From slope tomography to FWI: Is the conventional workflow viable in complex settings?

Ultra-long offset seabed acquisitions implemented with sparse array of ocean bottom nodes (OBN) are emerging as the goto strategy for velocity model building. These stationaryrecording geometries provide the flexibility to record a plethora of wave arrivals and in particular diving waves that undershoot the deepest targeted structures. These seabottom acquisitions are also amenable to the recording of frequencies as low as 1.5Hz. Full-Waveform Inversion (FWI) can be fed with this wide variety of low-frequency wave types to build broadband velocity models. Ultra-long offset surveys provide also a suitable framework to revive well proven tomography methods such as first-arrival slope+traveltime tomography to build kinematically-accurate initial velocity model for FWI. Here, we revisit the challenging 2004 BP salt benchmark by assessing a workflow combining first-arrival slope tomography and FWI against an ultra-long OBN survey involving a maximum offset of 97 km. Starting from a crude starting velocity gradient model, we manage to capture accurately the geometry of the salt and the subsalt structure down to 7km depth using 1.5-Hz starting frequency. Below 7km, the deficit of wide-angle coverage near the ends of the targeted structure combined with the inaccuracy of the tomography model in these areas prevents accurate reconstruction of subsalt structures by FWI when frequencies smaller than 1.5Hz are not used. The remedies to reconstruct subsalt structures in area suffering from deficit of wide-angle illumination should come from sophisticated sparsifying regularization in FWI and/or by the combination of first-arrival and short-spread + post-critical reflection slope tomography. The velocity model built by firstarrival slope tomography should be also beneficial for any variant of FWI involving more convex distances than the leastsquares norm of the differences.