ISMIP HOM full-Stokes momentum error at corners

pf4d

Hello,

I have successfully ran the ISMIP HOM experiement using the first-order "HO" model without error, but when applying full-Stokes equations, I get an anomaly near the edges :

~~ideas?~~

mathieumorlighem

Hi EVan,

we treat periodic boundary conditions using penalties and FS is extremely sensitive to penalties and overconstraints (LBB condition, pressure cannot be constrained, etc). You will not be able to run FS with periodic boundary conditions using this set up. The only way to get reasonable results is to not use periodic boundary conditions but set a Dirichlet constraint:

md.stressbalance.spcvx(find(md.mesh.vertexonboundary))=0;
md.stressbalance.spcvy(find(md.mesh.vertexonboundary))=0;
md.stressbalance.spcvz(find(md.mesh.vertexonboundary))=0;

your results should look better. Since we never use periodic boundary conditions in practice, we did not try and find a better way to make FS happy.

I hope that helps

Mathieu

soroush

Dear Matheiu and Evan,

I was actually trying something similar; I had a setup similar to ISMIP-F (isothermal transient, with prescribed friction therefore sliding) but with full-Stokes flow equation, and as Mathieu noted here, it did not work because of the periodic BC issue.

After Mathieu's suggestion I tried Dirichlet on boundary vertices to zero, but quite naturally, all the velocity was forced to zero at the end of the transient run. I also tried defining an ARGUS icefront.exp to apply Neumann on the front and spc the other vertices (as in SetMarineIceSheetBC module), but I was getting highly unstable results (e.g. strange stripe patterns of extreme high velocity).

Is there any straightforward way to setup a transient FS run similar to ISMIP-F settings?

thanks,

Soroush