open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I am performing a TELEMAC-3D (v7p2) simulation with thermal exchange with the atmosphere using the WAQTEL module. When diffusion of tracers is activated (SCHEME FOR DIFFUSION OF TRACERS=1), the diffusion step for the temperature tracer fails (I get the following error: GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS : 200 RELATIVE PRECISION: NaN). The computation continues but the temperature tracer becomes NaN.

I have tried to reduce the time step and precision and to increase the maximum number of iterations for the diffusion of tracers but it doesn’t work.
If diffusion of tracers is not activated (SCHEME FOR DIFFUSION OF TRACERS=0), the model works but the temperature tracer is only advected.

Is there a way to make the diffusion of tracer works with the heat exchange with the atmosphere?

Thank you,

Hello

Very often, people complain with lots of numerical diffusion. Are you sure you want to activate additional diffusion for tracer ?
If yes, the only way to make it work, is to decrease time step and/or increase number of iteration for the solver.
You can also use the GMRES sover (sover =7) and then increase the Krylov subspace dimension (Option for solver= 3 (default)). In all cases, this will be costly in CPU time.
I see that you are using a time step of 5 s with 2 planes, which is a big time step.

I hope that this helps

with my best regards

Riadh

Hello Riadh,

I am a bit confused by your previous post:
>Very often, people complain with lots of numerical diffusion. Are you sure you want to activate >additional diffusion for tracer ?

Documentation for the 7.2 says:

6.1.2 Diffusion step
Whether the diffusion terms are taken into account or not is established by means of the logical keyword DIFFUSION STEP (default value YES).

Am I right assuming that diffusion is ON by default for velocities as well as for tracers?

Best regards,

Alexander

Hi,

I am really interested in investigating the problem relating to "GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS : 200 RELATIVE PRECISION: NaN"

There are many reasons for this. Could you please post your fortran file and wind file here?

Best Regards
Huy

Hello Alexander

Yes Diffusion is on by default for tracers and for velocities. However, when dealing with tracers in case of temperature plume for instance, modeller wants most often to reduce diffusion (numerical and other types) in order to be able to achieve calibration. If diffusion is too low, calibration can be made using the viscosity (molecular, or turbulent).

I'm sorry if my previous post was not very clear.
I hope that this helps

with my best regards

Riadh ATA

Hello,

Reading these posts about tracer diffusion activation or not when simulating thermal exchanges between atmosphere and sea*, I have the following general question:

If diffusion of the Temperature tracer is turned off, are the related boundary conditions ATABOS%ADR... & BTABOS%ADR... (flux law for thermal exchanges at the sea surface) still taken into account in the calculation?

I think they are not (cf. CVDF3D.f) because diffusion and these specific boundary conditions for tracers (+ source terms related to solar ray penetration in BORD3D.f) are treated together in the diffusion step. So that thermal exchanges can only be modelled if diffusion is turned on (and it seems the WAQTEL steering file expects a value for the coeff. of dispersion along/across the flow).

Thank you in advance for any clarification.
JBS.


* I am also experiencing NaN values after some time in temperature simulations with air-sea thermal exchanges - not using WAQTEL though, but having harcoded similar formulations in BORD3D.

Hello,

Yes, the diffusion step includes other terms like boundary conditions (introduced as essential boundary conditions of the diffusion equation) and various source terms. I strongly advise to never turn off diffusion but instead to set the diffusion coefficient to 0 if you want no diffusion.

With best regards,

Jean-Michel Hervouet