Enhanced elastic wavefield separation using local orthogonalization filtering with applications in elastic modelling and inversion

P-wave and S-wave separation methods have been widely applied to recorded and modelled multi-component wavefields in the time- and space-domains. Traditional methodologies, which employ divergence and curl operators, have given reasonable results; however, substantial residual energy or crosstalk remains. Application of the divergence and curl operators will alter the physical properties of the respective separated wavefields in terms of amplitude and phase information. Numerous studies have suggested more robust methods to separate P- and S-waves, such as decoupled wave equations to preserve the physical properties of the decomposed wavefields. These formulations produce intrinsic artifacts from S-wave reflections and conversions at where the non-smoothed velocity exists. In this paper, we introduce a filtering methodology using signal and noise orthogonalization to reduce unwanted artifacts occurring during the P-wave and S-wave separation. By combining robust wavefield separation and orthogonalization filtering, the residual reflection events, mainly from the S-wave conversions and reflections, are attenuated. We demonstrate using synthetic and real data examples a decoupled elastic wave propagation modelling scheme followed by local orthogonalization filtering and show its application in elastic full waveform inversion (FWI), which attempts to handle interference noise between mode-converted waves. Consequently, from the numerical examples presented, it is shown that using cleaner and separated P- and S-wavefields obtained from the filtered elastic wave propagation modelling and inversion, the solution converges faster and reduces the risk of falling into local minima during elastic multi-parametric inversion.