Dynamic characterization of geologic CO₂ storage aquifers from monitoring data with ensemble Kalman filter

Monitoring the evolution of the CO₂ plume during geologic storage is essential for conformance, verification, and risk assessment and mitigation. Monitoring data also play a critical role in characterizing the storage formation and improving the reliability of predictive models. We investigate the feasibility of using the ensemble Kalman filter (EnKF) data assimilation framework to estimate the hydraulic properties of storage formations and to predict the migration of CO₂ plume from monitoring measurements, including transient pressure and saturation data at scattered wells and time-lapse seismic data (modeled as vertically-averaged saturation differences in time). To properly account for the uncertainty in the knowledge about saline aquifer properties, the initial ensemble of formation properties is generated based on uncertain statistical model (variogram) parameters. While integration of data from scattered wells provides limited improvement in reducing the uncertainty in the initial ensemble, assimilation of time-lapse seismic measurements (represented by vertically-averaged saturation differences in time) with the EnKF leads to more noticeable uncertainty reduction and reasonable estimates of the general connectivity trends in aquifer hydraulic properties. The estimation and sensitivity analysis results suggest important differences in filter performance during and after CO₂ injection. This difference is attributed to the change in flow behavior and the dominant forces before and after injection (pressure versus gravitational forces, respectively). Additionally, when prior model realizations miss essential flow-related elements (e.g., fractures) in an aquifer, the filter provides out-of-range updates, which could be interpreted as a systematic problem in the filter design, in this case possible inconsistency in the prior models.