Assessing the potential to use repeated ambient noise seismic tomography to detect CO₂ leaks: Application to the Aquistore storage site

The Aquistore project in Saskatchewan, Canada provides carbon dioxide (CO₂) storage for the world's first combined commercial power plant and carbon capture and storage (CCS) project. CO₂ has been injected at a depth of 3.2 km since April 2015 and a permanent near surface geophone array provides passive seismic monitoring. The ability to identify any containment breach is a vital part of risk management and reduction for CO₂ storage sites. We therefore investigate the potential to monitor seismic velocity changes following a hypothetical leak of CO₂ from the reservoir using passive monitoring methods. We estimate the expected shear-wave velocity change with CO₂ saturation, and using data from the geophone array we investigate whether ambient noise interferometry (ANI) and a tomographic inversion for Rayleigh wave group-velocity maps could provide a suitable CO₂ leakage detection tool. To assess the repeatability of the method, we conduct, for the first time, a time-lapse ambient noise tomography survey of a CO₂ storage site to cover time periods preceding and following injection start-up. Sensitivity analysis results indicate that usable surface wave data derived from the current array configuration are sensitive to depths of ∼400 m and shallower. We do not expect to observe any changes due to CO₂ migration at such shallow depths and the estimated seismic velocities pre- and post-injection agree to within 60 m s⁻¹, which is on the order of double the predicted velocity change with CO₂ saturation. Therefore, due to uncertainties in travel-time picks (5–15%) and variations in the obtained velocity structure between consecutive days (up to 20%), we would be unable to resolve the expected seismic velocity change with an influx of CO₂ at 400 m (∼3–4%). Additionally, the noise source variability does not allow stable velocity estimates to be made in the time-frame of currently-available data. Consequently, in the event of a CO₂ leak at the Aquistore site, using the standard ambient noise analysis methods applied herein, Rayleigh wave tomography could be deployed to detect velocity changes due to CO₂ saturation only if (a) a wider aperture surface array was in place to allow longer period surface waves to be used, providing sensitivity at greater depths, (b) arrival times of interferometrically-synthesised surface waves could be picked with increased accuracy, and (c) there is stability of the noise source distribution between repeated surveys. However, a map of three-dimensional near surface velocities, as obtained in this study, could nevertheless be useful for near surface static corrections when using active-source seismic reflection surveys to image and monitor the reservoir. More generally, further similar studies are required to assess the applicability of ANI for leak detection at other CO₂ storage sites.