Time-domain seismic forward and inverse modeling for a dissipative medium is a vital research topic to investigate the attenuation structure of the Earth. Constant Q, also called frequency independence of the quality factor, is a common assumption for seismic Q inversion. We propose the first- and second-order nearly constant Q dissipative models of the generalized standard linear solid type, using a novel Q-independent weighting function approach. The two new models, which originate from the Kolsky model (a nearly constant Q model) and the Kjartansson model (an exactly constant Q model), result in the corresponding wave equations in differential form. Even for extremely strong attenuation (e.g., Q = 5), the quality factor and phase velocity for the two new models are close to those for the Kolsky and Kjartansson models, in a frequency range of interest. The wave equations for the two new models involve explicitly a specified Q parameter and have compact and simple forms. We provide a novel perspective on how to build a nearly constant Q dissipative model which is beneficial for time-domain large scale wavefield forward and inverse modeling. This perspective could also help obtain other dissipative models with similar advantages. We also discuss the extension beyond viscoacousticity and other related issues, for example, extending the two new models to viscoelastic anisotropy.

Seismic-wave attenuation is an important component of describing wave propagation. Certain regions, such as gas clouds inside the earth, exert highly localized attenuation. In fact, the anisotropic nature of the earth induces anisotropic attenuation because the quasi P-wave dispersion effect should be profound along the symmetry direction. We have developed a 2D acoustic eikonal equation governing the complex-valued traveltime of quasi P-waves in attenuating, transversely isotropic media with a vertical-symmetry axis (VTI). This equation is derived under the assumption that the complex-valued traveltime of quasi P-waves in attenuating VTI media are independent of the S-wave velocity parameter υ S0 in Thomsen's notation and the S-wave attenuation coefficient A S0 in Zhu and Tsvankin's notation. We combine perturbation theory and Shanks transform to develop practical approximations to the acoustic attenuating eikonal equation, capable of admitting an analytical description of the attenuation in homogeneous media. For a horizontal-attenuating VTI layer, we also derive the nonhyperbolic approximations for the real and imaginary parts of the complex-valued reflection traveltime. These equations reveal that (1) the quasi SV-wave velocity and the corresponding quasi SV-wave attenuation coefficient given as part of Thomsen-type notation barely affect the ray velocity and ray attenuation of quasi P-waves in attenuating VTI media; (2) combining the perturbation method and Shanks transform provides an accurate analytic eikonal solution for homogeneous attenuating VTI media; (3) for a horizontal attenuating VTI layer with weak attenuation, the real part of the complex-valued reflection traveltime may still be described by the existing nonhyperbolic approximations developed for nonattenuating VTI media, and the imaginary part of the complex-valued reflection traveltime still has the shape of nonhyperbolic curves. In addition, we have evaluated the possible extension of the proposed eikonal equation to realistic attenuating media, an alternative perturbation solution to the proposed eikonal equation, and the feasibility of applying the proposed nonhyperbolic equation for the imaginary part of the complex-valued traveltime to invert for interval attenuation parameters.

The eikonal equation is utilized across a wide spectrum of science and engineering disciplines. In seismology, it regulates seismic wave traveltimes needed for applications like source localization, imaging, and inversion. Several numerical algorithms have been developed over the years to solve the eikonal equation. However, they suffer from computational bottleneck when repeated computations are needed for perturbations in the velocity model and/or the source location, particularly in large 3D models. Here, we employ the emerging paradigm of physics-informed neural networks (PINNs) to solve the eikonal equation. By minimizing a loss function formed by imposing the validity of the eikonal equation, we train a neural network to produce traveltimes that are consistent with the underlying partial differential equation. More specifically, to tackle point-source singularity, we use the factored eikonal equation. We observe sufficiently high traveltime accuracy for most applications of interest. We also demonstrate how machine learning techniques like transfer learning and surrogate modeling can be used to massively speed up traveltime computations for updated velocity models and source locations. These properties of the PINN eikonal solver are highly desirable in obtaining an efficient forward modeling engine for seismic inversion applications.

No abstract

Summary The generalized standard-linear-solid model, also called the Zener model, is widely used in viscoacoustic/viscoelastic wavefield forward and inverse modeling, because the wave equations in this model can be written in differential equation form, which can be solved efficiently by time-domain numerical methods such as finite difference method, spectral element method, etc. For this model, however, two different expressions for the relaxation function (or complex modulus) appear in the literature somewhat confusingly. In addition to this confusion, the time- and frequency-domain versions of the wave equations for the generalized standard-linear-solid model are scattered throughout the literature. Here, we revisit the generalized standard-linear-solid model and seek to overcome the confusion concerning the expression for the relaxation function (or modulus). We present a unified approach to derive the viscoacoustic wave equations. We start with the time- and frequency-domain formulations separately to derive two sets of viscoacoustic wave equations. All these viscoacoustic wave equations are expressed in a simple and compact form. The two sets of viscoacoustic wave equations are equivalent to each other. The proposed method to derive the appropriate viscoacoustic wave equations can be extended to derive wave equations for other dissipative media.