Hydrogeological characterization of the Weyburn carbon dioxide project area and gradient-free inverse conditioning of heterogeneous acquifer models to hydraulic head data

I. Pressures and hydraulic heads indicate mostly lateral flow through aquifers in the Weyburn Project area in the north-central Williston Basin. The wide range of TDS (<5 to >350 g/L) is a result of mixing between Ca-SO4 water of meteoric origin, high salinity Na-Cl brines, and hypersaline Ca-Na-Cl brines. The Watrous Fm. separates the deep aquifer systems from a shallow active hydrogeologic system. The Weyburn Field is a good site to store CO2 from a hydrogeological perspective.; New insights and interpretations with respect to the hydrogeology of the Williston Basin include: (1) The absence of Ca-Cl brine compositions in the Mississippian invokes the debated hypothesis of secular variation of global seawater chemistry through geological time; (2) Patterns of brine migration suggest the occurrence of paleo-hydraulic gradients significantly higher than those observed at present; and (3) The oil reservoirs along the Mississippian subcrop are likely not filled to capacity. The anhydritic seal rock fades underlying the subcrop likely failed hydraulically during periods of elevated paleo-fluid pressures.; II. The gradient-free calibration method works without the need for expensive or technically-involved computations of sensitivity coefficients of hydraulic head to model grid block permeability. A simple and effective approximation procures a matrix of sensitivity indicators as a substitute for the computation of the Jacobian matrix. Formulation of the minimization problem is based on an approximation of the gradient of the model response with respect to the parameters.; The gradient-free method solves the inversion problem in a similar manner as proven gradient-based calibration methods, but with a straightforward and transparent approach. As a result, some key points are elucidated with respect to inverse conditioning to head data. A calibration data point "samples" a heterogeneity feature when it occurs along a flowline that intersects the feature. The permeability field is adjusted only within sensitive areas, seeding locally-precise permeability perturbations, resulting in consistent identification of key heterogeneities and a corresponding improvement in local precision from conditioning stochastic permeability field realizations to head data.