Assimilation of Radar Data and Short-Range Prediction of Thunderstorms using 3DVAR, Cloud Analysis and Ensemble Kalman Filter Methods

Results assimilating radar data using 3DVAR+cloud analysis and using an ensemble Kalman filter method into a storm-scale NWP model, for several tornadic thunderstorm cases are sythethized. With both procedures, data assimilation cycles at intervals comparable to the radar volume scan frequencies are used. With the former procedure, conventional as well as radar radial velocity data are analyzed using a 3DVAR scheme while the radar reflecitivty data are assimilated via a complex cloud analysis procedure that produces the analyses of cloud and hydrometeteo fields and perform in-cloud tempearture adjustments. This procedure is computationally efficient and can be implemented for real-time storm-resolving high-resolution predictions. It will be shown to produce reasonable analyses of thunderstorms for several cases, and reasonable predictions of individual storm cells for up to 3 hours can be obtained from the analyzed state. Significant adjustment usually occurs, however, immediately following each analysis time due to the lack of good consistencies of the analyzed state with the model dynamics and physics. Various analysis and assimilation configurations with this procedure and their impact on the analysis and forecast will be discussed.

Results using a theoretically suporior ensemble Kalman filter method (EnKF), for at least one of these cases will also be shown. In general, the EnKF method is able to produce much better analyses, especially of the state variables not directly observed. The accuracy of analysis and prediction can be significantly impacted by the model errors and the errors in the definition of the storm environment. Results using a combination of the above two procedures, in which the first EnKF analysis cycle starts from the 3DVAR+cloud analysis will also be presented, so will be selected results from realtime short-range forecasts that assimilate radar data.