open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

It seems that every time I try to make a custom modification to the vertical layer spacing of T3D I run into non-convergence issues.

I am currently trying to modify the spacing of the layers to increase their resolution towards the surface using MESH TRANSFORMATION = 2 with user defined spacing of the Sigma layers based on percentage of the water depth (TRANSF_PLANE%I(IPLAN) = 2) in the USER_MESH_TRANSF subroutine. The model runs but after about 0.5 hrs exceeds the number of iterations for the propagation step solver.

What I find strange is that keeping the same mesh transformation (Opt. 2) but manually specifying the layers to be evenly spaced gives me the same problem. I would have thought this would be more or less the same as selecting MESH TRANSFORMATION = 1 which converges without issue.

Reducing the timestep of the model does not seem to have any influence.

Have attached my custom USER_MESH_TRANSF subroutine for inspection. Model is setup with NUMBER OF HORIZONTAL LEVELS = 7.

I will try to do some more testing on this shortly - I know that this option is used by many people without issue so there may be something specific to my model setup. I am prescribing time and spatial variable water levels and currents at the boundary through custom modifications in BORD3D so that will be one of the first places I will look.

Will provide some more info as I get updates but would be interesting to know if anyone has any insights into why an even vertical distribution specified with MESH TRANSFORMATION = 2 would lead to almost immediate non-convergence while the default option doesn’t.

After some further testing on this issue, I have observed the following:

• Non-convergence is also apparent with the standard sigma scheme as the number of layers increase, even in the case where there are no vertical density gradients. Including the HYDROSTATIC INCONSISTENCIES FILTER with the recent bug fixes helps resolve this issue but non-convergence of the PPE solver is still common.
• Changing the PPE solver to the previous default of Method of Characteristics (SOLVER FOR PPE = 1) and preconditioning to method 34 (PRECONDITIONG FOR PPE = 34) as per some examples greatly improves the convergence of the solution

In general, stability of the dynamic pressure has been one of the more challenging aspects for me recently while trying to setup a baroclinic T3D model for modelling ocean currents. However, with the above changes to the PPE numerics, suppressing strong vertical currents due to hydrostatic inconsistencies and applying the GOTM k-e turbulence model over standard vertical mixing length is has been mostly resolved.