Filling missing suspended sediment data by artificial neural networks

Cigizoglu H.

14th International Conference on Computational Methods in Water Resources, DELFT, Netherlands, 23 - 28 June 2002, vol.47, pp.1645-1652 identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 47
  • City: DELFT
  • Country: Netherlands
  • Page Numbers: pp.1645-1652
  • Istanbul Technical University Affiliated: No


Establishing sediment monitoring instruments on rivers is a costly operation. There might be some gaps on the continuous data due to several reasons like disturbance of the sediment measurement device by the sediment material on the river or a mechanical problem on the computer recording the sediment measurements. The methods available in literature for sediment concentration estimation are complicated, time consuming and necessitate cumbersome parameter estimation procedures. 1 this study, artificial neural networks (ANN's) were employed to forecast and estimate the missing sediment concentration values. The data employed for training and testing the ANNs were the suspended sediment and flow data of different rivers in the nearby catchments. The forecasting results obtained using the previous observed sediment values were close to the real ones. The sediment concentration estimation, on the other hand, using the observed river flow values as input provided realistic approximations in terms of mean squared error (MSE) and total sediment amount. The ANN estimates are compared also with those of the corresponding sediment rating curve and found significantly superior.