Hi Mathieu,
I've tried solving the FO model for the Greenland mesh from the other tutorial, it worked right off the bat. I'm next going to look at the transient problem again, which didn't work last week and was my motivation to look at the stressbalance equation to start with. But I definitely understand more about ISSM now, so fingers crossed 🙂
For completeness: I've tried your code for relaxing the problem. If I run the stressbalance problem directly after with the damping term on and md.settings.solver_residue_threshold = 1e-3
, the solver converges and the solution looks similar-ish to the SSA and HO solutions. For 1e-4 and smaller I get the same error as before. With damping turned off, 1e-3 runs out of iteration steps with a vastly different solution at the end, and has the same error for 1e-4. I'll not pursue this any further for now. However, since 1e-3 wasn't working initially and I had to go to 1e-2, I expect that running the damping step for longer initially might help more. The reference was definitely helpful, especially knowing that the resolution needs to be that fine for the floating ice part.
Another question: You mentioned before
mathieumorlighem They are MUCH faster than FS, a lot more stable, and give 99.9% of the time the same results as FS.
is the 99.9% from experience, or are you aware of any references that do a detailed comparison by any chance? Most papers that I've seen so far for real-world ice geometries use SSA, and since I'm just getting started in the whole ice sheet topic, I don't have much of an overview yet.
Thank you so much for your help!
Best,
Nicole