Time-lapse FWI: Toward integration into reservoir monitoring

Abstract There has been an increasing application of time-lapse full-waveform inversion (FWI) to field data sets. Time-lapse FWI has key advantages over other conventional approaches for reservoir monitoring. First, FWI can invert the full wavefield (diving waves, primary reflections, and multiples). Second, it allows estimation of time-lapse models of the subsurface from minimally processed data, potentially allowing reduction of turnover in reservoir monitoring. Third, it can improve the description of physical phenomena occurring in the subsurface (e.g., elastic effects). Fourth, nonrepeatability can be addressed, at least in principle, in its formulation. Nonetheless, we discovered that applying time-lapse FWI into field studies has its own particularities that are distinct from those typically found when applying FWI to velocity model building. We show that unphysical features can be introduced even when baseline and monitor models have correct kinematics and perfect repeatability conditions are honored. Our examples suggest that illumination of the reservoir with diving waves is critical for accurate reconstruction of its time-lapse information even in the case of honoring those conditions. In addition, our field data example suggests that wide-band time-lapse FWI can unlock information on a reservoir that would be unknown otherwise. This paper aims to show that (1) applying time-lapse FWI to reservoir monitoring goes beyond the estimation of baseline and monitor models of velocity; (2) degraded conditions of illumination of a reservoir may introduce misleading spurious features in estimated time-lapse models even if perfect repeatability is met and if the velocity model is accurate within the typical conditions necessary to conduct FWI; and (3) time-lapse FWI is inherently connected to geosciences and reservoir integrated disciplines, hence requiring cross-disciplinarity for its successful application. A field data example and relevant synthetic examples support those assertions.