open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi Telemac users,

Attempting to develop a tidal model of The Red Sea the moment. The model is setup using the wave equations (Treatment of Linear System = 2) and forced with TPXO tidal constituents in the Gulf of Aden (see Figure Below). Turbulence model is Smagorinsky and bathymetry is mostly based on GEBCO. CAS File is attached for reference.

The issue I am running into is that the tidal amplitudes are generally underestimated throughout the entire basin. Phasing is typically alright depending on the choice of friction coefficient, however, inclusion of astral potential tends to push the water levels out of phase (better agreement with just WL boundaries). I suspect that most of the attenuation is occurring at the Bab-el-Mandeb Strait (narrow southern strait of the Red Sea) as I am getting pretty good agreement with observations in the Gulf of Aden between the model boundary and the strait.

The main things I've tried so far are:

--> Decreasing the bed friction which alleviates the issue to some extent but using unrealistically low values then starts to cause other issues such as phasing problems.

--> Applying amplitude calibration factors to the tidal boundaries of between 5 to 20%. Although this helps with the overall amplitude it modulates the tides in a strange way. As the M2 constituent is the most dominant in the Red Sea, perhaps better results would be achieved if the individual constituents could be separately calibrated - BTW, this would be a great feature to add to the model.

Hoping that somebody might know if there are any other specific numerical parameters or schemes I could potentially look at to try and decrease the attenuation. It could be bathymetry related however I've seen that some authors have had relatively good success with the COHERENS model using the same TPXO boundaries and GEBCO bathymetry. Although they don't mention having to calibrate the boundary constituents I guess its plausible that they have. The results they are presenting are quite impressive compared to others that I've seen.

Link here: www.sciencedirect.com/science/article/abs/pii/S2352485517302347

Thanks,

Also attached is a plot of the observed and modelled water levels for a run with a constant bed friction Chezy Coefficient = 100 which has given me the best agreement thus far - just ignore the first few days of model spin-up.

If you are using astral potential (TIDE GENERATING FORCE = YES) you also need to define "LONGITUDE OF ORIGIN POINT" try setting this to near the middle of the domain and see if the phase issue is better.
LONGITUDE OF ORIGIN POINT = 38.54
LATITUDE OF ORIGIN POINT = 20.39

As I'm running in spherical coordinates I believe these parameters just set the (0,0) point of the Mercator projection that Telemac uses for computation.

I've done previous test with these parameters when using meshes in spherical coordinates and it doesn't really make any significant changes. In fact, there is a thread i started on this matter only a month or so ago.

Hello Toby,

If activating TIDE GENERATING FORCE = YES and SPHERICAL COORDINATES = YES, CORIOLIS is automatically activated to YES as you did in your steering file.
In that case, a correct LATITUDE OF ORIGIN POINT is expected so that the Coriolis force is correctly set.

With TIDE GENERATING FORCE = YES, LONGITUDE OF ORIGIN POINT is also expected to be set correctly, otherwise the time for tide generating force is not good.

I would advise to set them correctly (the coordinates greeve wrote seem to sound good or something like that).

Anyway, if it is not good enough, have you tried TPXO9 or TPXO9v5?

Looking at your steering file, I have a few questions:
- why using Smagorinski turbulence model?
- why using quasi bubble elements for velocity? You can try the default choice DISCRETIZATIONS IN SPACE = 11;11 + FREE SURFACE GRADIENT COMPATIBILITY = 0.9 or less if you have spurious oscillations,
- have you tried to only prescribe water depth for maritime boundary and look if your water level results are better or not? Do you currently both prescribe water depth and velocity components at the maritime boundary (566 in the boundary condition file)?
- you can simplify your steering file by deleting lines when default choices are used (e.g. ADVECTION = YES, PROPAGATION = YES, DIFFUSION OF VELOCITY = YES), the reading of the steering file is easier,
- why using TREATMENT OF NEGATIVE DEPTHS = 1 rather than the recommende option 2 (with CONTINUITY CORRECTION = YES + MASS-LUMPING ON H = 0 + SUPG OPTION = 0;0 or 2;0)?
- have you tried to define 2 (or more) areas with different friction coefficient values?

