open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

In cases where we set TREATMENT OF FLUXES AT THE BOUNDARIES = 2 (i.e. giving priority to prescribed fluxes of tracers), should we expect the observed/computed values at the boundary to approach the prescribed values as mesh element sizes tend to zero? This would seem logical to me but is not obvious from the derivation.

For example, in cases where I see substantial differences between the prescribed tracer value at the boundary (given in the steering file) and the observed value in the selaphin file, should I always expect to see improvements if I refine the mesh at the boundary?

I am not aware that documentation for other advection-dispersion models discusses this incompatibility between Dirichlet boundary conditions and boundary fluxes (e.g. Delft3D or MIKE21/3). Does anyone know if there are similar issues with these advection-dispersion models, and if so, how the problem is treated?

Thanks in advance.

Enda

Hello,

Normally the tracer value, when you use this option for fluxes, must be between the previous value and the precribed value. Second thing, it is true that both options should coincide when you refine the mesh, but this must be done with changes of the time step also, and after some refinement truncation errors may complicate the process.

Now does everybody has the same problem? Think of a very very large element near the boundary, going deep into the interior of the mesh, with an arbitrary small time step, the latter so that the flux should be arbitrary small. Then if you impose a prescribed value, the quantity you add in the mesh (through linear interpolartion over the large element) will correspond to an exagerated or even infinite flux. Everybody has this problem, but you realise it only when you carefully check (and print) the conservation of mass in the form : final mass = initial mass + fluxes.

With best ergards,

Jean-Michel Hervouet

Thank you Jean-Michel for the explanation.

Just to clarify - when you say the tracer value must be between the previous value and the prescribed value, what exactly do you mean? Do you mean the computed value at the time step before the prescribed value is applied?

Thanks again.

Enda

Yes, if you start from a value at 0 and prescribe 100, the value that gives the correct flux will be between 0 and 100.

Regards,

JMH