Statistical analysis of the spatial variability of very extreme rainfall in the Mediterranean area

U.S.- Italy Research Workshop on the

Hydrometeorology, Impacts, and Management of Extreme Floods

Perugia (Italy), November 1995

STATISTICAL ANALYSIS OF THE SPATIAL VARIABILITY OF

VERY EXTREME RAINFALL IN THE MEDITERRANEAN AREA

P. Furcolo, P. Villani, and F. Rossi

Department of Civil Engineering

University of Salerno, Italy

Natural hydrometeorological disasters in the Mediterranean region have occurred in the past essentially by outlying storm events characterized by considerable rainfall intensity and rare frequency. The characterization of this type of event is a crucial point in risk mitigation. Because of the rare occurrence of these extreme events they can be analyzed only on a regional frequency basis in order to reduce the uncertainty associated with parameter estimation at gauged sites and for risk assessment at ungauged sites. A statistical regional model includes: (i) a probabilistic model that can describe the extraordinary high floods and rainfall observed in the past (outliers) and (ii) a regionalization model that can take into account the observed spatial variability of the parameters of the probabilistic model. A regional model based on the TCEV distribution along with a geostatistical analysis of its parameters is presented here. The regional model assumes the observed variance stems from two sources: sampling variability due to uncertainties in at-site estimates, and spatial variability due to effective differences between sites. Traditional geostatistical techniques refer to the exactitude property in gauged sites whereby at-site estimates are affected by sampling uncertainties that can be predominant for high order parameters. An iterative geostatistical procedure is implemented, which makes it possible to obtain the spatial structure of the noiseless variate. Some initial results are shown with reference to a case study for an Italian region while the objective differentiation between areas with different risks constitutes one of the most important findings of the proposed regionalization procedure.