Bayesian Seismic Inversion Meets History Matching: Rock Facies Modeling for CO2 Storage with 4D Seismic Constraints

Summary Accurate reservoir characterization is crucial for ensuring the long-term security of geological carbon dioxide (CO2) storage. While assisted history matching (AHM) with 4D seismic data is a key technology for reducing geological uncertainty, conventional workflows often oversimplify the inherent uncertainty in the seismic observations, potentially biasing the results. In this paper, we introduce an integrated AHM framework that accounts for seismic data uncertainty to produce more reliable geological models. The workflow leverages Bayesian seismic inversion to obtain a probabilistic estimate of the 4D acoustic impedance change and its associated uncertainty. These probabilistic data are then assimilated via the ensemble smoother with multiple data assimilation (ES-MDA) algorithm, with the goal of updating the parameters of the underlying geological facies model. The proposed methodology is successfully applied to a reservoir interval of the Sleipner CO2 storage site, using 4D seismic data from the 2016 and 2020 vintages. The history-matched posterior facies models show the emergence of a north/south sandstone facies trend system, interpreted as local channel features. This update leads to an improvement in the match between the synthetic and observed seismic data and a reduction in geological uncertainty. Our findings demonstrate that the probabilistic treatment of seismic data in AHM can lead to more geologically plausible reservoir models and therefore more trustworthy forecasts, representing a significant step toward more robust risk management for industrial-scale CO2 storage projects.