Ensemble retrospective assimilation for a global nonlinear shallow-water model

Ensemble filtering is becoming a potential data assimilation alternative strategy to variational and standard-like Kalman filtering methods. We have recently showed that an extension to the ensemble filter, namely the ensemble retrospective data assimilation, presents a viable approach to implementing fixed-lag smoothers, meant to further improve results from its filter counterpart. Our preliminary results were obtained based on an idealized strongly nonlinear dynamical model. Relevant questions associated to the ensemble retrospective assimilation methodology still remain to be answered for more realistic dynamics.

In this presentation discussed the results of an implementation of an ensemble retrospective assimilation system based on a global nonlinear shallow-water model. This provides a testbed for addressing typical questions related to primitive equation models, such as on balance and initialization, as well as questions arising on the smoother performance when an approximate algorithm as the one used here is taken in to consideration.