A Bayesian Framework for Joint Inversion of Multi-Physics Geophysical Data in the Niger Delta Petroleum System

This study develops a Bayesian probabilistic framework for joint inversion of multi-physics geophysical datasets in the Niger Delta petroleum system, providing an original contribution to quantitative subsurface characterisation with rigorous uncertainty quantification. The Niger Delta's complex shallow geology creates systematic imaging challenges that cannot be adequately resolved by any single geophysical dataset. Joint inversion exploiting complementary sensitivity of seismic, gravity, and electromagnetic methods can significantly reduce model non-uniqueness, but formal Bayesian frameworks for multi-physics joint inversion have not been developed or validated for Niger Delta conditions. This study formulates a hierarchical Bayesian joint inversion model linking seismic impedance, gravity, and CSEM resistivity data through a shared petrophysical parameter model. Markov Chain Monte Carlo sampling explores the posterior model space for a synthetic Niger Delta test case and a real 3D dataset from a deepwater development block. The petrophysical link model uses a Gaussian mixture rock physics framework calibrated from existing well data. Computational efficiency is achieved through gradient-based MCMC with adjoint-state gradient calculations. Findings reveal that the Bayesian joint inversion recovers reservoir property models with substantially lower uncertainty than individual inversions, with hydrocarbon saturation uncertainty reduced from 32 percent to 18 percent standard deviation in the test case. The posterior model ensemble captures geological variability consistent with sedimentological understanding of the turbidite system. Application to the real dataset identifies a low-resistivity gas zone missed by conventional seismic analysis. The study contributes an original Bayesian multi-physics inversion framework validated for Niger Delta conditions and recommends its adoption in deepwater development well decision-making.

Keywords: Bayesian inversion, multi-physics, Niger Delta, joint inversion, uncertainty quantification.