A review on reflection-waveform inversion

Abstract

Full-waveform inversion (FWI) utilizes optimization methods to recover an optimal Earth model to best fit the observed seismic record in a sense of a predefined norm. Since FWI combines mathematic inversion and full-wave equations, it has been recognized as one of the key methods for seismic data imaging and Earth model building in the fields of global/regional and exploration seismology. Unfortunately, conventional FWI fixes background velocity mainly relying on refraction and turning waves that are commonly rich in large offsets. By contrast, reflections in the short offsets mainly contribute to the reconstruction of the high-resolution interfaces. Restricted by acquisition geometries, refractions and turning waves in the record usually have limited penetration depth, which may not reach oil/gas reservoirs. Thus, reflections in the record are the only source that carries the information of these reservoirs. Consequently, it is meaningful to develop reflection-waveform inversion (RWI) that utilizes reflections to recover background velocity including the deep part of the model. This review paper includes: analyzing the weaknesses of FWI when inverting reflections; overviewing the principles of RWI, including separation of the tomography and migration components, the objective functions, constraints; summarizing the current status of the technique of RWI; outlooking the future of RWI.

Data-driven Full-waveform Inversion Surrogate using Conditional Generative Adversarial Networks

“…The major threat to the method is the usage of poor-quality training data. Despite the robustness of FWI velocity modeling, examples where FWI could not converge to an acceptable velocity model because of seismic acquisition limitations and low signal-to-noise ratio are common [34]. If this data were used to train the model, the pix2pix cGAN will also produce a poor-quality velocity model.…”

Adaptive Feedback Convolutional‐Neural‐Network‐Based High‐Resolution Reflection‐Waveform Inversion

“…As the initial model is normally smooth, at the first iteration of FWI, the gradient is mainly given by the interaction of the source wavefield with the reflection‐dominated residual wavefield at the interfaces. Thus, at the first iteration, reflected waves provide mainly the high‐wavenumber component of the FWI gradient at the interface (migration component) (Mora, 1989; Yao et al., 2020). In the following iterations, both the source and reflection‐related residual wavefields create corresponding scattered waves at the interfaces.…”

Pseudotime domain joint diving-reflected FWI using graph-space optimal transport

“…Reflection waveform inversion (RWI) relies on a reflective initial model to update the tomographic component of V p beyond the depths sampled by diving waves (e.g. Vigh et al, 2019;Yao et al, 2020). Since reflectivity is a function of V p , depthdomain RWI requires iterative re-migration of the model discontinuties (Brossier et al, 2015) and offset selection to attenuate the conflict between fixed reflectivity and evolving kinematics.…”

“…JFWI, as well as FWI (Virieux and Operto, 2009), suffers from phase ambiguity when the predicted traveltimes differs from the observed of more than half a dominant period (Brossier et al, 2015;Zhou et al, 2015). Therefore, it benefits from the use of objective functions robust to cycle-skipping, such as cross-correlation time-shift (e.g., Brossier et al, 2015;Wang et al, 2019;Yao et al, 2020); Graph-space optimal transport (GSOT, Métivier et al, 2019) in particular has been shown to perform well when starting from poor initial models in combination with JFWI, without loss of resolution compared to a L 2 -norm objective function (e.g., Li et al, 2019;Provenzano et al, 2020) In this paper, we cast JFWI in the pseudotime domain and combine it with a GSOT objective function. A realistic synthetic short-offset reflection dataset containing free-surface effects is used to demonstrate that: 1) JFWI in the pseudotime is more robust than in depth-domain, and reconstructs a macro-V p model suitable as starting model for FWI; 2) the pseudotime inversion strategy doesn't require offset selection, and dramatically reduces the need to iteratively re-migrate the reflectivity; 3) the advantages of pseudotime starting from a 1D initial model are maximised when using GSOT with respect to L 2norm.…”