The Damaging Downburst Prediction and Detection Algorithm (DDPDA) predicts the onset of damaging outflows from storm cells that form in an environment of high CAPE and weak environmental wind shear. This study examines twenty-six parameters derived from radar reflectivity, radial velocity, and environmental data sources. Data include 91 severe downburst events and 1247 non-severe storm cells from 64 days. This data set is randomly split into two parts in order to develop and evaluate a method to predict strong downburst events. These two parts are used for training and validating downburst prediction equations developed using linear discriminant analysis.

The process of randomly sampling the training and validation data sets is repeated 100 times in order to find the expected distribution of skill scores for the prediction equations as well as the most important parameters for use in predicting downburst events. A downburst prediction equation is developed for each of the 100 training/validation data set pairs. The distributions of skill scores as well as the average weights of the input variables are illustrated.

This is a continuation of work presented previously.