The New 4D-En-Var Regional Deterministic Prediction System at the Canadian Meteorological Center

The regional deterministic prediction system (RDPS) at the Canadian Meteorological Center will be replaced in the near future by the 4D-En-Var scheme, in which the background error covariances are a combination of climatological covariances and flow-dependent covariances derived from an ensemble-Kalman-filter global prediction system. The new approach is computationally less expensive than 4D-Var (currently operational) and has shown to be a promising technique in the context of global data assimilation. The RDPS uses a limited-area domain covering all of North America and a horizontal grid spacing of 10 km. Here we discuss the final stages of the development of this scheme, shortly before introduction into operations. Specifically, we show a comparison of the forecasting skill between the operational system and the new 4D-En-Var scheme, as well as the impact of several improvements related to the treatment of observations and the addition of new observation sources such as ground-based GPS.