Radar Data Assimilation using WRF 3DVAR: Radial velocity or reflectivity—which has a larger impact?

In recent years, demand has increased for high-resolution accurate forecasts of convective weather. NCAR's WRF-ARW model has played a critical role in real-time demonstrations of explicit convective forecasts showing skill in convective forecast guidance out to 36 hours. However, NWP models require further refinement to enable accurate forecasts of the timing and location of high-impact weather in the first 0-12 hours. Accurate specification of the initial state down to the convective scale through assimilation of high-resolution observations such as Doppler radar and densely spaced surface observations is one of the keys to improve NWP forecasts of high-impact weather events in the nowcasting range. Although the 4-dimensional variational (4DVAR) technique has been shown to have a great potential for radar data assimilation, the technique is still too expensive for day-to-day operational high-resolution analysis. An alternative is to use the less sophisticated 3DVAR technique. The 3DVAR radar data assimilation system was developed for the Weather Research and Forecasting (WRF) model and evaluated through case studies in the past. To further evaluate the capability and limitations of the WRF 3DVAR radar data assimilation system on multiple cases over consecutive days, a retrospective study was recently conducted using one-week IHOP (International H2O Project) data. 25 WSR-88D radars over the central United States were assimilated into the WRF model via the 3DVAR technique. Experiments with different initialization methods, including cold-start with large-scale model analysis, 3DVAR with rapid update cycles (RUC) without radar observations, 3DVAR with RUC and radial velocity or reflectivity and with both were conducted. The relative impact of the radial velocity and reflectivity on initialization and subsequent forecasts are examined through the statistical verification over the entire week as well as case studies. It is found that the reflectivity data assimilation accompanied by a cloud analysis results in significant positive impacts on the precipitation forecasts. It is also found that although the radar data assimilation improves the overall skill score, the results are case dependent. Diagnostic analysis is being performed to understand the convective processes that contribute to the success/failure of the radar data assimilation.