Troubles with grounding line simulations

Extension of Elmer in computational glaciology
Post Reply
rgladstone
Posts: 64
Joined: 15 Apr 2013, 16:23
Antispam: Yes

Troubles with grounding line simulations

Post by rgladstone »

Hi all, I am having trouble with grounding line simulations. More specifically, I tend to get very high mesh velocities in the vicinity of the grounding line, typically one or two elements upstream of the grounding line. This is mostly high vertical mesh velocities towards the upper free surface. And my simulations will typically fail at some point in either the upper or lower free surface solver. These high vertical velocities are at the spatial scale of individual elements.

So far I just have a simple domain with a linear downsloping bed and I'm spinning up from a slab of uniform thickness. I've tried this in both 2d (flowline) and 3d. I'm mostly running with resolution significantly finer than 1km and timesteps of a week to a month. Some simulations run for years or decades, but all fail at some point in the free surface solver.

Has anyone else had similar troubles?

Does anyone have a handy setup (sif + mesh/grd files) they would be happy to share, so I could check against my current setup?

Would it help if I post my sif and mesh files?

Do you think I should switch to using the internal mesh extrusion?

Cheers,
Rupert
gagliar
Posts: 79
Joined: 04 Sep 2009, 16:34
Location: LGGE - Grenoble
Contact:

Re: Troubles with grounding line simulations

Post by gagliar »

Hi Rupert,

Can you send me your 2D setup and I can have a look. From which test case did you start?

Regards
Olivier
rgladstone
Posts: 64
Joined: 15 Apr 2013, 16:23
Antispam: Yes

Re: Troubles with grounding line simulations

Post by rgladstone »

Hi Olivier, thanks for replying. I am attaching a tar file containing my sif file and mesh. This example has been running overnight on one processor on my laptop and hasn't failed yet. Note the small timestep and low nonlinear relaxation factor in the upper free surface solver.
flowlinegl.tar
contains sif and mesh for 2d flowline example
(800 KiB) Downloaded 769 times
This run previously failed with the following settings:
Timestep Sizes = 0.05
Nonlinear System Relaxation Factor = 1.0 (upper free surface solver)

I ran the cavity example before setting this up. It was very helpful to show the grounded solver and lower free surface evolution.

Cheers,
Rupert
gagliar
Posts: 79
Joined: 04 Sep 2009, 16:34
Location: LGGE - Grenoble
Contact:

Re: Troubles with grounding line simulations

Post by gagliar »

Hi Rupert,


I think your troubles came from the fact that you are using the Newton linearization of the Stokes equation. When the contact is tested, the boundary conditions are changed during the non-linear iterations of the Stokes solution (some nodes are released) and therefore if you want to use the Newton method, you should only activate it for a criteria lower than the criteria used to test the contact (10^{-3} here). Classically, we only use the Picard method which is safer.

I have also made some other minor changes in the sif, see the attached modified sif.

Doing so, I'm able to run the simulation using a one year timesetp (which seems more reasonable if one want to get a steady-state before the end of the week-end...).

Have a nice week-end
Olivier
Attachments
MIS1_flowline.sif
modified sif
(10.66 KiB) Downloaded 704 times
rgladstone
Posts: 64
Joined: 15 Apr 2013, 16:23
Antispam: Yes

Re: Troubles with grounding line simulations

Post by rgladstone »

Thanks so much Olivier, it would have taken me a looooong time to figure that out on my own.
Post Reply