3D targeted nullspace shuttle in 4D FWI: synthetic time-lapse FWI experiments for CO₂ monitoring configured for the Snowflake dataset

Jinji Li, Kimberly Pike, Kevin W. Hall, Kristopher A. Innanen

Time-lapse full waveform inversion (FWI) is a critical tool for monitoring subsurface CO₂ sequestration, an essential component of carbon capture and storage (CCS). However, interpreting time-lapse FWI results can be challenging due to artifacts arising from non-repeatability between baseline and monitor surveys. The targeted nullspace shuttle technique addresses this issue by enhancing structural consistency between baseline and monitor inversions without increasing the overall data misfit. Previous studies have demonstrated its effectiveness in 2D settings, showing improved reliability in the interpretation of time-lapse differences. In this study, we extend the targeted nullspace shuttle framework to 3D time-lapse FWI and assess it in 4D seismic monitoring scenarios through synthetic acoustic inversion experiments. The synthetic model is derived from the Snowflake vertical seismic profile (VSP) dataset, based on the Containment and Monitoring Institute (CaMI) Field Research Station (FRS) of Carbon Management Canada (CMC), jointly developed by the Consortium for Research in Elastic Wave Exploration Seismology (CREWES). The scenario simulates a realistic injection process in which 60-tonnes of CO₂ is injected. We evaluate the detectability of the injected plume using the nullspace shuttle method under varying acquisition sparsity and aperture conditions and quantify the relationship between CO₂ saturation levels and observable P-wave velocity changes. Our results indicate that the targeted nullspace shuttle enables effective plume delineation in sparse acquisition scenarios. However, the limited aperture inherent to VSP-based FWI restricts the resolution of finer-scale plume features, leading to a suboptimal or failed time-lapse recovery. The findings highlight the potential of the targeted nullspace shuttle as a focused and computationally efficient strategy for 4D FWI applications, particularly in improving the robustness of geo-monitoring.