CPL: Add field exclusions list for nonlinear maps.

[ambrad]

Add the user_nl_cpl namelist option nlmaps_exclude_fields. This is a list of fields that are to be excluded from the set that are nonlinearly mapped. For these excluded fields, the low-order linear map is used instead. The default is

nlmaps_exclude_fields = 'Faxa_rainc', 'Faxa_rainl', 'Faxa_snowc', 'Faxa_snowl'

[non-BFB] for cases using nlmaps
[NML] for all tests due to extra variable nlmaps_exclude_fields

[ambrad]

e3sm_integration on Chrysalis matches the dashboard except (1) that every test has a NML diff:

2023-12-01 17:43:37: NLCOMP
Comparison failed ... drv_in
  found extra variable: 'nlmaps_exclude_fields'

and (2) ERS_Ld3.ne30pg2_r05_EC30to60E2r2.WCYCL1850.chrysalis_intel.allactive-nlmaps, the stealth-feature test, DIFFs. Both of these are expected.

[ambrad]

@golaz Rob and I agreed that defaulting the namelist variable to include the four preciptiation fluxes is best. Is that OK with you? This has two implications:

1. Run scripts don't have to have any modifications to get the better wrain budget value.
2. A simulation won't be BFB with previous runs unless the namelist variable is set to the list of empty strings, like this:

cat << EOF >> user_nl_cpl
    nlmaps_exclude_fields = '','','','' 
EOF

[golaz]

@ambrad : thank you for working on this quickly. We discussed it during the meeting this morning. We should go ahead with this option for v3.

[rljacob]

I'm going to call this "non-BFB" all the v3 cases are using nlmaps now. Our testing is lagging in bringing in those cases.

[ambrad]

all the v3 cases are using nlmaps now. Our testing is lagging in bringing in those cases.

I see. Do I interpret this correctly as follows?

• e3sm_integration does not yet have this option on, which is why I see only one DIFF in my testing.
• But from the v3 campaign perspective, it's not BFB?

[rljacob]

right. Also when a PR makes any test non-BFB, I prefer to call it "non-BFB except for" or "non-BFB only in" instead of "BFB except for". Makes it easier to find the possibly non-BFB PRs in the git log.

[ambrad]

Re: testing, I checked that the following code modification makes the nlmaps test BFB with the current baseline, as expected and desired:

--- a/cime_config/testmods_dirs/allactive/nlmaps/user_nl_cpl
+++ b/cime_config/testmods_dirs/allactive/nlmaps/user_nl_cpl
@@ -1 +1,2 @@
 nlmaps_verbosity = 1
+nlmaps_exclude_fields = '','','',''

[rljacob]

@ambrad did you want to handle the integration?

[ambrad]

@rljacob Yes, I will do it unless you prefer someone else does.

Edit: I see I forgot to assign myself, prompting Rob's question. Fixed.