REGIONAL FREQUENCYANALYSIS FOR EXTREME RAINFALL IN SICILY

Abstract

The study regarded the regional frequency analysis of extreme rainfall in Sicily, using rainfall annual maximum series of duration 1, 3, 6, 12 and 24 hours in 235 sites, stationed in all Sicily and for the period 1928-1998. The applied regional models, after the appropriate verifications, were: the TCEV hierarchical model (Rossi et al.), the model based on the use of the linear moments LM (Hosking et al.) and the MGs parametric model (Maione et al.). For ungaged sites, as in a flood index approach, the index value of each model was evaluated through maps derived from spatial interpolation. The three models regional estimations were compared with the at-site estimation, computed by GEV and Gumbel distribution, for the stations with record length n≥45 and return period T =200 years. The comparison was carried out taking into consideration, for the regional models, the index value estimation in the two following cases: • Case 1: Index value estimated by historical series; • Case 2: Index value estimated by interpolated maps. The results indicate that, in Case 1, the MGs regional model performs better than the other two regional models for each duration t but, in Case 2, the performance of MGs model is weakly better than the other regional models studied only for t=1÷12 hours. Furthermore, we observe the same behavior of TCEV and GEV-LM models, also detected by Brath in center-northern Italy [4].

Keywords

• rainfall, regional model, MGs, TCEV, L-moments.