The fitting of the generalized distribution
Pareto join to declustering methods to
analyze to dataset of mean daily flow of
Artemis station, Piracicaba, São Paulo, Brazil

Renato Rodrigues SILVA[1]

Silvio Sandoval ZOCCHI1

§    ABSTRACT: One of the most used methodologies  in the context of the extreme value theory  is the fitting of the generalized distribution Pareto to data exceedances over a threshold and the fitting of generalized extreme value distribution. However,  to fit the generalized Pareto distribution  to a dataset, it must consider the data is a sequence independent and having a common distribution that it is usually an unrealistic assumption. An way of to solve this problem is to use methods of declustering propose by Leadbetter et al. (1989), that in general identify groups of ocorrence of extreme flow for after fitting of the generalized distribution Pareto to maximum of the groups. In this work, was mades the  fit generalized distribution Pareto  and expoencial distribution, particular case of GP, join to declustering methods to dataset mean daily flow  of Artemis station, Piracicaba, SP, Brazil and after was compared the estimates the return levels of 5, 10, 50 and 100 years. Concludes the intervals estimates of return levels of 50 and 100 years obtained through the fitting the exponencial distribution are more precision than obtained through the fitting the generalized Pareto distribution.

§    KEYWORDS: Generalized distribution Pareto; generalized extremes values distribution; return levels.



[1] Departamento Ciências Exatas, Escola Superior de Agricultura ``Luiz de Queiroz", Universidade de São Paulo, Caixa Postal 9, CEP 13418-900 , Piracicaba, São Paulo, Brasil. E-mail: rrsilva@esalq.usp.br /  sszocchi@esalq.usp.br