Nullspace shuttling versus conventional FWI time-lapse strategies

Kimberly Pike, Scott Keating, Kristopher A. Innanen

Time-lapse full waveform inversion is a powerful approach for detecting subtle subsurface changes, but it is highly sensitive to non-repeatable acquisition, noise, and differences in inversion convergence. Conventional time-lapse strategies attempt to mitigate these issues, but they can also introduce additional non-subsurface related time-lapse artifacts,especially when baseline and monitor surveys differ in source density. Nullspace shuttling reframes the problem by searching within the FWI nullspace for model adjustments that preserve the data misfit while minimizing the time-lapse difference. Using synthetic vertical seismic profile data based on the Carbon Management Canada Newell County Field Research Station, we compare five time-lapse full waveform inversion strategies against time-lapse nullspace shuttling under both fully repeated and sparse monitor acquisition geometries. Results show that while advanced time-lapse full waveform inversion methods reduce artifacts relative to simple parallel and sequential approaches, nullspace shuttlingm ost effectively isolates the true time-lapse anomaly, providing clearer recovery of the thin time-lapse anomaly and demonstrating robustness to sparse acquisition geometries. These findings highlight nullspace shuttling as a promising, computationally efficient alternative to multi-pass time-lapse full waveform inversion