Discriminant methods for radar detection of hail

A number of tools are available today – and more are currently being developed – to discriminate, within a particular storm, areas with hail precipitation using data provided by conventional meteorological radar systems. However, at a supraregional level most of the methods that have traditionally been employed to identify hailstorms often obtain ambiguous results. other hand, many currently available systems for data extraction and processing make it relatively easy to calculate a large number of variables derived from radar parameters for each storm analyzed and at different stages in its development. The questions are now: is it possible to select and/or classify these variables according to their ability to discriminate hailstorms from non-hail storms? And if so, would the combination of several of these variables enable us to develop new and improved discriminating tools? These questions have prompted the use of mathematical methods for discrimination, such as logistic regression and linear discriminant analysis. In both models the stepwise method was used to construct the mathematical equation automatically. Data from a C-band radar system installed in the NW of the Iberian Peninsula were used to set up the two models. o models share a number of variables (VIL, maximum reflectivity, height of the maximum reflectivity and ddBZ_max/dt.), but the discriminant model makes use additionally of the top and the tilt of the storm. The results have been assessed not only in terms of the precision indices (the models have a probability of detection of 84.9% and 86.8%, respectively), but also with respect to their accuracy, sufficiency relation, resolution and discrimination. y, the logistic model has been calibrated for S-band radar systems. The results show a very satisfactory probability of detection, demonstrating that these methods may also be effective in this type of system.