Author(s): Mozafar Ansari, Faridah Othman, Ahmed El-Shafie
Linked Author(s): Mozafar Ansari
Keywords: River discharge prediction, Artificial Neural Network (ANN), Jajrood River, Latyan Dam, Iran
Abstract: Predicting and modeling river discharge is one of the most important issues for any hydrological studies. Accurate discharge prediction leads to better performance in other operation or studies such as reservoir operation or flood plain mapping. Due to this importance, many studies have been done to improve the performance of hydrological models. Among of these models, artificial intelligences (AIs) are the state-of-the-art methods which bring proper results with limited variety of data. In this study, ANN model was applied on Jajrood River at Latyan Dam basin in Iran. Sixteen years of daily precipitation and discharge were the main inputs of the model. Beside these, due to seasonal rainfall in arid and semi-arid area, months of the year were also applied in the model. Four scenarios of inputs, which had different elements, were determined and results of these states were compared. Results showed that in the first one, which had months of the year and three days of rainfall and six days lagged data of previous discharge had the lowest root mean square error (RMSE) and the highest correlation coefficient (R) result among other scenarios. Therefore, months of the year could be alternative data for all elements which change during the year such as wet season, soil moisture condition, and etc. which are not available or not easy to be obtained
Year: 2017