Abstract:

Flood frequency analysis is very important in the design of water structures, planning and projecting of flood control. One of the important steps of flood frequency analysis is to estimate a suitable distribution model and its parameters for the data. In this study, optimization based DEoptim algorithm is proposed to estimate the parameters of probability distribution functions and is compared with the traditional maximum likelihood method (MLM) and Method of moments (MOM). Annual maximum flow data of 2 stations on the Eastern Black Sea Basin are used. Gamma, Weibull, Log-Normal, Logistic and Normal probability distributions were used to determine flood recurrences flows. To determine the most appropriate probability distribution, Kolmogorov-Smirnov (KS) test, mean square error (RMSE), Kling-Gupta efficiency coefficient (KGE), mean absolute error (MAE), Nash-Sutcliffe efficiency coefficient (NSE) and Akaike information criterion ( AIC) tests were used. According to the results, it was seen that the DEoptim method is a more effective method in estimating the probability distribution function parameters compared to the MLM and MOM methods. Finally, using the most appropriate probability models, flood estimates that may occur in different recurrence periods (5, 10, 20, 50, 100, 200 and 500) were calculated and the results were discussed.

Keywords: