Efficient modeling of seismic signature of patchy saturation for time lapse monitoring of carbon sequestrated deep saline reservoirs

Abstract

Various mechanisms controlling the multiphase flow in a real geological porous medium such as those associated with carbon dioxide (CO2) storage in a saline reservoir can lead to a patchy saturation distribution. Successful monitoring of CO2 plumes using time-lapse seismic data under these conditions is a challenge due to the degree of uncertainty in the relationship between CO2 saturation and elastic (seismic) responses. Moreover, efficient modeling of these responses is vital for practical bookkeeping of stored volumes. We investigate the potential of using seismic methods to monitor CO2 in the subsurface by using reservoir simulation data generated in two types of models. The first one consists of a random distribution of absolute permeability, not unlike typical representation of geostatistical models of permeability. A second model, more geologically meaningful, represents an eolian sand deposit containing bounding surfaces. By combining reservoir flow modeling with seismic modeling, we demonstrate that the patchy nature of the saturation distribution, resulting from small-scale multiphase flow features commonly neglected in reservoir simulation exercises, can be seismically modeled with an equivalent stack of homogeneous isotropic/anisotropic layers and the elastic properties of this equivalent stack of layers can potentially predict the actual CO2 saturation within the reservoir to reasonable accuracy. We conclude that using efficient waveform inversions to extract homogeneous equivalent layer properties from time lapse seismic data and relating them back to the CO2 saturations is the key to the development of an effective monitoring strategy for carbon sequestrated reservoirs. We also believe that such an effective monitoring will require integrating reservoir flow simulation with seismic simulation for the given reservoir so that appropriate and feasible seismic modeling assumptions (like including anisotropy) can be determined prior to monitoring.