open TELEMAC-MASCARET The mathematically superior suite of solvers

I have been experimenting with different options for SCHEME FOR ADVECTION OF TRACERS (this is still on V7 as I can't get v8 to work).

Originally I went with your suggestion Chi-Tuan of SCHEME FOR ADVECTION OF TRACERS = 5;5;5;5;5;5;5;5;5;5 + SCHEME OPTION FOR ADVECTION OF TRACERS = 4;4;4;4;4;4;4;4;4;4 as I believed this was part of the solution to the issue to the MAX_BASKET_SIZE size I was having. As it turns out after some days of simulations my model is unexpectedly crashing at a certain time point. Looking at the output I notice an issue with my salinity tracer which is set to a value of 35 at my ocean boundary at some points in the model (tidal flats and coastline) it goes above this up to around 40. Although there is no specific error in the output I believe the error is linked to the tracer as when tracers are removed the model runs fine. Please find attached the results file that goes to just before the crash.

I have since tried SCHEME FOR ADVECTION OF TRACERS = 13;13;13;13;13;13;13;13;13;13 or 14;14;14;14;14;14;14;14;14;14 + SCHEME OPTION FOR ADVECTION OF TRACERS = 1;1;1;1;1;1;1;1;1;1 as per your suggestions Chi-Tuan but my model won't start running with these at all. I have read the manual in detail and experimented with multiple options with tidal flats as detailed in the manual to try to get these to work but with no success. I also tried SCHEME FOR ADVECTION OF TRACER = 15 but again the model won't run. I have attached a typical output/error file but it doesn't suggest much to me.

The only thing that seems to work is reverting back to SCHEME FOR ADVECTION OF TRACERS = 1;1;1;1;1;1;1;1;1;1. I understand that this is not recommended for tidal flats but the results at least look OK.

Any comments very welcome,

Many thanks,

Neil

Hello,

I think shivanesh.rao did not add the 3 subroutines as I did not suggest him.
As he did not use Thompson BC, only changing the advection scheme to another one different from the characteristics prevents from the error of MAX_BASKET_SIZE in parallel, but you can try to ask him directly.

The main problem in you case is also that you use Thompson BC (for MAX_BASKET_SIZE). If willing using so many tracers simultaneously, I would suggest you to try OPTION FOR LIQUID BOUNDARIES = 1 for every liquid boundary (and change your BC file with e.g. 5 4 4 instead of 5 6 6). The results will change but if your extent is large enough, it can be OK.

For the choice of advection scheme, characteristics method is not recommended for tracers because not conservative. Mass conservation for tracers is a main property for me.

Anyway, the last line is suspicious. I do not how your cluster/computer works:
"slurmstepd: error: task_p_post_term: rmdir(/dev/cpuset/slurm16647178/slurm16647178.4294967294_0) failed Device or resource busy"

Can you try to run your computation in sequential and see if you still have the same error message?

Looking at your steering file in #35999, if using schemes 13, 14, 15 or LIPS (5 + option 4), you have to change the keyword TREATMENT
OF NEGATIVE DEPTHS to 2 or 3.
If you have not, an error message should have been written in the temporary directory in at least one of the PE*.LOG file. This may lead to your error.

You can also have a look at the latest TELEMAC-3D user manual (enclosed) only available by compiling the document from the trunk. There are additionnal informations/tips compared to v8p1r1. Be careful! There are some updates because the sources have changed since v8p1r1 (preperation of next release v8p2).

Hope this helps,

Chi-Tuan

Hello again Neil,

With distributive schemes (e.g. 13, 14, LIPS...), you may have to decrease the time steps compared to the ones used with characteritics (after changing the mandatory options such as TREATMENT OF NEGATIVE DEPTHS).

Hope this helps,

Chi-Tuan