Seung-Jong Baek;
Joël Bédard;
Peter Houtekamer
Meteorological Research Division, Environment and Climate Change Canada
The 8th EnKF Data Assimilation Workshop
Recentering of a regional ensemble Kalman filter
Talk:
REnKF_hybrid.pdf
A global ensemble Kalman filter is running at Environment and Climate Change of Canada not only to produce the initial conditions for the global and regional ensemble predictions (GEPS and REPS) but also to provide the ensemble of short-range forecasts to the ensemble variational (EnVar) data assimilation suite in the global and regional deterministic predictions (GDPS and RDPS) for the estimation of flow dependent background covariances.
For a better representation of the flow dependent covariances, we are examining two different recentering approaches for the ensemble analyses in the context of an experimental regional EnKF. One approach is to couple the regional EnKF and the cycling EnVar via a control member around which the ensemble of analyses are recentered. In the other approach, an EnVar takes the mean background trajectory of the ensemble as unique background trajectory to arrive at a unique analysis. The ensemble of analyses are then recentered around this unique analysis. In both approaches, the EnVar assimilates more observations than the EnKF and employs the hybrid covariance approach where the static and flow dependent components of the covariances are added using appropriate weights. In this presentation, the performance of the REPS will be compared for the two approaches.
A global ensemble Kalman filter is running at Environment and Climate Change of Canada not only to produce the initial conditions for the global and regional ensemble predictions (GEPS and REPS) but also to provide the ensemble of short-range forecasts to the ensemble variational (EnVar) data assimilation suite in the global and regional deterministic predictions (GDPS and RDPS) for the estimation of flow dependent background covariances.
For a better representation of the flow dependent covariances, we are examining two different recentering approaches for the ensemble analyses in the context of an experimental regional EnKF. One approach is to couple the regional EnKF and the cycling EnVar via a control member around which the ensemble of analyses are recentered. In the other approach, an EnVar takes the mean background trajectory of the ensemble as unique background trajectory to arrive at a unique analysis. The ensemble of analyses are then recentered around this unique analysis. In both approaches, the EnVar assimilates more observations than the EnKF and employs the hybrid covariance approach where the static and flow dependent components of the covariances are added using appropriate weights. In this presentation, the performance of the REPS will be compared for the two approaches.
Thanks to McGill's Atmospheric and Oceanic Sciences department for hosting this web-site.