Flood frequency analysis – Part 1: Review of the statistical approach in South Africa

Abstract

Statistical flood frequency analyses of observed flow data are applied to develop regional empirical and deterministic design flood estimation methods, particularly for application in cases where no, or insufficient, streamflow data are available. The soundness of the statistical approach, in the estimation of flood peak frequencies, depends on the availability of long records with good-quality observed flow data. With flood frequency methods currently under review in South Africa, a sound statistical approach is considered essential. This paper reviews the statistical flood frequency approach in South Africa, which includes an appraisal of the capability of the most commonly used probability distributions in South Africa to properly cope with the challenges encountered in a flood frequency analysis, based on extended experience in flood hydrology. All the distributions tend to perform poorly when lower probability frequency events are estimated, especially where outliers are present in the dataset. Research needs are identified to improve flood peak frequency estimation techniques, and practical pointers are suggested for the interim, in anticipation of updated methods. The importance of a visual interpretation of the data is highlighted to minimise the risk of not selecting the most appropriate distribution.