Some improvement strategies for the time-domain dual-sensor full waveform inversion

Abstract

Full waveform inversion (FWI) is an effective method in retrieving high-resolution subsurface parameters from seismic data. In this paper, we review the dual-sensor FWI previously studied and present some improvement strategies. We first modify the low-pass Wiener filter to band-pass Wiener filter for time-domain FWI to reduce the overlapped regions. Secondly, we use spectral conjugate gradient method to update the models instead of the conventional conjugate gradient method to speed up the convergence and improve the inversion accuracy. At last, we further improve the efficient boundary storage method which is suitable for the first-order velocity-stress acoustic wave equation to reduce the storage burden. After that, we validate these new strategies for the modified Marmousi model, and the synthetic example shows the feasibility and robustness of these new strategies in terms of improving computational efficiency and alleviating large amounts of data storage.