The ARPS ensemble Kalman filter (EnKF) system is enhanced and applied to the 8 May 2003 Oklahoma City tornadic supercell storm case, for the assimilation of data from WSR-88D radars and high-resolution surface networks. The latter include the Oklahoma Mesonet with data available every 5 minutes.

Two one-way nested ensemble data assimilation (DA) systems are used, with a 1 or 1.5-km horizontal-resolution system nested inside a larger 3-km system. Perturbations representative of meoscale forecast errors are generated by analyzing perturbed pseudo soundings extracted from the first guess using the ARPS 3DVAR. Such perturbed analyses become the initial conditions of the 3-km ensemble. A similar procedure is used to construct perturbed boundary conditions for the 3-km system, which assimilates conventional observations, including surface mesonet data.

The nested storm-scale EnKF system starts from the 3-km ensemble at a later time, and assimilates additional level-II WSR-88D radar data at 5-minute intervals. Storm-scale perturbations are introduced into the initial conditions of this ensemble. The analysis of surface data is found to significantly improve the low-level storm environment, and the 3-km ensemble provides spread in the storm-environment that helps improve the performance of the EnKF system.

The results of the analyses and subsequent forecasts will be presented, and compared to the results from our previous successful study in which the ARPS 3DVAR and cloud analysis combination for radar data assimilation was used. The forecasts will be directly verified against radar observations. Progresses and outstanding problems will also be discussed.