Lawrence Berkeley National Laboratory

One strategy to mitigate potential climate change associated with elevated atmospheric CO$_ 2 $ concentrations is the sequestration or storage of anthropogenic CO$_ 2 $ in deep geologic formations. While the purpose of geologic carbon sequestration is to trap CO$_ 2 $ underground, the potential exists for CO$_ 2$ to migrate away from the intended storage site and pass from the subsurface to the atmosphere along permeable pathways such as well bores or faults. Therefore, to ensure the success of geologic carbon sequestration projects, the long-term storage of CO$_ 2 $ must be verified. Although numerous technologies are available to measure near-surface CO$_ 2 $ concentrations and fluxes, storage verification may be challenging due to the large variation in natural background CO$_ 2 $ fluxes and concentrations, within which a potentially small CO$_ 2 $ anomaly will have to be detected. To detect and quantify subtle CO$_ 2 $ leakage signals, we present a strategy that integrates near-surface measurements of CO$_ 2 $ fluxes or concentrations with an algorithm that enhances temporally- and spatially-correlated leakage signals while suppressing random background noise. The algorithm consists of a filter that highlights spatial coherence, and temporal stacking (averaging) that reduces noise from temporally uncorrelated background fluxes. We assess the performance of our strategy using synthetic data sets in which the surface leakage signal is either specified directly or calculated using flow and transport simulations of a variety of leakage source geometries one might expect to be present at sequestration sites. These simulations provide a means of estimating the number of measurements required to detect a potential CO$_ 2 $ leakage signal of given magnitude and area. Our results show that given a rigorous and well-planned field sampling program, subtle CO$_ 2 $ leakage may be detected using the statistical algorithm; however, leakage of very limited spatial extent or exceedingly small magnitude maybe difficult to detect with a reasonable set of monitoring resources.