Rapid and accurate characterization of time-lapse seismic data is important to enable operational adjustments to be made and provide a guide for future drilling. A facies-based Bayesian inversion offers some advantages over traditional simultaneous prestack inversion, primarily avoiding the laborious construction of low-frequency models. To implement a facies-based inversion method, we adjust the model parameterization to be the ratio of monitor to baseline elastic properties. With this parameterization, the set of facies is reduced to those corresponding to specific production scenarios (production facies) that characterize expected subsurface changes between the monitor and baseline acquisitions. Production facies’ elastic properties are generally modeled through rock physics relationships. The inversion operates on the difference of the angle stacks directly, and hence requires properly calibrated and registered baseline and monitor data. The result is a rapid workflow that can image changes in elastic properties accurately. We demonstrate the technique on a synthetic example, and also on field data from an operating oil sands thermal recovery project in Alberta, Canada.
Many custom approaches exist for 4D seismic reservoir characterization: from rapid relative inversion analysis to a fully coupled seismic-to simulation project (Tian et al., 2014). The inverse problem is highly non-linear (Thore, 2012) and seismic inversion for elastic properties from time-lapse data can be accomplished with varying levels of constraint, usually incorporating seismic amplitudes but also with time shifts between monitor and baseline (Zhan et al., 2017). The goal in many applications is to balance the often conflicting speed of the process against the accuracy and reliability of the results, enabling timely operational decisions to be made.
