Plan to continue run with Baltic fixes and UM GC5 central tuning (from @jemmajeffree). @sofarrell suggests checking these changes in an AMIP run too.
@sofarrell is preparing notes on penetrating SW radiation.
Evaluation
Using the multi variable atmosphere files is very slow. @MartinDix to look at switching to single variable files in the netCDF conversion and to creating a catalog of the CM2 0.25 runs (which used OM3 diag table).
Discussion on how to best handle OM3 scripts modified for CM3. Currently on a branch in OM3 repo (some variable names had to be changed) @dougiesquire keen to avoid divergence and propagating errors. Papermill allows switching datastores. CM3 should use OM3 diag table and file naming so that output is compatible.
MartinDix
(Martin Dix ACCESS-NRI Associate Director for Model Development)
2
Date: 5 Nov 2025
Continuing CM3 O025 run
Run with updated Baltic bathymetry (truncated OM3 file to avoid having to update model grid) and GC5 central atmospheric parameters restarted. Run shows a ~ 1.5 W/m^2 worsening of the net TOA radiation balance GC5 central configuration · Issue #53 · ACCESS-NRI/dev_coupling · GitHub. Possibly due to missing some GC5 central parameters that are not available in the standard vn13.0 code (Met Office used tests that made some internal parameters namelist variables).
TODO: Check whether this run did improve Baltic salinity drift and which components of the radiation balance changed.
Code version updates
ACCESS rAM3 uses UM vn13.5 and there’s now an update of both AM3 and CM3 to the same version. Runs ok but not fully tested yet. With this version we should be able to use the full set of GC5 central tunings.
TODO: @MartinDix to test GC5 parameters in AM3 AMIP experiments
Datastore:
@CharlesTurner’s work on zarr virtualisation has improved speed of reading atmospheric data enormously, though sea ice still seems slow. @anton suggests checking whether CM3 writes grid information every time (OM3 doesn’t). Virtualisation requires a higher level of file consistency and is not yet automated.
Question of which CM2 runs to use. Ideally use @wghuneke’s run but need to check where it started. 100 km ocean runs also worth adding (PI control will aready be in ESGF catalog but could add PD control).
@cbull to look at updating the evaluation template to include CM2 runs.
MartinDix
(Martin Dix ACCESS-NRI Associate Director for Model Development)
3
Date: 19 Nov 2025
Sea-ice volume in NH:@sofarrell says there’s a problem with the ice distribution in the Arctic.The Alaskan side gets too warm in summer while the Russian side which should have melting retains thick ice. Possible problem with pressure pattern and direction of surface winds? TODO: compare winds with AMIP run.
Continued spinup run: Baltic salinity increased after bathymetry changes and now looks similar to CM2 0.25 level.
GC5 central AMIP experiments: AMIP experiments using a Met Office GC5 central configuration show a decrease in net TOA radiation compared to GC5 (UM vn12.2 suites u-ct505 and u-du539). Suggests it was missing parameters that caused the increase in CM3. Next will try these parameters in a vn13.5 AMIP run and then CM3.
Consider adding CM2 piControl to the notebook template?
@kieranricardo reports loading CM3 ice data is slower than OM3. @anton suggests this may be because files have coordinate variables (which were removed from OM3). However files like `/g/data/zv30/non-cmip/ACCESS-CM3/cm3-run-11-08-2025-25km-beta-om3-new-um-params/archive/2000/ice/access-cm3.cice.1mon.mean.2000-01.nc` don’t have grid variables, though do have angle and anglet which are probably unnecessary.
