Estimation of suspended sediment concentration using regression analysis and artificial neural networks: A case study from the Eastern Black Sea Basin, Turkey


BAKİ O. T. , ERDOĞAN İ., BAYRAM A.

The International Civil Engineering & Architecture Conference (ICEARC 2019), Trabzon, Turkey, 17 - 20 April 2019, vol.-, pp.886-893

  • Publication Type: Conference Paper / Full Text
  • Volume: -
  • City: Trabzon
  • Country: Turkey
  • Page Numbers: pp.886-893

Abstract

In this study, it is aimed to estimate the suspended sediment concentration (SSC) from the stream flow rate (Q) and turbidity (T) measurements. The SSC, Q, and T were monitored seasonally in situ during a period of three years from 2015 to 2017 in the Aksu, Söğütlü, Karadere, Solaklı, İyidere, and Fırtına streams in the Eastern Black Sea Basin, Turkey. The conventional regression analysis (CRA) and artificial neural networks (ANNs) were used for modeling. The CRA was performed using three different equations. In addition to different learning rate and momentum, two different modeling algorithms, namely Levenberg-Marquardt (LM) and gradient descent with momentum, were used in the ANNs. In all models, four stations were used in the model training, one station was used in the validation phase and the remaining station was used to test the model. The estimation models were compared with root mean square error, mean absolute error, and determination coefficient. It has been seen that the ANNs are the best result in applied regression analysis and the ANNs models. The best model has succeeded in realistically reflecting the change of the peak values. It was concluded that it has applicable results on different streams.