Optimization of satellite radiance assimilation at the Canadian Meteorological Center

The Canadian Meteorological Center (CMC) is currently planning to implement a hybrid ensemble-variational (EnVar) data assimilation system. The ensemble approach provides a flow dependent background covariance matrix. The analysis remains variational, but is no more done in 4D-var. Several elements of radiance assimilation are reviewed in that context. The choice of bias correction (BC) predictors has very significant impacts, notably in the Tropics. It is found beneficial to use the same predictors for all variances (microwave and infrared). In order to give more weight to radiosonde and GPS radio-occultation data in the BC procedure, BC coefficients are updated every 6-h from an off-line analysis which does not assimilate radiances. The number of assimilated AIRS (149) and IASI (124) channels is nearly doubled. The observation error is defined as a function of innovation statistics pertaining to each channel. Assimilation cycles allow assessing the impact of these changes against the currently operational 25-km global 4Dvar system, as well as against independent analyses available from ECMWF.