I’ve noticed that changing the cycling frequency from 1 year to 1 month when running ACCESS-CM2 leads to different results. For example, the plots below show globally averaged ocean metrics in a simulation run at yearly cycling frequency (red) and monthly cycling frequency (blue).
I only noticed this because I was trying to identify where/why the kinetic energy was blowing up in this particular run. Are others aware of this and is this the expected behaviour?
Hi Hannah
Thanks for raising this. As I understand it, others have identified this problem in the past where results are not reproducible across different run lengths. We are working to make sure this doesn’t happen in newer models we develop but don’t plan to fix this in CM2. I’ll chat to the ACCESS-NRI folk and confirm. It may make sense to add this (as a known issue) to the hive docs page for CM2.
I raised an issue to include this as a known issue with restart reproducibility in the documentation:
The CM2 version of CABLE doesn’t include all the fields in the restart file that are required to give restart reproducibility.
This has been fixed in the version being tested in ACCESS-AM3 but the changes required are too extensive to backport to ACCESS-CM2.
