Suspended Load Modeling of River using Soft Computing Techniques (Case Study: Dareh-rood River)

The movement and deposition of the suspended load of rivers cause different problems; such as sedimentation in reservoirs, changing of river morphology based on sedimentation in the river bed, reducing the capacity of channels and water conveyance structures and changing of water quality for drinking and agricultural usage. In this study, adaptive neuro-fuzzy inference systems (ANFIS), gene expression programming (GEP) and support vector regression (SVR) was utilized for modeling and forecasting of suspended load for Dareh-rood catchment in Ardabil province. For this purpose, water discharge and sediment load of Moshiran hydrometric station located on Dareh-rood River (upstream of Emarat dam) was used. After evaluation of different input combinations (i.e. 8 scenarios) using the SVR, finally the model whose inputs consist of current discharge and one pervious discharge and suspended load was selected as the best scenario. The mentioned input combination was applied for ANFIS and GEP models. The results indicated that the SVR model with the highest R2=0.97, the lowest RMSE=1734, NS=0.97 and WI=0.98 was superior to the other models during the validation phase. Furthermore, the frequency distribution and boxplots of forecasting errors of applied the datadriven models confirm the efficiency of the SVR model among others. Meanwhile, the performance of the ANFIS model was somewhat better than the GEP model. The coefficients and functions used to calibrate the intelligent models that utilized in this study can be very helpful in estimating the suspended sediments of ungagged nearby stations with the similar tectonic and hydrological conditions over the study region.