An efficient multiparameter acoustic anisotropic full-waveform inversion depending on parameterization

Abstract

The pressure-based acoustic approximation of the elastic wave equations in anisotropic media has advantages corresponding to computational efficiency and numerical stability. However, the numerical scattering potentials from the anisotropic parameter perturbations for the pressure wavefield are not consistent with the elastic scattering theory. In multiparameter full-waveform inversion (FWI), choosing a suitable parameterization, considering the acquisition parameters (e.g., the offset-to-depth ratio and frequency band) and the accuracy of the anisotropy information in the background initial velocity model, is an important component to a successful anisotropic parameter estimation, because the parameterization determines the trade-off between inverted model parameters and their resolution power. However, because it is difficult to perform multiparameter FWI with various types of parameterization using the pressure-based acoustic wave equation, inaccurate scattered wavefields cause the gradient direction to lose its unique properties with respect to each model parameter. To overcome these issues, we adopt the combination of pressure- and vector-based acoustic wave equations converted vector virtual sources, which preserves the computational efficiency and the angular dependency of the partial derivative wavefields in elastic media. With the proposed method, we generate the numerical PP scattering patterns for various parameterizations, which are consistent with the elastic scattering theory. Through the numerical tests using the synthetic anisotropic Marmousi-II models and a real ocean bottom cable dataset from the North Sea, we conduct acoustic FWI with three different parameterizations using the proposed method and verify that the modified scattering patterns accurately reflect the characteristics of the anisotropic parameter perturbations.