Time-lapse Full-Waveform Inversion for CO2 Sequestration Monitoring: Strategies and Applications to Field Accelerometer and DAS Data

He Liu

Time-lapse full-waveform inversion (FWI) is a powerful technique for seismic analysis, enabling high-resolution imaging of subsurface physical properties to monitor reservoir changes during injection, production, and long-term CO₂ storage. However, accurate time-lapse analysis remains challenging due to the requirement for highly repeatable seismic surveys, including consistent acquisition geometry, stable ambient noise conditions, and other environmental factors. To address these challenges, this thesis develops a target-oriented common-model strategy (TO CMS) to mitigate non-repeatability issues that conventional parallel strategies (PRS) fail to overcome. TO CMS combines the strengths of target-oriented FWI (TO FWI)|which improves inversion convergence within the reservoir region|and the commonmodel strategy (CMS)|which reduces time-lapse artifacts by guiding both baseline and monitor inversions along similar convergence paths. Additionally, a multi-source amplitud-encoding method is employed to significantly reduce computational cost without compromising inversion accuracy. In the context of field-scale CO₂ monitoring, vertical seismic profile (VSP) surveys provide higher vertical resolution and improved signal-to-noise ratio (SNR) compared to surface seismic methods. When integrated with FWI, VSP data further enhances the detectability of subtle time-lapse anomalies. Despite these advantages, the application of FWI to field VSP data has remained limited due to its sensitivity to nonrepeatable acquisition and noise. This thesis presents a field experiment utilizing time-lapse walkaway VSP data and FWI to monitor long-term subsurface changes associated with a small-scale CO₂ injection. The workflow demonstrates that FWI can successfully detect reservoir changes resulting from the injection of less than 60 tons of CO₂ into a shallow, 7-meter-thick reservoir at a depth of approximately 300 meters. The results conrm that even in low-injection-volume scenarios, time-lapse FWI can deliver high-resolution imaging and effectively capture small-scale velocity changes. Building on this, the study further investigates the use of Distributed Acoustic Sensing (DAS) for time-lapse FWI. DAS offers key advantages such as high spatial sampling density, long-term deployment potential, and reduced operational costs, making it a promising alternative to conventional receivers. A second field experiment is conducted at the same site using time-lapse DAS-based VSP data. Despite limited acquisition geometry and the presence of strong near-surface noise, the application of TO CMS enables successful detection of CO₂-induced changes. The timelapse results show strong agreement with the synthetic test, conrming the robustness of the proposed workflow.