Water resources are needed to maintain the human life and the management the ecologic system for many areas. The most economical use, protection and development of water resources have a great importance for hydrological studies. Variable such as stream flow data are commonly used in hydrology. Accurate stream flow estimation is very important in terms of planning and management of water resources and minimizing the effects of natural disasters such as drought and flood. Monthly river flow data obtained from the Sakarya basin on Porsuk River between 1970-2000 years were used for the estimation study. For this purpose, forecasting performance has been analyzed using Adaptive Network Based Fuzzy Logic Inference System (ANFIS) and Artificial Neural Networks (ANN) models and performances of these two models were compared. In addition, the average monthly stream flow data, standard deviation values of these data were also used in the forecasting study and applied as an input to ANFIS and ANN models. For a one ahead estimation, models have been developed with different input combinations of 1-3 past value of stream flow data and standard deviation values. In this study, mean square error (mse), mean absolute error (mae) and correlation coefficient parameters were used to evaluate the performance of the models. According to the obtained results, it is seen that the ANN model has better forecasting performance for two inputs according to mse and mae parameters and for three inputs according to R and R2 parameters. Also, it is seen that the ANFIS model has the best performance for two inputs according to mse, mae, R and R2 parameters. There has been some improvement in the forecast performance if the monthly average river stream flow data as well as the standard deviation data has been applied as an input to the model
Stream flow forecasting, Adaptive Network Based Fuzzy Logic Inference System Method (ANFIS), Artificial Neural Networks (ANN)