Bayesian AVOAz Inversion of Fluid and Anisotropy Parameters Using Gibbs Sampling With AISM Algorithm

Fracture discrimination and reservoir fluid identification guide the characterization of the spatial distribution of reservoir oil and gas and the comprehensive evaluation of sweet spots, which are crucial for the exploration and development of unconventional hydrocarbons. However, the reflection coefficient equation derived under the anisotropic medium assumption contains numerous parameters, thereby augmenting the uncertainty of amplitude variation with offset and azimuth (AVOAz) inversion. In this study, an anisotropic reflection coefficient equation, including Gassmann fluid term and fracture parameters, is rewritten based on the horizontal transversely isotropic (HTI) medium assumption, which reduces the equation’s dimension and ill-posedness. In addition, the traditional Markov Chain Monte Carlo (MCMC) algorithm is complicated to stably sample from the high-dimensional posterior probability distribution function due to the large number of inversion parameters. To overcome the problem and evaluate the uncertainty of AVOAz inversion, the adaptive independent sticky MCMC (AISM) strategy combined with the Gibbs algorithm is introduced. Under the Bayesian framework, the enhanced algorithm constructs a proposal distribution from a non-parametric distribution and employs support points for adaptive updating of the proposal distribution. The method of equation rewriting, coupled with the refined AISM algorithm, significantly improves the accuracy and robustness of inversion. The proposed method is validated through model testing and application to real data from a shale exploration area in Southwest China. The inversion results demonstrate a high concordance with well data, affirming the reliability and applicability of the inversion method.