Burkely Gallo-November 3

Using Convection-Allowing Ensemble Attributes to Derive Operationally Relevant Tornado Probabilities


Gallo, Burkely
Ph.D. Student


November 3, 2017 - 3:00 pm


November 3, 2017 - 4:00 pm


120 David L. Boren Blvd. Room 5600, Norman, OK 73072   View map

Using Convection-Allowing Ensemble Attributes to Derive Operationally Relevant Tornado Probabilities

Convection-allowing model (CAM) ensembles generate high-resolution fields of storm environment attributes, such as the Significant Tornado Parameter (STP), along with storm-scale attributes unavailable in convection-parameterizing ensembles, such as hourly maximum updraft helicity (UH). This work uses UH and STP from CAM ensembles to generate operationally relevant tornado probabilities on a daily basis. In addition to using solely model-derived fields, empirical frequencies of a tornado given a right-moving supercell and a value of STP are incorporated into a set of probabilities. Using the empirical frequencies shifts the paradigm from a threshold exceedance approach in which a tornado would be assumed should some value of UH and/or STP be reached to a more realistic probabilistic approach. Through the probabilistic approach, each time and location has a probability of a tornado occurring based on the empirical frequencies. Magnitudes of the probabilities generated using the empirical frequencies are similar to operational Storm Prediction Center forecasts, whereas the threshold exceedance approach often generates probabilities with high magnitudes.

Probabilities were implemented in the NSSL-WRF ensemble and the HREFv2 ensemble and evaluated real-time by participants in the 2017 Spring Forecasting Experiment. Participants found the probabilities to be useful but also to have a false alarm problem associated with UH generated by nocturnal mesoscale convective systems (MCSs). Thus, a method was developed to use the time of model UH occurrence and the diurnal report distribution in the forecasts, lessening the nocturnal false alarm. Incorporating the time of UH occurrence lowers the overall forecast performance according to attributes such as the ROC area, but the diurnal cycle of the average probability more closely matches the diurnal cycle of the observations. When applied to individual days the time-dependent probabilities focus much more strongly on areas with right-moving supercellular tornadoes than methods that do not consider the time of UH occurrence.