open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi all,

I am currently experiencing a strange phenomenon in my TELEMAC (v6p3) simulations and require some hints what could be the cause of the problem.
First, I did transient simulations with a tracer using the MOC advection scheme, which worked fine.
Afterwards I activated water sources in my model (99 in total, so I had to icnrease the MAXKEYWORDS), which should act as groundwater abstraction wells. Because I am not interested in their impact on my basin, I declared their VALUES OF THE TRACERS AT THE SOURCES : 0 and later close to 0 (0.0000001)

This caused the first problems with the MOC advection scheme. I got Inf and -Inf values for the tracer distribution in my model. Now after changing the advection to the explict Psi scheme, the model is working fine but the tracer disappears after one time step.

Here a mass balance of the MOC tracer run, which had very high and low values:

---

DIFFUSION OF TRACERS STEP
GRACJG (BIEF) : 1 ITERATIONS, RELATIVE PRECISION: 0.3027964E-06

BALANCE OF Cl-g/m3 (UNIT: * M3)

INITIAL QUANTITY : -0.1556273+133
FINAL QUANTITY : -0.1710617+133
BOUNDARY 1 FLUX: -431.7459 ( >0 : ENTERING <0 : EXITING )
BOUNDARY 2 FLUX: 5.600279 ( >0 : ENTERING <0 : EXITING )
BOUNDARY 3 FLUX: 172.0894 ( >0 : ENTERING <0 : EXITING )
BOUNDARY 4 FLUX: 127.4840 ( >0 : ENTERING <0 : EXITING )
QUANTITY CREATED BY SOURCE TERM : 0.1720257+130
TOTAL QUANTITY LOST : 0.1560636+132
RELATIVE ERROR : 90.72110

---

Is there the possibility that the water sources could neglect tracers at all? Without these sources the model was running fine, so I didn't predict any problems when activating them. Or am I missing something obvious when dealing with tracers and water sources?

Thanks a lot in advance,
Robert

Small update:

I got tracer and source terms working when using the advection scheme 14 (MURD) for both flow and tracer. But after about 180 simulated days (about 21,600 simulation steps), the tracer abruptly vanishes in the whole modeling domain

I added a figure showing the tracer concentration on one point in the model domain: lines represent the simulation, the circles represent field data.





Can somebody help me in understanding what happened? Does this point to some numerical instabilities because all my source points have negative discharge and a low tracer value so acting as boundary conditions at the edge of the model domain? But why does my tracer suddenly disappear after 180 days?

Cheers,
Robert

Does the point go to dry at the 180 simulated days?

Does the point have shallow water while you are sampling tracer concentration?

Does your field level/velocity testing (hydrodynamics, not tracer) match model outputs?

Does the mass balance look right from the sortie file?

Otherwise, we need to see the model to provide further comments.

Hailiang

No, the water depth stays the same as well as the velocity profiles. Also the water balance is fine. Only the chloride mass balance is going completely nuts.

I noticed that the tracer input of the sources first stays constant and after a period of 15 days is increasing over time. Then it disappears completely (no QUANTITY CREATED BY SOURCE TERM in the mass balance output). But then suddenly it increased by the factor 110...

I set up the sources with a abstraction WATER DISCHARGE OF SOURCES : -0.02 and a tracer value of 0. (VALUES OF THE TRACERS AT THE SOURCES : 0.00). So I thought they would stay constant over time, but in my model the water abstraction by the sources stays the same, whereas the tracer first disappears and then "comes" back and completely empties the model.

day 201

day 202

Hi
Are you able to make a hot start at day 201 and reproduce the problem?

regards

I hotstarted the simulation some days before the initial tracer depletion. Now the tracer got depleted right at the start of the simulation and randomly the source terms are depleting the whole model domain with values like this

QUANTITY CREATED BY SOURCE TERM : -0.6009270E+13

In this run I changed VALUES OF THE TRACERS AT THE SOURCES : 0.00 assuming that this would cause a non-existent tracer depletion at these source terms. Nonetheless, the model tracer gets depleted very fast but the water abstraction by the source terms stays constant.

As I understand it, the sources are abstracting the correct quantity of tracer: discharge*(tracer value)*dt at every time step.

I attached my cas file and model output to this post.

Best

Hello
First task to do is to upgrade to a newer version. Since v6p3, there are lots of improvements and bugs which were fixed.
Moreover, some recommandations:
1- to have complete mass conservation, you should use TYPE OF SOURCES =2 (dirac option). with default option 1, there is no strict mass conservation
2- improve the accuracy of the solver of tracers, put it to 10-14 for instance. Defautl value is 10-6 which is not enough to have fully mass conservation especially for small values like yours.

I hope that this helps

with my best regards

Riadh

Thank you all! After updating to v7p2 and applying option 2 for sources the model is now working fine without the sporadic depletion of tracers.

Cheers,
Robert

Unfortunately, the sudden tracer depletion is happening again...
I am currently running a 15 year simulation (v7p2r3) to calibrate the model and after roughly 1300 days the tracer depletion is beginning. The inflow terms suddenly become NaN and after some more steps, there is no tracer left in the model domain:

DIFFUSION OF TRACERS STEP
CVTRVF_NERD (SCHEME NERD, 13 OR 14): 10 ITERATIONS
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 60 RELATIVE PRECISION: NaN

BALANCE OF CHLORIDE

---

(UNIT: ----G/M3

---

* M3)

INITIAL QUANTITY : NaN
FINAL QUANTITY : NaN
BOUNDARY 1 FLUX: NaN ( >0 : ENTERING <0 : EXITING )
BOUNDARY 2 FLUX: NaN ( >0 : ENTERING <0 : EXITING )
BOUNDARY 3 FLUX: 213.4925 ( >0 : ENTERING <0 : EXITING )
BOUNDARY 4 FLUX: 69.92463 ( >0 : ENTERING <0 : EXITING )
TOTAL QUANTITY LOST : NaN

I guess that the solver is not converging and therefore the model is falsely "losing" tracer (GRACJG is not solved in the amount of steps).
I changed the advection scheme (currently 14), the diffusion accuracy, the time step, the max. amount of iterations (MAXIMUM NUMBER OF ITERATIONS FOR SOLVER, MAXIMUM NUMBER OF ITERATIONS FOR ADVECTION SCHEMES) and tried different setups for tidal flats, but nothing helped. At the end, only deactivating all my source terms (in total 99) helped, but these sources are representing abstraction wells that are important for water balance.

Have I overlooked something? Why is the solver not converging and how can I fix this? Or, most important, what's the correct keyword to increase iteration steps of CRACJG?

Thanks a lot in advance!
Robert

Hello,

You should try to increase MAXIMUM NUMBER OF ITERATIONS FOR DIFFUSION OF TRACERS, which is 60 by default!

Best regards

Sara