Ensemble assimilation of stratospheric temperature and ozone observations in a chemistry-climate model

The focus of this presentation is to study the applicability and the advantages of ensemble data assimilation in the under-observed stratosphere. In particular, we investigate whether the limited number of ensemble members can accurately produce spatial, temporal and cross-species background error covariances in order to propagate information from the rare observations to the full state system. The study is performed using ensembles of simulations from a three-dimensional chemistry-climate model (IGCM-FASTOC), assimilating synthetic satellite ozone or temperature retrievals that mimick ENVISAT-MIPAS observations, in the context of a "perfect twin" experiment.

It is found that our ensemble Kalman filter constrains well the stratospheric chemistry and dynamics, given proper localization of the background error covariances, even with a limited amount of observations. The cross-species background error covariances are critical in this achievement, particularly the ones representing the advection of the dynamical and chemical variables. However, significant but non-deleterious noise is observed in the analysis ensembles, in contrary to other ensemble data assimilation systems, where an under-variability is typically observed in the ensembles. An interpretation of these results will be presented.