USE OF AKAIKE INFORMATION CRITERION FOR SELECTION OF FLOOD FREQUENCY-DISTRIBUTION

When conducting a conventional single-station flood frequency analysis, an appropriate distribution must be selected. Typically, sample statistics, probability plots, goodness-of-fit tests, etc. are used to facilitate the decision process. For the predominate case of a relatively short record length of flood series, this standard approach leads to undue emphasis on goodness of fit and virtually no consideration of the uncertainty due to additional parameters. The information criterion suggested by Akaike (AIC) is a measure to evaluate the "benefit" of goodness of fit and the "cost" of parameter uncertainty. The criterion is tested for 42 long-term hydrometric stations across Canada and its applicability and limitations are demonstrated in eight samples. The AIC is recommended as an aid in selecting a flood frequency distribution.