Modelli probabilistici per la stima delle portate di piena di elevato tempo di ritorno

The present paper describes the results of a regional analysis of peak flood flows carried out on the basis of the historical series recorded in more than 12000 gauging station distributed across the five continents; the regionalization procedure has been performed by means of suitable standardizations of the peak discharges. If the site-to-site variability of the statistical properties of peak flows is described in terms of the variability of the first two order moments - and therefore of the average and of the coefficient of variation CV - this approach leads to the MG model, introduced by the writers in previous papers. Accounting also for the variability of the third order moment - and therefore of the coefficient of skewness g - a new probabilistic model (generalized MG model) has been obtained. With respect to the other probability distributions considered in the paper, this model provides a better interpolation of the empirical frequency distributions of the peak discharges in the range of high return periods, especially for high values of CV and g. In the same respect, also the original MG model performs better than the probability distributions that are currently used for the analysis of peak flows. Moreover, the peak discharges obtained from the empirical absolute maxima of the standardized variables can be regarded as upper bounds to be used for the design of flood control structures in the presence of particularly high risk level (e.g. dam spillways or embankments to protect nuclear sites).