open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,
I am trying to run the sediment transport model.
I got an error(error message is attached)
I also attach the cas file.

Would you point out what my problem is?
If you need more information, I will upload it ASAP.
Thanks in advance for your help.

Regards,

Hi
is your model running without the Gaia coupling?

Hello.

Thank you for your fast response.

Telemac3d is coupled with Gaia.
There is no problem with running Telemac3d. But if I run the Telemac3d coupled with Gaia, I got the error.

Thank you.

Hello,

Do you have dry areas? If yes, the PSI scheme cannot deal with them (SCHEME FOR ADVECTION OF SUSPENDED SEDIMENTS = 5 and SCHEME OPTION FOR ADVECTION OF SUSPENDED SEDIMENTS = 1), the error message is classic in such cases.
You can rather use LIPS scheme (SCHEME FOR ADVECTION OF SUSPENDED SEDIMENTS = 5 and SCHEME OPTION FOR ADVECTION OF SUSPENDED SEDIMENTS = 4) or NERD scheme which is more diffusive (SCHEME FOR ADVECTION OF SUSPENDED SEDIMENTS = 14 and SCHEME OPTION FOR ADVECTION OF SUSPENDED SEDIMENTS = 1).
You can also decrease the time step and see if there is improvement or not.

If not satisfying, please also upload your TELEMAC-3D steering file + possible Fortran files.

Hope this helps,

Chi-Tuan

Thank you for your prompt assistance in resolving the previous issue. I appreciate the provided solution. However, a new problem has arisen. The model runs smoothly without considering sediment, and the current appears stable and consistent with observations. Yet, when I activate the sediment module using GAIA for sediment transport, an issue arises, causing the current to become unstable. I have attached figures for the sediment and current. I also attached the Cas files for your reference.

Regards,
-JH

Hello JH,

In your TELEMAC-3D steering file, there is one option which is currently not recommended: TREATMENT OF NEGATIVE DEPTHS = 1. With this choice, you may have negative depths locally which is not physical and the velocities and tracers/sediments may not be good.

It is recommended to use TREATMENT OF NEGATIVE DEPTHS = 2 (with then the mandatory choice MASS-LUMPING FOR DEPTH = 1.).

I am not a sediment transport specialist nor a GAIA specialist, but there is also one thing that may have to be changed: SOLVER FOR DIFFUSION OF SUSPENSION = 3 rather than 1 as the system may be not symmetric with settling velocty. What about the convergence of each operation? Can you upload a sample of the listing file?

Hope this helps,

Chi-Tuan

Hi Chi Tuan,

I have recently run into a similar problem with non-convergence of the default Conjugate Gradient solver for diffusion of the suspended sediments which seemed to be happening as water depths approached zero on tidal flats nearby the tracers. I'm using the recommended tidal flats options but even adjusting some of these to non-conservative options didn't improve the solution.

I tried multiple options before checking the 't3d_tidal_flats' example where I noticed the solver for diffusion was set to OPT. 3. This seems to resolve the issue, even though convergence of the solution is still not great at about 30-60 iterations on the EQUNOR solver.

As you also suggested this in the previous post, I was wondering whether this is something that is recommended/common when dealing with tidal flats. It seemed to be the only solution.

Thanks,
Toby

After further testing, even Solver Opt. 3 was giving non-convergence problems.

Seems that for this particular model, the only way to resolve it with GAIA coupled to T3D was to use the GMRES solver for diffusion of suspended sediments.

Hello Toby,

Thanks for your feedback.
Indeed, for the diffusion of sediment, due to the settling velocity, the matrix to invert may be non-symmetric and the conjugate gradient is not efficient anymore.

If you have a look at the TELEMAC-3D dictionary, you can see that default value for SOLVER FOR DIFFUSION OF THE SEDIMENT is 3 contrary to solvers for diffusion of other variables (velocities, tracers, k-epsilon) for which 1 = conjugate gradient is usually efficient.
Anyway, you are right, when the matrix is non symmetric, I would recommend to use GMRES. Be careful in that case if you use LIPS advection scheme for sediment, there may be oscillations close to tidal flats, Alexander Breugem noticed that in a specific case.

Hope this helps,

Chi-Tuan

Hi Chi-Tuan,

Thanks for the feedback. Its interesting you mention the oscillations near the tidal flats for the the PSI LIPS/GMRES combination. In fact Alexander also mentioned this to me in the past but I had forgotten. Indeed, I'm seeing strange unphysical spikes in SSC nearby the tidal flats with this combination so will try changing to a NERD scheme (Opt. 13) for the suspended sediment advection scheme. . I know this will increase diffusion quite substantially but I'd rather that at the expense of unphysical results.

I've actually had a lot of problems with stability of the PSI LIPS scheme - more so in T3D than T2D. In the past I was using it for the advection scheme of velocities but was getting extremely unphysical results around tidal flats. I have now gone back to the method of characteristics for velocity advection on the recommendation of Alexander and for the most part, these kinds of problems are gone. It seems these issues may also affect the tracers as well?

Also, are the numerical options for sediments in TELEMAC-3D a carry-over from the SEDI-3D days? Just wondering as these are now mostly defined in the GAIA CAS file.

Regards,
Toby