open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear telemac and tomawac users,

I have set up a model that runs ok for tomawac-only or telemac-only.
Telemac model reproduces the effect of a strong wind coming from the south and pushing water toward the north. The western and eastern boundary are solid [222 in the cli], and the northern and southern are free velocity but fixed elevation (0m, MSL) [544 In the cli].
The increase of water level in telemac-only model due to the wind is similar to what has been observed. It takes ~6h to reach a stable solution.
Tomawac model reproduces the propagation of a southerly big wave (created under the same strong wind). Again the predicted wave heights are similar to the observed ones. Here again, it takes ~6h to reach the final solution.

I now struggle as I want to couple telemac and tomawac:
I manage to have a coupled model running until the end of the simulation, but I experience several changes in the water depth (I guess due to the telemac-tomawac coupling).

Do you know why the water depth is changing so dramatically in the coupled model, and what can be done to avoid that?

Also, could you please provide some indication for an appropriate coupling period to simulate both
*the wave set-up in telemac and
*the change in water depth due to the wind in tomawac?
I plan to couple the models every 6h, and have a full run duration of 12h. Is that suitable?

Please find attached the steering file of
the telemac-only
the telemac and tomawac coupled a
the associated bathymetry and boundary conditions

Thanks fro any advice or recommendations.

Claire

Hello Claire

first of all, which version are you using? Irememeber that we had several issues in the coupling between telemac and tomawac and we fixed them (hopefully). As far as I know, all the fixes are already in v7p3r1.
On the ohter hand, when you couple telemac2d and Tomawac, you have to take care to not double the forcing on hydrodynamics. Telemac2d is forced by wind (Wind=yes) and by the coupling with Tomawac (which brings fxwave and fywave from tomawac to force telemac2d). Is it exactly what you want? if so, you have to be more smooth with your numerical options (smaller time step, more diffusive options) but also you can play with the coupling period.
The best option for coupling is 1 to 1, which corresonds to the reality. But if your physic is not rapid in time, you can relax this period to more than 1.

I hope this helps

with my best regards

Riadh

Hi Riadh,

Thanks a lot for your reply and recommendations.

I have used the versions from v7p2 to v7p3r0.
I have tested today the v7p3r1 (it was not fully installed before).

For the forcing of the model, yes, ideally, we do want to reproduce the water pushed by the wind near the coast (coastal inundation), and then propagate the waves there and reproduce the wave run up.
Do you think there are other ways to simulate those processes altogether?

I have halved the time step of both Telemac and Tomawac, and have added some diffusivity with the following keyword:
DIFFUSION OF VELOCITY = YES
OPTION FOR THE DIFFUSION OF VELOCITIES=2 (I have also tried 1)
VELOCITY DIFFUSIVITY=0.0001

But unfortunately I get the same results with the later amendments.

Would you have any comment or recommendation on those amendments?
Claire

Hello Claire

Option 2 for diffusion of velocities is not recommended if you have tidal flats (since there is a division by h in the non conservative term of diffusion).
For the diffusivity, you can increase the value to 0.1 or more is you want. This will stabilize more the model but will smear local features (shocks or wave curvature).

If you want to focus on propagation and run up aspects, you can use Finite volumes with HLLC or Kinetic scheme (first order) and even second order if CPU time is not prohibitive. Your model is not that big, and CPU time will remain reasonble.
With second order you will capture more realistic behavior in the propagation step. You can even couple with Tomawac now (from v7p3) But this coupling was never tested for real big cases and I will be happy to have some feedbacks on it from users.

The other solution also, is the use of Boussinesq in order to capture dispersive effects. This will increase CPU time in a significant way also.

Conclusion: to improve results, you have to pay

kind regards

Riadh