My results above do not quite reflect what the paper shows. 1. Velocity on the free surface is nearly the same over the main channel and the floodplain 2. The observed secondary current is different. (Note: this was visualized in ParaView as the lateral and vertical velocity on the channel cross-section) I also tried a coarser mesh with cells elongated in the streamwise direction. Surprisingly, that case shows a strong secondary current. However, I cannot tell if that result is physically correct or a numerical artifact of the coarse, anisotropic mesh. What should I check or change first? Is there a known rule of thumb for mesh resolution for resolving this kind of secondary current in LES? Honestly, I do not have a clear design principle for what mesh resolution and cell shape is appropriate here. I would appreciate any hint on this. Many thanks!