Estimation of sediment discharge using a tree-based model

ABSTRACT The model tree (MT) approach, a data mining technique used to analyse relationships between input and output variables in a disordered and large database, was adopted in this study to predict sediment discharge with field measurement data. The derived models were analysed for accuracy according to the goodness of fit based on training, testing, and modelling processes. When the flow velocity, depth, water surface slope, channel width, and median bed material were selected as the river’s system variables, the model results of sediment discharge resembled the measured values. The results demonstrate that developing and using the sediment discharge estimation with the MT constitutes the most effective method if long-term sediment data are of sufficient validity.