Artificial Neural Networks Potential in Multi-Station Modeling of Suspended Load in Comparsion with Sediment Rating Curve Method

Sediments transported by river may cause damages to cultivated land and hydraulic structures. Accurate estimation of sediment load for hydraulic structures (e.g. dam) can prevent extra costs. Because of the existence of many rivers, our country, Iran, has high potential for dam construction. On the other hand, flood disaster causes huge damage every year. The main reason for magnifying the effects of this disaster can be related to the reduction of water conveyance capacity of the rivers because of sediment deposition. Therefore, the correct estimation of the transported sediment will be highly important. Prediction of the suspended sediment load can be accomplished by the Artificial Neural Networks (ANNs). In this study, ANNs are used to estimate suspended sediment load in Akhola station, located on the Ajichay River in East Azarbaijan, Iran. The available data for this station were daily discharge and sediment load The ANN sensivity for these parameters was examined in the modeling. In order to evaluate the effect of the upstream stations load, the data of Markid and Vanyar stations were also used to train the network, which led to more accurate result. The classic rating curve method was also used to estimate the sediment load at this station. To optimize the coefficients of the rating curve, the genetic algorithm was employed, its result of caerse did not show superiority on the classic optimization method. Regarding these results, multi-station estimation using ANNs has better efficiency.