Joint Bayesian stochastic AVA inversion of well-log and seismic data for facies estimation

Abstract

Summary Due to the inherent band-limited character of seismic data, conventional deterministic inversion methods cannot identify thin reservoir information beyond the seismic resolution. Besides, traditional Monte Carlo-based stochastic inversion requires a large number of iterative sampling, which is computational inefficiency. Hence, we develop a joint Bayesian stochastic AVA inversion method based on the linear inverse Gaussian theory and geostatistics. It directly integrates seismic data, well-log data and geostatistical information into a unified expression under the Bayesian framework, and uses the sequential Gaussian simulation to efficiently sample the joint posterior probability density function. The synthetic data example verifies the advantages of better consistency at the locations of har data and the reduction of the inversion uncertainty compared to the classical Bayesian linearized AVA inversion. The field data example shows the validity of this method in the quantitative estimation of facies.