open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear users and professors,

I am trying to simulate a freshwater discharge model in a shallow water bay. In order to analyze the plume, I set salinity as the only tracer. The results on the surface are very close to the observed salinity distribution.

But the results of the 3d layering are not good. It seems that the vertical numerical diffusion and mixing process are too fast, resulting in no obvious salinity stratification. The layer surface and bottom are almost same.

I will attach my steering file and fortran file of meshtranf.f. In my case, this area is shallow (the deepest point is 4m), and its has limited capacity of water exchange with the open sea, forcing by local wind a lot.

I dont't know what is the better turbulence model and advection schemes to use and what parameters should I tune to make the salinity stratification results close to the reality. Or there is a problem with the parameter settings of my model, which leads to bad results.

Thank you for your Help
Regards
Yang

Hello!

Stratification is highly mesh dependant, so increasing the mesh resolution in the vertical direction can improve your results.

But the first I would try is to use k-epsilon (with appropriate diffusion coefficients, 1E-6) and lowering the solver precision for tracer diffusion (1E-12).

I you want to keep your turbulence settings, you need to use a damping function to help with stratification.

Hope it helps!

Kind regards
PL

Hi, PL

Thank you very much for your suggestions and guidance, I will try as soon as possible.

But for vertical layers that are highly dependent on mesh, which means that only need to increase the number of layers, it that true? Do I need to use meshtranf.f to make the bottom layer more dense? o other distribution.

Thanks you in advance
Best regards
Yang

Hi!

Yes, the idea is to increase the number of planes with a smaller vertical distance between them, especially in the zone of your halocline. But first thing to try would be the turbulence models / damping functions.

Kind regards
PL

Hi

Thanks for your explanation
i will optimize these two parameters first

Regards
Yang

Hi PL

I am trying to keep the mixing length model but adding damping friction = 3: Musk, also increased the number of vertical layers at the same time to make the bottom mesh is more dense.

At first it worked well, but after 1 hour the model always has exceeded the iterations. I have reduced the diffusion accuracy by 1.E-12 and the maximum number of iterations is also quite large. In addition, my time step is 60s. Is this normal? Increased damping friction will cause the model Is it difficult to converge?

Thanks for your helping
regards
Yang

Hi!

The convergence issue is probably due to the mesh refinement in both horizontal and vertical direction while keeping the same time step. I don't know the typical mesh size and flow velocities you have but 60 s sounds large if you are not modelling ocean with very large meshes...

Kind regards
PL

Hi PL

I tried many different configuration models, and indeed that considered the dumping function is useful, but because the water depth of the area is too shallow, the result still not very good, I was thinking that it is possible to modify the subroutine of drialg.f about dump function to optimize the vertical stratification of tracer. Do you know any examples or experiences that can help me achieve that？

Best regards
Yang

Hi Yang,

Unfortunately, I don't have experience for this. Maybe try to describe more into details what needs to be improved (text and figures) so that other users can hopefully chime in and suggest things to try!

Good luck!

Kind regards
PL

Hello Yang,

Have you tried the different damping function available in standard TELEMAC-3D? Viollet, Munk and Anderson?

You can implement your own damping function with the user subroutine USER_DRIUTI.

You can try to implement e.g. Toorman damping function for both velocity and tracers variables or only for one of them.

Hope this helps,

Chi-Tuan