Bayesian Sparse Seismic Inversion Method with Spatial Correlation Constraints

Abstract Seismic inversion plays an important role in seismic exploration by enhancing the accurate characterization of subsurface structures and lithologic parameters. However, conventional inversion methods often neglect the intrinsic spatial correlation of stratigraphic architectures formed by sedimentary processes, thereby reducing inversion accuracy and structural continuity. To address this limitation, we construct a spatial correlation function by integrating empirical geological information with theoretical formulations and propose a spatially correlation constrained Bayesian sparse inversion (SCCBSI) method, which uses the function as a spatial constraint and incorporates sparse inversion within a Bayesian framework. This method addresses two major issues: the insufficient spatial autocorrelation within individual parameters (both laterally and longitudinally), and the spurious linear correlations between different parameters induced by background trends and other factors. The approach enhances inversion accuracy and improves lateral continuity in the inversion results. Synthetic model tests and field data applications demonstrate that the proposed method provides stable and reliable estimations of P-velocity, S-wave velocity, and density. The inversion results are consistent with geological prior information, validating the robustness and effectiveness of the proposed method.