Skip to content

Add Thwaites gamma0 ensemble files - #1015

Draft
trhille wants to merge 5 commits into
MPAS-Dev:mainfrom
trhille:thwaites_gamma0_ensemble
Draft

trhille wants to merge 5 commits into
MPAS-Dev:mainfrom
trhille:thwaites_gamma0_ensemble

Conversation

@trhille

@trhille trhille commented Nov 15, 2026 •

Copy link
Copy Markdown
Collaborator

Add template and setup files for Thwaites gamma0 ensemble.

Note: For the GNN ensemble, the user will likely need to overwrite the job_script.sh files to use srun -n 32 -G 1 ./landice_model instead of compass run. I'm not sure that compass run can handle GPU allocation on Chicoma.

Checklist

  • User's Guide has been updated
  • Developer's Guide has been updated
  • API documentation in the Developer's Guide (api.rst) has any new or modified class, method and/or functions listed
  • Documentation has been built locally and changes look as expected
  • Document (in a comment titled Testing in this PR) any testing that was used to verify the changes

Add template and setup files for Thwaites gamma0 ensemble.
@trhille
trhille marked this pull request as draft November 15, 2026 05:38
@trhille

trhille commented Nov 15, 2026 •

Copy link
Copy Markdown
Collaborator Author

Testing

I am able to set up and run on Chicoma, but the melt rate is too low. I specified 81.9 Gt/yr in the .cfg file but I am getting ≤70 Gt/yr:
image

It turns out that the calculation uses time 0 from the TF file: TFocean = ff.variables['ismip6shelfMelt_3dThermalForcing'][0, :, :], so really this is an issue with the input files.

edit: This is fixed in 2934e75

trhille and others added 4 commits November 14, 2026 22:43
Turn off grounded basal melt because the GNN solver currently assigns
the depth-averaged velocity to all levels, which leads to too much
basal melt.
Add optional calibration_start_year and calibration_end_year options to
[spinup_ensemble]. When both are set, calc_mean_TF averages the 3D thermal
forcing over all time slices whose xtime year is in that inclusive range
before interpolating to the ice draft. This lets deltaT be calibrated
against a period mean while runs still use the time-varying TF_file_path.
Without the options, the first time slice is used as before.

Document the options in the spinup templates, the User's Guide and the
Developer's Guide.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Add separate templates for Thwaites MOLHO and GNN ensembles. GNN
currently requires running on branch trhille/mali/add_gnn_hooks.
Use a 4 layer, 512 hidden channel GIN, which we found worked better
than the GIN with 4 layers and 1024 hidden channels. Also use first-
instead of third-order numerics, which allows better agreement between
the simulations using GNN and MOLHO solvers.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant