Global sensitivity analysis for crosswell seismic and neutron-capture measurements in CO2 storage projects

Abstract

Quantification of reservoir uncertainty is an essential part of a monitoring design. A systematic approach that quantitatively links predicted uncertainties in a monitoring program to the underlying reservoir variability is, however, needed. We developed a methodology for quantifying uncertainty in crosswell seismic monitoring combined with neutron-capture logging and applied global sensitivity analysis (GSA) to compute and rank contributions of uncertain reservoir parameters to the predicted uncertainty of the measurements. The workflow is illustrated by a numerical study using a simplified model of a [Formula: see text] storage site where crosswell measurements have not actually been taken. Synthetic seismic responses are computed through the integration of multiphase flow, a new thermodynamically consistent fluid substitution model, and a fast marching eikonal solver. We quantified uncertainty in first-arrival times to illustrate the potential utility of crosswell seismic surveys and their limitation. Consistent with these calculations, uncertainties in neutron capture cross-section logs are also computed and related to predicted [Formula: see text] migration. The predicted uncertainty range for neutron-capture measurements indicated significant sensitivity to the uncertainty of the reservoir properties (standard deviations [STDs] of up to 6 c.u. in the injector and up to 3.5 c.u. in the monitoring well). However, the STD of predicted time-lapse crosswell seismic responses for two different source locations did not exceed 0.75 ms during the life of the project, suggesting limited value of first-arrival measurements for reservoir-parameter inversion in this case. With the time-dependent uncertainty of the predicted measurements, calculated GSA indices provided a quantitative basis for the monitoring program design. Practical implications of GSA results for model reduction and subsequent inversion were also evaluated.