Forecasting streamflow using base flow separation technique

In this study, daily rainfall-runoff simulation were discussed during 2000-2005 Khormaloo, Gharechay, Chehelchay Rivers in Golestan province using adaptive neuro-fuzzy inference system (ANFIS). First separate base and surface flow using a recursive digital filter. First separate base and surface flow using a recursive digital filter. Then using discharge in antecedent days and meteorological parameters such as temperature and rainfall, base and surface flow simulated separately using ANFIS and the best entries were determined. Total flow was determined by combination of best model of base and surface flow by ANFIS. The results of the combination of base flow and surface flow models and finally total flow modeling showed in all three stations, the model that combination of base flow and surface flow on before day has highest coefficient of determination and minimum error. Coefficient of determination in the test phase for Nodeh Khandooz, Ramian and Lazoore stations is 0.691, 0.991 and 0.999 respectively and error in stations is 1.413, 0.131 and 0.036 respectively. The results showed that the use of separation base flow technique increase the speed and accuracy of simulation and eliminates the problem of low speed execution in ANFIS and base and total flow predicted accurately.