Dual-Doppler wind retrieval is an invaluable tool in the study of convective storms. The resultant 3-D wind estimates both illuminate storm kinematics and permit dynamical retrievals, including parcel trajectory and vorticity budget calculations. Unfortunately, the nature of the errors in these kinematical and dynamical analyses is not thoroughly known, making it difficult to properly assign confidence in ensuing inferences about storm behavior. Using an observing system simulation experiment (OSSE) framework, this study characterizes these errors for the case of a supercell thunderstorm observed at close range by two Doppler radars. Synthetic radar observations are generated from a high-resolution numerical supercell simulation, then input to a three-dimensional variational (3D-VAR) dual-Doppler wind retrieval technique. Parcel trajectories and vorticity analyses are then computed from the retrieved storm-scale wind field. The sensitivity of the analyses to the dual-Doppler retrieval settings, errors in the hydrometeor fall speed parameterization, and the radar cross-beam angle and scanning strategy are examined.

Imposing the commonly adopted assumptions of spatially-constant storm motion and steady-state flow in the moving frame produces large errors in our analyses, especially at higher levels. These errors are substantially mitigated using shorter volume scan times, even at the cost of larger elevation or azimuth angle increments. Using a 30° cross-beam angle, typical of close-range mobile radar deployments, leads to systematic, locally severe underestimation of the wind fields. Near-ground parcel trajectories initiated around the main updraft and rear-flank downdraft are generally qualitatively accurate, with larger trajectory errors occurring at middle levels due to unaccounted flow-unsteadiness. In both cases, many of the more significant periods of vorticity tilting and stretching along the true trajectories are reproduced in the analyses, but often with much smaller magnitudes. The results suggest that, given a suitable dataset, patterns in dual-Doppler kinematical and dynamical retrievals are generally qualitatively reliable, but that caution should be used when making quantitative interpretations.