Contrast of five fitting methods of the general extreme values distribution from 31 historical records of annual maximum events

Abstract

In this work the results of five fitting methods of the general extreme values (GEV) distribution are compared, according to the standard and absolute fitting errors. Thirty one historical records, mainly flood flow ones, were used, and the results showed that all fitting methods give congruous values, and that the optimization method, minimizing the mentioned errors, is the best for fitting the GEV distribution. A computational subroutine is proposed for the estimation of confidence interval, on the basis of the fitting methods results for each return period. Lastly, several conclusions are drawn from the results' global analysis. The formulae and operative sequences of each utilized method or procedure are cited in four appendices.