A Robust Gradient for Long Wavelength FWI Updates

Ramos-Martinez, J.; Crawley, Sean; Zou, Z.; Valenciano, A.; Qiu, Lirong; Chemingui, Nizar

doi:10.3997/2214-4609.201601536

Cited by 24 publications

(6 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For this reason, the forward scattering is cancelled out, but the backward scattering is enhanced. This cancellation can occur only in the back scattering area if we change the parameterization from velocity type to the impedance type (Operto et al, 2013;Ramos-Martinez et al, 2016).…”

Section: Cp-i: Eight Seismic Velocities Plus One Thomsen's Parametermentioning

confidence: 99%

Elastic orthorhombic anisotropic parameter inversion: An analysis of parameterization

Alkhalifah

2016

GEOPHYSICS

View full text Add to dashboard Cite

CitationOh ABSTRACTThe resolution of a multiparameter full-waveform inversion (FWI) is highly influenced by the parameterization used in the inversion algorithm, as well as the data quality and the sensitivity of the data to the elastic parameters because the scattering patterns of the partial derivative wavefields (PDWs) vary with parameterization. For this reason, it is important to identify an optimal parameterization for elastic orthorhombic FWI by analyzing the radiation patterns of the PDWs for many reasonable model parameterizations. We have promoted a parameterization that allows for the separation of the anisotropic properties in the radiation patterns. The central parameter of this parameterization is the horizontal P-wave velocity, with an isotropic scattering potential, influencing the data at all scales and directions. This parameterization decouples the influence of the scattering potential given by the P-wave velocity perturbation from the polar changes described by two dimensionless parameter perturbations and from the azimuthal variation given by three additional dimensionless parameters perturbations. In addition, the scattering potentials of the P-wave velocity perturbation are also decoupled from the elastic influences given by one S-wave velocity and two additional dimensionless parameter perturbations. The vertical S-wave velocity is chosen with the best resolution obtained from S-wave reflections and converted waves, and little influence on P-waves in conventional surface seismic acquisition. The influence of the density on observed data can be absorbed by one anisotropic parameter that has a similar radiation pattern. The additional seven dimensionless parameters describe the polar and azimuth variations in the P-and S-waves that we may acquire, with some of the parameters having distinct influences on the recorded data on the earth's surface. These characteristics of the new parameterization offer the potential for a multistage inversion from high symmetry anisotropy to lower symmetry ones.

show abstract

Section: Cp-i: Eight Seismic Velocities Plus One Thomsen's Parametermentioning

confidence: 99%

Elastic orthorhombic anisotropic parameter inversion: An analysis of parameterization

Alkhalifah

2016

GEOPHYSICS

View full text Add to dashboard Cite

show abstract

“…To produce long-wavelength velocity updates, we adapted the equations for the velocity gradient (Ramos-Martínez et al, 2016) to work with the W2 misfit function. The dynamic weight implementation of the velocity kernel was translated to the equivalent expressions such that the first-order time derivatives of the source and residual wavefields are computed before the adjoint source backward propagation.…”

Section: Theorymentioning

confidence: 99%

“…However, the separation in the data space can be challenging. This, and the fact that for deep targets only reflections are available for the inversion, has motivated the development of FWI gradients that isolate the different wavenumber ranges in the velocity updates (e.g., Xu et al, 2012;Zhou et al, 2015;Ramos-Martínez et al, 2016).…”

Section: Introductionmentioning

confidence: 99%

“…Here, we have combined the quadratic form of the Wasserstein distance (W2 norm) to measure the data misfit with a robust implementation of the velocity gradient. Our numerical implementation introduces dynamic weights (Ramos-Martínez et al, 2016) in the velocity sensitivity kernel derived from the impedance and velocity parameterization of the objective function. It effectively separates the migration isochrones produced by the specular reflectivity from the components produced by transmitted arrivals.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Long-wavelength FWI updates in the presence of cycle skipping

Ramos-Martinez¹,

Qiu²,

Valenciano³

et al. 2019

The Leading Edge

View full text Add to dashboard Cite

Full-waveform inversion (FWI) has become the tool of choice for building high-resolution velocity models. Its success depends on producing seamless updates of the short- and long-wavelength model features while avoiding cycle skipping. Classic FWI implementations use the L2 norm to measure the data misfit in combination with a gradient computed by a crosscorrelation imaging condition of the source and residual wavefields. The algorithm risks converging to an inaccurate result if the data lack low frequencies and/or the initial model is far from the true one. Additionally, the model updates may display a reflectivity imprint before the long-wavelength features of the model are fully recovered. We propose a new solution to this fundamental challenge by combining the quadratic form of the Wasserstein distance (W2 norm) for measuring the data misfit with a robust implementation of a velocity gradient. The W2 norm reduces the risk of cycle skipping, whereas the velocity gradient effectively eliminates the reflectivity imprint and emphasizes the long-wavelength model updates. We illustrate the performance of the new solution on a field survey acquired offshore Brazil. We demonstrate how FWI successfully updates the earth model and resolves a high-velocity carbonate section missing from the initial velocity model.

show abstract

“…We assume that the δ field is obtained from other types of data (e.g., well logs, VSP) and we invert for v z and ε. The resulting v z sensitivity kernel for this parameterization is similar in form to one derived for inverting isotropic acoustic velocity (Ramos-Martinez et al, 2016). This velocity kernel targets longwavelength model updates (diving waves and "rabbit ears") because it suppresses the specular reflectivity in the gradient (migration isochrones).…”

Section: Introductionmentioning

confidence: 99%

An effective multiparameter full-waveform inversion in acoustic anisotropic media

Ramos-Martinez

Qiu

Valenciano

et al. 2017

SEG Technical Program Expanded Abstracts 2017

View full text Add to dashboard Cite

Multi-parameter Full Waveform Inversion (FWI) suffers from leakage among different medium properties. Here we discuss a practical multi-parameter FWI solution that minimizes leakage and retrieves the long-wavelength features of the anisotropic earth model. Our algorithm is based on a regularization of the objective function and a specific parameterization of an acoustic Transversely Isotropic (TI) medium consisting of the vertical velocity, and Thomsen parameters epsilon and delta. We show, by using synthetic data, that Total Variation (TV) regularization significantly reduces leakage of the vertical velocity in the epsilon model during the inversion. Finally, we show an application on field data from the Gulf of Mexico where the flatness of the common image gathers is significantly improved.

show abstract

A Robust Gradient for Long Wavelength FWI Updates

Cited by 24 publications

References 9 publications

Elastic orthorhombic anisotropic parameter inversion: An analysis of parameterization

Elastic orthorhombic anisotropic parameter inversion: An analysis of parameterization

Long-wavelength FWI updates in the presence of cycle skipping

An effective multiparameter full-waveform inversion in acoustic anisotropic media

Contact Info

Product

Resources

About