To improve surveillance of severe weather, the Center for Collaborative Adaptive Sensing of the Atmosphere (CASA) was established in 2003 by the National Science Foundation to develop low-cost, high-spatial-density and dynamically adaptive networks of Doppler radars for sensing the lower atmosphere. The first CASA testbed was deployed in southwest Oklahoma in 2006 and con-sists of four scanning polarimetric Doppler radars located on average 25 km apart with maximum ranges of 40 km.

The high spatial and temporal resolutions of the CASA radar data and the much improved low-level data coverage provide an unprecedented opportunity for producing detailed low-level wind analyses for realtime warning and prediction applications. Such analyses are further aided by the availability of dual- or even triple-Doppler wind coverage within parts of the network, and of the 5-minute-interval surface observations from the Oklahoma Mesonet. Towards this goal, the gen-eral-purpose ARPS 3DVAR analysis system was adopted to produce low-level wind analyses in realtime at 5-minute intervals during the spring of 2008, and the analysis products were presented for experimental use and evaluation at the NOAA Hazardous Weather Testbed located on the University of Oklahoma campus, in collaboration with the CASA End User Team. The project also cuts across all three research thrusts of CASA, namely, the analysis and prediction, sensing and distribution thrusts.

Specifically, from April through June of 2008, data from four CASA radars and two WSR-88D radars as well as those from the Oklahoma Mesonet were collected and analyzed using the 3DVAR. The NCEP NAM model forecasts interpolated to the analysis times were used as the analysis back-ground. Experiments using both CASA and WSR-88D radar data are compared to those using only WSR-88D radar data in realtime. Results indicate that using CASA radar data together with WSR-88D data can greatly improve the quality of the low-level wind analysis. The gridded wind analyses significantly increased the usability of the radar wind observations, especially when they are collected using adaptive scanning strategies. Several new features recently incorporated into the system include additional quality control, equation constraints, and a new anisotropic filter. Pre-liminary experiments using these new features show improved, more accurate wind analysis. Addi-tional tests will be performed on several selected cases collected during the 2008 spring experiment period and results will be reported at the conference.