open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi everyone,

I simulate tides over the Northwestern European Continental Shelf (see domain.png) and I compare model results with water level time series in the Belgian Coast area.

It works pretty well when working on a projected plan (UTM31N). But, according to the size of the domain, it could be interesting to try the effect of using spherical coordinates.

I tried two different configurations: 1) changing coordinates of the existing mesh to Mercator coordinates, and 2) doing a new mesh directly in Mercator coordinates (to avoid triangle distorsion).

Using both configurations, the results are not good anymore (the tidal oscillation of the water surface is damped of about 50%).

Looking at snapshots of the horizontal velocity norm, I noticed a different behavior at the open boundary when using the UTM31N grid (utm.png) and a Mercator grid (mercator.png). It is very smooth with the UTM grid and there is a jump with the Mercator grid.

Does someone has a suggestion to solve that problem? I also put the steering file in the attachment. I already tried unsuccessfully the following changes:

- free surface gradient compatibility = 0.8 (instead of 1)
- solver accuracy = 1.E-6 (instead of 1.E-4)

It is worth mentioning that I made the same comparison with a larger domain covering part of the Atlantic Ocean, and the result looks good in both situation (UTM or Mercator grid).

Actually, I noticed also a strange behavior at the open boundary of my larger model when using the Mercator coordinate system (large_mercator.png), while it is much better with UTM31N (large_utm.png). The figures represent the norm of the horizontal velocity.

Any suggestion would be very welcome.

Olivier.