How to run ACCESS-OM2-RYF from an existing control run?

COSIMA Experiments
,
1

Hi,
I want to run ACCESS-OM2-RYF9091 forced by jra55v1.3 (GitHub - COSIMA/01deg_jra55_ryf at ryf9091-on-gadi) from an existing control run. The base documentation that I followed is -run-access-om. I understand that the ACESS-OM2 runs are now managed by ACCESS-NRI, hence that tutorial may not be useful for my use case. Now I also tried to follow this tutorial - Running Model Experiments with payu and git. But I keep getting the below error -

 ice_read_nc_xy: Cannot find variable licefw_io

Which maybe is due to restart file incompatibility? The restart file I am trying to use is this one -

/g/data/ik11/outputs/access-om2-01/01deg_jra55v13_ryf9091/restart995/

Overall, I am a bit confused about all this, and I would very much appreciate if anyone could tell me what the most efficient way is, to run 01deg_jra55v13_ryf9091 setup from a well spun-up existing restart file.

Thanks!

2

Hi Sougata,

To run with old restarts, you need to follow the old COSIMA wiki instructions here:

Also the github config you link above is unlikely to match the config that the restart was run with. If that matters to you, best to ensure you find and use the correct config.

1 Like
3

Hiya. Yes, that is a way to do a restart from an existing run, using the same version of ACCESS-OM2. But the code has moved a long way from that original RYF run, which started in about 2019! Others will know more, but I think it will be difficult to use those restart files to run with the new ACCESS-NRI release of OM2. One option may be to take the T/S from the restart file to create a new initial condition file and run it from there? [It will be close to spun up, but may need a few years to get stable.]

We have been considering for some time whether we should rerun the RYF control, and have concluded that the use case may not be so urgent (and we definitely don’t have the spare CPU time this quarter).

1 Like
4

Hi @sb4233 - I’ve just moved this topic from General to COSIMA since it’d be more relevant there - I’ve also tagged it with help since this is a help request :slight_smile:

1 Like
5

In that case, is there any configuration of the ACCESS-OM2-01 (including interannual forced ones) that can be used from an existing restart file?
My purpose for this simulation is to take output of some additional diagnostics, which are not present in the control runs, and for that it does not matter whether the simulation is forced by interannual or repeat year forcing.

6

Thanks!!

7

Hi @sb4233 , this might be a little off-topic, but we’re currently preparing to run a new control run of the 0.1deg RYF configuration with BGC included. There’s a poll linked here for input.

As @AndyHoggANU mentioned, though, we don’t have spare CPU allocation for this quarter, so the run will be delayed.

8

I will review the restart files and get back to you if I’ve found something later this week.

1 Like
10

Hi @sb4233, if you want to continue an existing run without introducing additional changes (other than intended perturbations), you’ll need to run with the same (i.e. old) configuration, as Adele suggested. Did this work for you?

11

Hi @adele-morrison and @aekiss, I have tried the steps mentioned here: Tutorials · COSIMA/access-om2 Wiki · GitHub

For clarity I am writing the steps here as well -

#STEP 1: 
git clone --no-hardlinks https://github.com/COSIMA/01deg_jra55_ryf.git my_new_exp

#STEP 2: 
cd my_new_exp

#STEP 3: 
git remote -v

#STEP 4: 
git remote set-url origin https://github.com/COSIMA/01deg_jra55_ryf.git

#STEP 5: 
git branch -va
gitk -all &
git log

#STEP 6: 
git checkout 54d6340 #<commit hash for run 1115>

#STEP 7: 
git checkout -b my_new_exp

#STEP 8: 
vim sync_output_to_gdata.sh

#STEP 9: 
#change the GDATADIR= ... 

#STEP 10: 
payu setup && payu sweep

#STEP 11: 
ln -r /g/data/ik11/outputs/access-om2-01/01deg_jra55v13_ryf9091/output1115  </path/to/archive/dir/output>

#STEP 12: 
ln -r /g/data/ik11/outputs/access-om2-01/01deg_jra55v13_ryf9091/restart1115 </path/to/archive/dir/restart>

#STEP 13: 
module use /g/data/hh5/public/modules
module load conda/analysis3

#STEP 14: 
payu setup

#STEP 15: Make other changes as necessary, e.g., diag_table changes

#STEP 16: 
payu sweep
#STEP 17: 
payu run

But I believe the JRA forcing files have been removed, as indicated by the below error that I got:

Error - from NetCDF library
ncvar opening /g/data/x77/amh157/JRA55-do/RYF/v1-3/RYF.rsds.1990_1991.nc
No such file or directory

Could you please tell me what’s the updated path for the forcing file?

12

Hi @sb4233, basic question, are you a member of vk77?

To test: what happens when you do ls /g/data/x77/? (I’m not at the moment; so can’t test to see if this file exists.)

13

@sb4233 Here is the updated path of the forcing files

/g/data/vk83/experiments/inputs/JRA-55/RYF/v1-4/data

14

JRA55-do folder is not there in /g/data/x77/amh157/

1 Like
15

Also here: /g/data/ik11/inputs/JRA-55/RYF/v1-3/

1 Like
16

Thanks! and I need to change the path in ./atmosphere/forcing.json right?

17

@sb4233 change the path here and do payu sweep && payu setup to see if payu is able to identify the input files.

18

Thanks @ezhilsabareesh8, but I think the config you are referring to was forced by v1.4, but the config I am running is forced by v1.3.
My config.yaml looks something like this -

# Force payu to always find, and save, files in this short (or scratch) project directory
shortpath: /scratch/v45

# Model configuration
name: common
model: access-om2
input: /g/data/ik11/inputs/access-om2/input_08022019/common_01deg_jra55
submodels:
    - name: atmosphere
      model: yatm
      exe: /g/data/ik11/inputs/access-om2/bin/yatm_1bb8904.exe
      input: /g/data/ik11/inputs/access-om2/input_08022019/yatm_01deg
      ncpus: 1

    - name: ocean
      model: mom
      exe: /g/data/ik11/inputs/access-om2/bin/fms_ACCESS-OM_97e3429_libaccessom2_1bb8904.x
      input:
          - /g/data/ik11/inputs/access-om2/input_08022019/mom_01deg
          - /g/data/x77/amh157/passive/passive4
      ncpus: 4358

    - name: ice
      model: cice5
      exe: /g/data/ik11/inputs/access-om2/bin/cice_auscom_3600x2700_722p_d3e8bdf_libaccessom2_1bb8904.exe
      input: /g/data/ik11/inputs/access-om2/input_08022019/cice_01deg
      ncpus: 799

# Collation
collate:
  walltime: 10:00:00
  mem: 192GB
  ncpus: 4
  queue: normal
  exe: /g/data/ik11/inputs/access-om2/bin/mppnccombine

So I assume for my case I need to replace the input for atmosphere to

- /g/data/ik11/inputs/JRA-55/RYF/v1-3/

??

19

Yes that is correct.

20

The model is running now but got the same error:

gadi-pbs:
                                                                                                   Req'd  Req'd   Elap
Job ID                         Username        Queue           Jobname         SessID   NDS  TSK   Memory Time  S Time
------------------------------ --------------- --------------- --------------- -------- ---- ----- ------ ----- - -----
141271466.gadi-pbs             sb4233          normal-exec     01deg_jra55_ryf  3620603  108  5184   21tb 03:50 R 01:22:39
   Job run at Mon May 19 at 15:47 on (gadi-cpu-clx-2594:ncpus=48:mem=201326592kb:jobfs=102400kb)+(gadi-cpu-clx-25...

Also I don’t see any output file in the work directory. Is there a way to monitor the run ?