Modeling daily suspended sediment load of rivers using artificial intelligence approaches

Knowing the amount of suspended sediment loda (SSL) carried by rivers is an important factor in watershed management. Hence, it is necessary to measure or estimate its amount accurately. In this paper, it was attempted to apply three artificial intelligence approaches including artificial neural network (ANN), artificial neural network combined with particle swarm optimization (ANN-PSO) and long short-term memory (LSTM) to predict the daily SSL, using the data of Jamal-Beig hydrometric station in Kharestan watershed, Fars province, Iran. To achieve this goal, the daily data of SSL ( Q s ) and flow ( Q f ) were collected during 30 years (1992-2020). Eighty percent of data was considered for the training phase and 20% for the validation phase. The performance of the models was evaluated using the RMSE, R, NSE and PBIAS criteria according to the estimated and measured SSL values. The results showed that the ANN-PSO model with lower values of PBIAS = -1.048% and RMSE = 26.494 ton/day and high values of NSE = 0.827 and R = 0.912 provides the best performance compared to ANN and LSTM models for estimating daily SSL in Kharestan watershed.