Analysis of the forecast improvement from radar-data assimilation within the RUC, Rapid Refresh and HRRR

Since November 2008, the NCEP operational Rapid Update Cycle (RUC) has included a diabatic-digital filter initialization (DFI)-based radar reflectivity assimilation procedure. In this procedure, a latent heating-based temperature tendency is derived from 3D radar reflectivity mosaic data and applied during the diabatic digital filter initialization part of the RUC or RR to induce mesoscale air-flow patterns. The procedure has been ported to the corresponding Rapid Refresh (RR) mesoscale analysis and prediction system components (the Gridpoint Statistical Interpolation – GSI – analysis and the Advanced Weather Research – ARW – model. The RR system is in final testing for operational implementation at NCEP in September, 2011. The real-time High-Resolution Rapid Refresh (HRRR) has been run as a one-way nest within the RUC (through April of 2011) and recently as a one-way nest within the RR. The HRRR forecasts benefit from the RUC / RR radar-DFI assimilation procedure.

Based on statistical evaluation of controlled retrospective forecasts, we will quantify the impact of the radar-data assimilation on the RUC, RR and HRRR forecasts. Particular emphasis will be given to evaluating the forecast length, diurnal time frame, and regional aspects of the forecast improvement. This quantitative analysis will be supplemented by detailed case study analyses to illustrate various aspects the radar-data assimilation impact, including an examination of 1) how the application of the radar-DFI impacts the wind and thermodynamic fields within the mesoscale models, 2) how the radar-DFI impact projects onto the grid-scale and parameterized precipitation systems within the cycled mesoscale models (RUC and RR), 3) how the radar-DFI impact projects onto the (uncycled) 3-km HRRR forecasts (storm spin-up, etc.), and 4) what impacts the radar-DFI has on the mesoscale environmental fields for both the mesoscale RUC and RR and the HRRR models. We will also document forecast skill sensitivity to variations in the radar-DFI latent heating strength, hydrometeor and moisture specification, and other assimilation details.

Finally, we will show some preliminary results from assimilation of radial velocity data within the 13-km RR. A companion presentation (Alexander et al.) will describe ongoing work on more advanced radar-data assimilation techniques for the HRRR, including sub-hourly cycled application of radar-DFI at 3-km and use of radial velocity data at 3-km.