open TELEMAC-MASCARET The mathematically superior suite of solvers

I'm currently running a large TELEMAC-3D hydrodynamic model of the Canadian Arctic, and was wondering if there have been any other reports of water level instabilities when salinity is active? Specifically, I'm getting some weird "valleys" of lower water levels at various points throughout the domain, and associated eddy currents. What's particularly unusual is that the overall model remains mostly stable, and validates reasonably well in other locations, there are just these persistent surface variations.

Some notes:

1. Telemac 8p5, Linux, 64 partitions
2. 20 sigma layers, salinity values look reasonable throughout the domain
3. Low volume errors in the log; however, my tracer error is ~10% of the boundary flow
4. Tested both boundary types 1 and 2
5. Tested increasing the accuracy thresholds
6. Tested higher and lower timesteps
7. Tested alternate solvers
8. Tested a range of turbulence models
9. Cartesian grid due to issue #971
10. Tested density law #4
11. Tested both vertically varying and constant boundary salinity profiles. This produces results that are incorrect in different ways, suggesting that the bounds have something to do with it here.

I've spent a while struggling with this one, so really appreciate any perspectives here. I've attached my .cas and grid files, and happy to complete any additional testing.

Definitely not an expert in modelling flows with density gradients but here are a few suggestions that may be worth trying:

• Applying HYDROSTATIC INCONSISTENCIES FILTER = YES. I would suggest applying the following bug fix mentioned here though
• Try FREE SURFACE GRADIENT COMPATIBILITY = 0
• SCHEME FOR ADVECTION OF VELOCITIES = 1

I also noticed that you are using the hydrostatic assumption but have some setting for the PPE solver which is only used with non-hydrostatic equations. Have you tried with fully non-hydrostatic equations?

Thank you for the reply! I haven't tried activating the inconsistencies filter, so I'll give that go.

I have tried pushing the FREE SURFACE GRADIENT COMPATIBILITY down to 0.3. The manual makes it sound like I shouldn't push it to zero; however, I did notice it was set that way in some of the examples, so might be ok. I have tried with SCHEME FOR ADVECTION OF VELOCITIES = 1, so no luck there.

Your point about the PPE solver is interesting! Do you have any suggestions for a different solver for hydrostatic conditions? I'd rather not go fully non-hydrostatic for efficiency reasons, but it is worth etsting!