Feature/gaas fix for amip emissions

[vbuchard]

When the model runs in replay with GAAS turned on and AMIP emissions chosen, the model crashes after several days with a segmentation fault. AOD becoming completely off.
This PR updates the GAAS_GridComp_Extdata.yaml in AMIP/ similar to the one in OPS.

Fixes #281

[github-actions]

Label error. Requires at least 1 of: 0 diff, 0 diff trivial, Non 0-diff, 0 diff structural, 0-diff trivial, Not 0-diff, 0-diff, automatic, 0-diff uncoupled. Found: bugfix

[github-actions]

Label error. Requires at least 1 of: 0 diff, 0 diff trivial, Non 0-diff, 0 diff structural, 0-diff trivial, Not 0-diff, 0-diff, automatic, 0-diff uncoupled. Found: bugfix

[github-actions]

Label error. Requires at least 1 of: 0 diff, 0 diff trivial, Non 0-diff, 0 diff structural, 0-diff trivial, Not 0-diff, 0-diff, automatic, 0-diff uncoupled. Found: bugfix

[vbuchard]

Just a note: this file will need to be updated if the offset is no longer 450s (a match of the heartbeat_dt in cap.rc). @bena-nasa has open a ticket to have a more flexible solution in ExtData.

[mathomp4]

Just a note: this file will need to be updated if the offset is no longer 450s (a match of the heartbeat_dt in cap.rc). @bena-nasa has open a ticket to have a more flexible solution in ExtData.

Ooh. Yeah. In GEOSgcm v12 this can really not be true. The heartbeat is resolution dependent from like 1800 s to 225 or so.

[mmanyin]

Just a note: this file will need to be updated if the offset is no longer 450s (a match of the heartbeat_dt in cap.rc). @bena-nasa has open a ticket to have a more flexible solution in ExtData.

Ooh. Yeah. In GEOSgcm v12 this can really not be true. The heartbeat is resolution dependent from like 1800 s to 225 or so.

When we get to that point, I would recommend we use a token like HEARTBEAT in the yaml file that gets replaced by the run script.

[mmanyin]

Do we really want fail_on_missing_file to be FALSE by default? Won't this lead to confusion if the file path is wrong, but the user assumes that all is working properly? What would be the down-side of fail_on_missing_file = TRUE?

[mathomp4]

Do we really want fail_on_missing_file to be FALSE by default? Won't this lead to confusion if the file path is wrong, but the user assumes that all is working properly? What would be the down-side of fail_on_missing_file = TRUE?

I think @bena-nasa knows why that is there...and I think it needs to be?

[bena-nasa]

@mmanyin, @vbuchard we probably don't want fail_on_missing_file in AMIP mode is my best guess. However, that is 100% necessary when GAAS is run in OPS since they run the forecast as a part of the same run where the assimilation happens.

So we might want to eliminate that in the AMIP mode

[bena-nasa]

@mmanyin with regards to the HEARTBEAT issue. That's also on the to do list for ExtData

[mmanyin]

@bena-nasa Thanks for the clarifications. I will go ahead and approve the PR, with the big caveat to set fail_on_missing_file to TRUE if the model is running in non-forecast mode.

[mmanyin]

This PR is a step in the right direction. It depends on the proper assignment of REPLAY_ANA_EXPID and REPLAY_ANA_LOCATION in AGCM.rc, and some clever code in the run script, but it works. Two caveats, as mentioned in the discussion: PT450S will need to be changed if the timestep changes. And the 'fail_on_missing_file' flag for each export should probably be set to TRUE for hindcast simulations, to help guard against inadvertent mistakes.