Have compsets determine CCSM_BGC setting for CO2 in ending _BGC% setting

[ekluzek]

Right now there are several xml variables that have to be coordinated across CESM in different components.

Things that have to coordinate:

• CCSM_BGC – here in CMEPS
• DATM_CO2_TSERIES – in DATM
• CLM_CO2_TYPE – in CLM
• OCN_CO2_TYPE – in MOM
• CO2 settings: CAM_BLDNML_OPTS (-co2_cycle) and co2_cycle_rad_passive namelist -- in CAM

Right now they are treated independently and defaults are set by patterns from the compsets. But, the patterns watch different things, and are mostly tied to the compset period. The patterns checked for are also inconsistent, since they can evolve independently. But, this leads to problems in getting it setup correctly.

See for example this case in CDEPS:

ESCOMP/CDEPS#225

And I'm also realizing it's a problem for CLM as well, and CLM, DATM, and CMEPS currently aren't set consistently.

But, I think the problem goes away if we rearrange how these are set to do the following principles:

• Have CMEPS set the CCSM_BGC setting based on the components involved and the _BGC% setting
• Have components set their defaults based on what CCSM_BGC is set to (and the compset when necessary)
• Have components die if their CO2 setting is inconsistent with CCSM_BGC in CMEPS
• Have the CCSM_BGC setting, set by the _BGC% setting in the compset

Some potential steps to get there

• Update CTSM defaults to use CCSM_BGC and _BGC%B in the compset
• Update CTSM buildnml to check for consistency with CCSM_BGC and DATM_CO2_TSERIES
• Do similar steps for MOM
• Add _BGC%CONST type setting to CMEPS
• Update CTSM compsets to require _BGC% in them (and have a checker for this)
• Update MOM compsets
• Update CAM defaults and CAM compsets
• Could potentially remove some of the extraneous logic for setting CCSM_BGC in CMEPS

[billsacks]

Thanks a lot for thinking about this, @ekluzek ! I don't have my head around all of the relevant pieces, but I'd like to talk with you to better understand your suggestion.

[ekluzek]

Here's what this would look like for CLM:

CLM:

ESCOMP/CTSM#4184 (comment)

MOM:

would look almost identical to CLM above

DATM:

DATM would be a little different, because it also has to check the period. But, would be something like:

    <values match="last">
      <value compset="^SSP119_" ccsm_bgc="CO2A|CO2B|CO2C">SSP1-1.9</value>
      <value compset="^SSP126_" ccsm_bgc="CO2A|CO2B|CO2C">SSP1-2.6</value>
      <value compset="^SSP245_" ccsm_bgc="CO2A|CO2B|CO2C">SSP2-4.5</value>
      <value compset="^SSP370_" ccsm_bgc="CO2A|CO2B|CO2C">SSP3-7.0</value>
      <value compset="^SSP434_" ccsm_bgc="CO2A|CO2B|CO2C">SSP4-3.4</value>
      <value compset="^SSP460_" ccsm_bgc="CO2A|CO2B|CO2C">SSP4-6.0</value>
      <value compset="^SSP534_" ccsm_bgc="CO2A|CO2B|CO2C">SSP5-3.4</value>
      <value compset="^SSP585_" ccsm_bgc="CO2A|CO2B|CO2C">SSP5-8.5</value>
      <value compset="^HIST_"   ccsm_bgc="CO2A|CO2B|CO2C">cmip6_20tr</value>
      <value compset="^20TR"    ccsm_bgc="CO2A|CO2B|CO2C">cmip6_20tr</value>
      <value compset="^OMIP_DATM%IAF.*_MOM.%[^_]*ECO" ccsm_bgc="CO2A|CO2B|CO2C">omip.iaf</value>
      <value compset="^OMIP_DATM%JRA.*_MOM.%[^_]*ECO" ccsm_bgc="CO2A|CO2B|CO2C">omip.jra</value>
      <value compset="_DATM%CPLHIST" ccsm_bgc="CO2A|CO2B|CO2C">cplhist</value>
    </values>

Note

Note, this is a usability improvement, because now the settings change if the user happens to change CCSM_BGC. And it allows the user to run easily cases with everything but CO2 transient, because they just set CCSM_BGC to none.

[ekluzek]

For CMEPS for CCSM_BGC it would look something like:

    <values match="last">
      <!-- Always use CO2 from CAM when coupled to CAM -->
      <!-- Technically these are NOT needed if all compsets have _BGC% -->
      <value compset="_CAM">CO2A</value>   
      <value compset="_DATM">none</value>
      <!-- Set to CO2A when used with DATM for either CLM or OCN model with MARBL -->
      <!-- Technically these are NOT needed if all compsets have _BGC% -->
      <value compset="DATM%CPLHIST.*CLM">CO2A</value>
      <value compset="DATM%CPLHIST.*BLOM%ECO">CO2A</value>
      <!-- Set it regardless of the version of MOM used (The . any character after MOM)-->
      <value compset="DATM%CPLHIST.*MOM.[^_]*MARBL">CO2A</value>
      <!-- These two aren't needed anymore with above
      <value compset="^OMIP_DATM%IAF.*_MOM.%[^_]*MARBL">CO2A</value>
      <value compset="^OMIP_DATM%JRA.*_MOM.%[^_]*MARBL">CO2A</value>
      -->
      <!-- Rather than explicitly setting transient compsets to CO2A, rely on compsets having _BGC% to define behavior -->
      <!-- Use the highest level based on the compsets involved -->

      <!-- CO2C for fully coupled (B compsets) -->
      <value compset="CAM.*CLM.*MOM.*_BGC%BP">CO2C</value>
      <!-- CO2B when CAM coupled to CLM (F compsets) -->
      <value compset="CAM.*CLM.*(DOCN|SOCN|XOCN).*_BGC%B">CO2B</value>
      <!-- CO2A for CLM and MOM standalone (I, G or C compsets)-->
      <value compset="DATM.*CLM.*(DOCN|SOCN|XOCN).*_BGC%B">CO2A</value>
      <value compset="DATM.*(DLND|SLND|XLND).*MOM.*_BGC%B">CO2A</value>
    </values>

As stated above provided all compsets have been updated, the only part needed are the last 4 settings dependent on _BGC% being in the compset. So just these:

      <value compset="CAM.*CLM.*MOM.*_BGC%BP">CO2C</value>
      <value compset="CAM.*CLM.*(DOCN|SOCN|XOCN).*_BGC%B">CO2B</value>
      <value compset="DATM.*CLM.*(DOCN|SOCN|XOCN).*_BGC%B">CO2A</value>
      <value compset="DATM.*(DLND|SLND|XLND).*MOM.*_BGC%B">CO2A</value>

In order to transition to that, we'd want to include the top portion until all compsets are converted and then they could be removed.

[ekluzek]

For robustness it would also be best if _BGC% was a required part of the compset, and that there was the addition of an option to specify constant values. So something like

_BGC%CONSTANT

Then it can die with an error if a compset without _BGC is provided, which will ensure compsets are correct moving forward.

[ekluzek]

Some notes on another take on this is here: https://github.com/ESCOMP/CDEPS/pull/403/files#r3918465969