open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I am using TELEMAC 2D to model hydrodynamics in a large estuarine system in Ecuador. All of my first attempts resulted in severe numerical instabilities.

Therefore, I created a very simple case with a simplified domain and only one open sea boundary (TPXO). I did multiple simulations with various bathymetries. What I learned is that the bathymetry has a major impact on the numerical stability. Even when I apply a very smooth but deep bathymetry, some instabilities start to appear.

The real bathymetry at the open sea boundary ranges between -4000m to -2000m and raises to 0m when going inlands. However, only when I put the minimum bathymetry at -50 m at the open sea boundary and implement a linear gradient land inwards, my simulation remain stable.

My question is this: How can I get stable results without compromising too much on my real bathymetry? Do I use the right numerical settings?

The mesh is homogenuous with a cell size of 5 km.

I summarized my simulation results in a couple of slides. Also the steering file is shown in the first slide:
docs.google.com/presentation/d/1PbPA4HVE...U9I/edit?usp=sharing

Thank you in advance.

Kind regards,
Ignace

Hello Ignace,

What about the velocities on every maritime node, computed by TELEMAC-2D from your TPXO solution (are you using the regional Pacific Ocean solution? A local solution like the Gulf of Mexico does not seem to exist for your extent)? Are they OK (not too big) on every maritime node?

Have you tried to decrease your time step and/or refine your mesh when the bathymetry is deeper? It seems that bathymetry gradients are steep in this area.

Decreasing FREE SURFACE GRADIENT COMPATIBILITY from 1 to a value between 0 and 0.9 may suppress spurious oscillations. A usual value of 0.9 is generally advised, but you can decrease this parameter to 0.5 e.g. or even 0. if your results are better.

Have you tried to smooth your bathymetry with the keyword BOTTOM SMOOTHINGS?

Do you have dry zones in your models?

Have you tried other values for VELOCITY DIFFUSIVITY?

You can also try to use another finite element for velocity (quasi-bubble):
DISCRETIZATIONS IN SPACE = 12;11 if all the previous ideas do not improve your tests.

Hope this helps,

Chi-Tuan

Thanks for the presentation slides Ignace.


FIY
for my global model 10km resolution with tidal flats i.e. very large domain with only wind forcing (for now), lowering
did the trick

Thomas