Using a statistical efficiency methodology for predictors’ selection in the bedload transport problem: A high gradient experimental channel case

Abstract

Bedload transport rates for high-gradient gravel bed rivers has been studied through a physical model that replicated the typical features of these channels. A stepwise regression was performed to identify the best predictors from a set of independent variables. As independent variables channel slope, the ratio of area occupied by large particles to the total plan area, flow discharge, mean flow depth, mean flow velocity, water surface velocity, boundary shear stress, and shear velocity were considered. Different characteristic diameters (d16, d50, d84, and d90) were used to nondimensionalize the variables as well as to test the influence of grain size. A linear and a potential model were obtained for each characteristic diameter. Based on the correlation coefficients (R2) with the data used to build the models, the d50 and d84 linear and potential models were selected to perform further analysis. A set of independent data was used to verify the selected models. Better performance was observed for the potential models with 96% of the data falling within ½ order of the magnitude bands both for d50 and d84 . R2 for the d50 and d84 potential models were 0.63 and 0.76, respectively. Therefore, the d84 potential model can be selected as the present study representative model.

Keywords

• High gradient
• Bedload transport
• Akaike-information-criterion
• Best predictors
• Laboratory experiments