Title: Estimates of analysis error, forecast error, and predictability derived from the adjoint of 4D-Var

Authors: Andy Moore (Dept of Ocean Sciences, University of California at Santa Cruz, USA)
Hernan Arango (Institute of Marine and Coastal Sciences, Rutgers University, USA)
Gregoire Broquet (Laboratoire des Science du Climat et de l'Environnement, France)

The adjoint of the Regional Ocean Modeling System (ROMS) 4D-Var data assimilation system has been used to compute the expected analysis errors and forecast errors in linear functions of the California Current (CC) circulation. ROMS 4D-Var has been run sequentially for the CC for a period spanning several years, and error estimates will be presented for various linear functions that characterize different aspects of the CC coastal upwelling circulation, namely upper ocean transport and heat content. Our methodology is based on the recently proposed approach of ensemble 4D-Var for estimating the analysis error covariance. However, when using the adjoint of the entire 4D-Var system, the explicit generation of an ensemble of 4D-Var analyses is circumvented. In addition, the adjoint of 4D-Var can be used to quantify the impact of each observation on the predictability of the circulation resulting from data assimilation. Given that the analysis and forecast error covariance estimates are intimately related to the prior error covariances, a series of consistency checks will also be presented which allows the efficacy of the prior covariances to be assessed.

GMAO Head: Michele Rienecker Global Modeling and Assimilation Office NASA Goddard Space Flight Center

Curator: Nikki Privé Last Updated: May 27 2011