PROBABILITY DISTRIBUTION MODELS FOR ESTIMATING ANNUAL PEAK FLOW IN THE MKHOMAZI RIVER, SOUTH AFRICA.

Accurate computation of peak flow in rivers is vital in preventing environmental degradation. This study developed stochastic models for computing annual peak flows in the Mkhomazi River, South Africa. Six curves of selected functions (Normal, Log–Normal, Pearson III, Log–Pearson III, Gumbel and Log–Gumbel) were fitted to the annual duration series of the at–site annual peak flow of the river. Chi–square tests for goodness–of–fit were performed at 10% significance level to determine the appropriateness of all models in describing the observed data. The non–parametric significant differences in the performances of applicable models found was examined by applying the Kruskal–Wallis (H) test. A system of ranking based on aggregate of scores from statistical performance measures was adopted to suggest a most suitable model. The Normal, Pearson, Log–Pearson and Gumbel models were found suitable for application in the study area. These models are recommended for use in modeling peak flow in the river as they were found appropriate based on standard statistical test. Thus, the outcome of this work provides additional inputs for the improvement of hydrological researches in South Africa. [ABSTRACT FROM AUTHOR]