Estimation of sediment transport using artificial intelligence methods in the Kızılırmak basin

Accurate estimation of the amount of sediment is very important, as it leads to faster depletion of the economic life of dam reservoirs. In order to prevent the decrease in the economic life of the dead storage and to reduce the sedimentation in the dam reservoirs, it is necessary to determine the sediment which is threatens reservoirs sustainability, and carried by rivers. There are many geological, topographic and climatologic factors affecting the sediment. Because of the multiplicity of these factors and the complexity of their relationships with each other, it is quite difficult to calculate the amount of sediment carried by any river. Recently, Artificial Neural Networks (ANN) are widely used to solve the complex problems such as sediment. In this study, the data of flow (m3/sec) and the amount of transported sediment (ton/day) data of 4 different observation stations that are in the Kızılırmak River Basin is used for the prediction of the sediment. In the sediment estimation, 10 different input combinations consisting of different flow streams and SSL data were used. These neural network methods developed by using those data are compared with the results of the sediment rating curve method. It was also accomplished by applying discrete wavelet transform. The success of ANN and Wavelet-ANN models were determined according to Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Nash-Sutcliffe Efficiency Coefficient (ENash) and Determination Coefficient (R2) performance criteria. As a result of the study, it was determined that wavelet-ANN methods were generally more successful than sediment estimation in all stations except Kızılırmak-Söğütlühan Station. Keywords: Discrete Wavelet Transform, Kızılırmak River, Sediment, Sediment Rating Curve, Artificial Neural Network