An OSSE framework for assessing the utility of storm-scale ensemble forecasts to tornado warning operations

Under the envisioned warn-on-forecast paradigm (Stensrud et al. 2009), forecasters will make increasing use of short-term, storm-scale ensemble NWP output in order to increase tornado, severe thunderstorm and flash flood warning lead times. Extending tornado warning lead times may be particularly challenging, requiring (at minimum) the development of a real-time ensemble analysis-forecast system that produces useful probabilistic forecasts of intense low-level thunderstorm rotation beyond 15 or 20 min. In this study, we adopt an observing system simulation experiment (OSSE) framework to assess the achievability of such a system given current model and observational limitations and near-term computational constraints. Ensemble forecasts are initialized from ensemble Kalman filter (EnKF) analyses of multi-Doppler radar data, then processed to generate low-level vorticity forecast products. Special attention is paid to the impact on forecast quality of model error, radar-storm geometry, and the length of the assimilation period.