Hybrid variational/ensemble data assimilation for WRF

The accuracy of analyses produced by modern data assimilation systems depends strongly on the precision of forecast error covariances specified as input. In reality, these errors are typically synoptically dependent, anisotropic, and inhomogeneous, and should be represented as such in data assimilation system.

Two of the most popular ways to implement flow-dependence error covariances in modern data assimilation are a) 4D-Var, and b) Ensemble filters. Both techniques have shown considerable promise in recent years, but also have drawback in terms of development/maintenance cost, ensemble size required, suitability for highly nonlinear regimes, etc.

In this talk, we present a "hybrid" approach to advanced data assimilation that combines 4D-Var with an Ensemble Transform Kalman Filter (ETKF). Using the community "WRF-Var" data assimilation system, the ensemble of forecasts is used to provide flow-dependent error covariances via additional constraints within the variational framework. By varying the relative weight assigned to the flow-dependent (ensemble) and climatological components of forecast error, it is possible to find an optimal configuration that is superior to either the full variational or ensemble system for the size of ensemble likely to be possible in operational NWP in the next few years. An additional benefit of the technique used here is the relative simplicity of implementation into existing 3/4D-Var systems. Initial results from WRF applications will be presented.