Efficient frequency-domain reflection-based full waveform inversion using wavefield separation

Full waveform inversion (FWI) is a suitable algorithm to invert the subsurface velocity with high accuracy. The low frequency components and long offset seismic data are necessary for the FWI to invert the long wavelength velocity. To invert the long wavelength velocity from the reflection-dominant short offset seismic data, the reflection-based full waveform inversion (RFWI) which decomposed the gradient of the FWI into high and low wavenumber components was introduced. However, the conventional RFWI still contained the high wavenumber components which were generated between the primary wavefield and the scattered wavefield. Moreover, the true amplitude migration which needed additional computation was necessary. In this study, the new frequency-domain RFWI algorithm is introduced to overcome the drawbacks of the conventional RFWI. The new RFWI algorithm uses the up/down-going wavefield separation to remove the high wavenumber component and the two-step approach to enhance the computational efficiency. The effectiveness of the proposed algorithm is verified by the synthetic numerical test using short-offset Marmousi seismic data.