Contrast of Six Fitting Methods of the Log-Pearson Type III Distribution in 31 Historical Records of Annual Extreme Events

Abstract

ln this paper; six of the available methods to fit the Log-Pearson type III (LP3) distribution to a sample of máximum annual values are described with detail. Those methods are applied to each of the 31 historical records used in this study; most of the flood records have a length varying from 16 to 113 data. For the interpretation of the results, two numerical measures were used: goodness of fit, through absolute and standard errors, and one of predictive nature, through estimations associated with different return periods, as well as with the estimation ratio (ER). lt is concluded that the application of all the fitting methods of the LP3 distribution is highly convenient and even necessary since it allows the observation of variability of its predictions, the selection of the right fitting procedure based on the ER, and the generation of the variability estimation interval.