dust_emis_fact needs separate values for F and B cases

[dmleung]

Issue Type

Other (please describe below)

Issue Description

I am documenting our discussion in PR #1509 here as an issue. #1509 supports @cecilehannay 's spinup run in issue #1507 with multiple namelist parameter changes in the namelist_defaults_cam.xml. In PR #1509, @cacraigucar requested to change the dust tuning factor dust_emis_fact in the 1° or ne30 horizontal resolution from 4.0 to 3.2. I mentioned that 3.2 works better for B cases in 1°, while 4.0 works better for F cases (freeruns) in 1°. But 3.2 was not officially put inside the namelist defaults, and 4.0 has been in the namelist defaults since I did my latest dust tuning in PRs #1423 and #1426 a few months ago. @adamrher and I agreed to use a smaller dust_emis_fact = 3.2 to allow larger dust emission fluxes for Cecile's B cases.
The aerosol and chemistry communities (and ACOM) mainly work with aerosols in F cases, so I generally recommend maintaining 4.0 as the dust tuning factor (but I know there are people running B cases too). I told @cacraigucar that, if possible, it would be great to have dust_emis_fact values separately stored for F cases and B cases in the namelist defaults, but I am not sure if this is technically possible. I am not sure how to settle this, so I create this issue to see if @cacraigucar or anyone else has any proper idea.

The following is a few instances of what was merged in #1509. They were originally 4.0D0.

<dust_emis_fact                            phys="cam7"                  >3.20D0</dust_emis_fact>
<dust_emis_fact dyn="mpas"                 phys="cam7"                    >3.20D0</dust_emis_fact>
<dust_emis_fact dyn="se"                   phys="cam7"                    >3.20D0</dust_emis_fact>

@PeterHjortLauritzen @tilmes

P.S. Dust emission in the B case is lower mainly because CLM%BGC in B cases has generally higher LAI than that in CLM%SP in the F cases, thereby reducing global dust emission by 20–30 %. The land people like Will Wieder and Dave Lawrence are aware of this.

Will this change answers?

Yes

Will you be implementing this yourself?

Yes, but I will need some help

[adamrher]

@dmleung making the dust tuning dependent on the compset would introduce add'l complexity to our namelist logic that we would prefer to avoid. Would you be OK if we introduced FHISTC_LTso and FHISTC_MTso compsets that have BGC? For example, there is FHIST_BGC compset for cam6, so we could name it something like FHISTC_LTso_BGC.

[dmleung]

Thanks @adamrher for thinking about this!
@PeterHjortLauritzen and I had a conversation yesterday, let me summarize it here and try to make clear where the issue lies in to my understanding:
I. What Peter is currently trying to suggest is to minimize the number of scientifically supported compsets and simplify the namelist_defaults as much as possible. I think this is reasonable.
II. Meanwhile, Peter also says CESM should have the right tunings out-of-the-box for all the scientifically supported compsets.
III. The technical difficulty suggested by Adam is that the current namelist logic does not support the differentiation of tuning factors dependent on the F-case vs. B-case, or CLM%SP vs. CLM%BGC. This fits my understanding too.
IV. However, there are certain scientifically supported F compsets, like FHISTC_LTso and FHISTC_MTt4s, that are using CLM%SP.
V. This means that dust tuning will be off for these scientifically supported F cases.

So, the logically plausible solutions to me are for example:

1. CESM does not scientifically support compsets that use CLM%SP anymore and instead supports compsets that use CLM%BGC, such as FHISTC_LTso_BGC and FHISTC_MTt4s_BGC.
2. We bear with the CLM%BGC tunings for the scientifically supported CLM%SP compsets, and state in the CAM7 user guide the correct dust_emis_fact and other tuning factors for certain scientifically supported compsets that use CLM%SP. This might be suboptimal.

I hope the above lines make sense. I myself am open with having FHISTC_LTso_BGC and FHISTC_MTt4s_BGC as the scientifically supported F compsets, and I don't mind using CLM%BGC mode. I just hope that when the CESM community uses the scientifically supported F compsets, they have good tuning factors out-of-the-box.

Simone @tilmes says she wants to argue for
3. We spend some effort on making the namelist_defaults capable of differentiating between CLM%SP and CLM%BGC
when thinking from CMIP's perspective. She told me she will make a comment here so I will let her do so.

