Large-Scale CO2 Storage in a Deep Saline Aquifer: Uncertainties in Predictions Due to Spatial Variability of Flow Parameters and Their Modeling

Scarce data and uncertainties in the spatial variation of geological properties lead to different possible models of these heterogeneities. The aim of this study is to compare the pressure results and behavior of different permeability field models for a large-scale injection in a deep saline aquifer. Five ways of representing heterogeneities are tested and compared. The simplest representation defines homogeneous equivalent properties over the entire domain. A second representation is obtained by considering homogeneous layers. Two other models represent the lateral and vertical variations in permeability in greater detail by geostatistical methods with either a continuous model or a discontinuous model (discrete values). The last model is the semi-homogeneous model combining a heterogeneous area and a homogeneous area depending on the complexity of the flow process. Highly variable predictions arise from the heterogeneities, and significant differences in estimates are obtained using the various modeling methods. The optimum resolution depends on the type of response to be estimated. Averaged properties at large scale are not adequate to estimate the critical pressure propagation far from the well. Averaged properties in the injection area are not sufficient to assess the maximum increase in pressure or the extent of migration. Lateral and vertical connectivities, and reservoir compartmentalization modeling are required to obtain reliable results. But the resolution requirements are not to be at the finest scale: The discontinuous model (discrete values) gives satisfactory results compared to the continuous model. The way to represent spatial variability of porosity and pore compressibility is also studied. The influence of these two properties is far lower than that of permeability.