Help running recent CAM with RCP 4.5?

[jrvb]

Greetings! I am trying to set up a fully coupled simulation that follows one of the RCPs, but I am struggling to find a configuration that works. I am hoping someone here may be able to offer some pointers! :)

Specifically, I'm looking for the following:

• A version of CAM that is recent enough to include the contrail module src/physics/cam/ssatcontrail.F90
• A version of cime_config/config_compsets.xml (and related files) which includes the right defs for at least one of the RCPs (ideally RCP 4.5, but anything that works would be a useful start.)

Some things I've noticed:

• If I check out the CESM source starting with the CESM github repo (ie, following instructions here) the CAM source never includes ssatcontrail.F90.
• The most recent CESM version which includes RCP compsets is 2.1.0, which only has BRCP85L45BGCR and BRCP85C5L45BGC
• If I check out the CESM source starting with the CAM github repo (instructions here) the more recent versions include ssatcontrail.F90, but none of the config_compsets.xml seem to include RCP compset

With the CAM checkout, I've tried just using the full compset description RCP8_CAM60_CLM45%BGC_CICE_POP2_RTM_SGLC_SWAV instead of the alias BRCP85L45BGCR, but that fails due to missing POP-related files config files. I'm reluctant to just keep hacking bits from one github checkout into the other, since even if I get something that builds and runs, I'm concerned that it won't behave as expected. (And worst of all - I might not be able to tell easily.... :)

Does anyone have advice on putting together a configuration that will work for both an RCP compset and a more recent version of CAM? Are there safe ways to check things out to pull components with different git tags than what the default ./manage_externals/checkout_externals behavior does?

On a more general note, I was surprised that the CESM and CAM paths wind up checking out similar-but-incompatible sets of source code. I would love to understand more about which choices are made along each path and why. Does anyone know the history there?

Thanks!

-Rob

 

[eaton]

When you check out CAM from the ESCOMP/CAM github repo, you get a version which is used for development work and for research with configurations that don't include active ocean or sea ice models. The only way to run "fully coupled" simulations is to check out CESM from the ESCOMP/CESM github repo.

The compsets that supported the RCP scenarios were in CESM1 and were used to produce data for CMIP5. So I believe you'll need to work with CESM1.1.0 (which is downloaded from an svn server) to find an RCP4.5 configuration that works out of the box.

CESM2 was used for the CMIP6 experiments and the compsets for future scenarios used designators like SSP245 which combines the SSP2 and RCP4.5 scenarios. So I think the first decision is what kind of scenario you want to run, and whether CESM1 or CESM2 is the best starting point.

The decision not to hack pieces from different release versions together is a wise one. This is an extremely complex system and it's not plug and play software. The way CESM manages this complexity is through the compset mechanism which allows defining particular combinations of components with particular sets of build and runtime configurations. That said, the best way to run experimental configurations that have not already been set up is to first identify the compset that's closest to what you want to do, and then make necessary modifications.

The contrail parameterization in ssatcontrail.F90 was added to cam6_3_014 (you can figure this out by looking at the CAM ChangeLog file in cam/doc/ChangeLog). Note that this tag is only in the ESCOMP/CAM repo. The latest CESM production release is CESM2.1.3 which includes a version of CAM derived from cam6_0_002. I don't know whether or not it's possible (or realistic) to use the contrail parameterization in cam6_0_002, but that would be the path I'd recommend pursuing. The changes needed to enable the contrail code will be fould by looking at the differences between cam6_3_013 and cam6_3_014.

Also note that development versions of the cesm since cesm2_3_beta05 will include a version of cam that has the contrails. But these unreleased development versions have not been validated and can't be used to run emissions scenarios.

 

[jrvb]

Hi @eaton -

I just noticed that I missed your message when I was away on vacation --- thanks so much for the detailed reply! It sounds like running the contrail simulations along with an emissions pathway is likely to introduce a lot of unknowns, so I'll probably hold off on that.

Stepping back a bit, I am interested in looking at variability of near-term (2030-2040) climate behavior under various scenarios. As a baseline, I'm looking for something like this:

a. An ensemble setup that captures reasonable levels of climate variability in the 2030-2040 range
b. Something that allows following defined emissions pathways so I can account for a few possible human behaviors
c. Output that's fine-grained enough to allow some investigation of surface level extreme weather events (unusual heat waves, precipitation levels, etc.)

I would love some advice on how to best set up something like this. So far, I've been imagining this might work something like this:

* Start with ~modern restart files from CMIP6, LENS or similar so I can achieve (a) without re-doing a bunch of computations from 1850. Something that starts close to the present day seems ideal from a computation standpoint, but I'm open to suggestions!

* Use fully coupled runs (or something close?) to run forward to eg 2030. I would like to make sure that variation from things like different ENSO patterns are captured, so the prescribed SST of things like F2000climo don't seem like they would work?

* I would probably run these simulations at nominal 1-degree resolution, so the compute burden isn't too high.

* From whatever starting points I can find, I would probably want to add some variation (eg, by tweaking the initial ATM temperature field? Or maybe adding variation to the ocean state as well?) to broaden the ensembles and sample more of the overall variation by the target date.

* Assuming the above would yield reasonable variations in the large-scale climate behaviors, I was thinking of taking restart files from these ensembles and running forward for a shorter time with a finer simulation (maybe 1/4 degree for 2-3 months?) to get more realistic surface-level behaviors to analyze.

I would love to hear from folks with more expertise on how to set this up --- or if I'm thinking about this all wrong! :)

Best,

-Rob

 

[islas]

Hi Rob,

In case it's useful, I just want to make you aware that we do have ensembles under different scenarios that are available. Climate Variability & Change Working Group (CVCWG) | Community Earth System Model - here, under "Our Simulations", you can see the CESM2 large ensemble which uses SSP3-7.0 and has 100 members, and the CESM2 SSP2-4.5 ensemble which has 16 members. There will be a 15 member SSP5-8.5 ensemble becoming available any day now as well.

Isla

 

[jrvb]

Hi Isla -

This is very helpful - thanks! I see that there are quite a few output variables available via the web portal. I don't see any way to access restart files that could be used for continuing previous runs or that were used as initial conditions for the ensemble members described on these pages. Are those restart files available somewhere?

Also, are there details on the CESM configurations used for these? I would love to start with a known good configuration to reduce my chances of missing some critical settings and winding up with garbage output from my sims. :)

Best,

-Rob

 

[jrvb]

One other question --- Under the "Our Simulations" section I see the 16-member CESM2 SSP2-4.5 ensemble you mentioned, but I don't see the SSP3-7.0 simulations. Are those listed elsewhere perhaps?

Thanks!

-Rob

 

[islas]

Hi Rob,

The details of the SSP2-4.5 ensemble are here: CESM2 SSP2-4.5 Ensemble | Community Earth System Model
The details of the SSP3-7.0 ensemble are here: CESM2 Large Ensemble Community Project (LENS2) | Community Earth System Model

For the restart files for the SSP2-4.5 ensemble, if you have access to the NCAR supercomputer, you can find them in a sub-directory "rest" at the same location where you access the data e.g.,

/glade/campaign/cesm/development/cvcwg/cvwg/b.e21.BSSP245smbb.f09_g17/b.e21.BSSP245smbb.f09_g17.001/rest

If you don't have access, I don't think that we have the restarts available on the Climate Data Gateway but you may be able to get them by contacting the CVCWG liasons (Adam Phillips or Gary Strand).

I believe these were run with cesm2.1.4-rc.07 and you should be able to use the BSSP245smbb compset.

Isla

 

[jrvb]

Very helpful - thanks Isla!

-Rob