Title: Numerical solution for a time-parallelized formulation of 4D-Var

Author: Selime Gurol (ECMWF)
Mike Fisher (ECMWF)
Serge Gratton (ECMWF)

In order to allow 4D-Var to remain computationally viable on next-generation computer architectures, parallelization in the time dimension is necessary. It has been recently shown that a saddle-point formulation of 4D-Var allows such parallelization. Therefore, it represents an important step towards greater numerical efficiency since conventional 4D-Var requires many sequential computations. In this study, we will focus on the numerical solution of the saddle formulation of 4D-Var by using cost-effective preconditioners. Numerical experiments will be presented from Object-Oriented Prediction System (OOPS) platform developed by ECMWF.

GMAO Head: Steven Pawson Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: Feb 9 2015