Closed mnlevy1981 closed 3 years ago
6d72b75 allows us to select a different advection scheme for ecosystem tracers with virtual fluxes. Running with
tadvect_ctype = 'centered'
ecosys_tadvect_ctype = 'upwind3'
ecosys_vflux_tadvect_ctype = 'base_model'
Results in
------------------------------------------------------------------------
Advection options
------------------------------------------------------------------------
Tracer Name Advection Option
TEMP centered
SALT centered
PO4 upwind3
NO3 upwind3
SiO3 upwind3
NH4 upwind3
Fe upwind3
Lig upwind3
O2 upwind3
DIC centered
DIC_ALT_CO2 centered
ALK centered
ALK_ALT_CO2 centered
DOC upwind3
DON upwind3
DOP upwind3
DOPr upwind3
DONr upwind3
DOCr upwind3
zooC upwind3
spChl upwind3
spC upwind3
spP upwind3
spFe upwind3
spCaCO3 upwind3
diatChl upwind3
diatC upwind3
diatP upwind3
diatFe upwind3
diatSi upwind3
diazChl upwind3
diazC upwind3
diazP upwind3
diazFe upwind3
IAGE centered
The default for ecosys_tadvect_ctype
is upwind3
for the tx0.1v3
grid but base_model
the rest of the time; default for ecosys_vflux_tadvect_ctype
is always base_model
I've finished testing -- aux_pop
on cheyenne (intel only), and aux_pop_MARBL
(cheyenne for intel and gnu, hobart for nag and pgi) performed as expected: NLCOMP
failures for BGC compsets, but all tests ran successfully and were bit-for-bit with baselines. Here's what'll be in the change log:
Tag Creator: mlevy
Developers: mlevy, klindsay
Tag Date: 30 July 2020
Tag Name: cesm_pop_2_1_20200730
Tag Summary: Add new compset for high-res G case that is forced by JRA and
includes BGC (G1850ECOIAF_JRA_HR)
1. Newer version of MARBL (just a bugfix)
2. New feature: set tavg_contents_override_file in user_nl_pop
to replace fully resolved tavg_contents file with one created
by hand
3. Add a few more valid spacecurve pe layouts for high-res
4. Different default PE layout for new compset than high-res without BGC
5. Add BGC initial condition and forcing file namelist defaults for tx0.1v3
6. Clean up ecosys namelists (don't include variables that aren't used)
7. New testmoddir that lets us test new compset with CICE namelist
changes that appear in production runs
8. New ecosys_vflux_tadvect_ctype namelist var sets tadvect_ctype
for tracers with virtual fluxes; now existing ecosys_tadvect_ctype
only applies to those without
9. Additional tavg variables to allow some variables to appear in
multiple tavg streams and / or be truncated to top 150m
* TEMP_2, TEMP_z_t_150m
* SALT_2, SALT_z_t_150m
* UVEL_2, UVEL_z_t_150m
* VVEL_2, VVEL_z_t_150m
* SHF_QSW_2
* All passive tracers have _2 and _z_t_150m variants as well
Testing: aux_pop (cheyenne_intel) and aux_pop_MARBL (cheyenne intel & gnu plus
hobart nag and pgi); non-BGC compsets are bit-for-bit with no namelist
changes. BGC compsets fail NLCOMP (expected; see 6 above) but are still
bit-for-bit.
Description of changes:
Create a new compset set up to run a
G
compset withJRA
forcing and the MARBL tracers enabled on thetx0.1v3
grid. Will also automatically use CICE4 for existing_HR
compsets.Testing:
Test case/suite: I'll run
aux_pop
andaux_pop_MARBL
when this is ready to be merged and update the results in a later comment Test status: I expect this to be bit-for-bit, since the high-res compsets are not tested.Fixes #30 Fixes #29
User interface (namelist or namelist defaults) changes? None