open TELEMAC-MASCARET The mathematically superior suite of solvers

During my model validation, I have noticed that the friction factor becomes more effective with using less horizontal layers. I had to change the friction value (Strickler Coefficient) from 65 to 0.1 to get some accurate results (free surface and velocities). What could be the issue? is there is any other parameters that we can alter to enhance the results?

Hello,

What is the water depth?
If it is too shallow, you may have to reduce some hard coded value in the CALCOT subroutine.

Hope this help,

Chi-Tuan

water depth, at the lowest, is 2 m. but my question was why the friction becomes less influential when more layers are used? Thanks

A friction coefficient of 0.1 for a Strickler law is very low.

How many horizontal levels do you use?
Is it the same steering file as in the "Wind Error" post?

Chi-Tuan

Yes, I used 20 layers

I am having the some problem. The water depth in my model ranges between 1 and 5 m. I got an okay agreement when I prescribed number of horizontal layers = 5 (first figure). But when I increased it to 8 layers, the results are quite different (second figure).











I had a look at calcot.f. I don't believe I see anything hardcoded that I can modify.

Hello,

Can you upload your steering file + potential Fortran files please?

3 years ago, there were some hard coded variables that can be changed in CALCOT, but from v8p2, they can be changed with some new keywords:
- MINIMUM DISTANCE BETWEEN PLANES CLOSE TO THE BOTTOM,
- MINIMUM DISTANCE BETWEEN PLANES CLOSE TO THE FREE SURFACE,
- THRESHOLD HEIGHT BEFORE CRUSHED ELEMENTS.
Anyway, the may be interesting if having shallow waters.

Do you have any measurement to say your results are OK for 5 planes and not for 8 planes?

But it would be interesting to know your MESH TRANFORMATION and friction law you are using.

What do you visualise as "velocity UV". Is it the magnitude of velocity only with U and V components? Is it depth-averaged magnitude of velocity?

Chi-Tuan

For mesh transformation, I am using the classical sigma transformation.
For friction law, I am using Nikuradse. I have attached the steering file here.





I am comparing depth-averaged velocity for now before looking into the vertical distribution. The first figure below shows the transects that we took. The transect on the left side has the greatest response to changes in layers.




The red dots in the figures below represent ADCP results. The blue line represents Telemac model output.







Looking at the speed distribution map, it seems like the problem initiated from the downstream boundary.

Hello,

Do you have the same behaviour with VERTICAL VELOCITY DERIVATIVES = 1, which is the default value?

Anyway, you have to use the latest release of the TELEMAC-MASCARET system v8p2r1 prior to other answers as you use keywords which disappeared in release v8p1! Some bugs were solved in previous release and may solve your issue.

Hope this helps,

Chi-Tuan

1) changing VERTICAL VELOCITY DERIVATES didn't make any difference in the results.

2) running with v8p2r1 also didn't make any difference in the results.

3) I also tried the run with Scheme for Advection of Velocities =1 (instead of 5) and still have the same discrepancies between 5 levels and 8 levels.