Modeling the performance of large-scale CO2 storage systems: A comparison of different sensitivity analysis methods

In this study, we perform sensitivity analyses using a high-resolution basin-scale reservoir model developed for a hypothetical carbon sequestration project located in the Southern San Joaquin Basin in California, USA. We use the massively parallel version of the multiphase multicomponent simulator TOUGH2 to simulate CO2/brine migration and pressure buildup within the CO2 storage formation and overlying/underlying formations. We evaluate the impact of parameter uncertainty on risk-related performance measures, i.e., CO2 saturation and pressure buildup at multiple locations, and the extent of the CO2 plume and overpressure zone. We compare three sensitivity analysis methods: a local sensitivity method and the global Morris and Sobol'/Saltelli methods. The uncertainty of sensitivity indices in the global methods is evaluated so that we can interpret the results even when we have a limitation in the computational resources. Results show that the three methods provide complementary information for identifying important parameters and system understanding. All three methods give consistent interpretations and importance rankings, except when a parameter has a significant non-linear effect and/or strong interaction with some other parameters. In addition to the magnitude of parameter sensitivity, our analysis emphasizes the direction (i.e., favorable or adverse in the risk perspective), non-linearity and/or interaction effects, and physical interpretation of each parameter sensitivity trend. Parameter importance varies with time and space, and also depends on the CO2 plume or pressure behaviors. In this study, the reservoir permeability is among the most important parameters for all measures, although it has a large trade-off effect in risk such that a higher permeability would tend to reduce reservoir pressure but, at the same time, increase the size of the CO2 plume footprint.