open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all,
I am modelling flash floods for smaller catchments with TELEMAC-2D FE. For this purpose, I am using the "runoff_scs_cn"-subroutine (rainfall over the whole catchment) and therefore the option tidal flats as well.

The results are looking good (compared to other models with same DEM and input/rain data), but there is one big difference. For each cell in TELEMAC the water level rises up to approx. 2 cm and after reaching this threshold, the water flows normally. Over several hours the threshold gets depleted. It’s not affected by roughness (I used for testing uniform manning values of 0.2 and 0.04)

How could I erase this (wrong) threshold? In some other models, it's possible to set a "minimal depth for outflow". I read in one thread, that this threshold is occurring because of the option "tidal flats", but I need this option for the subroutine "runoff_scs_cn". Moreover, I don't want to use the FV-scheme, if it isn't necessary to solve this problem.

Thanks for any help/hints for this problem.
Thomas

Hello Thomas

2cm (10-2) looks as a high value for a threshold. We use values like 10-6. So, this is a bit weird.
What we can investigate, is in the direction of the wetting/drying algorithm of FE schemes. in extreme cases where we have steep bathymetries and small water depths (and I guess it is your case), the detection of wet/dry front, which is made by the following criteron Z_i > (Z_j+H_j) (i and j are node indices of a given element), can fail.
To confirm this assumption, choose one element in your model where you observe this issue and just verify by hand whether this condition is not satisfied.

kind regards
Riadh

Hello,

It would be interesting to see the result of the check Riadh proposed.

However, I have now done several similar applications, the main part were performed on steep bathymetries, and I came to the conclusion that FV are best for such cases. There are several papers from previous user conferences dealing with this topic (Clément TACCONE).

Yes, the model will be slower than FE but I still find it acceptable: CPU time max 24h for 6-10 hours of simulation on a 3 million element mesh (mesh size 0,5-3 m) using the HLLC scheme on a Windows 24 CPU workstation.

Best regards
PL

Hi Riadh and Pierre-Louis,
Thanks for your replies.

@ Riadh: I am using a mesh size of 5 m, for the wet/dry criteria (Zi > Zj+Hj) it means the slope must be greater than approx. 0.4 % (0.02 m / 5 m). This criterion fulfills largely in my headwater catchment, where water levels of 2 cm occur (no significant water inflow from other elements). Anyway, I checked some points and Zi was always greater than Zj+Hj.

@ Pierre-Louis: Do you observed such effects with FE as well? If I am using FV and switch off the "tidal flat" option I get an error, that for the "runoff_scs_sc"-subroutine "tidal flats" option must be set to YES (an error for the FE scheme). Is it possible just to comment this error warning (I would suggest the FE or FV-scheme doesn’t effect this subroutine)?

Best regards,
Thomas

Hi,

I guess I observed similar issues but the most limiting one was instabilities on steep slopes with flowing water. Fixing those would have required a dramatic refinement for the mesh (criteria mentionned my Riadh) and time step. FV deals with such situations without significant issues.

Which version are you using? I remember that up to a certain tag TIDAL FLAT was initialized to NO if using the FV kernel. So setting it to YES from the cas file did not help. From version 7.3 things are much more easy for rainfall on FV, I would suggest you to use this version. Beware that you need to have TYPE OF SOURCES = 2 if you intend to use culverts.

Good luck!

Best regards
PL

I am using v7p2r3 but I am updating to v7p3r1 in the next days. What is much easier for rainfall on FV in this new version, I can't find any hints in the manual or release notes?

Best regards
Thomas

Hi,

I never used 7.2.3 so it maybe also works fine but in early 7.2 releases there were still some small bugs when using FV and rainfall (rain not included in source terms, problem with TIDAL FLAT keyword you mentionned). I was using a 7.2 version with trunk developments and I don't have a clear view on possible differences with the official releases.

With 7.3.1 everything works like a charm.

If you are about to move to this version you might actually wait a couple of extra days as version 8.0.0 should be released before end of year I think.

Kind regards
PL

Hello ALL,

this is a follow up to Thomas question. He is a colleague of mine at TUM.

I found in meantime that the "Hidden Parameter" H0 in fricti.f has significant influence on flood propagation of overland flow. The picture enclosed shows a comparison of max. water depths for H0=3cm (standard) and H0=1mm for a flooding caused by rainfall using the FE-model (Cut-off is 1mm). The solution FE with H0=1mm is much closer to the FV-result. What we do not know is, whether there are side effects, if we use a smaller H0. All results look reasonable and the simulation runs smoothly without delay. Nevertheless we would like to ask:

Did anyone analyze the influence of H0 in fricti.f with respect to overland flow?

Best regards
Karl