Performance assessment of an iterative ensemble smoother with local analysis to assimilate big 4D seismic dataset applied to a complex pre-salt-like benchmark case

The use of 4D seismic (4DS) (or time-lapse seismic, TLS) in data assimilation (DA) makes the process more complex due to the higher amount of data to be assimilated, requiring more robust methods and better computational resources (processing capacity and memory). The development and application of permanent seismic monitoring technologies have increased in the last years, improving the overall 4D seismic quality, in terms of signal resolution and repeatability. However, a massive amount of data is generated from the multiple monitors, making the incorporation of 4DS data in the DA process much more complex. Therefore, robust DA methods capable of dealing with huge amount of data effectively and efficiently are essential. This paper aims to assess the performance of an iterative ensemble smoother method, named Subspace Ensemble Randomized Maximum Likelihood with a local analysis (SEnRML-LA), to assimilate a big data set. The method was applied in a challenging pre-salt-like benchmark case with eight seismic surveys, one base, and seven monitors. The 4DS data are the impedance ratios (between two consecutive monitors) in 15 seismic horizons, totaling 105 maps to be assimilated. To our best knowledge, this is state of the art in terms of practical applications in data assimilation. It was possible to assimilate all the data simultaneously: the 105 horizons for the 4DS data and the wells’ production and pressure data. The data assimilation was successful in terms of results quality and method performance. We also ran a case assimilating only well data for comparison purposes.