Two-step full waveform inversion of diving and reflected waves with the diffraction-angle-filtering-based scale-separation technique

SUMMARY Full waveform inversion (FWI) is a highly non-linear optimization problem that aims to reconstruct high-resolution subsurface structures. The success of FWI in reflection seismology relies on appropriate updates low-wavenumber background velocity structures, which are generally driven by the diving waves conventional FWI. On other hand, reflected mainly contribute updating high-wavenumber components rather than components. To extract information from addition waves, we propose two-step strategy separates given model into reflectivity and models then alternately update them using scale-separation technique based diffraction-angle filtering (DAF; was proposed effectively control wavenumber gradient). Our first inverts suppressing energy at large diffraction angles, necessary compute wave paths (i.e. rabbit-ears-shaped kernels) for subsequent stage. Then, due (banana-shaped (rabbit-ears-shaped gradient small angles. similar (RWI) it high- uses kernels updates. main difference between our RWI adopts DAF-based space domain, makes algorithm both banana-shaped computationally attractive. By applying synthetic data Marmousi-II real ocean-bottom cable North sea, demonstrate method properly reconstructs structures even if initial deviate true models.