open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

When modeling salinity dispersion with k-epsilon model, I've found a problem which puzzles me a lot: as the default setting of Telemac-3D, if ITURBV.EQ.3(K-Epsilon), then ITURBH also equals to 3. With this option, "horizontal viscosity(diffusivity) = vertical viscosity(diffusivity) + molecular viscosity(diffusivity)". In my case, calculated vertical viscosity(diffusivity) is rather small(1.0E-2~1.0E-6)which is normal for vertical direction. However, for horizontal direction, those values are too small and I think this is the reason of abrupt change of salinity filed(see 01.png). I want to use Smagorinski model for horizontal but really don't know how...
Has anyone expericed this problem, why does TELEMAC refuse to use a combination of k-epsilon and Smagorinski, thanks a lot !

With best regard.
Robinson

Hello,

So far we refused to mix k-epsilon on one direction and another model in the other, this is why all directions are done in k-epsilon if one of them is asked (the control is in lecdon_telemac3d.f). This is just for lack of background theory, to understand what we do. However we are in the context of high aspect ratio and I am not sure that k-epsilon has really been validated in this framework, where vertical gradients are much higher than horizontal gradients for purely artificial reasons of mesh size. It seems that users have suppressed the control and that it works with mixtures of turbulence models. In your case have you tried, to start with, constant diffusion on the vertical (model 1) and mixing-length model on the vertical (model 2) ? This one works. To my opinion, k-epsilon with high aspect ratio is not very far from mixing length.

With best regards,

Jean-Michel Hervouet

Hi,

Thanks a lot for your explanation.
In my case study, I have tried constant diffusion in the vertical and mixing length model in the vertical, with Smagorinski horizantally. And I've found the resulted horiztal salinity filed is quite different from the one with k-epsilon model(horizontal and vertical) and I guess it is because of those too small horizontal diffusion from k-epsilon. Maybe a mixture of k-epsilon and other turbulence models is lack in background theory, but many numerical models still do this and it seems work well. So I really want to try a mixture use of k-epsilon and other turbulence models to see how it works. I've tried to simply modify related codes in lecdon_telemac3d.f but really not sure. Could you tell if this is possible ...
Thanks again!

With best regards.

Hello,

I think I've found my problem in salinity simulation. It's about the Coriolis force. In FUDAA for Telemac-3D, it does not contain the key word 'CORIOLIS = YES/NO', but only 'CORIOLIS COEFFICIENT'. Add it to the CAS file, surface salinity field seems much better.

with best regards.
Robinson