Multisource Full Waveform Inversion with Topography Using Ghost Extrapolation

Abstract

Summary Standard finite-difference modeling of the wave equation for models with severe topographic changes in elevation generates artificial diffractions in the synthetic traces. This can lead to unacceptable inaccuracies in the velocity tomogram computed by full waveform inversion (FWI). To alleviate this problem we propose a polynomial interpolation of the field values at and near the irregular free surface. This improved finite-difference procedure is denoted as the ghost extrapolation method. To validate its effectiveness, synthetic data are computed from a Marmousi model with severe topographical variations of the free surface. These data are inverted using both the standard FWI and the phase-encoded multisource FWI methods. Results show the ghost extrapolation method is effective in eliminating noticeable artifacts and producing accurate tomograms.