Target-oriented Time-lapse Waveform Inversion using Virtual Survey

Abstract

Time-lapse seismic data are widely used for monitoring time-variant subsurface changes. Conventional analysis provides qualitative information by comparing results from consecutive surveys, whereas waveform inversion can retrieve quantitative estimates of reservoir properties through seismic waveform fitting. The quantitative evaluation of the physical parameters obtained by waveform inversion allows for better interpretation of fluid substitution and migration during processes like oil and gas production, and carbon sequestration. Since reservoir changes are localized and only part of the data are of interest, the time-lapse waveform inversion can be optimized in terms of computational cost and convergence rate. In this study, we propose a scheme of localized waveform inversion with computed datasets we refer to as virtual surveys. Both the model domain and trace duration in forward modeling are reduced by the reorganization of the data. We show a numerical example in which the recovery of the reservoir change is computationally faster and more robust to source-receiver locations than inversion with original survey.