open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello everybody,

I have a internally coupled T2D-TOM-SIS configuration, designed to model up to 30 years of sand banks migration under combined effects of tide and waves, thanks to HPC resource. As Tomawac is very slow and asks for a lot of CPU resources (even called every 5min to correctly deal with tidal level variations), I use a morphological factor of 10.
My issue is that seabed oscillations gradually appear and grow up during the run, as classically in many morphological simulations.
I consider only one class of sand with a total bedload formulation and advection scheme 14. Sisyphe is called every T2D time step (about 20s).
I already tested to introduce some artificial dissipative effects of bed slope, sliding, or smoothing of the bottom.. without satisfactory effect (too smooth results).
I also have tried the finite volume formulation of Exner equation, but it seems that there are some troubles with it: solid transport values are quasi null (except on dry points). Without Tomawac coupling, the VF formulation seems to work but oscillations are still present.

Any comments or suggestions are welcome.

Would anyone have a similar experience with this kind of long term morphological computation and how do you succeed dealing with it ?

Has anyone ever tested less dissipative scheme as WENO (weighted essentially non oscillatory) schemes or similar ones ? and could you give me some advices or papers please ?

Thank you in advance,
Laurent Leballeur

Hello Laurent,
In some cases, I realised that the correction by skin friction can strongly affect the behaviour of a morphodynamics computations, as it impacts directly on the bed shear stress values computed from the hydrodynamics module. Could you please try to see how this affect your computation, and let us know ?

Best wishes,

Pablo

Hi Pablo,

As you said, the skin friction correction has a real effect on the morphological computation, but it does not reduce bottom oscillations for long-term simulations.

Regards,
Laurent Leballeur

Hi Laurent,

Could you add some figures to explain the oscillations? I know these can occur in morphological models, but usually the inclusion of waves smooth these out.

Have you tried using a bedload formula combined with suspended transport calculations. This again might smooth the results a bit.

Hi Michiel,

I'm going to send you some unmarked pictures on your personal mail, as the project is confidential.
I didn't try the explicit formulation of the suspended transport, but it might actually be a good idea since the Exner equation would no longer be used.

Best regards,
Laurent Leballeur

Hello everybody,

I can easily reproduce my problem of numerical instabilities from the test case bosse-t2d. I just increase the duration of the simulation, without any use of morphological factor, nor other modification.

In attachment, you'll find an illustration that clearly shows the morphological instabilities during the simulation.
The first red line is the initial bottom profile (T=0), the first green line is the bottom profile at the end of the default test case (T=2h46min40s). Until then, everything is going well !
Following lines are the same profile for further time steps, the last green line standing for T=16h06min40s.

If anyone has ideas to improve this numerical behaviour, (s)he is welcome !

Best regards,
Laurent Leballeur

Hello,
This is a typical behavior for this kind of problem. Actually, this is a propagation problem similar to the solution of the Burger's equation: a wave propagates until a time in which a shock wave is developed (note in your plot the formation of the wiggle). Including slope effects helps to smooth this kind of behavior, as from a mathematical point of view it adds the diffusivity to your model.
Hope it helps,

Pablo

Adding slope effects is generally not strong enough to suppress this shock wave. The sediment slide option that is added to the code does a better job I think. (Thank you for that Pablo ).

Thank you both for your answers !

Indeed, slope effects are already turned on for the default test case.
I tried to add sliding with FRICTION ANGLE OF THE SEDIMENT = 10. and the picture below (T=8h36m40s) clearly shows the diffusive effect but which is not enough to make the oscillations disappear and which tends to increase the lowering of the bedform by diffusion.





Pablo, do you know a solution that can solve this kind of wave propagation problem: specific numerical scheme for example or anything else ?

Michiel, can you please give me the references for the fully coupled model for the water and sediment that you told me ? and for the 3D modelling over sustained periods ?

Best regards,
Laurent