Hello,
I'm attempting a simulation in which I reproduce the parallel solution for SSA, with a Weertman sliding law, with a setup which should introduce no longitudinal stress gradients (i.e. constant thickness, constant slope, constant sliding coefficient, periodic boundary conditions). I've set symmetry boundary conditions, spcvy=0, left sidewalls unconfined, and assigned periodic boundary conditions on the upstream and downstream boundaries. The friction law is given by taub = C u, (where m=1), so horizontal velocity should be given by u = taub/C.
I've parameterized driving stresses and friction such that the result should be u=99.5 m/yr, however results consistently show u = 99.6 m/yr, even after refining the tolerances. I'm not too concerned by this error, and it would definitely be close enough for most applications, but I was wondering if some aspect of my penalty method is introducing this drift. The penalty factor is set to the default 3, upstream and downstream boundary vertices are paired. Should I experiment with changing this number?
Alternatively, it might be some sort of numerical diffusion, artificially accounting for vertical deformation? Parameters are set such that there would be 2m/yr of vertical deformation in the higher order model. However, the HO model returns the same sliding velocity.
SSA results exhibit some lateral variation, on the order of machine precision.