Floods frequency analysis for Sicily (South Italy)

In this paper a regional flood frequency analysis based on the TCEV distribution is developed using flood data recorded in Sicily. The hierarchical approach, characterized by three investigation levels for estimating the parameters of the theoretical distribution is firstly discussed. The highest level of homogeneity hypothesis with regard to the skewness coefficient was verified by using Monte Carlo technique and taking account of the separation effect proposed by Matalas et al (1975). This analysis also showed the unability of the GEV model to reproduce the empirical CDF of the skewness coefficients and the convenience to use annual maximum peak flood (AMPF) data instead of annual maximum mean daily discharge (AMDD) records to estimate the parameters of the TCEV model. At the second level of regionalization, three different methods for identifying hydrologically homogeneous sub-regions are then performed: the first, based on the R-statistic test (Wiltshire 1986); the second, based on the hypothesis of statistical self-similarity of each selected area considering sampling statistical moments (Gupta et al. 1994); the third, funded on the same hypothesis of statistical self-similarity by using the sampling probability weighted moments (Kumar et al., 1994). The analysis showed that a further division into three sub-regions proposed by Cannarozzo et al. (1995) is not necessary and that a single theoretical growth curve is able to reproduce the empirical CDF of the standardized variable x’. Finally, the third level of investigation showed that two different empirical relationships proposed for estimating the index flood could be used for establishing the Q-T relationship anywhere in the region.