Efficient extended-search space full-waveform inversion with unknown source signatures

摘要: Full waveform inversion (FWI) requires an accurate estimation of source signatures. Due to the coupling between signatures and subsurface model, small errors in former can translate into large latter. When direct methods are used solve forward problem, classical frequency-domain FWI efficiently processes multiple sources for signature wavefield estimations once a single Lower-Upper (LU) decomposition wave-equation operator has been performed. However, this efficient formulation is based on exact solution wave equation hence highly sensitive inaccuracy velocity model due cycle skipping pathology. Recent extended-space variants tackle sensitivity issue through relaxation combined with data assimilation, allowing wavefields closely match from first iteration. Then, parameters updated by minimizing violations. jointly estimated approach, extended becomes dependent, making ineffective. In paper, we propose simple method bypass estimate during FWI. The proposed replaces each blended data-assimilated reconstruction make independent. Besides computational efficiency, additional degrees freedom introduced spatially distributing allows better at physical location when rough. Numerical tests Marmousi II 2004 BP salt synthetic models confirm efficiency robustness method.