No abstract

River flow estimation using records of past time series is importance in water resources engineering and management and is required in hydrologic studies. In the past two decades, the approaches based on the artificial neural networks (ANN) were developed. River flow modeling is a non-linear process and highly affected by the inputs to the modeling. In this study, the best input combination of the models was identified using the Gamma test then MLP-ANN and hybrid multilayer perceptron (MLP-FFA) is used to forecast monthly river flow for a set of time intervals using observed data. The measurements from three gauge at Ajichay watershed, East Azerbaijani, were used to train and test the models approach for the period from January 2004 to July 2016. Calibration and validation were performed within the same period for MLP-ANN and MLP-FFA models after the preparation of the required data. Statistics, the root mean square error and determination coefficient, are used to verify outputs from MLP-ANN to MLP-FFA models. The results show that MLP-FFA model is satisfactory for monthly river flow simulation in study area.

No abstract

Accurate simulation runoff process can have a significant role in water resources management and related issues. The inherent complexity of this process makes difficult the use of physical and numerical models. In recent years, application of intelligent models is increased a powerful tool in hydrological modeling. The aim of this study was the application of the Gamma test to select the optimal combination of input variables for runoff modeling in Sofi Chay. Streamflow modeling was performed based on the optimum number of the selected variables using the artificial neural network (ANN) and Support vector machine (SVM) methods .Gamma test results showed that monthly runoff with six antecedent runoff values provide better results to predict. Runoff simulation using support vector machines and artificial neural network models also showed that the best input structure will be delayed until six to predict of next month runoff. Among to models with the same input structure, support vector machine have relatively high efficiency compared to artificial neural network.

No abstract