Importance of using ensemble estimated background error covariances for the quality of atmospheric ozone analyses

Abstract

AbstractAn ensemble method combined with a four‐dimensional variational data assimilation system is used to provide monthly estimates of the background error covariance matrix for global stratospheric and upper tropospheric ozone for the year 2008. The system is composed of the Mocage chemical transport model and the Valentina assimilation system. The ensemble was obtained from parallel analyses of perturbed data from the Microwave Limb Sounder (MLS) instrument. The monthly estimates of background error covariances have then been introduced in the assimilation suite. To assess the separate contribution of each of its components, a number of analyses were realized, using only some estimated components of the background error covariances and a basic model for the others. The evaluation is realized by comparing the analyses with independent ozone profiles (from ozonesondes) and total ozone columns (from the Ozone Monitoring Instrument). It demonstrates that using the estimated statistics compared to basic models for the background error covariance matrix globally slightly improves the analysis quality; however, using the estimated statistics more largely improves the analysis quality for special situations encountered in April and October. In these situations, the most important parameter for the analysis quality is the use of estimated correlations. Copyright © 2011 Royal Meteorological Society