Prediction of Sedimentation in a Watershed Using RNN and SVM

Abstract

Sediment transport in rivers generally occurs at time of severe proceedings linked with strong precipitation and high flow of rivers. Traditional ways to collect data in high-risk conditions are dangerous and expensive when contrasted to measurement of water discharge. Because of a variety of controlling aspects on river sediment transport, shaping a suitable input arrangement to develop suspended sediment load (SSL) model for forecasting sediment capacity is extremely significant for water resources management. Present work emphasizes on applicability of support vector machine (SVM) and recurrent neural network (RNN) for its appropriateness in modelling relation amid river stage, discharge and sediment load. Model efficiency was assessed utilizing root mean square error (RMSE), mean square error (MSE) and coefficient of determination. Outcomes of SVM were contrasted with those of ANN and it could be seen that SVM can be utilized as a competent tool to predict sediment yield.