-
Notifications
You must be signed in to change notification settings - Fork 25
2022 MARBL Dev team meetings
Michael Levy edited this page Jun 7, 2022
·
14 revisions
- Prioritizing Mike's tasks
- New high-res BGC run (4p2z on frontera)
- machine is down for maintenance today
- frontera is supported machine in CESM 2.3, just need to port that my 2.2 sandbox
- Fred might have some CICE mods to make via SourceMods (I talked to him yesterday)
- Still need to finish backport of MARBL to CESM 1.3 after the high-res run begins
- Original high-res BGC run (4p1z on cheyenne)
- Yassir has 15 million core-hours for us to finish the run
- machine might be too busy to take advantage of economy queue (first job sat for a week, then started when machine issues took down a bunch of jobs... but crashed an hour in because machine issues were taking down a bunch of jobs)
- FEISTY
- walked Colleen, Zephyr, and Kristen through latest python interface (Using
dask-mpiand submitting a script to the queue instead of going through Jupyter Hub) - still to do: add documentation for new interfaces I created, and update docs for interfaces I changed; may still have a few feature requests from the walk-through to add as well.
- walked Colleen, Zephyr, and Kristen through latest python interface (Using
- MARBL in MOM: no progress in a several weeks, but I have a lengthy list of remaining tasks
- Rob L from UNH has been trying to port CESM 2.1.3 to a cluster at UNH and has asked for help
- Other topics? (Diagnostics? Other work in progress?)
- New high-res BGC run (4p2z on frontera)
- Two open issues re: diagnostics that have multiple potential solutions
-
#401:
photoC_NO3_TOT_zint_100mis defined as a diagnostic but never computed- Add code to compute the diagnostic? (it's a trivial addition to
marbl_diagnostics_mod.F90) - Remove from diagnostic list? (it's not a CMIP diagnostic;
photoC_NO3_TOT_zintis and that's computed correctly)
- Add code to compute the diagnostic? (it's a trivial addition to
-
#400: it would be useful for FEISTY if
pocToFloorwas in POP output- Adding
pocToFloortomediumstream by default means every CESM run is producing additional output - Telling Kristen she can modify
ecosys_diagnosticsto add it to a stream means that only runs Kristen does will have the field
- Adding
-
#401:
- FEISTY parallelization
-
xr.map_blocksprovides some speed up, but not very efficient
-
- SEWG and OMWG meetings during the CESM workshop are both happening Wednesday morning
- My plan: attend OMWG for MOM6 and BGC talks, spent rest of time in SEWG meeting
- Bill Sacks asked about an update from all of the different diagnostic projects -- thoughts on what I should include in a bullet point or two for him?
- FEISTY: talking with Deepak about parallelizing via
xr.map_blocks()
- Focusing more on FEISTY at the moment, but task list before merging MARBL driver to
dev/ncaris in a project board
- I don't think we talked about #400 last meeting: updates to MARBL (and POP) to output necessary FEISTY forcing
- Reviving MOM vs POP comparisons (caseroots are subdirectories of
/glade/work/mlevy/codes/CESM/cesm2_3_beta08+mom6_marbl/long_runs)- MOM is using NUOPC cap and calling KPP nonlocal subroutines
- 24 years of POP run from
cesm2_3_beta08with MCOG disabled:g.e23b08.TL319_g17.G1850POPECO_JRA.no_mcog - MOM companion run is still in progress, have ten or eleven years (aiming for twenty):
g.e23b08.TL319_t061.G1850MOMMARBL_JRA.002 - Running MOM in four-year chunks, so first eight years are archived
- I talked with Kristen last week, and that led to #400: providing the output FEISTY needs out-of-the-box rather than relying on post-processing
- FEISTY update
- Haven't walked Zephyr and Colleen through the code yet, but I think that's scheduled for Thursday? Will also happen at next Tuesday's group meeting
- I think documentation is in good shape, though I'll solicit feedback on that after the walkthroughs
- Still need to make code mods from when Matt and I reviewed code together
- I got an email about running an RCP simulation with just ocean / ice, forced from a
BSSP126cmip6run that I'm not sure how to answer
- Matt discussing work on additional alkalinity forcing options
- FEISTY update (time permitting)
- Anything to discuss? Diags / FEISTY / etc?
- Ann B reported an issue with
ladjust_bury_coeff=.true.on GreenPlanet; I think it's a memory issue, but might be compiler related at the root? (GP is using intel 2018, and I can't reproduce the issue on cheyenne)
- Will start the backport of MARBL to CESM 1.3 tomorrow or Thursday
- Waiting on one last round of tests, then KPP Nonlocal stuff will get merged to
dev/ncarand I can work on merging it into the MARBL driver branch
- Bringing SPECTRA into
development- Jessica opened up #396, which has changes to
zoo_loss; currently answer-changing, but we should discuss whether we want the ability to keep the old formulation - Currently, SPECTRA runs require
user_nl_marblto modify PFT settings; this file is generated by an R script in jessluo/gen_spectra. Future MARBL PR: setuse_SPECTRA = .true.and have MARBL python scripts generate correct settings (default diagnostics may also change based on value ofuse_SPECTRA)
- Jessica opened up #396, which has changes to
- Porting MARBL to CESM 1.3
- Email from Matt last week
- FEISTY update
- Any diagnostic updates to discuss?
- Nothing to discuss, but Gustavo laid out a solid plan to get KPP nonlocal work onto
dev/ncarwhich will let me merge it back onto the MARBL branch- Will continue to test MARBL with MCT driver, but this might also be a good time to verify NUOPC cap is behaving as expected (e.g. run 20 year runs with MCT for comparing to POP, but also run 1 month with NUOPC and make sure MOM is getting all the fields it needs to send to MARBL)
-
There is a placeholder for a POP tag for CESM 2.1.4 that fixes an issue in setting some share stream variables in certain compsets
Bugfix: add additional OCN_TRANSIENT option(s) to prevent improper riv_flux_shr_stream_year_last values (Some WACCM SSP runs were done with year 1900 river fluxes rather than year 2000; perhaps other CMIP runs were also affected)I have a note to talk to Keith about this during our weekly check-in on Thursday, but it may get pushed back a week or two depending on how we prioritize other work
-
SPECTRA tag: want to use
io_fallbackto set initial values of SPECTRA tracers (e.g.spCis replaced withppCand 4mp#C; all can be initialized to 0.2*``spC``). May also need to increasemax_avail_tavg_fields.