On the identification of an optimum flood frequency model

Abstract

Although a number of nonparametric analysis have appeared in the literature, traditional flood frequency analysis has been approached primarily as a problem in parametric statistical inference. Peak annual streamflow data are assureed to come frore a parent population whose distribution function is known7 is analytically expressable and contains a finite nurr.bre of parameters. A large number of peak flow distributions have been studied, for example, t.he normal, the lognormal) the Gumbel, the Carnrna and Weibull distributions. Goodness of fit procedures then test whether or not the data do indeed fit the assumed distribution with a specified degree of confidence. However, the use of the convetional goodness of fit procedures in flood frequency analysis has several disadvantages. Firstly, is the lack of power of these goodness of fit tests with respect to the typically skewed flood Peak distributions. This generates considerable variability in the estimation of design, events. Secondly, the conventional goodness of fit tests are subjective in that the final results drawn from such tests depend very much on the level of confidence utilized. This me ans that different levels of confidence can often lead to conflicting results. Lastly, and probably the most serious disadvantage is