A numerical study of 3D elastic time-lapse full-waveform inversion using multicomponent seismic data

Abstract

A common assumption in wave-propagation problems is that the subsurface is approximately an acoustic medium. Under this assumption, important wave phenomena such as S-waves are not included. Due to the increase in computational power in recent years, the acoustic assumption may be left behind and replaced by the more physically correct elastic assumption. Time-lapse seismic data contain information about changes in the subsurface due to the production of hydrocarbons or injection of [Formula: see text]. Full-waveform inversion (FWI) is an inverse method that can be used to quantify these time-lapse changes in the subsurface. Using a 3D isotropic elastic implementation of the FWI method, we studied two strategies for performing time-lapse FWI. We used synthetic ocean-bottom multicomponent seismic time-lapse data to estimate changes in the P- and S-wave velocity models. A sensitivity analysis in which the sensitivities with respect to the magnitude and physical size of the time-lapse anomalies and the noise level in the data was performed. The strategy focusing on explaining the data differences between the baseline and monitor data sets provided fewer artifacts in the inverted elastic models than the strategy that tried to explain the full monitor data set, and it was therefore preferable. The data-difference strategy depends on good repeatability in the time-lapse data sets and sufficient convergence of the inversion of the baseline data set.