Numerical viscosity due to the MUSCL and HLL coulpled scheme
 

Hi, all

I'm now in the investigation of numerical diffusion(viscosity) due to the numerical scheme itself. In more detail, I'm using 4-th order MUSCL-TVD scheme with HLL Riemann solver on my code, but recently found that they inherently have the diffusive errors.
For example, if I run that code for inviscid fluid(nonlinear shallow water equation without any dissipation term), the results show vorticity generation with varying depth, which shouldn't be for inviscid fluid. So my guess is that such vorticity generation has come from numerical diffusion(viscosity).

My problem is that, I have not enough idea to minimize or theoretically interprete 'numerical viscosity' because the scheme(MUSCL + HLL) is a coupled one and further it has limiter function which is not explicitly expressed in truncation error analysis.

Could any of you give me the idea to identify(or quantify) 'numerical viscosity'?
Thanks,


sonsiest