Uncertainty quantification of CO 2 leakage through a fault with multiphase and nonisothermal effects

The potential for CO 2 leakage through a permeable fault is a key concern for geologic CO 2 sequestration (GCS) in saline formations. If CO 2 migrates vertically upward through a fault from the storage reservoir to an overlying fresh‐water aquifer, phase change can occur because temperature and pressure decrease with decreasing depth. The decrease in CO 2 density during phase transition causes an additional reduction in temperature. In this paper, we present a computational model for simulating the behaviour of a leaky fault connecting a saline CO 2 storage reservoir and an overlying fresh‐water aquifer. We address phase transition, considering the nonlinear CO 2 enthalpy and viscosity functions. The model results indicate that the CO 2 leakage rate initially increases when CO 2 migration is driven by both buoyancy and overpressure during the period of injection. In the post‐injection period, CO 2 leakage is only driven by buoyancy and the leakage rate decreases. The influence of nonisothermal conditions is more pronounced during the first stage. The deterministic model of this faulted reservoir system is used within an uncertainty quantification (UQ) framework to rigorously quantify the sensitivity of the brine and CO 2 leakage in response to the uncertain model parameters. The results demonstrate that fault permeability is the most sensitive factor affecting both CO 2 and brine leakage rate. Reduced‐order models of CO 2 and brine leakage are developed for the emulation of a large number of sample points, from which probability distributions are derived and can be incorporated in risk assessment of groundwater contamination resulting from CO 2 and brine leakage. © 2012 Society of Chemical Industry and John Wiley & Sons, Ltd