An Efficient Approach for Quasi-Continuous Monitoring of CO2 Sequestration Reservoirs

We present an approach for quasi-continuous, time-lapse reservoir monitoring with sparse seismic data. The approach is well suited for monitoring geologic reservoirs such as those used for CO2 sequestration, where a good knowledge of the fate of the stored CO2 is critical. Changes in the seismic properties of the reservoir make seismic monitoring appropriate for determining the fate of sequestered CO2 after injection. The approach we present takes advantage of the small changes in the seismic property of the geologic reservoir that are expected to occur in a small time interval. We efficiently reconstruct high temporal resolution images of the reservoir without demanding a larger data volume than that required using conventional time-lapse seismic monitoring strategies. Conventional time-lapse monitoring approaches emphasize spatial resolution over temporal resolution. With conventional approaches, to improve temporal resolution, dense seismic datasets must be acquired frequently, making the monitoring project uneconomical since a high spatial density dataset has to be recorded frequently in time. With our approach, we acquire spatially sparse data at small time intervals (frequently). A spatially sparse dataset refers to that dataset which is a small fraction (as little as 10%) of what would be acquired to reconstruct a high spatial resolution geophysical image of the subsurface. The high spatial resolution obtained by the proposed approach occurs because unrecorded data are estimated from data recorded before the date of interest and data recorded after the date of interest, i.e., from past and future data. The estimated data combined with the acquired sparse data produce the dense data necessary for reconstructing high spatial resolution images of the reservoir. With high temporal and spatial resolution, early detection of important reservoir changes is more likely to occur. We use stochastic properties of the seismic data from the recorded datasets as prior information. The dense baseline dataset, acquired before the injection of the CO2 into the reservoir provides prior information about the background geology, which stays the same throughout the life of the reservoir. We use prediction error filters (PEFs) as the estimation tool. To handle the inhomogeneity and non-stationary property of the time-lapse seismic data, we allow the PEFs to vary smoothly in space and time. Because a consistent time interval between data acquisition campaigns in a reservoir monitoring project is not guaranteed, future work involves incorporating a parameter and time varying weighting function in the estimation process.