ABSTRACT
In this chapter, we discuss a comparative analysis of six annual maximum (AM) flood heights time series models at two sites: Chokwe and Sicacate in the lower Limpopo River basin of Mozambique. The six AM time series models considered were the annual daily maximum (AM1), annual two-day maximum (AM2), annual five-day maximum (AM5), annual seven-day maximum (AM7), annual ten-day maximum (AM10), and annual 30-day maximum (AM30). A generalized extreme value (GEV) distribution was fitted to each of the six AM time series models. The goodness-of-fit of the GEV distribution at each site was assessed using Anderson-Darling (A-D) and Kolmogorov-Smirnov (K-S) statistics. An analysis of variance (ANOVA) was performed to check for significant differences between the six AM models at each site with reference to skewness, coefficient of variation (CV), excess kurtosis, A-D statistics, and K-S statistics. The results revealed no evidence of significant differences at the 5 percent level of significance among the six AM time series models in terms of skewness, CV, excess kurtosis, and A-D and K-S statistics. A correlation analysis was also performed to check for significant correlations among the time series models: the results revealed high correlations among all six time series models at both sites. These findings suggest that any one of the models at each site can be used in place of the other five annual maximum time series models in flood frequency analyses of the lower Limpopo River. Without losing generality, the annual daily maximum flood heights time series model is used for further analysis to obtain flood frequency curves mainly because of its simplicity and relative ease of use.