Hope this helps,

Chi-Tuan

Hi Chi-Tuan,

Thanks for the feedback. Answers to your comments/suggestions below:

-- why using Smagorinski turbulence model?

--> No particular reason. Easy to specify and takes the guess work out of choosing a coefficient. Would you recommend an alternative?

-With TIDE GENERATING FORCE = YES, LONGITUDE OF ORIGIN POINT is also expected to be set correctly, otherwise the time for tide generating force is not good.

--> Will give this another go to see if it makes any difference - in the past I haven't noticed any difference when using spherical grids. Seems like an odd programming choice when the grid is given in Lat/Long to require this. I thought the Coriolis and astral forces would be simply calculated based on the lat/long grid coordinates provided.

- why using quasi bubble elements for velocity? You can try the default choice DISCRETIZATIONS IN SPACE = 11;11 + FREE SURFACE GRADIENT COMPATIBILITY = 0.9 or less if you have spurious oscillations,

--> Not too sure. Have always used 12;11 with TELEMAC2D without too many issues but have never thought to change back to the default. Will give it a try.

- have you tried to only prescribe water depth for maritime boundary and look if your water level results are better or not? Do you currently both prescribe water depth and velocity components at the maritime boundary (566 in the boundary condition file)?

--> Boundaries are specified with 544 (Depth only). I have had pretty limited success in the past with TPXO and 566 boundary conditions. On this particular model I was getting large boundary jets with 566, even with Thompson boundaries enabled so gave up on the idea.

- you can simplify your steering file by deleting lines when default choices are used (e.g. ADVECTION = YES, PROPAGATION = YES, DIFFUSION OF VELOCITY = YES), the reading of the steering file is easier,
--> Agreed. I tend to include most options even if I'm not using them just as a self-reminder but I agree it makes it difficult to read.

- why using TREATMENT OF NEGATIVE DEPTHS = 1 rather than the recommended option 2 (with CONTINUITY CORRECTION = YES + MASS-LUMPING ON H = 0 + SUPG OPTION = 0;0 or 2;0)?

--> I was under the impression that these were only recommended when using distributive schemes which are compatible with tidal flats. I'm using the characteristics method and haven't run into many issues even with tidal flats enabled, but its worth trying with these settings.


- have you tried to define 2 (or more) areas with different friction coefficient values?

--> Not really. I have run a huge range of friction coefficients of both manning/strickler and chezy constant over the domain which minor changes in the phasing and amplitudes of the tides, but nothing of the order of error I'm seeing. Also tried to locally reduce friction in the southern strait to the Red Sea without much success

Hi Chi-Tuan,

As a follow up to the LONGITUDE OF ORIGIN POINT and LATITUDE OF ORIGIN POINT concern, I have made some comparisons of simulations as follows. All simulations were performed with a constant Strickler of 60 and a tidal amplification factor of 1.4. All other model parameters are the same as the .CAS file I initially posted for now.

- Scenario 1: No specification of LATITUDE OF ORIGIN POINT and LONGITUDE OF ORIGIN POINT (default values) and TIDE GENERATING FORCE turned off

- Scenario 2: Specification of LATITUDE OF ORIGIN POINT and LONGITUDE OF ORIGIN POINT given as per recommendation and TIDE GENERATING FORCE turned off

- Scenario 3: Specification of LATITUDE OF ORIGIN POINT and LONGITUDE OF ORIGIN POINT given as per recommendation and TIDE GENERATING FORCE turned on

It seems to me that there is no real difference in the results between Scenario 1 and 2. The specification of the latitude and longitude doesn't make any discernable difference, at least in this case. With the addition of the TIDE GENERATING FORCE enabled (Scenario 3) tides are slightly modulated but phasing is worse.

I am of the belief that tides are over-attenuated through the narrow strait - just don't know if this is mostly numerical of bathymetry related. Will investigate trying to find a better bathymetric data source in this location to try and narrow down possibilities.

In the mean time, welcome to any other suggestions that others may have.

Regards,
Toby