Acoustic full-waveform inversion in an elastic world

Abstract

Full-waveform inversion (FWI) is a technique used to obtain high-quality velocity models of the subsurface. Despite the elastic nature of the earth, the anisotropic acoustic wave equation is typically used to model wave propagation in FWI. In part, this simplification is essential for being efficient when inverting large 3D data sets, but it has the adverse effect of reducing the accuracy and resolution of the recovered P-wave velocity models, as well as a loss in potential to constrain other physical properties, such as the S-wave velocity given that amplitude information in the observed data set is not fully used. Here, we first apply conventional acoustic FWI to acoustic and elastic data generated using the same velocity model to investigate the effect of neglecting the elastic component in field data and we find that it leads to a loss in resolution and accuracy in the recovered velocity model. Then, we develop a method to mitigate elastic effects in acoustic FWI using matching filters that transform elastic data into acoustic data and find that it is applicable to marine and land data sets. Tests show that our approach is successful: The imprint of elastic effects on the recovered P-wave models is mitigated, leading to better-resolved models than those obtained after conventional acoustic FWI. Our method requires a guess of V P ∕V S and is marginally more computationally demanding than acoustic FWI, but much less so than elastic FWI.

Addressing Viscous Effects in Acoustic Full-waveform Inversion

“…Analogously to the method for elastic effects, the present method also involves the computation of matching filters. Nevertheless, their characteristics are different to those computed for elastic effects (Agudo et al, 2017) given the different characteristics of each type of medium and, consequently, different characteristics in wave propagation. The inversion of the resulting data is then performed using the acoustic wave equation without including viscous effects, and we show that this results in better quality of the recovered models than those obtained after acoustic FWI of the observed data.…”

“…The proposed method consists of suppressing viscoacoustic effects from the observed data by computing matching filters that match modeled viscoacoustic to modeled acoustic data generated from estimates of the subsurface P-wave velocity and Q models, followed by the application of the filters to the observed data. This is analogous to the method proposed by Agudo et al (2016Agudo et al ( , 2017 to mitigate elastic effects in acoustic FWI, but with three main differences:…”

“…Here, we implement an alternative scheme by adapting the method of Agudo et al (2016Agudo et al ( , 2017, originally designed to mitigate elastic effects, to address viscous effects. The aim of this modified method is not to obtain an intrinsic attenuation model of the subsurface, but to mitigate viscous effects and improve the estimated P-wave velocity models from FWI.…”

Multiscale reflection phase inversion with migration deconvolution

“…However, the isotropic acoustic assumption neglects elastic, attenuation, and anisotropic effects, which can severely compromise the inversion results. In some cases, a matching filter can be used to reduce the non-acoustic effects of the seismic data (Agudo et al, 2018), but the matching filters need to be iteratively updated and are prone to errors. Sun and Schuster (1993) proposed amplitude replacement as a one-step method for compensating for some of the non-acoustic effects in the data.…”

Broadband Q-Factor Imaging for Geofluid Detection in the Gulf of Trieste (Northern Adriatic Sea)

“…Thus, it is not a direct indicator of fluids' presence, unless there is a borehole calibration or geological information that may single out one or part of these possible contributions (Jannsen et al, 1985;Tonn 1991). Trying to distinguish them is still an open challenge and a subject for current research (Wu 1985;Best et al, 1994;Hackert and Parra 2003;Helle et al, 2003;Picotti and Carcione 2006;Mulder and Hak 2009;Cheng and Margrave 2012;Ba et al, 2015;Bai et al, 2017;Agudo et al, 2018;Song et al, 2020a, among others). In this paper, we apply a recent method introduced by Lin et al (2018) to marine seismic data, producing high-resolution, broadband images of the anelastic absorption contrasts, and the related Q factor.…”