The Use of Ensemble-Based Sensitivity with Observations to Improve Predictability of Severe Convective Events

Ensemble sensitivity can reveal important weather features early in a forecast window relevant to the predictability of high-impact events later in time. Sensitivity has been shown on synoptic scales with simulated observations to be useful in identifying ensemble subsets that are more likely than the full ensemble mean, which may potentially add value to operational guidance of high-impact events. On convective scales, with highly nonlinear ensemble perturbation evolution and very non-Gaussian distributions of severe weather responses (e.g., simulated reflectivity above some threshold), it becomes more difficult to apply linear-based ensemble sensitivity to improve predictability of severe events. Here we test the ability of ensemble sensitivity to improve predictability of a severe convective event through identifying errors in sensitive regions of different members early in a forecast period using radar and surface-based observations. In this case, through the inspection of a number of operational models, an overnight mesoscale convective system (MCS) and its associated cold pool appeared to strongly influence whether or not severe convection would occur the following afternoon. Since both the overnight MCS and next-day convection are associated with strong nonlinearity and non-Gaussian distributions in the ensemble, this case allows a rigid test of using ensemble sensitivity and related techniques with observations for convective events. The performance of the sensitivity-based technique will be presented, and integration into an operational tool for severe convection will be discussed.