Attempting different bed shapes

[glugeorge]

First, I tried moving the sill such that it's actually in the range of where the ice is (not 2000km away, instead 100km away). Couldn't get the solver to converge to a steady state solution. I have a hunch this is potentially because it's a pretty sharp change.

In the midst of trying to use Schoof's bed function which is a lot smoother.

[glugeorge]

I was finally able to use the Schoof bed function -there's some interesting behavior for sure. I took a while trying different numbers to get the solver to converge and also to fully iterate through time evolution. The interpolation issue #5 still hasn't been fixed - a bit trickier than I thought, figuring out how to explain why. Here is the evolution of perturbing A from 4.9e25 to 0.9e25:

We see the acceleration of the grounding line once it reaches the sill. Here are some profiles evolving over time:

[glugeorge]

Also, it seems like there is some sort of hysteresis like observed in the paper - the solver needs to build out towards the sill portion, it cannot resolve it by assuming steady state. I'm wondering what direction I can take this with, now that we have this more realistic bed working to an extent

[glugeorge]

I've added in a retreat scenario with Schoof's bed. We can see the retreat slow going up the sill, and then accelerate after moving past it. Weaker perturbations lead to a steady state on top of the sill. Regardless of the scenario, if I run it without coupling and the same initial steady state N (kept static), the uncoupled model halts retreat right away.

[jkingslake]

Great! Where did the noncoupled simulation stop? On the retrograde or the prograde part?

On Wed, Nov 30, 2022, 5:06 PM George Lu @.***> wrote:

[image: image] https://user-images.githubusercontent.com/25989736/204917593-f0a565bb-1622-4f0e-ac9a-0c1a98d43582.png I've added in a retreat scenario with Schoof's bed. We can see the retreat slow going up the sill, and then accelerate after moving past it. Weaker perturbations lead to a steady state on top of the sill. Regardless of the scenario, if I run it without coupling and the same initial steady state N (kept static), the uncoupled model halts retreat right away.

— Reply to this email directly, view it on GitHub https://github.com/glugeorge/coupled_ice_hydrology/issues/3#issuecomment-1332789720, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALTXJ3LATBLCYMTKHVJHUHDWK7FM3ANCNFSM6AAAAAAR6SIWPE . You are receiving this because you were assigned.Message ID: @.***>

[glugeorge]

The noncoupled simulation stops on the prograde part of the sill, pretty much meters inwards from the initial condition. We can see how steep that initial effective pressure slope is so when held static, it prevents the noncoupled simulations from moving much. I'm trying to devise a scenario where retreat starts on the retrograde portion, so maybe that could somehow encourage the noncoupled model to move past the peak in effective pressure and exhibit some more interesting behavior rather than reaching a standstill within the first few timesteps

[glugeorge]

Also, I was thinking, would it be useful to include the uncoupled model but allow the constant (but nonuniform) hydrology system stretch with the grounding line (what I had before)? I figure it could be a good intermediary step that shows how the coupled model has additional impact on the dynamics on top of the stretching hydrology. Because I've been thinking that having the static system stop any kind of movement in the first few timesteps might not be interesting (and expected due to the effective pressure shape and the nonzero boundary condition). Could there be a physical basis for the hydrology system stretching?

[jkingslake]

The noncoupled simulation stops on the prograde part of the sill, pretty much meters inwards from the initial condition. We can see how steep that initial effective pressure slope is so when held static, it prevents the noncoupled simulations from moving much. I'm trying to devise a scenario where retreat starts on the retrograde portion, so maybe that could somehow encourage the noncoupled model to move past the peak in effective pressure and exhibit some more interesting behavior rather than reaching a standstill within the first few timesteps

I think that's a good idea.

I am not sure about the stretching of the drainage system idea. I think our experimental design is somewhat inspired by the inversions people use to characterize bed properties when they are trying to simulate ice streams/sheets 'realistically' (i.e. attempting to recreate the observed velocities). Maybe (re)read some of those papers and we can talk about how the fixing-the-effective-pressure simulations align with those approaches from ice-sheet modelling.

[jkingslake]

...I think stretching doesn't really have a physical basis that I can think of. Maybe there is an argument for shifting the N profile with uniformly with the GL as it moves, as an intermediate complexity approach that could be useful to people, but I am not sure about the stretching.

I think it is important to think about how realistic our N profiles are - as in, do we see an area of high N upstream of the GL aligned with the steepest dh/dx? And if not, maybe it's because h has a different shape, perhaps down to a lower $C$. And this brings me back to the idea that investigating a lower $C$ regime would be interesting.

[jkingslake]

here's a paper to take a look at: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2014JF003239