[tilmes]

Thanks, Danny. I think, we need to understand what the scientifically supported compsets are used for:

• B-comsets use BGC by default; land and ocean include biogeochemistry that is interactive. This is important for a fully coupled system.
• F-compsets have different purposes. There are some RFMIP and AerChemMIP and maybe other CMIP experiments that require a fixed ocean (F-case) for some of their experiments, but the same forcings as the B-case, so I assume that includes BGC. Since the B-case doesn't tune dust, we should also not do this for these F-cases with BGC.
• There are other F-compsets that the chemistry community especially uses, which prefer using CLM%SP, because of an LAI, etc., that fit observations, therefore various land emissions (biogenic) and dust, are more realistic. Since these compsets are used for present day and for example airquality studies, it makes a lot of sense to have dust tuning applied.

So, it seems like having 2 types of F-cases (compsets) makes sense, one with CLM%BGC, for various CMIP simulations, also for potential F-case simulations in the future, and F-cases with CLM%SP that are important for present-day studies, nudged compsets, and air quality studies.

Based on that, we could possibly have F and B compsets used the same land setting (CLM%BGC) for CMIP purposes, it would a lot sense to have the same namelist settings. And then have also FHISTC_LTso_SP compsets that would have different namelist variables, including dust tuning and potentially other tunings for the present day. This is basically what Adam is suggesting, but have BGC the default for both F and B cases. Thoughts?

[dmleung]

Thanks Simone. See responses below:

Based on that, we could possibly have F and B compsets used the same land setting (CLM%BGC) for CMIP purposes, it would a lot sense to have the same namelist settings.

Agreed. No matter it is F or B case, using CLM%BGC mode will yield smaller dust than CLM%SP mode. Dust in F-BGC and B-BGC cases are rather close to each other, so having pretty much the same dust_emis_fact for F-BGC and B-BGC cases is fine.

And then have also FHISTC_LTso_SP compsets that would have different namelist variables, including dust tuning and potentially other tunings for the present day.

I think this is the difficulty Adam mentioned – we don't have the namelist logic that can assign different tunings for FHISTC_LTso_SP and FHISTC_LTso_BGC compsets. I personally would love to have FHISTC_LTso_SP compset with different namelist variables than the FHISTC_LTso_BGC compset, yes.

This is basically what Adam is suggesting, but have BGC the default for both F and B cases. Thoughts?

Yes, again, I am fine with having CLM%BGC mode as the default scientifically supported compsets for both F and B cases. It really comes down to whether CESM wants to scientifically support any F compsets with CLM%SP.

• There are other F-compsets that the chemistry community especially uses, which prefer using CLM%SP, because of an LAI, etc., that fit observations, therefore various land emissions (biogenic) and dust, are more realistic. Since these compsets are used for present day and for example air quality studies, it makes a lot of sense to have dust tuning applied.

So, it looks like to me that we might eventually settle in either one of these three scenarios:

1. We take away FHISTC_MTt4s_SP and only scientifically support FHISTC_MTt4s_BGC. We ask the chemistry community to start using CLM%BGC mode.
2. CESM scientifically supports both FHISTC_MTt4s_SP and FHISTC_MTt4s_BGC but only includes the CLM%BGC dust tuning in namelist_defaults, so dust in FHISTC_MTt4s_SP is by default biased high. We put the recommended dust_emis_fact = 4.0 for FHISTC_MTt4s_SP in the CAM7 user guide and ask the chemistry community to refer to the user guide.
3. Adding namelist complexity and having tuning factors for the scientifically supported FHISTC_MTt4s_SP and FHISTC_MTt4s_BGC compsets, so the chemistry community can choose to run SP mode with the right tunings out-of-the-box. This is what Adam and Peter currently want to avoid, but it looks like this is what Simone wants.

Since the B-case doesn't tune dust, we should also not do this for these F-cases with BGC.

Not so sure what you mean by this. Currently, the dust_emis_fact = 3.2 (put in by Sheryl in PR #1509) is used for the dust tuning for B cases and for Cecile's spinup run. The B case does tune dust.
3.2 also (roughly) works for FHISTC_MTt4s_BGC even though it is an F case.

• There are other F-compsets that the chemistry community especially uses, which prefer using CLM%SP, because of an LAI, etc., that fit observations, therefore various land emissions (biogenic) and dust, are more realistic. Since these compsets are used for present day and for example airquality studies, it makes a lot of sense to have dust tuning applied.

My (naive) question to Simone and the chemistry community is, does the chemistry community need to run CLM%SP mode? Maybe they can move on to compsets with CLM%BGC mode if CESM doesn't scientifically support F cases with CLM%SP and manage the corresponding tuning factors anymore.

[dmleung]

A similar discussion happened a while ago. Cecile attached a plot showing how SP TLAI and the BGC TLAI are different:
NCAR/cesm_dev#30