Discriminant analysis applied to the forecasting of thunderstorms

Abstract

The combination of several thermodynamic variables based upon the data provided by a radiosounding can be useful for the forecasting of thunderstorms. As a matter of fact, there are many indices that allow the establishment of a storm risk prediction once they have been gauged. The problem comes when not all indices lead to the same prediction. In these cases, it is necessary to establish one single function based on the information provided by all the variables employed, which should be able to determine a two-fold prediction: risk or no risk. This article presents a statistic model for the short tem prediction of thunderstorms in the region of Leon (Spain). To reach this aim 15 meteorological variables were selected. These variables were easy to handle by non-expert staff, and they allowed the characterisation of the preconvective environment early in the morning on thunderstorm days. The variables have been properly combined and gauged with the help of a dense network of meteorological observers. The result has led to the construction of a reliable model. The discriminant quadratic model has been easily applied to determine in an objective and binary way the risk/no risk for the occurrence of thunderstorms.