Changes for the CalvingMIP submission #70
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Comparing to results from other groups, we found that CISM has too much retreat
during the retreat phase of Experiments 2 and 4, mainly because of the method
used to compute cf_length. I was computing cf_length by reconstructing the CF
along a path connecting cell centers. This path can become noisy (i.e., with
spurious projections and indentations), giving excessive retreat.
I tried two new approaches:
(1) Compute the CF length in a cell based on the number of cell edges bordering the ocean.
Cells with two ocean edges have their thinning rate multiplied by sqrt(2),
reflecting the longer CF (but not by a factor of 2, since the calving direction
is not normal to the two edges).
(2) Compute the CF length based on the secant of the angle between the calving direction
and the x and y axes. The thinning rate is multiplied by 1 for the compass directions,
by sqrt(2) for the diagonals, and by intermediate values in between.
Method (1) does not fix the problem of excessive retreat.
Method (2), however, gives more accurate retreat and is now the default for CalvingMIP.
For realistic ice sheet geometries, the calving rate is not radially symmetric,
so method (2) is not appropriate. Method (1) is now the default for subgrid CF schemes
other than CalvingMIP. It is a bit cleaner than the previous method.
Other changes:
(1) Modified the method of estimating the CF location along the diagonal trajectories
in the circular domain. This method uses the x0 and y0 coordinates and is a bit
more accurate than the old method; it is less prone to getting stuck for several
years at the same value.
(2) Added a subroutine to estimate the CF location along the Halbrane and Caprona
trajectories for the Thule geometry. The Caprona method is approximate and might
not be as accurate as offline methods using 2D interpolation schemes.
(3) The 1D fields x0 and y0 are now loadable at startup, like x1 and y1.
(4) To deal with I/O involving x0 and y0, I added some code in the distributed get_var
and put_var functions in parallel_mpi.F90.
(5) Added a config option, which_ho_calvingmip_domain, to specify whether the user
is running on the circular or Thule domain. This parameter is set automatically
by the CalvingMip setup script.
Note: This commit will be followed by a commit that includes a modifed CalvingMIP setup script.
That script creates a grid with the origin at a cell corner instead of a cell center.
This commit modifies the CalvingMIP setup script. The main change is that the grid origin is now located at a cell corner instead of a cell center. This means that on a 5-km grid measuring 1600 km along each edge, there are now 320 cells in each direction instead of 321. This change makes it easier to initialize the calving front at a distance of 750 km from the origin, instead of being extended or falling short by half a grid cell. Other changes: * Fixed a bug that gave the wrong retreat rate for Experiment 4 * Added the which_ho_calvingmip_domain variable to the config files * Edited the README file for clarity and accuracy
This commit extends the Circular and Thule grids by one cell in each direction. E.g., for a 5km run, where there were previously 160 cells in each direction, there are now 162. This extension makes it easier to interpolate scalars to the CalvingMIP results grid, which corresponds to the CISM (x0,y0) grid. I re-ran the experiments, which should provide our final submitted results. Compare to the old grid, answers change at the machine roundoff level, presumably because the grid decomposition is different.
This is the version used for the CISM CalvingMIP submission in Nov. 2024. This commit changes the way thck_effective (the effective ice thickness) is computed in calving front cells. For any CF cell with an interior neighbor, the calculation is unchanged; thck_effective is set to the maximum thickness of the cell's interior edge neighbors. For CF cells without interior edge neighbors (i.e., cells whose only edge neighbors are other CF cells), thck_effective is now set to the maximum thickness of the CF neighbors. Previously, we set thck_effective = thck, giving values that are too low. I also fixed a minor bug in the way the CF location is computed along the diagonal axes for the circular domain. I set dthck_dx_cf = 0.0005 in the config template file for CalvingMIP. This value gives slightly more accurate results than the default value of 0. Setting dthck_dx_cf = 0 gives a little too much retreat. If desired, we could modify the python setup script to control this parameter at the command line. We ran the final CalvingMIP runs with dthck_dx_cf = 0.0005.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This branch include several changes to improve accuracy for the CISM CalvingMIP submission of Nov. 2024: