open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear Telemac forum,

I have a model where one of the boundaries (the only outlet) has a stage-discharge curve that when trying to simulate a steady-state case (to be used as a hotstart) the result show some areas near that boundary that are "artificially" flooded because the nodes are below the water level imposed in the river and even some points are even below the thalweg of the river.





Things that I tried without success:

• Double check the interpolation/elevation mapping, those areas are indeed below the river "normal" water level.
• Discarded the Stage-discharge cruve and used a Thompson boundary type. The results were "Ok" for the steady state but weren't in the unsteady ones.
• Re-calibrated the stage discharge curve with more information. Didn't make any real difference for the hotstart flow value.

Possible solutions:

• Use a different type of boundary? (something that i missed here? like a Newmann type e.g. water level gradient = 0 through the boundary?)
• Somehow force the flow (in that boundary) to always "exit" the domain.The flow value is always negative in that boundary.
• Use a "composite" boundary.Stage-discharge curve for the near-river area and thompson for the rest? is this well-posed?

Are any of those alternatives somehow possible within telemac? Is there any other alternative that I didn't consider?

Regards,

José Díaz.

Hello!

I guess you want to model overland flooding from the river? In that case I think you maybe should shift your outlet boundary further downstream to avoid having such problem.
Another thing is that imposing a stage-discharge curve over such a long boundary is probably not correct (I assume the curve is valid mainly for the channel part).
If you can't shift your model boundary maybe try to have two liquid boundaries, one with the Q-h curve and the second with free outlfow (4 4 4, not sure it works though, never tried it but read several posts about it on the forum).

Good luck!

Best regards
PL

Dear PL,

Yes!, my model will be used for Dam-break routing and the outlet boundary is essentially a massive flood plain which is almost the same case at any point downstream.

Free velocities in the outlet (in the floodplain boundary) and keeping the S-D curve for the river seems more logical, but somehow I remember reading in the forum that this kind of boundary is prone to make the model unstable (?) and should be avoided (due to the infinite solutions there and telemac picking one perhaps?) and because this model will be use for extreme cases I would like to minimize any source of numerical instabilities .

Anyways, I'll definitely try this setup and report back.

Thank you!

regards,

José Díaz.

Dear PL

For the hotstart case simulation run like a charm : No artificial flooded areas, but when routing the Dam-break Hydrograph at some point the Free flow boundary (444) started to create some instabilities that led to the final crash.

in the meantime I'll try to test it using a Thompson for the boundary that was free flow and see what happens...

Thank you anyways.

José Díaz.

Dear José,

Thanks for the feedback! Just another comment. I have seen a calculation made by a colleague with roughly the same configuration, outflow on a land with no constant downstream water level that could be defined along the boundary. The model was run with finite volumes and setting a water level below the bottom evelation produced what looked like acceptable results. I did not test it myself but I could not detect anything obviously wrong.

I don't remember but I think it was a classic prescribed level condition (5 4 4), not 100% sure though.

Give FV a try, it is furthermore well suited for dam break cases!

Good luck!
PL

Dear PL,

We agree on FV being more suited for Dam-break studies! (that was our initial setup), but this particular case has over 2 Mill. elements and over 12 cases with their corresponding scenarios giving us overall 30+ cases to analyze. FV was 2-3 times slower compared to FE with some numerical tunning, and since both schemes were producing similar results (being FV more stable of course) we decided to stick to FE.

On the other hand, the tests with the Thompson boundary for the floodplain (and keeeping it with a 444 condition) were succesful! The case even ran like 5-8% faster because the "force positive depths" subroutine iterations decreased significantly! (we didn't expect that).

Thank you so much for your suggestions, they were really helpful!

Regards,

José Díaz.