We present an iterative approach for quasi-continuous time-lapse seismic reservoir monitoring. This approach involves recording sparse data sets frequently, rather than complete data sets infrequently. In other words, it involves acquiring a completely sampled baseline data set followed by sparse monitor data sets at short calendar-time intervals. We use the term “sparse” to describe a data set that is a small fraction of what would normally be recorded in the field to reconstruct a high-spatial-resolution image of the subsurface. Each monitor data set could be as little as 2% of a single, complete conventional data set. The series of recorded time-lapse data sets is then used to estimate missing, unrecorded data in the sparse data sets. The approach was tested on synthetic and field crosswell traveltime data sets. Results show that this approach can be effective for quasi-continuous reservoir monitoring. Also, the accuracy of the estimated data increases as more sparse data sets are added to the time-lapse data series. Finally, a moving estimation window can be used to reduce computational effort for estimating data.
Evaluating parameters for ligand-based modeling with random forest on sparse data sets
