Developing Initial Model for Seismic Full Waveform Inversion Using Conventional Data Processing Tools

Abstract

We believe Full waveform inversion (FWI) is capable to be used as a part of seismic data processing routines. In order to check the possibility of using FWI in conventional data processing sequences, we evaluated the effect of different initial models on inversion results. We developed new initial models for full waveform inversion using horizon-guided well interpolation and compared it against initial velocities converted from stacking velocities,with and without dip move-out (DMO) correction. Acoustic full waveform inversion results from Marmousi2 model showed that when the subsurface structure has strong dips, interval velocities which are converted from stacking velocity without dip corrections fail to initialize FWI properly. However, applying dip move-out correction on the seismic data will relax the dip complexities in the velocity analysis stage and a good initial model for FWI could be developed. Alternatively, horizon-guided well interpolation uses velocities derived from well logs and makes an interpolated velocity model along the picked horizon by a constrained post-stack inversion. This makes an initial velocity model for full waveform inversion which ensures the convergence to the correct solution.