Stochastic Seismic Inversion Framework for Quantifying Reservoir Uncertainty in the Paleogene-Neogene Niger Delta: A Full Bayesian Approach

Quantitative uncertainty characterisation in Niger Delta reservoir description is a frontier problem of direct commercial consequence, as probabilistic resource volume estimates derived from seismic inversion are routinely biased by the non-uniqueness of the inversion problem and by the inadequate propagation of geological prior information into the inversion objective function. This research develops a full Bayesian stochastic seismic inversion framework tailored to the depositional architecture, rock physics properties, and seismic data characteristics of Paleogene-Neogene Niger Delta reservoirs. The methodology incorporates a spatially variable prior model derived from stratigraphic forward modelling of Niger Delta sequence stratigraphy, enabling prior distributions of acoustic impedance that reflect the expected lateral continuity of channel belt and lobe architectures. The Bayesian inversion is implemented using a Markov Chain Monte Carlo sampler with efficient proposal distributions designed for the high-dimensional parameter space of 3D seismic inversion. Application to a producing field in OML 58, Rivers State, generates an ensemble of 500 equiprobable acoustic impedance realisations from which net sand thickness distributions and volumetric uncertainty bounds are derived. The P10-P90 range of net sand thickness from the Bayesian method is 38 percent narrower than that from a conventional geostatistical approach, demonstrating improved uncertainty quantification. Validation against production history data confirms that the Bayesian-derived ensemble captures actual performance outcomes within the P10-P90 bounds at 91 percent of monitored well locations. Keywords: stochastic inversion, Bayesian, Niger Delta, reservoir uncertainty, MCMC