Farklı Yapay Sinir Ağı Modelleri Kullanarak Nehir Akımı Tahmini

Accurate estimation of streamflow is crucial for water resources planning, design and management, determining flood drought management strategies, minimizing their adverse effects. In this study, the usability Artificial Neural Network (ANN) models to estimate monthly was investigated. For purpose, data two stations located in Seyhan Basin south Turkey were used. The Sarız River-Şarköy observation station (No: D18A032) meteorology 17840) precipitation flow used belong period 1990-2017. Nine input combinations consisting lags obtained ANN models. We techniques, namely Multilayer Perceptron (MLP) Radial Basis Networks (RBNN) streamflow. MLP technique, three learning algorithms with gradient descent momentum adaptive rule backpropagation (GDX), Levenberg-Marquardt (LM) resilient (RBP) parameters each different model by using nine trial error. success evaluated five performance metrics. Which more successful decided according combination highest Nash Sutcliffe efficiency coefficient (NSE) value test period. Although similar results MLP-GDX, MLP-RBP, MLP-LM RBNN models, (except MLP-LM) slightly than most MLP-GDX-M6 model. model, MAE=1.148 m3/s, RMSE=1.815 R2=0.724, NSE=0.717, CA=1.069 testing novelty study that we have examined credibility including predicting natural rivers.