Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Norvege → Regional Flood Frequency Analaysis for Southern Africa

University of Oslo (UiO) (2011)

Regional Flood Frequency Analaysis for Southern Africa

Haile, Alem Tadesse

Titre : Regional Flood Frequency Analaysis for Southern Africa

Auteur : Haile, Alem Tadesse

Université de soutenance : University of Oslo (UiO)

Grade : Master Thesis in Geosciences 2011

Extreme floods are natural disasters often associated with losses of life, and severe impact to agricultural production and infrastructures. However, efficient estimations of the magnitude of such extreme events with their non-exceedance probabilities, either for design or risk management planning purposes, are often limited by the data availability (i.e., both in quality and quantity). In this study, regional frequency analyses of annual maximum series (AMS) of flood events from unregulated rivers of southern Africa were conducted. This includes preliminary data analysis (data screening and outlier analysis), sensitivity analysis, identification of homogenous regions and suitable regional distribution models for the regions, development of regional growth curves and regression models to estimate the quantile floods for unguaged catchments. The study area comprises five countries (459 stations) : Namibia, Malawi, Zambia, Zimbabwe and South Africa. The AMS derived for each station were examined for validity, dependency, and the existence of outliers. After thorough examinations of the AM flood events, 122 gauging sites were selected for further analysis. The study area was divided into nine possible homogenous regions based on the geographical grouping method together with the heterogeneity tests. The AMS from Namibia, Zimbabwe, Zambia and Malawi were grouped as regions R1, R2, R3 and R4, respectively, while the South African catchments were further classified into five possibly homogonous regions. The choice of an appropriate regional flood distribution model was performed based on L-moment approaches together with the index flood procedures and goodness-of-fit (GOF) tests. The Generalized Pareto (GPA), Pearson type III (PE3), Three-parameter lognormal (LN3) and the Generalized Extreme Value (GEV) distributions were found to be suitable models for AMS of floods in southern Africa catchments. Regional flood frequency curves were constructed based on the best regional distribution for the nine regions and design floods estimated for the return periods of 2-500 years. Based on assessments the accuracy of the derived quantilequantile plots, it was concluded that the performance of this regional approaches was satisfactory and also confirmed when validated against sites not included in the regional analysis. Keywords Outliers ; index flood ; regional flood frequency analysis ; L-moment ; Southern Africa


Version intégrale (3,13 Mb)

Page publiée le 11 octobre 2012, mise à jour le 13 février 2018