TY - JOUR
T1 - Conditioning the full-waveform inversion gradient to welcome anisotropy
AU - Alkhalifah, Tariq Ali
N1 - KAUST Repository Item: Exported on 2020-10-01
PY - 2015/4/23
Y1 - 2015/4/23
N2 - Multiparameter full-waveform inversion (FWI) suffers from complex nonlinearity in the objective function, compounded by the eventual trade-off between the model parameters. A hierarchical approach based on frequency and arrival time data decimation to maneuver the complex nonlinearity associated with this problem usually falls short in anisotropic media. In place of data decimation, I use a model gradient filter approach to access the parts of the gradient more suitable to combat the potential nonlinearity and parameter trade-off. The filter is based on representing the gradient in the time-lag normalized domain, in which small scattering-angles of the gradient update are initially muted out. The model update hierarchical filtering strategy include applying varying degrees of filtering to the different anisotropic parameter updates, a feature not easily accessible to simple data decimation. Using FWI and reflection-based FWI, when the modeled data are obtained with the single-scattering theory, allows access to additional low model wavenumber components. Combining such access to wavenumbers with scattering-angle filters applied to the individual parameter gradients allows for multiple strategies to avoid complex FWI nonlinearity as well as the parameter trade-off.
AB - Multiparameter full-waveform inversion (FWI) suffers from complex nonlinearity in the objective function, compounded by the eventual trade-off between the model parameters. A hierarchical approach based on frequency and arrival time data decimation to maneuver the complex nonlinearity associated with this problem usually falls short in anisotropic media. In place of data decimation, I use a model gradient filter approach to access the parts of the gradient more suitable to combat the potential nonlinearity and parameter trade-off. The filter is based on representing the gradient in the time-lag normalized domain, in which small scattering-angles of the gradient update are initially muted out. The model update hierarchical filtering strategy include applying varying degrees of filtering to the different anisotropic parameter updates, a feature not easily accessible to simple data decimation. Using FWI and reflection-based FWI, when the modeled data are obtained with the single-scattering theory, allows access to additional low model wavenumber components. Combining such access to wavenumbers with scattering-angle filters applied to the individual parameter gradients allows for multiple strategies to avoid complex FWI nonlinearity as well as the parameter trade-off.
UR - http://hdl.handle.net/10754/552157
UR - http://library.seg.org/doi/abs/10.1190/geo2014-0390.1
UR - http://www.scopus.com/inward/record.url?scp=84928235748&partnerID=8YFLogxK
U2 - 10.1190/geo2014-0390.1
DO - 10.1190/geo2014-0390.1
M3 - Article
SN - 0016-8033
VL - 80
SP - R111-R122
JO - GEOPHYSICS
JF - GEOPHYSICS
IS - 3
ER -