Application and validation of numerical codes in the prediction of compound channel flows

This paper attempts to compare the accuracy of a two-dimensional (2D) depth-averaged code (TELEMAC-2D) and a three-dimensional (3D) code (SSIIM) in the numerical simulation of free surface flows. Assessment was conducted using benchmark data collected from a meandering compound channel at the UK EPSRC Flood Channel Facility. For both codes a standard k–ε turbulence model was employed to simulate the flow field. Assessment of both codes was based on the prediction of depth-averaged velocity traverses around a meander bend, flow features inferred by velocity vectors in the horizontal plane, and water surface elevations for both low and high relative flow depths. In addition, predictions with the 3D code of velocities parallel and perpendicular to the main channel direction are compared with those observed in the experimental programme. Both codes predict the lateral distribution of depth-averaged velocity in the main channel with reasonable accuracy at low relative flow depth, with the 3D code predicting better than the 2D code (maximum percentage errors in depth-averaged velocity are 16% and 22% respectively). At a higher flow depth no extra accuracy in the prediction of depth-averaged velocity is to be gained from the 3D model than from the 2D model, although the 3D model can predict the correct location and direction of secondary currents. For all flow conditions better accuracy was achieved for both codes at the apex section where the mean flow direction is more aligned with the local main channel direction. An accuracy of within 20% can be expected for predictions of depth-averaged velocity magnitude from a 3D model using a standard isotropic representation of turbulence.