Free surface sensitivity - drag coefficient & mesh sizes
 

Hello,

I'm calibrating the free surface in a river channel for different 3D triangular meshes. I'm using the drag coefficient (CD) to do so. I notice that on my rough mesh my free surface calibrates with a realistic CD, however using my refined meshes the free surface is a lot lower and is also less sensitive to changes in CD. This results in a calibrated free surface for my refined meshes with a high CD above what is realistic. I expected there to be differences in elevation between different meshes as the model became independent, however the free surface seems to be lower and less accurate for the refined mesh.

How does the changes in mesh size affect the sensitivity of the free surface to changes in the CD?
Why would the free surface be lower for the refined mesh?

The geometry is very similar for the rough and refined meshes, with the largest differences being 0.01m in some of the pools. The boundary conditions are unchanged. Testing the solvers doesn't change the elevation, rather than how "smooth" it is. Neither does changing the boundary conditions for the free surface.

I'm struggling to think what is at play here and I'm wondering if there's some theory behind free surface simulation I've overlooked? Could it be due to the time-step?

I'm using Telemac 3D (v7p2r3).

Many thanks.

if you change your time-step as they refine the mesh size? Remember Courant's theory.