Non-smooth weertman basal velocity

Extension of Elmer in computational glaciology
Post Reply
ChristianB
Posts: 10
Joined: 28 Apr 2014, 22:02
Antispam: Yes

Non-smooth weertman basal velocity

Post by ChristianB »

Hi all,

I have been getting some sliding velocities that look a bit strange to me using the Weertman relation. They are however not completly unrealistic and wanted to hear if other users are seeing similar sliding patterns. It looks similar to the post: Small basal velocity value patch from Adjoin inverse method. In that post however a inverse problem is solved where I am only working on a forward problem solving a diagnostic setup.

The picture below shows norm basal velocity (left) and norm surface velocity (right). I find it strange that sliding velocities are not smooth. At one spot sliding is >100 m/yr while right next to it almost zero. The total domain is 40x20 km with 200 meters between nodes. I would expect sliding velocity to be more like the surface velocity (however smallere in magnitude of course) when nothing but surface and bed topography is given to the model.

Image


I am using the navier-stokes solver (no parstokes) with the following boundary condition,

Code: Select all

!! bedrock:
Boundary Condition 3
  Name = "bedrock"
  Target Boundaries  = 5
  Flow Solution Name = string "Flow Solution" ! and that is its name

  Normal-Tangential Velocity = Logical True
  Flow Force BC = Logical True

  Velocity 1 = Real 0.0

  Vxb = Equals Velocity 2
  Vyb = Equals Velocity 3

  Slip Coefficient 2 =  Variable Coordinate 1
    Real Procedure "ElmerIceUSF" "Sliding_Weertman"
  Slip Coefficient 3 =  Variable Coordinate 1
    Real Procedure "ElmerIceUSF" "Sliding_Weertman"

  Weertman Friction Coefficient = Real 0.0283
  Weertman Exponent = Real 0.5
  Weertman Linear Velocity = Real 0.00001
End
I have tried different Friction Coefficients and Exponents.

To extract basal values I have tried a couple of different methods. Passing the raw vtk data in python and extracting all bed nodes and using the ExportVertically solver to get velocity values at the basal boundary condition. Both methods give the same results.

I am however not sure if the "Normal-Tangential Velocity" call in the BC results in permanent rotation of all velocities components along the bed. Do I have to rotate them back to align with the global coordinate system or are values in the VTK files already aligned with the global coordinate system?

Is this a "normal" sliding pattern for the weertman sliding law in Elmer or do I need to postprocess the solution in some way?

Cheers,
Christian
rgladstone
Posts: 64
Joined: 15 Apr 2013, 16:23
Antispam: Yes

Re: Non-smooth weertman basal velocity

Post by rgladstone »

Hi Christian, the "Normal-Tangential Velocity" setting is local to the boundary condition in which you apply it. You don't need to "undo" this.

Are you using a spatially uniform basal drag coefficient? If so then the non-smooth basal velocity might arise from a non-smoth bedrock. I am guessing your upper surface geometry is quite smooth from the fact that the surface velocity looks quite smooth.

If the non-smooth basal velocities arise from some kind of numerical instability you could test for this by running at several different resolutions and seeing whether the pattern is consistent across resolutions.

When you say "norm basal velocty" and "norm surface velocity", what does the "norm" mean? It looks like you have plotted velocity magnitudes in metres per year. The normal velocity should (certainly at the bottom boundary) be zero.

BTW, which glacier is it? :)

Cheers,
Rupert
ChristianB
Posts: 10
Joined: 28 Apr 2014, 22:02
Antispam: Yes

Re: Non-smooth weertman basal velocity

Post by ChristianB »

Hi Rupert,

Thanks for the reply.

I am using a constant basal drag coefficient. You are right that on a small scale the bedrock is not smooth, however, I use 200 meter spacing between nodes (square grids for the moment) so on that scale it should be "smooth". I think I will try your suggestion and refine the grid resolution. If nothing else I can see if it is numerical or model related.

When I say norm I mean L2-norm which is equal to the magniude (without the vz component).

It is actually not a "real" glacier but a synthetic landscape with a home grown glacier on top. I needed a controlled enviroment to test some things :)

Cheers,
Christian
Post Reply