open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all!

I am looking for insights regarding the influence of bathymetry variations with wind driven currents in 3D. The resulting currents are normally higher in shallow areas due to the fact that the ratio between wind stress to gravity term is larger than in deeper areas (where gravity term is larger). Such a difference is observed not only for the depth averaged current but also for the surface current which is somewhat less intuitive (one might expect the surface current to be less influenced by bathymetry).

I made some quick tests using a closed canal with a parabolic bottom profile so that the max depth is 10 m whereas the depth near the banks is approx. 1 m. The aim was to reproduce a recirculation current in the model. I tested two different turbulence models: 1) k-epsilon and 2) Smagorinsky (UV) + Tasnis’ mixing length (W). The wind blows parallel to the canal at 10 m/s. See results below.

With both set-up we obtain stronger currents near the banks which was expected. But there is a strong difference regarding the surface current in the deeper part of the section (greater with set-up 2). By looking at the viscosities, I see that in the central part they are much greater with k-epsilon than with Smago + Tsanis which explains why the surface current is lower with k-epsilon.

Does someone have experience of similar set-up? The advantage of k-epsilon is that the wind stress is used to compute the turbulent energy at the water surface (see soukep.f) but the generated turbulence means that the surface current can differ significantly from what can be obtained if using other turbulence models. It is also interesting to observe that with k-epsilon the viscosity descreases in the zone where the ccurrent is reversing (depth of approx. 4 m) whereas it is the opposite with Smago + Tsanis. Another thing is that I would have expected to have no recirculation current along the banks (at depth shallower than 4 m) but the results show that such a current is present on the whole section.

In an ideal world there would be field measurements available in order to help in the decision, but without it, which approach can be considered the more appropriate from a theoretical point of view?

Thank you in advance for any feedback!

Best regards

PL

Hello Pierre Louis

I will not answer to the main question of your post, but in a perfect (numerical) world, if you refine both 2D and z meshes, you will converge to a solution that could be considered as trustable reference solution, independtly of turbulence model you use.

I hope you will find lab/field measurements that could help to validate your investigations.

with my kind regards

Riadh

Hello Riadh,

Thank you for you reply! I wish our world could be a bit more ideal sometimes
Unfortunately, us consultants almost exclusively deal with rather limited material and resources which means that we very often have to make some "educated guesses"...

Best regards
PL