Closed eclare108213 closed 1 year ago
Ran the QC, it fails. The plots suggest the ice is a bit thinner in the new simulation. Images below are baseline, new modifications, difference. I tested with CICE cb58257857d429c4c2d and Icepack main plus changes up to commit #236023f above.
Thank you, @apcraig. I would expect the snow grains to be larger and the resulting ice thinner with the parameter change. The magnitude of the difference (generally less than 0.5 m) does not concern me, especially considering how thick the ice is in the southern hemisphere. That's a bit surprising. Out of curiosity, do you have a QC run with just the default parameters, i.e. with snwgrain and snwitdrdg not turned on?
The ice thickness of the baseline run (snw options off) is much thinner than either of the runs with snw options on. The two snw runs are closer to each other than to the baseline. I attach figures for the baseline (no snw options), baseline snw configuration, and new snw configuration (i.e. including mods in this PR) as well as one diff, the baseline - new snw configuration.
Thank you. Very interesting, and these plots are generating lots more questions. I will move this conversation into an issue. In the meantime, I think we should merge this PR to get the bug fixes into main, and then migrate them into the E3SM fork of Icepack. Anyone opposed?
One important comment I might add. The snw cases are starting with the "icdefault" option because restart files are not available. The standard case starts with a proper initial condition. Also, I'm not sure whether these plots are means or something else, maybe someone else knows.
The magnitudes of the changes look reasonable to me.
One more data point, ran a QC for the standard run + icdefault. This is with snw options off. 3 images follow, standard run + icdefault, current snw options run + icdefault (required), difference. The runs still fail QC but the differences in the thickness are reduced relative to comparisons with the standard run from a restart file.
Will merge this now and create a PR to migrate this to the E3SM-Project branch.
160 measured results of 160 total results 156 of 160 tests PASSED 0 of 160 tests PENDING 2 of 160 tests MISSING data 2 of 160 tests FAILED
Both runs with -s snwgrain have different answers, as expected: FAIL conda_macos_smoke_col_1x1_debug_run1year_snw30percent_snwgrain compare ibased33 different-data FAIL conda_macos_restart_col_1x1_snwgrain_snwitdrdg compare ibased33 different-data
This was a change in a recent PR, after the baseline was run - not a problem here: MISS conda_macos_smoke_col_1x1_debug_fsd12_run1year_short compare ibased33 missing-data MISS conda_macos_restart_col_1x1_fsd12_short compare ibased33 missing-data
Snow parameter: Oleson et al eq 3.68:
This PR removes a factor of 100 from f_liq in the current code and changes the constant to S_wet = 10^18 * C1 = 4.22e5.
Snow drainage: this change allows wet metamorphism to function when liquid water associated with the snow scheme is not used directly in melt ponds, i.e.
use_smliq_pnd = .false.
It also fixes a bug in the computation of drained water mass.Brine volume: this change is only to the before-and-after calculations used to check brine volume conservation. There is an extra factor of 10 which ought to be removed, if possible, but I'm leaving it in for now.
All bugs were found during the E3SM/Icepack merge process.