Investigating the role of precipitation variable in the performance of daily suspended sediment load modeling (case study: Saied Abad Chai watershed)

Knowledge of the exact amount of daily suspended sediment loads can be used to identify the erosion and sedimentation status of the watersheds. In this research, intelligent artificial neural network and Gene Expression Programming was models used to estimate the daily suspended sediment load. Also, due to the importance of the watershed response to the input variables of the models, in addition to the flow discharge variable, the dynamic precipitation variable was also selected for entering the models due to its influential role in creating erosion and sediment production. The results of this study showed that all the models that used the precipitation variable along with the flow discharge had higher NSE and R2 statistic and lower RMSE and MAE statistics compared to the models that only the flow discharge variable to estimate the suspended sediment load. Also, GEP model with input variable combinations including instantaneous flow, average, average daily flow discharge with a delay time of three days, average daily precipitation and average daily precipitation with a delay time of three days, the most efficient model for estimating the daily suspended sediment load with the highest amount of statistics NSE was 0.90 and R2 was 0.92 and the lowest value of RMSE was 2282.42 (ton/day) and MAE was 750.38 (ton/day) compared to artificial neural network models. In general, the results of this research showed that the flow discharge variable, alone could not properly explain the variance of river sediment. Using precipitation variable as input variable to intelligent models played a significant role in increasing the precision of estimation of suspended sediment load and using precipitation variables In addition to the flow discharge variable, during the modeling process, the efficiency of the models was increased.