Development of a machine learning model for river bed load

Abstract

Prediction of bed load sediment transport rates in rivers is a notoriously difficult problem due to inherent variability in river hydraulics and channel morphology. Machine learning (ML) offers a compelling approach to leverage the growing wealth of bed load transport observations towards the development of a data-driven predictive model. We present an artificial neural network (ANN) model for predicting bed load transport rates informed by 8117 measurements from 134 rivers. Inputs to the model were river discharge, flow width, bed slope, and four bed surface sediment sizes. A sensitivity analysis showed that all inputs to the ANN model contributed to a reasonable estimate of bed load flux. At individual sites, the ANN model was able to reproduce observed sediment rating curves with a variety of shapes without site-specific calibration. This ANN model has the potential to be broadly applied to predict bed load fluxes based on discharge and reach properties alone.