open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear TELEMAC-2D experts / Users,
I'm looking for some advices from the Telemac community.
I'm trying to run a tidal model over fjords / channels so I cannot use any of the tidal databased (e.g. TPXO) as boundary conditions. thus I'm using surface elevation time series obtained from harmonic analysis performed using measurements at tidal stations, combined with the Thompson method to obtain the velocity.
Some channels have a steep bottom along the cross-section.
I have managed to run the model for at least 2 days, but I have noticed at some nodes (not all) along some boundaries that have parallel (along the boundary) velocities, plus unrealistic velocities (too high).
Also, changing some parameters / numerical schemes I can get quicker or slower these unrealistic velocity conditions.
These instabilities and how much the results change when changing some numerical parameters make me believe that even if I get my model to run stable, I cannot trust on the results.
The most relevant settings I have changed so far, trying many combinations, are:
VELOCITY PROFILES
TYPE OF ADVECTION
SCHEME OPTION FOR ADVECTION OF VELOCITIES
SUPG OPTION
IMPLICITATION FOR DEPTH
IMPLICITATION FOR VELOCITY
SOLVER
PRECONDITIONING

I'm attaching my case files (now is not running more than 10 hrs) if someone can take a look to them and point me to the right direction.
Thank you very much
Ignacio

Hello Ignacio,

I cannot see your geometry file if you tried to upload it. Anyway, what about your mesh refinement close to the steep bottoms? Are there steep slopes around the boundaries where you can see high velocities? Can you upload some figures to show your issue?

Have you tried to decrease your time step? What about the Courant number in the whole area? What about the minimum size of elements of your mesh and the maximum velocity expected?

Anyway, looking at the steering file you sent, there are a lot of keywords activated, some are default values, some are useless as not activated with other choices done (e.g. GEOGRAPHIC SYSTEM, ZONE NUMBER IN GEOGRAPHIC SYSTEM, keywords related to tracers as none tracers...). I think you should simplify this steering file to be clearer. Have you tried starting from a minimal configuration with the less number of keywords (the other let to default values)? Can you see the same unrealistic velocities?

What about not using Thompson conditions (but the classical boundary conditions) nor Spalart-Allmaras turbulence models (but another one, just to start a constant viscosity or later a more complex one)? Same issue with caracteristics (= 1 for advection schemes)

Hope this helps,

Chi-Tuan

Hi Chi-Tuan
Thank you very much for your response.
Because of the large files, I uploaded the entire case files into a google drive folder: drive.google.com/drive/folders/1g8WTUf1h...7-kz6EVO?usp=sharing

would be awesome if you can take a quick look at them.

I meshed the domain following some rules that I made up: the mesh is finer at steeper slopes, at shallower waters, closer to the open boundaries and nearer the zone I'm interested in (although the weight to the steep slope rule is not as high as I wanted because I was going to end up with too many nodes).

The smallest distance between nodes is 10 m and the longest is 200 m.

The bathymetry near the boundary has a special treatment. I smoothed it a bit and truncated it to have depths of at least 10 m. I smoothed the transition from the boundaries to the 'real' bathymetry in 500 m.

I started with a very simple steering file, but as I wasn't getting good results I started adding keywords that I thought might help with precision and / or stability as I was reading more in depth the user manual and the forum.

I have also tried not to use Thompson (OPTION FOR LIQUID BOUNDARIES : 2;2;2;2;2;2 and 1;1;1;1;1;1) and the different options for the velocity profile (VELOCITY PROFILES: 1;1;1;1;1;1 or 4;4;4;4;4;4 or 5;5;5;5;5;5) with no success.

I tried different turbulence models, including constant viscosity (TURBULENCE MODEL : 1) and realized that this option is not the one that is making my model unstable at the boundaries.

I was trying to avoid TYPE OF ADVECTION = 1 because it is not conservative (or less conservative than other methods) and also I did try it with no success.

I'm feeling like I'm running out of options, but I'll keep trying. I'll incorporate your suggestions.





Thank you!!

PD: I'm attaching some figures showing the issue.