Looking at the SWE equations and at some discretization I found that it was argued here and there that the diffusion part can be neglected since it is somehow contained in the bottom roughness part. Clearly, turbulent diffusion takes energy out of the system but in contrast to the roughness term this is marginal. I would rather say that turbulent diffusions is mimicked by numerical diffusion of lower order schemes and that in the absence of numerical diffusion higher order schemes can be terribly wrong if turbulent diffusion is neglected. Moreover, I would really questions all this simplifications and rather go for even solving for the transport part of turbulence since this is important in 2d models for the horizontal advection of turbulence, which is quite important for many applications like sediment transport or simulation of 2ndary currents. Otherwise, I think that eddying regimes cannot be really reproduced by such crude simplifications. However, I did now wrote several SWE codes on unstructured grids during the last year and I am just now starting to take a deep look into it and therefore I would be thankful for some discussion in this direction.