This paper studies laboratory open-channel confluences using a 3D,
elliptic solution of the Reynolds-averaged Navier-Stokes equations,
including a method for approximating the effects of water surface
elevation patterns and a renormalization group modified form of the
k-epsilon turbulence model. The model was tested by comparison with
laboratory measurements of an asymmetric tributary junction. This
suggests that although the model is unable to reproduce the
quantitative detail (notably upwelling velocity magnitudes) of the flow
structures as measured in laboratory experiments, statistically
significant aspects of the experimental observations are reproduced.
The model is used to (1) describe and explain the characteristic flow
structures that form in a confluence with one of the tributaries angled
at 45 degrees, both with and without an elevation difference Cbed
discordance) in the angled tributary; and (2) investigate the relative
importance of junction angles (30 degrees, 45 degrees, and 60 degrees),
bed discordance, and ratio of mean velocities in the tributary channels
upon flow structures. This shows that bed discordance significantly
enhances secondary circulation because of the effects of flow
separation in the lee of the bed step, which significantly increases
lateral pressure gradients at the bed and reduces water surface
superelevation in the center of the tributary and water surface
depression at the downstream junction corner. Extension to
consideration of a number of junction angles, levels of bed
discordance, and velocity ratios suggests that a small (10%) reduction
in tributary depth can significantly increase the intensity of
secondary circulation, albeit in a relatively localized manner.
Simulations involving a numerical tracer illustrate the importance of
bed discordance for mixing between the two flows and question the use
of simple 2D parameterizations of mixing processes that do not consider
bed discordance when the latter is present.