ABSTRACT
Turbulence within rivers plays a significant role in determining the mean flow velocity field and is integral to the overall energy loss mechanism. With the continued advance of computational hardware, rivers are now frequently modelled in 2D (rather than 1D) using the depth-averaged Navier-Stokes equations (i.e. Shallow Water Equations), often with higher order schemes and sometimes at cell resolutions less than the flow depth. With this trend, some treatments of turbulent eddy viscosity can lead to mesh size dependency in the results and significant errors in the prediction of inundation levels on the surrounding floodplains. Investigations into the issue of mesh-size sensitivity and the convergence of model results to physical test data for five turbulence models were performed, utilising three benchmark case studies from laboratory to real-world scale. In each case model results were compared with calibration data, with emphasis on head loss predictions.
