|
Abstract
|
In this study, a wide statistical enhancement was performed in the prediction of bed load transport by conducting an extended analysis on 783 measurements taken from the field in various river systems. Numerical performances of three widely used equations were investigated in which the van Rijn equation was giving the best accuracy with R2 = 0.823 and RMSE = 0.673 while the Einstein-Brown equation gave R2 = 0.684 and RMSE = 1.124, whereas Meyer-Peter and Müller equation gave R2 = 0.762 and RMSE = 0.845. A new relationship has been developed incorporating temperature sensitivity and width-depth ratio effects that have not been taken into consideration in any conventional formula. The developed formula has been calibrated by a regression analysis in which eight coefficients were estimated to be highly significant with a probability of p < 0.001. The performance of the model has also been verified at various scales where a similar level of accuracy was obtained for large, R2 = 0.916, medium, R2 = 0.928, and for small rivers, R2 = 0.919. With systematic sensitivity analysis, the dominant factor was Shields parameter with a sensitivity coefficient of 0.845, followed by temperature 0.634 and Froude number 0.567. The new equation improved its intervals of prediction to ± 0.5 log units for 90% of the measurements from the previous equations that lay between ± 0.7 to ± 1.0 log units. It resulted in an overall enhancement of the prediction accuracy by 38.8% compared to the MPM equation and 25.4% relative to the van Rijn equation.
|