Computer Program for Statistical Analysis of Annual Flood Data by the Log-Pearson Type III Methods

Abstract

The Federal Water Resources Council has recommended the adoption of the Log-Pearson Type III method of establishing flood flow frequencies. The computer program developed in this study was written in Fortran IV language to facilitate Log-Pearson Type III method computations. Annual floods are sorted in decreasing magnitude and then logarithms, mean, standard deviation and skewness of the logarithms are computed. The magnitude of the 100, 50, etc., year flood are determined with the aid of tables. The initial computer printout consists of sorted values of the floods, empirical values of recurrence interval and probability, and the logarithms of the floods. Application of the method to selected streams in the United States indicates that difficulties may be encountered when a given set of data contains one or more very low floods or outliers. The Log-Person Type III distribution appears to have a substantial advantage over the Gumble and log-normal distributions that have been used for flood frequency analysis because it can be used for data having either a plus or a minus skewness. Also, it reduces to the log-normal distribution for zero skewness where the data fit this distribution. However, it will require a data screening procedure and sufficient use to indicate desirable restrictions on skewness values for short records and perhaps for various regions.