JOURNAL OF ENGINEERING RESEARCH, cilt.2, sa.1, ss.47-70, 2014 (SCI-Expanded)
Regionalflood frequency analysis for the SeyhanBasin in Turkey is done by two different approaches,and generalized growth curves to be used at un-gauged sites are developed byeither method. First, the homogeneity of the annual flood peaks series recordedat 11 un-regulated gauging stations in the Seyhan Basin is analyzed by theconventional Cv (coefficient of variation) test; and second, discordance,homogeneity, and goodness-of-fit by the Z tests by the method of L-moments areperformed. The Seyhan Basin is subdivided intothree homogeneous sub-basins from the aspect of annual flood peaks based on theCv test, which detects regions having recorded series with variationcoefficients close in magnitudes. The probability distributions of Log-Logistic,Pearson-3, log-Pearson-3, Gumbel, and Wakeby, whose parameters are computed bythe method of probability-weighted moments, are taken as candidatedistributions. The goodness-of-fit tests of Kolmogorov-Smirnov and Crammer VonMisses are applied, and the Wakeby distribution is found to be superior to theothers. According to the analyses by theL-moments method, the Seyhan Basin as a whole turned outto be a homogeneous region. By inspection of the L-moment ratio diagrams of L-variationversus L-skewness and L-skewness versus L-kurtosis of those 11 series, and bythe Z-statistic analysis of the L-moments method, the Log-Logistic distributionis found to represent the recorded series of annual flood peaks more closely.By regression, an expression relating average peak discharge (Qav)to the catchment area is obtained, and growth curves for the frequency of floodpeaks (Qmax.) by both the Cv-based and the L-moments methodsare developed using the data of those 11 gauged sites.