The Development of a Hybrid Enkf-3dvar Algorithm for Storm-scale Data Assimilation

Many studies have been performed during the past decade aiming to use Doppler radar data for initializing cloud-resolving models. At the Center for Analysis and Prediction of Storms (CAPS), the ARPS Data Analysis System (ADAS, Brewster 2003a, b), with incremental analysis updating capabilities, was developed as the first step towards assimilating radar data and other conventional and remotely-sensed data into a nonhydrostatic NWP model, the ARPS (Xue et al. 2000; 2001). A limitation of ADAS is that observations that differ from the analysis variables cannot be directly analyzed and are handled only by generation of pseudo-observations (e.g., via single-Doppler velocity retrievals; Weygandt et al. 2002a, b). Examples of such observations include radar radial velocity and reflectivity, GPS precipitable water, and satellite radiances. All of these data are very important to storm-scale data assimilation. To more effectively use these data, advanced DA techniques, including variational and ensemble Kalman filter methods are needed. In the context of large-scale hydrostatic flows, the 3DVAR method of analysis has reached a considerable state of maturity at operational NWP centers (Derber et al. 1991; Parrish and Derber 1992; Andersson et al. 1998; Courtier et al. 1998; Rabier et al. 1998; Wu et al. 2002, Barker et al. 2004). Non-conventional data, including those from satellite and radar, can be directly analyzed by 3DVAR. In current operational 3DVAR systems, analysis increments are often divided into balanced and unbalanced parts, with the balanced parts often linked together via statistically derived balance relations. For example, in the NCEP system (e.g., Wu et al. 2002, Purser et al. 2003a, b), the temperature increment is a sum of the unbalanced and balanced parts, with the latter computed from the stream_____________________________