open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi!

I am running a 3D-model of a bay with open boundary with fixed elevation towards the sea, and a varying wind as driving force. My total simulation period is 698400 s. If I run the model only for 6984 s (but with everything else the same) I get results that looks realistic. BUT when I run the whole period, the model seems to compute the currens almost as a 2D model - the current direction is varying in the horizontal but almost constant in the vertical (but when I run the shorter period the results are the contrary). I have attached a picture showing the UVW velocity in the surface layer at the same time step and with the same view properties for the long period (to the left) and the short period (to the right). I have also attached my other files.

I am running version 7.0.1.

Can anyone help me with this very strange problem?

Thanks in advance,
Charlotta

I had to modify the file type for these files...

Hi Charlotta!

Looking at your set up it does not seem so strange since the wind speed at t = 6984 s is more than 20 m/s in U direction while it is significantly lower at t = 698400 s.
You should run with constant wind speed if you want to compare the two results (MY_OPTION = 1 i meteo.f).
A last thing: you have two different turbulence models with k-eps on the vertical and smagorinsky on horizontal. Beware that telemac switches automatically to k-eps on both horizontal and vertical.

Cheers!
PL

Thanks, but the compared results are from the same time step (at 3600 s) and the only thing that is different is that the computations continue for a shorter or longer period after that. Since they use the same input file for the wind, the wind speed equals at equal time step.
I am aware that the smagorinski is changed to k-epsilon (but in MIKE 3 this combination is the usual setup).

/Charlotta

Hi,

Sorry, i did not read carefully your first post. This is strange, and I have tested it just to try, and I do have similar results wether I use DURATION = 698400 or 6984.
The only difference with you is that I stopped the long simulation after 3600 s and I look at the result on the local submesh. But as you can see from figures below, I do have the same surface current pattern at t = 3600 s.





By checking the 3D result file I obtained by running your case, it looks like what you depict on your 698400 s picture does correspond to the current at the bottom (level 1). Can it be that you were actually looking at the bottom current instead of at the surface?

Good luck!
PL

Hi again,

No, it is the surface layer (but it almost resembles the bottom layer) the strange thing is that the 3D simulation result almost appears to be calculated as a 2D simulation (with the long period simulation).

I have shortened the time step to 1 s instead of 10 s and then it works well for a few more time steps before it turns into the "2D"-mode. I am now testing a version where the wind varies less, it seems to work better. So it could be that something goes wrong when the wind input varies much. And I am quiet sure that it is something in the program that is wrong. The total period should not affect the result in the early time steps.

How many processors do you run on and what version do you use?

Thanks,
Charlotta

I have used the same set up with 4 proc and version 7.0.1. I used your steering file by just changing DURATION from 698400 to 6984. I don't understand why we get different results...

Maybe try with constant wind (only keywords, no fortran file) and in scalar mode.

But there's no obvious reason why you obtain the results you have apart from a bug, so maybe other users/developpers might have a look at this.

Looking forward to see if you can solve this issue.

Good luck!
PL

Okey. I only get the strange results when I have DURATION = 698400, with DURATION = 6984 I also get good results (the same as yours).

Thanks for taking your time,
Charlotta