Modeling of suspended sediment concentration using conventional and machine learning approaches in Thames River, London, Canada

Issam Mohamed and Imtiaz Shah


Water resource management, planning, hydraulic design, environmental conservation, reservoir management and operation all require reliable information and data about Suspended Sediment Concentration (SSC). To predict such data, direct sampling and Sediment Rating Curves (SRC) are commonly utilized. Since direct sampling can be risky during extreme weather events and SRC cannot provide satisfactorily dependable results, engineers are trying to propose new precise forecasting approaches. Various soft computing techniques have been applied to model different hydrological and environmental problems and have showed promising results in this regard. Although many studies have been performed to simulate the phenomena of SSC at numerous rivers and creeks in the literature, the SSC is a site-specific problem. In this study, Adaptive Neuro Fuzzy Inference System (ANFIS) and Artificial Neural Network (ANN) models were proposed and compared with the conventional SRC and linear regression methods. Using different combination of measured data from 1993 to 2016 of SSC and simultaneous Stream discharge, Water Temperature, and Electric Conductivity for Thames River at Byron Station, London, Canada, several models were trained. Goodness of each model was evaluated using Mean Absolute Error, Root-Mean Square Error and Nash-Sutcliffe Efficiency Coefficient. Results show that ANN models are of a superior accuracy if compared with other approaches in predicting SSC for this river.

Permanent link: