Toward Data-domain Waveform Inversion of Reflected Waves

Full Waveform Inversion is becoming a powerful tool for quantitative seismic imaging from wide-azimuth seismic data. The method is today mainly used as a high-resolution tomography in the Earth zones sampled by both reflected and diving events. The classical formulation however prevents the reconstruction of the low-part of the wavenumber spectrum of the velocity model, from reflection-only data. The used of reflections in waveform inversion is a crucial challenge for deep imaging. This study, through a simple canonical example, first analyses the issue of classical waveform inversion formalism when used with reflection data. It is then shown that if a prior knowledge of reflectivity can be known, reflected waves can be be exploited in a data-domain approach to retrieve the long wavelength of the velocity, in particular when we used appropriate misfit measurement as the cross-correlation that overcome cycle-skipping issues, and inversion domain as the pseudo-depth domain that preserves the invariant property of zero-offset time.