Ensemble Sensitivity Analysis for Targeted Observations of Supercell Thunderstorms

In situ observations of supercells have the potential to improve the accuracy and lead times of storm scale forecasts. Ensemble sensitivity analysis (ESA) can identify in which storm regions variations to the initial state in a model are most likely to affect prediction of storm strength and severity. To demonstrate this technique, two supercells were forecast in an ensemble of idealized simulations, one ensemble producing a strong supercell with the other generating a much weaker storm. ESA was performed to relate perturbations in state variables (temperature, moisture, pressure, and the three-dimensional wind) to proxies for storm strength and severity. Although ESA typically calculates the slope between the perturbations and forecast responses, a robust statistical technique is used that substitutes rank correlation in place of slope. This provides some advantages including allowing a non-linear relationship between perturbations and forecast responses and allows the collective impact of all the perturbation variables to be estimated with multiple regression.

For updraft helicity, the strongest statistical relationship was with low-level perturbations near the mesocyclone region at lead times on the order of 5-15 minutes and with the storm environment, forward flank gust front, and rear flank gust front at 45-60 minute lead times. Much weaker statistical relationships were present at intermediate lead times. For low-level wind speed, the strongest sensitivity was to perturbations in the storm environment and forward flank gust front regions, increasing with longer lead times. These and several other relationships were evident with the strong supercell, with generally weaker relationships found in the weak supercell.