Use of Regression Techniques to Predict Hail Size and the Probability of Large Hail

Abstract

Multiple regression and logistic regression equations were derived for the prediction of hail size based on data from 1992 through 1994. The multiple regression equation was formulated to predict hail diameter. The logistic regression equation was developed to predict the probability of hail size greater than or equal to 1.9 cm in diameter. Variables used for this study consisted of vertically integrated liquid (VIL) computed from the Weather Surveillance Radar-1988 Doppler (WSR-88D) radar and convective parameters derived from the skew-T/Hodograph analysis and research program. Data were obtained from the Baltimore, Maryland/Washington D.C., WSR-88D radar and Dulles, Virginia, upper-air soundings. Numerous parameters were tested; however, only VIL, 85-kPa temperature, freezing level, and mean storm-relative inflow in the lowest 2-km were retained for the multiple regression equation. These four parameters were used as a base to derive the logistic regression equation, and this derivation process added no additional terms or interactions to these four terms. Both of the equations were verified using statistical techniques with hail occurrence and diameter data from May 1994 through May 1995. The multiple regression equation was found to be of limited use in predicting hail diameter. However, the logistic regression equation did considerably better in predicting the probability of hail size greater than or equal to 1.9 cm.