Reflection waveform inversion with variable density

Abstract

Full waveform inversion (FWI) is considered as an effective technique for high-resolution velocity model building. It requires a good starting model to avoid the cycle-skipping problem when the low-frequency data are unavailable. To provide a background staring model for FWI, reflection waveform inversion (RWI) is successfully applied to emphasize on low-wavenumber velocity model updating. Given a proper reflectivity model, the low-wavenumber gradient of RWI is obtained by cross-correlating the background wavefield and the reflection wavefield. While explicitly generating the reflection wavefield from the reflectivity model is time-consuming. Therefore, in order to mitigate the time-consuming problem, we have proposed an efficient RWI approach by alternatively updating the density model and the low-wavenumber velocity model. This approach doesn't require explicitly generating the reflection wavefield. We use synthetic and field data to demonstrate that, the proposed method could effectively provide a good starting model for FWI.