Benchmark modeling of the Sleipner CO2 plume: Calibration to seismic data for the uppermost layer and model sensitivity analysis

Comparison of simulated time-series CO2 plume with observed plumes derived from seismic amplitude data. Upper panel: CO2 plume from seismic data (Boait et al., 2012). Middle panel: Simulated plume from this study. The black outlines are the observed areal extents of CO2 plume; Bottom panel: Simulated plume from Singh et al. (2010). •Applied two multi-phase compositional simulators to the Sleipner Benchmark model for the upper most layer in Utsira Sand.•Calibrated the model against time-lapsed seismic monitoring data from 1999 to 2010.•Approximately matched with observed plume after introducing permeability anisotropy and CH4 in the CO2 stream.•Performed hundreds of simulations of parameter sensitivity on the plume migration and fate of CO2.•Even with uncertain parameters, the predicted fate of CO2 is within a narrow band of structural and solubility trapping. An important question for the Carbon Capture, Storage, and Utility program is “can we adequately predict the CO2 plume migration?” The Sleipner project in the Norwegian North Sea provides more time-lapse seismic monitoring data than any other sites for tracking CO2 plume development, but significant uncertainties still exist for some reservoir parameters. In order to simulate CO2 plume migration and assess model uncertainties, we applied two multi-phase compositional simulators to the Sleipner Benchmark model for the uppermost layer (Layer 9) of the Utsira Sand and calibrated our model against the time-lapsed seismic monitoring data at the site from 1999 to 2010. Approximate match with the observed plume was achieved by introducing lateral permeability anisotropy, CH4 in the CO2 stream, and adjusting reservoir temperatures. Model-predicted gas saturation, thickness of the CO2 accumulation, and CO2 solubility in brine – none of them used as calibration metrics – were all comparable with interpretations of the seismic data in the literature. Hundreds of simulations of parameter sensitivity (pressure, temperature, feeders, spill rates, relative permeability curves, and CH4) showed that simulated plume extents are sensitive to permeability anisotropy, temperature, and CH4 but not sensitive to the other analyzed parameters. However, adjusting a single parameter within the reported range of values would not reproduce the north–south trending CO2 plume. It took a combination of permeability, CH4, and temperature adjustments to match simulated CO2 plume with seismic monitoring data. On the other hand, even wi