, generates massive data, resulting in an even greater computational cost. To reduce the computational burden of FWI, we have developed an FWI algorithm using plane-wave data. Using this approach, plane-wave gathers transformed from shot gathers are used as input data in inversion. Because the number of plane-wave gathers is generally far smaller than that of common shot gathers for the same data set, we can significantly reduce the computational cost and efficiently handle a massive data set. 2-D numerical testing showed that the developed FWI algorithm was more efficient than conventional FWI using common shot gathers. However, estimation of the source wavelet is essential for successful FWI in most cases. The developed FWI algorithm here excludes the source effect by using the normalized wave field instead of estimating the source wavelet through inversion processing. When the FWI algorithm was applied to data generated by slightly different source wavelets, a good velocity model was reconstructed without any artefacts from the effects of the different source wavelets. Furthermore, it presented a stable inversion result with only small artefacts, even though we used random noise-added data. Finally, in a numerical experiment with data from the SEG/EAGE 3-D overthrust model, which includes complex structures and thin layers, the developed FWI algorithm yielded a reasonable reconstructed velocity model.

KW - Acoustic properties

KW - Computational seismology

KW - Inverse theory

KW - Seismic tomography

KW - Wave propagation

KW - Wave scattering and diffraction

UR - http://www.scopus.com/inward/record.url?scp=84925424228&partnerID=8YFLogxK

U2 - 10.1093/gji/ggu498

DO - 10.1093/gji/ggu498

M3 - Article

AN - SCOPUS:84925424228

VL - 201

SP - 53

EP - 60

JO - Geophysical Journal International

JF - Geophysical Journal International

SN - 0956-540X

IS - 1

ER -