Title: A smoother-based strategy to estimate system error covariances

Authors: Ricardo Todling (NASA/GMAO)

Starting from sequential data assimilation arguments, this presentation discusses how the use of residual statistics from filtering combined with smoothing residuals allow inferring components of the system (model) error covariance matrix that project onto a dense observing network. The residuals relationships involving the system error covariance matrix are similar to those available to derive background, observation, and analysis error covariance information from filter residual statistics. An illustration of the approach is given for two low-dimensional dynamical systems: a linear harmonic oscillator and the nonlinear Lorenz (1995). Recast of the sequential approach into the language of variational data assimilation leads to a procedure for estimating system error in 4D-Var that involves two interconnected experiments: one being a cycling 6-hour 4D-Var, and another a non-cycling 12-hour 4D-Var started every 12-hour from the corresponding 6-hour 4D-Var solution. Preliminary results using the European Centre for Medium-Range Weather Forecasts (ECMWF) 4D-Var system are also presented.

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

Curator: Nikki Privé Last Updated: May 27 2011