open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear all,

I made a 2D model of the Scheldt estuary as it was in 2009 and in 1955. For 2009 we have a lot of data for calibration, but for 1955 we only have water level data.
I need salinity values in these models as input for a benthos model. So I calibrated salinity as a tracer for the 2009 model with the tracer diffusivity.

The model of 1955 has a different mesh resolution ( it has a 40m resolution compared to 150m of the 2009 model). I wanted to use the same value as the 2009 model for the tracer diffusivity of the 1955 model, but the results were unrealistic. I solved for now by using the same mesh for both models, but I want to know how the mesh resolution affects this parameter? I can imagine it also affects the velocity diffusivity?

If I have a coarse model of a very large area and I want to have a closer look at a specific part of the model, I can locally refine the mesh without needing to recalibrate the entire model. But this local mesh refinement will also affect some parameters or not?

Thanks

Sven

Hello,

Every physical term will have a different reaction to mesh size, so it is difficult to give a general answer. Artificial diffusion of advection schemes is an important issue, for example the artificial diffusion of a simple 1D upwind scheme in finite differences is (U*DX)/2, where U is the velocity and DX the mesh size, and you directly see here the effect of the mesh. This artificial diffusion may completely mask the values given by turbulence models, especially in free surface flows where the mesh size may be several kilometers, and when you change the mesh size you change the total (natural+artificial) diffusion.

With best regards,

Jean-Michel Hervouet

Jean-Michel,

thank you for the quick reply.

If I just consider it the artificial diffusion is also highly dependent on the velocity?
So if I keep the same mesh, but I change the bathymetry and because of this the flow velocities in the whole estuary are much lower, this will affect the artificial diffusion? So the original value for the diffusion of the tracer is no longer valid and I need a calibration to be sure of the results?

Sven

Hello,

Yes, I'm afraid, though generally diffusion has less influence than friction for calibration.

Regards,

JMH

Jean-Michel,

a last remark. A finer mesh should give a more accurate solution and the artificial diffusion should be less. In my estuary model representing the situation of 2009 the mesh resolution is 150m and I calibrated the salinity succesful. I refined the mesh to 40m and used the bathymetry of 1955 together with boundary conditions of 1955 but the upstream discharge input stayed the same. In 1955 the estuary was not so deep as it is today and flow velocities are much lower. So the mesh resolution is higher and the velocities are lower,so I would expect the diffusion to be less. I used exactly the same parameters as the model for 2009, but I got salt intrusion much further than in 2009, so a much higher diffusion. If I use the mesh of the 2009 model for the 1955 situation, I get more or less normal or expected values for salinity. So I think it must be the mesh resolution.

Any idea why this happened?

Hello,

The time step has also its importance, and normally if you reduce the mesh size you reduce also the time step, to get the same CFL number. Then the advection schemes may behave differently. For example the method of characteristics is less diffusive in one step than in two half steps. We have found recently that schemes 13 and 14 have also this behaviour.

REgards,

Jean-Michel Hervouet

This is very interesting. I have a mesh resolution of 10m at the upstream boundaries that's why I never needed to change my time step.
I will run a simulation with a smaller time step to test the impact of this parameter!

Thanks