Deriving Operationally Relevant Tornado Probabilities from Convection-Allowing Ensembles
Hourly maximum fields of simulated storm diagnostics from experimental versions of convection-allowing models (CAMs) provide valuable information regarding severe weather potential. The focus of this work is to extract operationally relevant tornado probabilities from the CAM-based Weather Research and Forecasting (WRF) ensemble initialized daily at the National Severe Storms Laboratory (NSSL-WRF). Probabilities are derived in three main ways: by using updraft helicity (UH), UH filtered by model-derived environmental parameters, and through combining UH and model-derived environmental parameters with observed climatological tornado frequencies. The set of probabilities generated using observed climatological frequencies move beyond the binary paradigm of the other forecast methods. Rather than using a specific threshold of UH as a proxy for tornadogenesis and relying on the ensemble to generate probabilities, this method treats each point as having a certain probability of producing a tornado, depending on the surrounding environmental conditions. Additionally, these forecasts are compared with the 0600 UTC Storm Prediction Center (SPC)’s tornado probabilities, to determine whether these forecasts approach the skill of expert forecasters. While the methods derived solely using ensemble parameters overforecast tornado probability magnitude, the probabilities that incorporate climatological frequency information perform much more reliably.
For the probabilistic forecasts to be operationally relevant, cooperation with forecasters is critical in their development. A database of right-moving supercells developed by SPC forecasters was used to generate the climatological frequencies on which the final set of probabilities is based. Through NOAA’s Hazardous Weather Testbed Spring Forecasting Experiment (SFE), subjective daily evaluation of the probabilistic forecasts will take place during SFE 2017, running from 1 May–2 June 2017, supplementing the feedback received during SFE 2015 that led to the development of the climatological frequency probabilities. The dataset of right-moving supercells is also used to test the common supposition that the rotation of simulated storms correlates with the rotation of observed storms. Contrary to conventional wisdom, it is found that the strength of the UH does not correlate well with the observed rotational velocity. Thus, specific values of UH are not particularly indicative of given rotational velocities in supercells.