Accelerated 3-D full-waveform inversion using simultaneously encoded sources in the time domain: application to Valhall ocean-bottom cable data

We use the method of simultaneously encoded sources to perform computationally inexpensive full-waveform inversion (FWI) on fixed-spread, marine seismic data. A workflow is proposed whereby both data- and model-based preconditioning strategies are enforced to mitigate the non-linearity of the seismic inverse problem. Artificial crosstalk, introduced by the false correlation of forward and adjoint wavefields of simultaneously simulated sources, is minimized by simulating supershots of random linear combinations of data with iteration-varying encoding. Using encoded sources with partial-source assembly, crosstalk is furthermore suppressed by randomizing the locations of encoded subsources. Synthetic case studies verify our basic workflow approach, demonstrating accurate model reconstruction in the most extreme case of a single supershot. Application to real data from the Valhall oilfield in the North Sea demonstrates reconstruction of near-surface features with one to two orders of magnitude speedup per FWI iteration. Such an efficiency gain can be incorporated into a seismic data processing workflow both for tomographic inversion and for quality control measures.