Potential of iterative ensemble methods for solving the nonlinear state and parameter-estimation problem
Iterative ensemble smoothers improve the assimilation results in cases with modest nonlinearity. This presentation explains the properties of two much-used methods, the Iterative Ensemble Smoother (IES by Chen and Oliver, 2012, 2013) also named or Ensemble Randomized Maximum Likelihood, and the Ensemble Smoother with Multiple Data Assimilation (ESMDA by Emerick and Reynolds, 2013). We explain the derivation and formulation of the two methods and discuss their properties with nonlinear dynamics. We show that we can use the techniques for sequential data assimilation as well as for parameter estimation and that we can easily account for model errors. Finally, we present a new stable and numerically efficient formulation of IES that exploits the rank of the ensemble. For nonlinear problems, we claim that iterations will give a more significant improvement than just increasing the ensemble size.
Prof. Geir Evensen NORCE