EnKF analyses of two tornadic supercells using rapid-scan phased array radar data

The first Phased Array Radar (PAR) adapted to observe weather via its unique scanning capabilities became operational in spring 2004 as part of the United States National Weather Radar Testbed located in Norman, Oklahoma. Using a storm-scale Ensemble Kalman Filtering (EnKF) data assimilation methodology suggested by Dowell and Wicker (2009) PAR velocity and reflectivity data are assimilated to generate storm-scale analyses representing the thermodynamic and kinematic structure of two tornadic supercells.

The first tornadic supercell, on 29 May 2004 (previously presented at SLS 2008 by the authors), traversed central Oklahoma, producing 10 tornadoes and significantly impacting the greater Oklahoma City metropolitan area. This data set represents the first tornadic supercell ever observed using a volumetric rapid-scan 10 cm radar. Our current analyses assimilate radial velocity and reflectivity data every minute during a 90 minute period when the storm was less than 100 km from the radar.

The second tornadic supercell, on 13 May 2009, produced an EF-0 tornado in the Oklahoma City metropolitan area. The mesocyclone was located less than 40 km from the radar. This data set used a more sophisticated scanning strategy which includes data at the lowest two elevations every 43 seconds. This case also has increased spatial resolution in both azimuth and elevation compared to 29 May 2004. The cross-beam winds from the EnKF analyses will be verified by using the Doppler data from the Twin Lakes WSR-88D radar, as it is located nearly optimally at right angles to the PAR radar beam.

The 2004 case compared analyses produced from assimilating the PAR data approximately every minute to the analyses produced from assimilating PAR data approximately every five minutes. Results indicate that an increase in temporal resolution improves the analysis of the storm. The storm “spin-up” time (the time needed to generate a mature storm) is reduced approximately in half, from 20 minutes down to approximately 12 minutes. The 2009 case will show the impact of improved spatial and temporal resolution and compare those results to those from the 2004 case. We will also show the impact from the higher spatial and temporal resolution on the ensemble forecasts generated from the analyses.