The study of multiple effects of a number of variables, and the assessment of the corresponding environmental risks, may require the adoption of suitable multivariate models when the variables at play are dependent, as it often happens in environmental studies. In this work, the flood risks in a given region are investigated, in order to identify specific spatial sub-regions (clusters) where the floods show a similar behavior with respect to suitable multivariate) criteria. The reason of the work is three-fold, and the outcomes have deep implications in the hydrological practice: (i) such a regionalization (as it is called in hydrology) may provide useful indications for deciding which gauge stations have a similar (stochastic) behavior; (ii) the spatial clustering may represent a valuable tool for investigating ungauged basins present in a given ‘‘homogeneous’’ Region; (iii) the estimate of extreme design values may be improved by using all the observations collected in a cluster (instead of only single-station data). For this purpose, a Copulabased Agglomerative Hierarchical Clustering algorithm – a key tool in geosciences for the analysis of the dependence information – is proposed. The procedure is illustrated via a case study involving the Po river basin, the largest Italian one. A comparison with a previous attempt to cluster the gauge stations present in the same spatial region is also carried out. The sub-regions picked out by the clustering procedure outlined here agree with previous results obtained via heuristic hydrological and meteorological reasonings, and identify spatial areas characterized by similar flood regimes.
