Issue with atmospheric CO2 concentration received by ocean

[tjiputra]

@mvertens , @DirkOlivie , @JorgSchwinger , @adagj , @kjetilaas

In all concentration-driven NorESM3 simulations (n1850, nhist, n1pctCO2, nCO2x4, etc., the atmospheric CO2 concentration is read from respective flbc files, which contain latitudinally-varying CO2 concentrations. The ocean, however, receives globally uniform values. The temporal values look correct, i.e., seasonally- and time-varying. So there may be some issues with the coupler when sending the values from atm to ocn.

The land component seems to receive spatially varying values, but good to check as well.

Here's the globally averaged values from ocean and land from recent nhist run.

[adagj]

Thank you @tjiputra
@kjetilaas Can you check land?
@mvertens How is this done in the coupler for the ocean?

[mvertens]

@tjiputra - I was on vacation and just got back. Can you please point me to the case and run directory that you used to get your global averaged plots so that I can try to replicate this behavior. Also - were you using blom history output to determine that the co2 was not global - or mediator output? If it was blom output - if there is a problem in the blom cap? At any rate - I would like to look at mediator output that shows exactly was is being sent to the ocean and land as the first step - and for that I'd like a case to start from.

[tjiputra]

@mvertens , here's the case:
/cluster/projects/nn9560k/adagj/NorESM/cases-noresm3dev/nhist.ne16pg3_tn14.pr-noresm3_0_beta19.Base_i446F.Olivia.2026-06-07/

and run directory:
/cluster/work/projects/nn9560k/adagj/noresm/nhist.ne16pg3_tn14.pr-noresm3_0_beta19.Base_i446F.Olivia.2026-06-07/run/

The atmosphere should be reading this latitudinally varying CO2 file:
flbc_file = '/cluster/work/projects/nn9560k/inputdata/atm/waccm/lb/cmip7_ghg_version20260607/ghg_cmip7_CMIP_CR-CMIP-1-0-0_1750-2023_monthly_gnz_0p5degLat_version20260207.nc'

I have used the blom history output. I checked the land history outputs and there, the CO2 values are also not latitudinally varying.
Thank you!

[mvertens]

@tjiputra @mvdebolskiy @DirkOlivie @gold2718 - I have replicated the case that @tjiputra has pointed me to. I have run for 5 days and have asked for mediator history files to be written daily. What I am finding is that cam is sending a constant value for CO2_diag to the mediator and that this is then sent to clm and blom. @DirkOlivie @gold2718 - it's surprising to me that there is no latitude dependence in the CO2 that cam is sending to the mediator given that as @tjiputra points out flbc_file has a latitude dependence.
My case is here:
/cluster/work/projects/nn9560k/mvertens/noresm/nhist.ne16pg3_tn14.pr-noresm3_0_beta19.Base_i446F.Olivia.2026-06-07
My run dir is here:
/cluster/work/projects/nn9560k/mvertens/noresm/nhist.ne16pg3_tn14.pr-noresm3_0_beta19.Base_i446F.Olivia.2026-06-07/run
The cpl data is here:
/cluster/work/projects/nn9560k/mvertens/archive/nhist.ne16pg3_tn14.pr-noresm3_0_beta19.Base_i446F.Olivia.2026-06-07/cpl/hist

[JorgSchwinger]

@mvertens

I can confirm that this was also the behavior in NorESM2. I tried once to read in a modified LBC-file with a latitudinal dependence, but HAMOCC only received a constant value. The reason for this in NorESM2 was that CAM would always do a mean value (in mo_flbc).

[gold2718]

@JorgSchwinger, CAM only reads a global mean if that field (CO2_LBC_MEAN) is on the file. Otherwise, it reads and uses CO2_LBC which should have latitude dependence (the current file used for CMIP7 does).

[oyvindseland]

I have not followed the full structure of co2 so I do not know what is used for the radiation but even when the model reads values for latitude bands it still calculates an average.

if (flbcs%has_mean) then
   global_mean_vmr = flbcs%vmr_mean(last) &
        + dels * (flbcs%vmr_mean(next) - flbcs%vmr_mean(last))
else
   do lchnk = begchunk, endchunk
      ncol = get_ncols_p(lchnk)
      vmr_arr(:ncol,lchnk) = flbcs%vmr(:ncol,lchnk,last) &
           + dels * (flbcs%vmr(:ncol,lchnk,next) - flbcs%vmr(:ncol,lchnk,last))
   enddo
   call gmean (vmr_arr, global_mean_vmr)
endif

endfunction global_mean_vmr
end module mo_flbc

[JorgSchwinger]

Yes, I agree, I have also checked this, (what I think is) the relevant part of the code has not changed since NorESM2. It calculates a global mean even if a latitudinal dependent vmr is read in as shown above.

[mvertens]

Don't we want to bypass using the global mean in terms of what we send to the land and ocean?

[gold2718]

Sorry, I did not correctly address what happens in values sent to the surface.

It's not the calculation that is the issue but in the CAM cap, it is using the globalized value for the Sa_co2diag field sent to the mediator. Should we turn this off (i.e., use the value read from the file)?