Evaluation of radar data assimilation methods used to initialize precipitation systems within the RUC, Rapid Refresh, and High Resolution Rapid Refresh (HRRR) models

A diabatic digital filter initialization (DFI)-based procedure for assimilating radar reflectivity data was developed in 2007 for the Rapid Update Cycle (RUC) and was adapted for use with Rapid Refresh (RR) model in 2008. Fields from the RUC and RR (both with the radar assimilation procedure) have been used to initialize real-time High Resolution Rapid Refresh explicit convection resolving model forecasts (Weygandt et al, this conference). Despite a relatively simplistic design and small computational costs (compared to advanced schemes such as 4DVAR and EnKF), the DFI-based procedure has produced impressive results for both the parameterized convection scale applications (the 13-km RUC and RR) and for convection resolving models (the 3-km HRRR) initialized from the RUC and RR. It is important to note that for the HRRR, significant improvement has occurred for forecasts at a 3-km scale from assimilation at a 13-km scale. The key to this forecast improvement at both the parameterized convection scale and the explicit convection scale has been project the reflectivity information on to the divergent component of the model wind field. This is done through a radar reflectivity-based latent heating rate that is applied within the DFI.

In this presentation we will describe the radar reflectivity assimilation algorithm as it is implemented for both the RUC and the Rapid Refresh, and illustrate the forecast improvement from the radar assimilation. Next we will document the related improvement in HRRR convective forecasts from use of the RUC / Rapid Refresh fields with the radar assimilation. We will provide details on a minor, but important, modification to the HRRR initialization procedure that provides the most direct transfer of radar information from the RUC / RR grid to the HRRR. We will also illustrate in detail how this radar information evolves from the RUC / RR scale (13-km) to the HRRR scale (3-km) during the first few hours of the HRRR forecast.

More recent work with Rapid Refresh system has focused on evaluating the impact of hourly assimilation of radial velocity data and consideration of adjustments to the assimilation procedure to maximize the forecast improvement from the radial velocity data. In the final portion of the talk will provide an update on this ongoing work