MOM6 PanAntarctic @ 0.1°

[AndyHoggANU]

Following our first attempt at a high-resolution PanAntarctic, we aiming to retreat back to 0.1° resolution for the time being. The advantages of this are that:

• We can run faster, and tune open boundaries, etc, with less computational cost;
• We can make a more direct comparison with ACCESS-OM2-01.

The first step in doing this was to regrid the runoff (and fix the mistake we made first time around, where we attempted to interpolate runoff). To do this, we took the JRA RYF V1.3 runoff file and used Alistair Adcroft's regrid_runoff script as follows:

./regrid_runoff.py /g/data/ik11/inputs/access-om2/input_20201102/mom_01deg/ocean_hgrid.nc /g/data/ik11/inputs/access-om2/input_20201102/mom_01deg/ocean_mask.nc -m mask -r friver -p /g/data/ua8/JRA55-do/RYF/v1-3/RYF.runoff_all.1990_1991.nc RYF.runoff_all.1990_1991.panan01.nc

This does most of the job, but note that it takes a looong time with such a high-resolution dataset. Also, it doesn't get the attributes quite right, and we need to add:

ncatted -O -a cartesian_axis,i,a,c,"X"  -a cartesian_axis,j,a,c,"Y"  RYF.runoff_all.1990_1991.panan01.nc

Finally, note that the above uses the ACCESS-OM2-01 global grid files, while MOM6 expects the runoff file to exactly match the model grid. I did this with:

ncks -d j,,844 -d JQ,,845 RYF.runoff_all.1990_1991.panan01.nc -O RYF.runoff_all.1990_1991.panan01_cropped.nc

These outputs are all currently stored in /g/data/ik11/inputs/mom6/regrid_runoff but we may have to be more organised about how we save these files once we have a good workflow going.

Note that it would seem more logical to use the actual model grid to do this, but (a) it took longer and (b) it seemed to remap everything north of 37S onto the northern boundary. An alternative may be to chop down the JRA runoff file to the model domain, and then regrid using model grid files. This may be needed for the 0.05° case when we get back to it.

In the meantime, we have spun up the first year of a simulation called panant-01-zstar, and it looks OK. Next, we will add a panant-01-hycom1 simulation.

[AndyHoggANU]

We are trying to better understand how the hycom1 vertical coordinate is working here. So we will use the panan-01 case for a few new tests.

Note that in each of these new cases we have increased the resolution in the rho2 diagnostics be specifying a file-based scheme:
DIAG_COORD_DEF_RHO2 = "FILE:diag_rho2.nc,interfaces=rho2"
This file was made here. This seems to work OK.

Test 1: Have done a short simulation with
ALE_COORDINATE_CONFIG = "HYBRID:hycom1_75_800m.nc,sigma2,FNC1:1.1,4400,2.65,.01"
This is acting to thicken the surface layers more rapidly, and will be closer to the standard KDS75 case that we use in the z* version.
This test is called panant-01-hycom1-v2. The thicker layers in the upper ocean mean that there is a larger z* region, which will hopefully be addressed in the second test.

Test2: Repeating this test, but now with an altered file for target densities
ALE_COORDINATE_CONFIG = "HYBRID:hycom1_panant.nc,sigma2,FNC1:1.1,4400,2.65,.01"
These target densities are denser for the same layer, aiming to particularly improve resolution around layer 30. The notebook for producing them is here. This notebook also shows the effect of the FNC1 function used in Test1 above. Will start with a year of this and then evaluate, calling it panant-01-hycom1-v3.

[AndyHoggANU]

Major Update

Over the last few months we have been working to improve the vertical resolution in the hycom1 coordinate, and in the rho2 density diagnostics. Ultimately, we were sidetracked by a change in the default configuration of SIS, which meant we were restarting (with zero sea ice) every 3 months. This is now handled in payu, and we are able to run again.

While tracking this bug down, we conducted numerous tests, from panant-01-hycom1-v2 through to panant-01-hycom1-v12. All of these tests have been deleted.

The latest authorised runs are now panant-01-hycom1-v13 and panant-01-zstar-v13. The difference between these cases and the original runs primarily come down to:

• In the hycom1 case, we changed ALE_COORDINATE_CONFIG from "HYBRID:hycom1_75_800m.nc,sigma2,FNC1:1.08,4000,4.5,.01" to "HYBRID:hycom1_panant.nc,sigma2,FNC1:1.1,4400,2.9,.01". The changes in the FNC parmeters act to deepen the near-surface geopotential layers, leaving us free to have more density layers in hycom1_panant.nc.
• In both cases, we changed DIAG_COORD_DEF_RHO2 to be file-based: "FILE:diag_rho2.nc,interfaces=rho2" so as to have finer control on deep isopycnal layers.

Otherwise, runs are similar. Preliminary output tells us that the changes in the coordinate have acted to reduce dense shelf water outflow in the hycom1 case. While we investigate this issue, we recommend focussing on the zstar case.