Fluid Front History Matching Using an Iterative Ensemble Smoother

This study investigates the effect of approximating the front arrival time in the 4D seismic time-parameterization. This parameterization requires ensemble prediction of arrival times at the front location; this means that it is necessary simulating beyond the seismic acquisition time, ideally until all the cells at the front in all ensemble members have been flooded. Due to the computational effort, sometimes the exact front arrival time is approximated by an arrival time equal or larger than the simulation end time. The objective is to understand if an inexact ensemble prediction is giving worse results than a precise but computationally expensive ensemble prediction. Results show that the impact of not having a correct ensemble prediction for all the front cells in all realizations is not big. The quality of the results from the case where ensemble predictions have been approximated by an arbitrary big arrival times is comparable to the case where the simulation have been prolonged until all the front cells have been flooded (correct predictions). This leads to the conclusion that if the 4D seismic front arrival time parameterization is used, the use of a proxy or long simulation is not needed.