Title: Which matters more in Hybrid Ensemble 4D-VAR, variances or correlations?

Authors: David D. Kuhl (Naval Research Laboratory, Washington DC)
Tom Rosmond (Science Application International Corp., Monterey, CA)
Craig H. Bishop (Naval Research Laboratory, Monterey, CA)
Elizabeth Satterfield (Naval Research Laboratory, Monterey, CA)

Recently there has been growing interest in the use of covariance models that linearly combine static and ensemble flow dependent covariances in 4D-VAR. The performance differences between the Hybrid and standard 4D-VAR can be partially attributed to a change in the forecast error variance field and partially attributed to a change in forecast error correlations. In addition, the sensitivity of Hybrid Ensemble 4D-VAR to ensemble size and/or the quality of the static covariance used in the Hybrid has not been thoroughly tested. Indeed, the theoretical results of Bishop et al's presentation suggest that a superior static covariance model for use in a Hybrid would be one based on a very large historical collection of ensemble perturbations. Here, using NRL's recently developed observation space Hybrid Ensemble 4D-VAR, a series of careful experiments are performed to determine the extent to which differences in performance are associated with either (a) changes in the forecast error variances or (b) changes in the forecast error correlations. Results from a low-resolution version of the system show a high degree of sensitivity to the specification of variances. By the time of the conference, we hope to be able to compare these low resolution results with higher resolution counterparts and to have also explored sensitivities to changes in ensemble size, and changing to a static covariance model based on historical ensemble perturbations.

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

Curator: Nikki Privé Last Updated: May 27 2011