The use of the simple 1D steady-state convective cloud model in the decision tree for determining the probability of thunderstorm occurrence

Abstract

The algorithm of a binary decision tree [1–3] determining the probability of thunderstorm occurrence was applied to a set of predictors including the instability indices and results of the 1D steady-state convective cloud model. Predictants A, B, C, D=1/0 correspond to various temporal and spatial limits for thunderstorm occurrence recorded at synoptic stations in an area 100km around Prague. The predictors were found on the basis of TEMP 12.00 GMT data from the Prague Libus station for the months V–VIII in the years 1981–1985. The first tests were carried out to compare the decision quality for various types of predictants and for sets of predictors consisting only of the instability indices, only of model predictors and of both quantities. The best results were obtained using a decision tree constructed on the basis of sets of all the predictors, where the model predictors were employed primarily in the unstable branch of the tree. The best decision quality was obtained for predictant B, in agreement with the concept of the strongest connection between the thermodynamic state of the atmosphere and the area of the extent of thunderstorm occurrence.