Abstract
Nine different geometries with asymmetric rectangular compound cross-sections were tested in a laboratory flume in order to investigate effects of the geometry on the values of three different dimensionless velocity ratios. A generalized single variable regression model has been derived to predict each of the three dimensionless velocity ratios with high accuracy. Another set of the multiple-variable regression models has been derived using one additional dimensionless parameter which is the ratio of the upstream channel width to the main channel width. The application of several key statistics and validation procedures has indicated that the developed models predict the three mean dimensionless velocity ratios with high correlation coefficients.