Efficient scattering-angle enrichment for a nonlinear inversion of the background and perturbations components of a velocity model

Research output: Contribution to journalArticlepeer-review

30 Scopus citations

Abstract

Reflection-waveform inversion (RWI) can help us reduce the nonlinearity of the standard full-waveform inversion (FWI) by inverting for the background velocity model using the wave-path of a single scattered wavefield to an image. However, current RWI implementations usually neglect the multi-scattered energy, which will cause some artifacts in the image and the update of the background. To improve existing RWI implementations in taking multi-scattered energy into consideration, we split the velocity model into background and perturbation components, integrate them directly in the wave equation, and formulate a new optimization problem for both components. In this case, the perturbed model is no longer a single-scattering model, but includes all scattering. Through introducing a new cheap implementation of scattering angle enrichment, the separation of the background and perturbation components can be implemented efficiently. We optimize both components simultaneously to produce updates to the velocity model that is nonlinear with respect to both the background and the perturbation. The newly introduced perturbation model can absorb the non-smooth update of the background in a more consistent way. We apply the proposed approach on the Marmousi model with data that contain frequencies starting from 5 Hz to show that this method can converge to an accurate velocity starting from a linearly increasing initial velocity. Also, our proposed method works well when applied to a field data set.
Original languageEnglish (US)
Pages (from-to)1981-1992
Number of pages12
JournalGeophysical Journal International
Volume210
Issue number3
DOIs
StatePublished - Jul 7 2017

Fingerprint

Dive into the research topics of 'Efficient scattering-angle enrichment for a nonlinear inversion of the background and perturbations components of a velocity model'. Together they form a unique fingerprint.

Cite this