Assessment of peak flow estimations through regional analysis in the north ege river basins

Annual peak flow estimates of given return periods, T, are commonly used for design and economic evaluation of various water resources projects. Design quantiles, if it is possible, are often supplied from the at-site flood frequency analyses of the peak flow data available at the hydrometric station(s) as close as the project site. Recently, various regional flood frequeny analysis (RFFA) models for the estimation of design quantiles have been developed. The two major objectives of the regional analysis are to estimate the T-year event magnitude at a given location, mainly at ungaged sites, in a statistically homogeneous region, and to improve at-site estimates with large sampling errors because of the limited data. In this research, an index type model for the regionalization of peak flows (as annual maxima) recorded at 45 gauges in the North Ege Region of Turkey, which was proven that it can be clustered into two statistically homogeneous subregions, were developed. In the scope of this study, four alternative regionalization approaches for the regional estimation of dimensionless flood quantiles over the North Ege Region of Turkey were proposed. All the approaches were based on the assumption that dimensionless floods recorded at the stations were subsamples drawn from a twoparameter lognormal population. The first approach was based upon the regional mean coefficients of variation (BCvk; k=1,2) while the second one on the regional weighted means of the logaritmic standard deviations ( yk Bs ; k=1,2). The weighted averages of the dimensionless T-year at-site quantile estimates in the third approach, while their medians in the fourth approach were used as the ordinates of the regional growth curves for the two homogeneous subregions. Average performances of the alternative regionalization approaches relative to the at-site quantile estimators of T=2, 5, 10, 20, 50 and 100 years have been assessed by comparing the BIAS (relative bias) and the RMSE (relative root mean square error) measures of the alternatives. Numerical results of the case study North Ege Region revealed that the fourth approach (based on the medians of the T-year at-site quantiles estimated by the Lmoment parameters) produces the more reliable regional estimates for the Sub Region-I, where spatial variation of Cv is high (a heteregenous subregion). Whereas, all four approaches with classical moment parameters produce beter regional quantile estimates for the Sub Region-II, where spatial variation of Cv is rather low (a homogeneous subregion). Finally, the mean flood-drainage area, and the standard deviation-drainage area relationships of power type, which were required in deriving dimensional flood magnitudes for the entire North Ege Region, were developed by ordinary least squares (OLS), and by nonlinear least squares (NLS) as well. Keywords: Regional Flood Frequency Analysis (RFFA), index methods, annual maximum flows, mean flood-drainage area relationships.