Comparison of Log-Pearson Type III and Gumbel Probability Distributions for Flood Frequency Analysis of Ikpoba River Catchment at Benin City

Abstract

In the engineering design of hydraulic structures for flood alleviation and storm water management in or near a river, the magnitude of design flood is required. For the purpose of evaluating the flood discharge for various return periods (FFA) for a particular river, several methods are utilized depending on the availability of data and amount of discharge details required and where hydrometric measurements are limited as is the case  of the catchment under study, estimates are made by more than one method and engineering judgment used in deciding design values.
In this study, the extreme value Type 1 (Gumbel) and Log-Pearson Type III probability distributions have been alternatively utilized to perform flood frequency analysis on the peak annual series discharge data of Ikpoba River at Benin City for the water years 1989 to 2000. The predicted design floods by alternative models for return periods of 2yrs, 5yrs, 10yrs, 25yrs, 50yrs, 100yrs, 200yrs and 1000yrs were obtained and compared.  Our results indicate that the river peak flows can be satisfactorily modeled by any of the two methods of analysis and that at lower return periods of up to 5yrs, the Gumbel distribution predicts lower peak discharge values and for higher return periods of 10 yrs and above, the Log Pearson Type III distribution predicted lower discharge values with a percentage deviation ranging from – 0.62% to -12.78%.
It is however recommended that the design values obtained by the Gumbel distribution be adopted to assure safe design in view of the available limited hydrometric data utilized for the study.

Keywords: Flood frequency analysis, Log Pearson Type III, Gumbel distribution, Peak discharge, return period, Annual series.