open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello, I have been troubleshooting a regional model for a while and would appreciate some ideas/input/expertise, if anyone is able and willing.

Aim: Regional model of Patagonian shelf

What I'm doing and why

1. Telemac-2D: have used before without issue, but not in this region.
2. I am applying the model to various palaeo-timeslices, and so it is a big computational task
3. First, I sort my domain and model out for TPXO forcing, then I calibrate it for the present-day forcing from the lower resolution global model (which I use for the palaeo-forcing). I write my own boundary conditions file for this (h,u,v) and have successfully done so in another region.

Issues

1. The tide does not propagate properly on the shelf - too much energy, and I suspect the system should be losing energy to the internal tide at the steeply-sloping shelf edge (particularly in the south).
2. A smaller on-shelf only model of the Patagonian shelf works fine with TPXO9, high resolution forcing.
3. However, when I take my boundaries off-shelf, the on-shelf tide is too energetic.
4. I have to take my boundaries off-shelf because of the relatively lower resolution palaeo-tidal forcing that I need to use, to give it time to propagate sufficiently into my region of interest.

Options + specific questions

1. Implement a baroclinic tidal dissipation parameterisation into the 2-d model (as per the work done here. It's still somewhat of a bodge job as buoyancy frequency in palaeo-timeslices unknown (but perhaps that issue would stand with Telemac-3D?). A lot of work if it might not fix the issue.
2. Move to Telemac-3d - massively computationally expensive given the task ahead. Would it deal better with the steep continental slope?
3. Try forcing the model internally and on the shelf (rather than at the boundaries) because my palaeo-tidal forcing dataset is reasonable on the shelf (but relatively low resolution, 1/8th deg). Is there a way to do this, with open boundaries at the edge of the model domain, and applying u,v,h internally?
4. Something else I've not thought of.
5. Give up and use a different model.

I'd be so grateful if anyone has any ideas or expertise they'd be willing to share. Because it's a palaeo-study it doesn't need to be perfect, it needs to be good enough (I'm working with large error bars!).

Many thanks in advance