Impacts of Microphysics and PBL Physics Schemes on Tornado Outbreak Prediction

The use of numerical weather prediction (NWP) has brought significant improvements to the forecasting of severe weather outbreaks. Tornado outbreaks can now be predicted up to four to eight days in advance; nonetheless, the precise location and timing of a tornado outbreak are often missed in forecasts prior to the onset of the outbreak. In addition, many physics parameterization schemes have been developed, each with its individual strengths and weaknesses.

This research examines the effect of microphysics and planetary boundary layer (PBL) physics schemes on tornado outbreak forecast accuracy. Twenty synoptically driven United States tornado outbreaks and twenty synoptically driven nontornadic outbreaks from 2008 to 2011 were simulated in a 12 km/4 km nested configuration in the WRF-ARW model. These model runs each encompassed 42 hours, from 1800Z on the day before the outbreak day to 1200Z on the day after. Each of the cases was modeled with a physics parameterization ensemble of 15 possible combinations: 5 microphysics schemes and 3 PBL physics schemes, all of which were designed to perform well in convective weather conditions.

A learning machine technique known as a support vector machine (SVM) was used to predict the probability of a tornado outbreak for each run. This algorithm classifies data into binary categories based upon characteristics of predictor values given as input. Meteorological parameters determined to be statistically significant in association with tornadoes were extracted from the WRF-ARW simulations at the peak time of each case. The covariates were input to the SVM, which then issued a yes/no prediction of a tornado outbreak. Contingency statistics were calculated for each of the ensemble member types over the 40 outbreaks.

The results indicate that the combination of WRF Double Moment 6-Class microphysics and Mellor-Yamada-Janjic PBL physics generates the highest skill – 0.69 – of all 15 combinations studied, significant at the 95% confidence level. This combination produced the highest probability of detection and false alarm ratio as well, though both of these statistics were within the margins of error of other physics combinations. PBL physics variations proved to be more significant than microphysics in influencing the modeling of the important covariates, with the Asymmetric Convection Model PBL scheme performing significantly more poorly than the other two examined PBL schemes in the modeling of shear and helicity parameters.

Global- and synoptic-scale models are run with a physics configuration tested to perform sufficiently for a variety of weather phenomena. However, local National Weather Service offices, many of which now run NWP models for their own forecast regions at high resolutions, could benefit from statistics regarding the forecasting usefulness of custom parameterization schemes for specific types of expected weather.