Algorithms for the analysis and characterization of convective structures relative to extreme rainfall events

Among many weather phenomena, convective storms are one of the most dangerous since they are able to cause, in a relatively small time window, great damages. Convective precipitations are in fact characterized by relatively small spatial and temporal scales, and as a consequence, the task of forecasting such phenomena turns out to be an elusive one. Nonetheless, given their dangerousness, the identification and tracking of meteorological convective systems are of paramount importance and are the subject of several studies. In particular, the early detection of the areas where deep convection is about to appear, and the prediction of the development and path of existing convective thunderstorms represent two focal research topics. The aim of the present work is to outline a framework employing various techniques apt to the task of monitoring and characterization of convective clouds. We analyze meteorological satellite images and data in order to evaluate the potential occurring of strong precipitation. Techniques considered include numerical, machine learning, image processing. The techniques are tested on data coming from real convective events captured in the last years on the Italian peninsula by the Meteosat meteorological satellites and weather radar.