Multivariate statistical analysis of flood variables by copulas: two italian case studies

Multivariate statistics are important to determine the flood hydrograph for the design of hydraulic structures and for the hydraulic risk assessment. In the last decade, the copula approach has been investigated in hydrological practice to assess the design flood hydrograph in terms of flood peak, volume and duration. In this paper, the copula approach is exploited to perform pair analyses of these three random variables for two Italian watersheds, in the Apennine and the Alps respectively. The criterion to separate continuous flow series into independent events is discussed along with its implications on the dependence structure. The goodness-of-fits of the proposed copulas are then assessed by non-parametric tests. Marginal distributions to derive joint distributions are briefly suggested. The possibility of generating flood events according to the proposed model and potential applications to hydraulic structure design and flood management are finally examined.