A Comparison of Suspended Sediment Load Methods and MLP Neural Network Model: A Case Study of Kashkan- Poldokhtar Station, Lorestan Province, Iran

In this study, to estimate the daily suspended sediment of Kashkan Poldokhtar hydrometric station in Karkhe river basin, the discharge and sediment data between the years 1972 to 2011 were collected. To this end, after generating the missing data, the normality of the data was assessed, and finally estimation by ANN models using multilayer perceptron (MLP) was done. The results of ANN method were compared with those of some models such as linear, median of the data set, monthly and seasonal, hydrological similar period and finally decreasing and increasing periods. The results of ANN method showed a good correlation for evaluating sediment rating curve approaches. Among the methods used for sediment curve, median of the data set was more accurate, for this method gives more weight to the higher flows and has the highest correlation (R2 = 0.98) which is even higher than the artificial neural network method. Moreover, by decreasing the number of points to the least possible, the error caused by conversion of logarithmic functions that depend on the number of distribution points somehow reduces (MSE=0.012).