Predictability and Dynamics of a squall-line and bow-echo event during BAMEX

An ensemble of cloud-resolving Weather Research and Forecasting Model (WRF) forecasts initialized with perturbations from an ensemble Kalman filter (EnKF) analysis is used to explore the predictability of a bow echo event during the Bow Echo and MCV Experiment (BAMEX) on 9-10 June 2003. The success of multiple WRF deterministic forecasts during the BAMEX campaign suggested that deterministic numerical weather prediction of convective-scale processes had come to fruition. Large variability in storm evolution, mode, and flow regime proves the contrary, highlighting the limit of practical predictability given realistic initial condition uncertainties from the EnKF analysis. Most members forecasted broad areas of severe convection. Some produced bow echoes very similar to that observed while others perform poorly, either producing the wrong convective mode or intensity. Given strong spatial and temporal variability in the environment, it was found that commonly used severe storm indices based on single environmental soundings (e.g., CAPE, CIN, low-level shear) had limited success in forecasting this bow echo event. Nevertheless, averaging 10 good and 10 poor members provided a clear difference in storm evolution with good members having stronger CAPE, shear, and a similar placing for an upper level shortwave that aided in the development of the squall line. Two distinct storm modes that formed within the ensemble indicate a bifurcation point between two regimes associated with upscale error growth due to moist convection. To further explore the bifurcation and the event's intrinsic predictability, a perfect model assumption is made in conjunction with initial condition uncertainties an order of magnitude smaller than the current observational analysis. The averaged 10 good and 10 poor members were linearly reduced by 1/2, 1/4, and 1/8 to produce the initial condition uncertainties. In essence, the ensemble forecast and additional sensitivity experiments demonstrate that: (1) this storm has a smaller predictability limit than first hypothesized under the given circumstances of realistic initial condition uncertainty and can be improved with more accurate initial conditions; (2) if the storm is near bifurcation points, there may be an intrinsic limit. The limit of both practical and intrinsic predictability highlights the need for probabilistic and ensemble forecasts for cloud-resolving weather prediction.