default CMIP7 CO2 files doesn't appear to be working?

[wwieder]

Brief summary of bug

WIEMIP simulations revealed that the original WIEMIP co2 file and the CMIP7 co2 file fail after 6 months in runs from 1850 to 2000:
#3936
For the WIEMIP simulations we seem to have solved the problem with a new file as posted here:
#3936 (comment)

General bug information

Default CMIP7 file that fails:
/glade/campaign/collections/gdex/data/d651077/cesmdata/inputdata/atm/datm7/CO2/co2_input4MIPs_GHGConcentrations_CMIP_CR-CMIP-1-0-0_gm_1750-2022_CMIP7_c20250228.nc

Default CMIP6 files still works:
/glade/campaign/cesm/cesmdata/inputdata/atm/datm7/CO2/fco2_datm_global_simyr_1750-2014_CMIP6_c180929.nc

Sample directory (kind of hacky for WIEMIP), this is a CLM6 HIST case
/glade/work/wwieder/wiemip/clm6_hh_ukesm_cou

playing around with setting in user_nl_datm_stream for co2tseries.cmip6_20tr

[slevis-lmwg]

Investigating by first creating a new case:
./create_newcase --case ~/cases_sp/ctsm54042_hist --compset IHistClm60Sp --res f10_g37 --run-unsupported

In the env_run.xml I see this relevant info:

    <entry id="CLM_CMIP_ERA" value="cmip7">
      <type>char</type>
      <valid_values>cmip6,cmip7</valid_values>
      <desc>Whether to use CMIP6 or CMIP7 fsurdat/landuse files.
    As of ctsm5.4 we do not have CMIP7 PFT raw datasets for SSPs, so default to CMIP6 for these. For other compsets default to CMIP7. User may modify to a non-default value in env_run.xml.
Caveats:
1) We supply only CMIP7 C13/C14 isotope datasets, so these get used regardless.
2) We supply only CMIP7 population density with ctsm6 in non-SSP cases, because the fire model is calibrated to that; conversely, for pre-ctsm6 and for SSP we supply only CMIP6 population density.
3) We supply only CMIP6 nitrogen deposition (ndep), so this gets used regardless.
4) For DATM we supply only CMIP6 aerosols.
5) For DATM we supply only CMIP6 CO2</desc>
    </entry>

...which answers my earlier confusion about why we hadn't realized that the CMIP7 CO2 file was problematic. It hasn't been used.

Now adding to my case's user_nl_datm_streams the CMIP7 CO2 file:

co2tseries.cmip6_20tr:datafiles=/glade/derecho/scratch/slevis/temp_work/WIEMIP/co2/co2_input4MIPs_GHGConcentrations_CMIP_CR-CMIP-1-0-0_gm_1750-2022_CMIP7_c20250228.nc
co2tseries.cmip6_20tr:year_first=1850
co2tseries.cmip6_20tr:year_last=2000
co2tseries.cmip6_20tr:year_align=1850
co2tseries.cmip6_20tr:datavars=co2 Sa_co2diag
which fails after 6 months as expected.

Now the question how to fix it...

[slevis-lmwg]

Starting with the cmip6 file (that works), I will extend time from 265 to 273 elements, i.e. from ending in 2014 to ending in 2022, using matlab.

UPDATE: Success, new cmip7 co2 file allows me to run a full year.

• Now I will use

import xarray as xr

# 1. Open the source file (to get attributes) and target file
ds_source = xr.open_dataset('co2_input4MIPs_GHGConcentrations_CMIP_CR-CMIP-1-0-0_gm_1750-2022_CMIP7_c20250228.nc')
ds_target = xr.open_dataset('co2_input4MIPs_GHGConcentrations_CMIP_CR-CMIP-1-0-0_gm_1750-2022_CMIP7_c20260601.nc', decode_times=False)

# 2. Append/update the target dataset's global attributes
ds_target.attrs.update(ds_source.attrs)

# 3. Save the updated target file (overwrites the original)
ds_target.to_netcdf('co2_input4MIPs_GHGConcentrations_CMIP_CR-CMIP-1-0-0_gm_1750-2022_CMIP7_c20260601.nc')
ds_source.close()
ds_target.close()

to copy and append the global attributes from the old failing cmip7 co2 file to the new cmip7 co2 file.

• Trying the simulation again to make sure nothing broke.
• ./rimport the final file, but also the one before I appended the global attributes.
• Then I will push the new script to the cmip7 repo.