Title: Towards model reduced 4D-Var in application to GMAO’s data assimilation system

Author: Joanna S. Pelc (NASA/GSFC)
Ricardo Todling (NASA/GSFC)

The NASA Global Modeling and Assimilation Office (GMAO) recently implemented operationally a 3D-Var-hybrid version of its data assimilation system (DAS). This system will rapidly progress towards a 4D-EnsVar. However, the machinery to exercise conventional 4D-Var is in place and it is desirable to have a comparison of these two 4D approaches and examine how they perform in the Goddard Earth Observing System DAS. As conventional 4D-Var is computationally more expensive than the data assimilation methods currently available at GMAO, the purpose of this work is to implement the model reduced 4D-Var approach (Vermeulen and Heemink, 2006; Pelc et al. 2012) in order to explore the possibilities of making the minimization problem more computationally affordable. This is addressed by targeting the implementation of the adjoint of the tangent linear approximation of the original model. Using an ensemble of simulations from the original model, a proper orthogonal decomposition is used to build a reduced model. Then the adjoint of the original model is approximated by the adjoint of the tangent linear approximation of the reduced model essentially replacing a high-dimensional minimization problem with a problem that is orders of magnitude smaller. This presentation will focus on preliminary results on the evaluation of the performance of the tangent linear and adjoint models approximated in the reduced space. The results will be compared with the currently available adjoint of the original model, Goddard Earth Observing System.

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

Curator: Nikki Privé Last Updated: Feb 9 2015