River confluences are complex hydrodynamic environments where convergence of incoming flows produces complicated patterns of fluid motion, including the development of large-scale turbulence structures. Accurately simulating confluence hydrodynamics represents a considerable challenge for numerical modeling of river flows. This study uses an eddy-resolving numerical model to simulate the mean flow and large-scale turbulence structure at an asymmetrical river confluence with a concordant bed when the momentum ratio between the two incoming streams is close to 1. Results of the simulation are compared with field data on mean flow and turbulence structure. The simulation shows that the mixing interface is populated by quasi-two-dimensional eddies. Successive eddies have opposing senses of rotation. The mixing layer structure resembles that of a wake behind a bluff body (wake mode). Strong streamwise-oriented vortical (SOV) cells form on both sides of the mixing layer, a finding consistent with patterns inferred from the field data. The predicted mean flow fields show that flow curvature has an important influence on streamwise variation of circulation within the cores of the two primary SOV cells. These SOV cells, along with vortices generated by flow over a submerged block of sediment at one of the banks, strongly influence distributions of the streamwise velocity and turbulent kinetic energy downstream of the junction. Comparison of the eddy-resolving simulation results with predictions from the steady Spalart-Allmaras RANS model shows that the latter fails to predict important features of the measured distributions of streamwise velocity and turbulent kinetic energy because the RANS model underpredicts the strength of the SOV cells. Analysis of instantaneous and mean bed shear stress distributions indicates that the SOV cells enhance bed shear stresses to a greater degree than the quasi-two-dimensional eddies in the mixing interface.