Evaluating model complexity in simulating supercritical CO 2 dissolution, leakage, footprint, and reservoir pressure for three-dimensional hierarchical aquifer

Abstract A hierarchical fully heterogeneous aquifer model (FHM) provides a reference for developing and testing 3 facies-based hydrostratigraphic models (HSMs) each representing a CO 2 storage aquifer with reduced permeability (k) heterogeneity resolution: 8-unit, 3-unit, and 1-unit homogeneous models. Under increasing aquifer lnk variances (0.1, 1.0, 4.5), flow upscaling was conducted to calculate equivalent permeabilities for the HSMs. Within a Design of Experiment uncertainty analysis framework varying geothermal gradient, salinity of formation water, caprock permeability, and injection rate, CO 2 injection coupled to convective mixing was simulated by all models. In addition to the injection phase, all simulations were carried out for 2000 years using PFLOTRAN, a massively parallel, multiphase, multicomponent numerical simulator that ran on the NCAR-Wyoming Supercomputing Center's Yellowstone supercomputer. Simulation outcomes of the HSMs were compared to those of the FHM within their full parameter space, and four performance metrics were evaluated: dissolved CO 2 , CO 2 leakage, plume footprint, and pore pressure evolution in response to injection and migration. Results suggest that aquifer variance, heterogeneity resolution, and salinity can all affect the development of fingering and convective mixing, and therefore the amount of dissolution storage. For the modeling choices and assumptions made in this study, the 3-unit HSM was found to be an all-around optimal model by capturing both the sensitivity of the FHM and the performance metrics under different reservoir storage or operational conditions. Implications for modeling long-term CO 2 storage in data-poor systems are discussed and future research indicated.