Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
New options to force retreat, remove isthmuses, and expand calving mask
I added several options to support ISMIP6-based Antarctic experiments with ice-shelf collapse.
First, I modified the force_retreat option, which uses a real mask with values in the range [0,1]
to force ice to thin or retreat. Previously, force_retreat was a logical option (T or F).
Now, it is an integer option (0, 1, or 2).
Option 0 => no forced retreat
Option 1 => Ice is thinned or removed wherever ice_fraction_retreat_mask > 0.
Option 2 => Floating ice is removed in cells with ice_fraction_retreat_mask exceeding
a threshold value (0.01 by default), provided these cells are connected to the ocean
through other cells with ice_fraction_retreat_mask above the threshold.
Grounded ice is unaffected.
Option 1 is the same as the old option force_retreat = T, used for ISMIP6 Greenland experiments
with forced retreat of the calving front. However, the old convention was to remove ice from
cells with a mask value < 1; the new convention is to remove ice from cells with a mask value > 0.
The new option 2 is appropriate for ISMIP6 Antarctic experiments in which a hydrofracture mask
is read in as a forcing file.
Initial experiments using a retreat mask from CESM2 21st century simulations showed that several shelves
become unstable, for example when a shelf is divided into two regions connected by a narrow isthmus
one cell wide. I addressed this problem by adding an option to remove ice isthmuses.
An isthmus is define as a floating or weakly grounded cell with either ice-free ocean or thin floating ice
on each side: e.g. if cell(i,j) is bordered by open ocean in cells (i-1,j) and (i+1,j).
When isthmus cells are removed after applying the retreat mask, a potentially unstable shelf is divided
into two separate regions. Subroutine remove_icebergs then removes the unstable region downstream,
leaving the more stable region upstream.
Also, I added a new logical option 'expand_calving_mask', to be used in conjunction with the calving_mask option.
If the new option is set to true, then calving_mask is expanded at model initialization to include
all floating ice in selected basins. This floating ice will then calve immediately, leading to
acceleration of grounded ice.
By default, this option is applied to 13 of the 16 ISMIP6 Antarctic basins. It is not applied to
the three basins that include most of the Ross, Filchner-Ronne and Amery ice shelves.
To change the basins where floating ice is calved, the user must insert basin numbers by hand.
Note that this option differs from the ABUMIP float-kill option, in that all ice is calved in cells
that are initially filled with floating ice, but not in cells where the ice subsequently floats.
In multi-century transient experiments at 8 km and 4 km, I found that when forcing retreat using
the ISMIP6 shelf-collapse mask, velocities can reach several tens of km/yr, but usually return to
reasonble values of < 10 km/yr after a time step or two.
In experiments using the expanded calving mask, ice speeds in the first few years can exceed 100 km/yr.
These usually return to realistic values within a few simulation years, but sometimes the simulation
can crash. More work would be needed to make this option truly robust.
This commit is BFB except when using the new options.
For earlier experiments with force_retreat = T in the config file, we will need to set
force_retreat = 1 and replace ice_fraction_retreat_mask with (1 - ice_fraction_retreat_mask)
to reproduce the results.- Loading branch information
1 parent
db18175
commit c1319da