Correlated interchannel observation error statistics for radiances: estimation and impact in a near operational context at Environment Canada

Model background and observation error statistics are key inputs of modern data assimilation systems used in Numerical Weather Prediction. For a long time, it was often assumed in operational context that the observation covariance error matrix is diagonal. The neglected errors correlations were, in principle, accounted for indirectly via for example data thinning or error inflation. In the case of radiances from vertical sounders, the advent of hyperspectral infrared sounders such as AIRS (Atmospheric Infrared Sounder), IASI (Infrared Atmospheric Sounding Interferometer) and the recently launched CrIS (Cross-track Infrared Sounder) with their thousands of channels represented an important challenge for the data assimilation community. Recently, inter-channel observation error covariances matrices were estimated for these instruments by various authors (e.g. Garand et al. 2006, Bormann et al. 2010) using different methods which gave consistent results. The purpose of this work is first the estimation of radiances observations error statistics including inter-channel correlations and then the study of the impact of their use in a near operational context in Environment Canada's Envar global assimilation system.