A diabatic digital filter initialization (DFI)-based procedure for assimilating radar reflectivity data within the hourly updated Rapid Update Cycle (RUC) was developed in 2007 and is now in final testing at NCEP, with a planned operation implementation in Fall of 2008. This will mark the first operational use of WSR-88D mosaic reflectivity data in an NCEP operational model. Ongoing evaluation of real-time parallel RUC forecast cycles (at ESRL GSD and NCEP EMC) continues to indicate significant improvements in the prediction of mesoscale precipitation systems compared to the operation RUC without the radar assimilation procedure.

Current radar assimilation work at ESRL GSD has focused on two areas. First, we are using hourly updated initial forecast fields from the radar assimilating RUC to initialize a 3-km explicit convection resolving nest, known as the High Resolution Rapid Refresh (HRRR). Run each hour out to 12-h over domain covering a region of high aviation impact (Chicago to New York City corridor), the HRRR has yielded further significant improvement for short-range prediction of precipitation systems and other mesoscale features.

A second focus area has been porting of the radar assimilation procedure to the Rapid Refresh (RR) system. This has required changes to both the Gridpoint Statistical Interpolation (GSI) package (used for the data assimilation portion of the RR) and the WRF ARW system (used for the model forecast portion of the RR). For the data assimilation part, a generalized cloud analysis has been incorporated into GSI, in which NSSL radar reflectivity mosaic data are used to compute a latent heating-based temperature tendency. Within the WRF model, the latent heating-based temperature tendency replaces the temperature tendency from the cumulus scheme and the explicit microphysics in the forward integration portion of the DFI. The ability of the assimilation package to accommodate lightning data (converted to proxy reflectivity data) will allow the RR to initialize ongoing convection over radar void regions such as Alaska (see details in Hu et al, 4MALD).

At the conference, we will describe the work in all of these areas and provide quantitative verification and case studies examples to illustrate the forecast improvement from the radar assimilation for the RUC, RR, and HRRR model forecasts.