Comparing and Combining the EnVar and EnKF Methods in a Limited-Area Deterministic Context
ECCC is currently developing a high resolution (grid spacing of 2.5 km) deterministic NWP system tailored for short-range forecasting based on an Ensemble-Variational (EnVar) data assimilation algorithm where the flow-dependent background-error covariances are provided by a companion ensemble Kalman filter (EnKF) system at a lower resolution (grid spacing of 10 km). To facilitate a large number of experiments, the same resolution as the EnKF is adopted for the deterministic forecast. This also permits the comparison of the forecast performance from the EnVar analysis and the EnKF ensemble-mean analysis in a deterministic context.
Various hybrid gain approaches (Houtekamer et al., 2018; Bonavita et al. 2015; Penny 2014), where a weighted mean of the EnKF ensemble-mean and the EnVar analysis is used to recenter the EnKF analysis ensemble, were tested. Their respective impacts on the forecast performance of both the EnKF ensemble mean analysis and the EnVar analysis will be shown. The benefits of using background-error covariances derived from a limited-area EnKF instead from the operational global EnKF will also be discussed.
Dr. Jean-Francois Caron Environment and Climate Change Canada (ECCC)