Monthly streamflow predıctıon of euphrates basin by using FFNN, ANFİS and LSTM models

In order to use water resources effectively, it must be controlled and manage accurately. If we think of the rapidly increasing urbanization, water structures and their surroundings, population growth and the demand for water, streamfllow estimate must be investigated in detail and accurate data should be obtained. İn short, to control the water, it is necessary to know the current and future potential of it. In this study, calculations had been done using past months streamflow as input. Generally, two types of models are used for streamflow prediction, conventional and artificial Intelligence (AI). Unlike traditional methods, prediction of nonlinear and nonstationary streamflow with AI models can give more accurate results. In this study, for the monthly streamflow estimation of the Euphrates Basin three different Aİ models; Feed Forward Neural Networks (FFNN), Adaptive neuro fuzzy inference system (ANFIS) and Long Short Term Memory Networks (LSTM) had been used. For this purpose, data between 1981-2011 at four stations in the Euphrates Basin were used. To run FFNN and ANFIS models, Matlab program was used and to rum LSTM model, Payton program was used. İn the end, the results were compared and it was found out that ANFIS was superior compared to LSTM and FFNN, Secondly, LSTM showed better performances than FFNN. To compare the performances of the models, Three of the most recommended model performance criterias; Mean Absolute Error (MAE), Correlation Coefficient (R) and Nash Sutcliffe Efficiency (NSE) were used. Keywords: ANFIS, FFNN, LSTM, Streamflow Prediction