Comparisons between analysis and prediction of convection using conventional and rapid-scan weather Doppler radar data

During the past 9 years, the National Severe Storm Laboratory (NSSL) has been collecting Doppler radar data using a new prototype next-generation operational radar system. NSSL's 10 cm Multifunction Phased Array Radar (MPAR) can electronically scan a 90 degree sector both horizontally and vertically which enables it scan the entire volume between 4-6x more rapidly than typical WSR-88D volumes. While the faster MPAR scanning has already shown considerable promise toward improving severe weather warnings (Heinselman et al. 2008) there is also considerable interest in demonstrating the use of such a capability in convective-scale numerical weather prediction (a.k.a., Warn on Forecast, Stensrud et al. 2009). While an MPAR observing systems experiment (Yusouff and Stensrud 2010) suggested that rapid-scanning would improve convective-scale analyses and forecasts in ensemble Kalman filter experiments, benefits from using MPAR data for storm-scale NWP have been somewhat more ambiguous. Two cases will be presented attempted to highlight the benefits and difficulties using rapid-scan radar data. The first case occurred during the early evening of June 14 2011 when Norman, OK was struck by a wet downburst wind event which produced damage across a ~7 km wide and ~20 km long swath in the northern and eastern portions of the city. The storm formed in an environment characterized by deeply mixed adiabatic boundary layer, weak vertical wind shear. In this case, MPAR was able to collect 19 elevation scans in a 90 degree sector every minute during a one hour period before the severe storm went over the radar site. The second case is a classic tornadic supercell from 24 May 2011 southwest of Oklahoma City, OK. The storm environment was typical for significant tornado outbreaks, having large CAPE and significant low- and mid-level wind shear. The storm was sampled for over an hour with continuous one minute volume scans during the lifetime the storm's EF-5 tornado near Piedmont OK. In both cases the temporal and spatial density of the radar data, relative to the model grid, needs to be carefully considered prior to its use within the data assimilation system (Lu and Xu 2009, Lu et al. 2011). At current model grid resolutions (Δx ~ 1 km) the background covariance is too smooth and too coarse to use effectively even conventional radar observations. Therefore some form of “superobbing” must be applied prior to use in the assimilation system. It is hypothesized that for rapid-scan data, higher temporal resolution may require temporal superobbing and/or some form of asynchronous data assimilation in order to fully realize the benefits from the MPAR radar in storm-scale NWP.