Prediction of zonal and total discharges in smooth straight prismatic compound channels using regression modeling

Experimental results are presented concerning the zonal and total discharge distribution and characteristics in a compound channel cross-section comprising one rectangular main channel and two symmetrical floodplains. The discharges in the main channel, floodplains, and total compound channel are found to be highly correlated to several dimensionless parameters defined using the compound channel cross-section geometry. Multi-variable regression analysis was utilized for developing predictive models that can estimate the main channel, floodplains and total discharges as a function of four different dimensionless parameters. The developed models to predict the zonal and total discharges in compound channels are found to be highly significant according to several major statistics including the model standard error, coefficient of determination (R2), and F-statistic. The developed multi-variable regression-based models are also tested for validity using available experimental data. Several statistical tests applicable to the analysis of residuals have indicated the effectiveness of the developed predictive models.