| Summary: In this study, a regional flood frequency analysis (RFFA) is applied to annual maximum discharges of 33 gauging stations in the Eastern Black Sea Basin, Turkey. Homogeneity of the region is determined by discordancy (Di) and heterogeneity measures (Hi) based on L-moments. Extreme-value type I, generalized extreme-value, lognormal, log Pearson type III, and generalized logistic distributions are fitted to the flood data of the homogeneous region. Chi square and probability plot correlation tests are used for the determination of best fit distributions for each station. Based on the appreciate distributions for each site, flood quantiles are estimated for the return periods of T=5, 10, 25, 50, 100, and 500 years. A non-linear regression model is developed for the estimation of design floods for ungauged catchments in the region. Drainage area, main stream slope, elevation, stream density, mean annual rainfall, and rainfall intensities are used as independent variables in the regression model. Mean relative error, root mean square error, and mean absolute error values are applied to the model in order to evaluate the performance of regression analysis. Artificial bee colony algorithm (ABC) and teaching-learning based optimization (TLBO) models are developed to compare the results with regression analysis. The analysis has concluded that TLBO and ABC show a reasonable performance and they are superior to the regression analysis. Finally, error values indicate that TLBO method yields better results than ABC for estimation of flood quantiles for different independent variables. Furthermore; artificial neural networks (ANN) and multiple non linear regression analysis (RA) models are developed using drainage area, main stream slope, elevation, stream density, mean annual rainfall and return periods. ANN is found to give more reliable results than RA for forecasting the maximum possible discharges.
Key Words: Flood Frequency Analysis, L-Moments, Artificial Bee Colony Algorithm, Teaching-Learning Based Optimization, Artifical Neural Networks
|
