Source encoding in full waveform inversion: comparison of encoding schemes and misfit functions

Abstract

Full waveform inversion (FWI) is a method of seismic tomographic inversion based on modeling the entire wave eld to extract information. This allows for the inversion of lower frequencies and longer o sets to produce more high quality imaging, but the non-linearity of the problem means that the computation costs can be prohibitive. A proposed method to reduce the computational cost involves combining multiple encoded source shots with different signatures into a single \supershot" for inversion. There are several key issues to be addressed with this method. Importantly, interference between sources or "cross-talk" may create unwanted noise. We examine FWI in its conventional and supershot incarnations, and test various methods of source encoding to reduce cross-talk noise. We also look at the issue of missing data in the supershot.