Time-lapse full-waveform inversion using simultaneous sources

He Liu, Xin Fu, Daniel O. Trad, Kristopher A. Innanen

Full waveform inversion (FWI) has been used to estimate high-resolution subsurface velocity models. It has become a powerful tool for time-lapse seismic inversion, which is promising to monitor reservoir profile changes with injection and production, and potentially long-term storage of CO₂. To overcome the challenge of expensive computational costs for FWI process, shot subsampling methods and source-encoding strategies have been used to make the full waveform inversion efficient while maintaining the quality of the inversion results with minimum sacrifice. The cyclic method subsamples the shots at a regular interval and changes the shot subset at each iteration step. Using this method, we can suppress the aliasing noise present in regular-interval subsampling. FWI using source-encoding strategies has been investigated using different methods. In previous work, we have used an amplitude-encoding strategy with different bases to accelerate the FWI process. In this work, we incorporate amplitude-encoding strategy with cyclic subsampled data scheme, which first subsamples the data cyclically and then compose blended during the iterations. In this way, we can directly eliminate much more crosstalk terms introduced by encoded individual shot gathers and reduce the data dimension to improve FWI efficiency. We have applied this strategy to acoustic and elastic time-lapse FWI in time domain, and the synthetic inversion results recovered the velocity profile changes in the time-lapse models very well with reduced computation efforts.