Authors: Ali Osman Pektaş
Addresses: Department of Environmental Engineering, Bahcesehir University, 34349 Besiktas, Istanbul, Turkey
Abstract: Appropriate prediction of sediment load concentration being carried by streams has a vital importance of water resources quantity and quality studies. In most studies some dimensionless parameters are derived by using observed variables of sediment system and then used as inputs of predictive models. In this study, instead of deriving new variables, widely used non-dimensional sediment model parameters have been compiled and examined. The main purpose of the study is to decide the essential parameters to establish effective models in predicting for both bed load and suspended sediment load. Cluster analysis, principal component analysis, multiple regression analysis and sensitivity analysis in artificial neural networks are used to determine the most influential parameters. The results of all methods are evaluated together and the parameters that are found significant are detected as the most relevant parameters.
Keywords: bed load; suspended sediment load; cluster analysis; principal component analysis; PCA; multiple linear regression; artificial neural networks; ANNs; water resources; water quality; sediment modelling; predictive models.
International Journal of Global Warming, 2015 Vol.8 No.3, pp.335 - 359
Received: 01 Mar 2014
Accepted: 09 Apr 2014
Published online: 23 Oct 2015 *