The ARPS 3DVAR system and its complex cloud analysis have been used to assimilate data from a mesoscale convective system in which a line-end vortex (LEV) developed and moved through much of southwest/central Oklahoma on 8-9 May 2007. This LEV spawned at least five mesovortices, two of which were tornadic. The MCS was observed by several WSR-88D radars, in addition to four X-band radars comprising the Engineering Research Center for Collaborative Adaptive Sensing of the Atmosphere (CASA) integrated project I (IP-1) network. The ARPS is used within frequent (5 min) assimilation cycles and for the subsequent forecast. Experiments are conducted using two one-way nested grids, with 2-km and 400-m horizontal grid spacing, respectively. The impact of assimilating radial velocity (Vr) and reflectivity data from CASA radars in addition to the WSR-88Ds is examined. Assimilation window lengths of 60 and 80 min are used for the 2-km and 400-m grids, respectively. An additional set of experiments examines the sensitivity of the model forecast to assimilation window length and end time on the 400 m grid.

Results suggest that the impacts of CASA data are clearly present for this tornadic convective system, especially on the 400-m resolution grid. On the 2-km grid, all experiments that assimilate radar data have a highly accurate forecast evolution of the MCS and LEV, with a slight improvement in the forecast LEV track in experiments where CASA Vr data are assimilated. Much more significant improvements in the forecast of mesovortices in the case are found from the assimilation of CASA Vr data on the 400-m resolution grid. These improvements are caused by the ability of the CASA radars to better observe the low-level shear profile, which is critical in the development and intensification of strong, long-lived mesovortices. A similar, but not as large, impact on the accuracy of the low-level shear profile is realized when 5-min data from the Oklahoma Mesonet are assimilated instead of CASA Vr data on the 400-m grid.

Preliminary results from assimilation window length experiments suggest that after 40 min of data assimilation, additional cycles add little to quality of the analysis and forecast evolution of the mesovortices. Experiments that vary assimilation window end time show a large sensitivity to earlier end times, as highly degraded forecasts of mesovortices are obtained when assimilation end time is reduced by as little as 10 min. Most recent analysis of the cause of this forecast degradation will be presented.