A Comparison of EnKF and EnPAT Inverse Methods: Non-Gaussianity

Abstract

The EnKF has been extensively used for real-time data assimilation in areas such as reservoir/groundwater modeling. One of the big challenges of the EnKF is how to handle the non-Gaussianity of aquifer properties, particularly for channelized aquifers where preferred flow conduits are encountered. EnPAT is a pattern-based inverse method and was developed to deal with the non-Gaussianity of model updating. In this work, we compared the EnKF and EnPAT on a benchmark example. The results show that EnPAT can better reproduce the curvilinear geological features and thus has a better transport prediction.