TELEMAC-MASCARET Forum: Question about scheme 14 (1/1)

Welcome, Guest

TOPIC: Question about scheme 14

Question about scheme 14 11 years 2 months ago #10489

Lufia

Hi,

I have some simple questions about the scheme. In the Telemac release notes of v6p0 (3.6 Discussion in the wiki) stands:

A possible drawback is a sensitivity to the mesh, probably by construction because we use the edges as a way of transit of tracer. An obvious odd property is that the result certainly depends on the numbering of edges.

I read the paper form the (IAHR). However, I'm not sure if these information are up to date. As I understood, the flux over a segment is only limited when the there is a negative depth. An iterative procedure is used to compute the fluxes, so there should be no dependency for the hydrodynamic simulation.

Has the numbering only an influence on the transport of tracer or also on Sisyphe?
Will the results of the hydrodynamic simulation differ between serial and parallel runs or a different mesh numbering?
If there is a dependency on the mesh, it will be only relevant if the flux must be limited at some nodes?

With best regards,

Leo

The administrator has disabled public write access.

Question about scheme 14 11 years 2 months ago #10493

riadh

Hello Leo

Actually these informations are nomore up-to-date. First, when it was implemented, this scheme was dependant to node numbering, but later, this problem was overcame and this scheme is one of the best ones of the FE kernel of Telemac-2D. It is positive, mass conservative, with low numerical diffusion, and recommanded for models having wetting and drying passages.
It's only drawback is the fact that it is explicit in time, thus a little bit more expensive in CPU time. Note that, although it is explicit, this remains completely transparent for user, Telemac makes internal sub-iterations to take into account the stability criterion.
In a nutshell, this scheme is the one to use for severe unsteady models and especially for tracer transport. As far as I know, there is no difference between serial and parallel results. The only differences could be observed only when looking at machine precision level and these differences are mainly due to round-off error.

I hope that this is helpful
With my best regards,

Riadh

The administrator has disabled public write access.

Question about scheme 14 11 years 1 month ago #10496

Lufia	Hello Riadh, thank you very much. Looking at pros of the scheme, the larger execution time does not matter. It is nice to know that Telemac takes care about the stability criterion. Maybe it's time for me to look in the source code to see the details of this superior scheme With best regards, Leo
	The administrator has disabled public write access.