A Weather Research and Forecasting (WRF) model based mesoscale ensemble Kalman filter data assimilation and forecast system, developed at the National Severe Storms Laboratory is analyzed using a Hierarchical Clustering Approach (HCA). The ensemble system is useful for providing improved short-term guidance of severe weather events. In addition to initial and boundary condition uncertainties, the ensemble system accounts for uncertainties in model physics by varying the planetary boundary layer, convection, land surface, radiation, and microphysical parameterizations across the ensemble members. The HCA technique, which is helpful for understanding the impact of sampling different sources of uncertainties in the ensemble system, is applied to several severe weather forecast parameters from the ensemble system for 24 selected severe weather events in springs 2007-2009. Results indicate that the ensemble members largely cluster by cumulus parameterization (CP) for the severe weather parameters. In particular, the Kain–Fritsch (KF) members have a tendency to form a separate cluster away from the Betts–Miller–Janjic (BMJ) and Grell–Devenyi (GD) members. Some secondary sub-clustering associated with microphysical, planetary boundary layer and shortwave radiation schemes are also observed. Accumulated precipitation forecasts show a repeatable clustering pattern based on CP.