ACCESS-ESM1.5 in pacemaker mode

Earth System Earth System Model
, ,
1

I would like to run the new release of the ACCESS-ESM-1.5 model in ‘pacemaker’ mode. In this case, where SSTs in a particular area are prescribed, and all other aspects of the model evolve freely.

From what I understand and in conversation with people who have done this with the CM, this requires

  • the SST restore files to tell the model what SST should be in my restoring region (1 per year) (I have made these, or attempted to)
  • a masking file with a linear transition from ‘full restore’ to ‘no restore’ at the edges of the prescribed area (I have made this, or attempted to)
  • some edits to the model so it knows what to do (Holger Wolff has made a start on this)

Full disclosure: I am not a modeller so I’m starting from essentially a zero knowledge base (sorry). Holger was helping me with this at the end of last year (through CLEX) but we ran into some difficulties and now he has moved on. At least one issue was to do with how I made the restore files but we didn’t manage to pinpoint the specific problem.

Help would be very much appreciated - sorting this out is probably quite a long play but I’m posting here because surely I won’t be the only person running pacemakers in the ACCESS-ESM

Other details: I made the SST restore files in python in ARE, although I think others who have done similar things with the ACCESS-CM have made their restore files using NCL (which I don’t speak).

1 Like
2

Hi @georgyfalster, thanks for your post.

It would be great to have pacemaker capability in ESM1.5. Based on some of the earlier discussions on CM2, my initial impression is that similar changes should be possible with ESM1.5.

I’d be happy to try and help with changes to the ESM1.5 configuration, but I think expertise from others with CM2 pacemaker experience will be really useful too.

Do you have much information about the changes Holger was making to the model?

4

Wonderful, thanks @spencerwong

From what I remember/understood at the time, the structure was largely in place already to run the ESM1.5 in pacemaker mode. It was just something along the lines of:

  • modify the config.yaml file in the ocean model telling it where the restore mask is and making it read in a pre-run script to update the SST file each year
  • modify some parameters in the ocean/input.nml sbc namelist

Ariaan Purich had done this with a previous release of the ESM1.5 and has some example files that should speed things up a lot - particularly with someone on board who knows MOM5.

9

Hi @georgyfalster, I’ll add a brief summary about the setup now that these experiments are complete.

The pacemaker experiments described here have now been successfully run by adapting the setup from @ariaan. There were a few steps in getting the setup correct, and the most difficult part was creating a temperature restoring file, temp_sfc_restore.nc, which MOM5 wouldn’t complain about. There were few important points noticed through trial and error:

  • It’s possible to run a long (e.g. 500 year simulation) using a single continuous temp_sfc_restore.nc restoring file. In this case, it was important to remove the modulo attribute from the time dimension and pad data at the start of the very first month and end of the very last month so that the time interpolation worked properly.
  • When processing the restore file with xarray, it was easy to inadvertently lose attributes needed by MOM5. In particular, the model required the following:
    • time needs to be a record dimension
    • The temp variable should have the dimension order TIME, DEPTH1_1, GRID_Y_T, GRID_X_T
    • The following coordinate attributes appeared to be necessary for the model to work out which axis is which: DEPTH1_1:units = "m" ,GRID_Y_T:units = "degrees_north" , and GRID_X_T:units = "degrees_east"

@georgyfalster has uploaded payu configurations for the experiments here

Feel free to add any other notes or information!

1 Like
10

Perfect, looks great! Only things to add are that I’ll be adding example scripts and restore files to those same github repos.

Also for anyone else considering something similar: for these long runs, we also used the ‘diagnostic’ suite of outputs to keep the output size reasonable.

The experiment proposal is described in more detail here: Experiment Proposal: Indo-Pacific pacemakers

2 Likes