Comparison of Different Methods for Estimating of Monthly Discharge Missing Data in Grand Karoon River Basin

Acceptable statistical data is the main basis for hydrological studies. Because thereare lots of continuous and disperse blanks in most of hydrological data such as riverdischarge, it is necessary to estimate and forecast these data by suitable methods. Theseblanks are caused by different factors such as loss of data record, elimination ofincorrect data and disordered function of measurement instruments. There are manyprocedures to estimate and regenerate these data, and depending on the condition of agiven station, a particular procedure may produce the best results. In this study, themethod of artificial neural networks has been compared to other methods includingnormal ratio, graphical, simple linear regression, multivariate linear regression and timeseries (auto regression) methods, in order to regenerate the monthly and annualdischarge data for hydrometric stations in Grand Karoon river basin. After eliminatingobserved data, their values were estimated using mentioned procedures. Then, thepriority of each procedure was assessed by means of rooted mean square of errors(RMSE). The results of monthly data regeneration indicated that artificial neuralnetworks was the best procedure in most stations with a frequency value of 59.26%.