Models for Predicting River Suspended Sediment Load Using Machine Learning: A Survey

Abstract

Suspended sediment load (SSL) prediction study is critical to water resource management. This paper presents studies related to the prediction of SSL using machine learning (ML) algorithms over the last 13 years. This research gives a survey of current studies that are used machine learning techniques to predict sediment load on several rivers in different reign. Also, it aims to find a performance model to predict the SSL. This is done by making comparisons between several studies that used machine learning techniques to predict sediment load on several rivers using different time scales. Several metrics were used to determine the best prediction model. Most of the metrics used are: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), R-Squared (R2) and Nash-Sutcliffe Efficiency Coefficient (NSE). The results of comparisons using different ML algorithms to predict the SSL have shown that the Multilayer perceptron (MLP) algorithm is the best compared to other algorithms.