Abstract

Full waveform inversion (FWI) is a powerful tool to quantify Earth properties of the subsurface from seismic data, and has become a major area of research. Because of very high computational cost, FWI has so far been used for either 2D full elastic or 3D acoustic media. However the Earth is three-dimensional, elastic and highly heterogeneous, and therefore, one would require a full 3D elastic wave equation for accurate modeling of amplitudes within inversion process. Acoustic approximation could impact significantly on the final inversion results due to the amplitude variation with offset (AVO) effect. Here we propose an efficient way to perform 3D elastic FWI for time-lapse seismic data using grid injection method, which allows for efficient calculation of synthetic seismograms after model alterations within a localized area. The technique allows for significant reductions in computational time and memory requirements without any significant loss in accuracy.