Skip to content

2019 MARBL Dev team meetings

Jump to bottom
Michael Levy edited this page Jun 18, 2019 · 46 revisions

July 2, 2019

General discussion

MARBL Software Updates

  1. Kristen's PR (#330) to bring in explicit calcifiers:
    • Up to date with development
    • Reviewed with Keith and Mike
    • Mike made some updates to general infrastructure, now Kristen has some work to do (after OCB)

POP Software Updates

  1. Updates needed to go with explicit calcifier work:
    • Add a new test to the test suite
    • Question: what's the best way to initialize data? Three options, ranked from Mike's most- to least-preferred
      1. Create a new initial data set that is just a copy of the current init_ecosys_init_file (for both x1 and x3 grids) but also containing cocco tracers (that are copies of the sp tracer values), use it when running with Kristen's tracers
      2. Same as above, but make it the default init file for running regardless of whether user wants cocco tracers or not
      3. Use io_read_fallback to initialize the cocco tracers, don't modify initial condition file at all
  2. Updates on marbl-diags
    • Worked on it at the early June hackathon

MOM Software Updates

  1. Mike started work on the MOM driver
    1. Alper's fork of CESM now includes MARBL as an external
    2. The buildlib script in the MOM6 interface builds MARBL
    3. Setting USE_MARBL_TRACERS = True runs marbl_instances%init() and registers the 32 resulting tracers
      • No initial conditions provided, so tracers are all set to 0
      • Need a namelist flag somewhere to set num_PAR_subcols (current init sets value to 1)
      • Need to figure out how many active cells can be passed to surface_flux_compute() at a time (i.e. the MOM6 equivalent to the number of active cells on a POP block)

No Meetings

  • June 18, 2019 (CESM Workshop)
  • June 4, 2019 (Python hackathon at NCAR)
  • May 21, 2019 (Mike out of town)

May 7, 2019

General discussion

  1. Diagnostics: I have a notebook to compare JRA and CORE: a few time series of global means
    1. What variables would be useful to look at in this framework? (adding additional 2D fields is trivial, adding 3D fields will just require a vertical integral that I assume esmlab can do)
    2. Should we generate these plots in marbl-diags?
    3. I converted Who's single cycle of the JRA-forced hindcast to time series in order to make these plots -- should I copy them from my scratch space into /glade/p/cgd/oce/projects/JRA55/IAF so they don't get scrubbed? It's roughly 1 TB of data.
  2. MOM driver as a priority ahead of A. Shao visit this summer?

MARBL Software Updates

  1. Single column test progress

    • more robust build system (for turning on netCDF support if needed)

    • not sure of the best way to store input data, tempted to just keep it in the MARBL repo (current version is a 2-column file and 98 KB, so even adding 4 or 5 more columns is probably not a burden for git). Test currently fails on TravisCI due to lack of file to read:

      (run_exe): Running following command:
(run_exe): /home/travis/build/mnlevy1981/MARBL/tests/driver_exe/marbl.exe -n test.nml
Beginning compute_cols test...
MARBL ERROR (marbl_io_mod:netcdf_check): netCDF error: No such file or directory
MARBL ERROR (marbl_io_mod:marbl_io_open): Error reported from nf90_open(marbl.nc)
MARBL ERROR (marbl_compute_cols_drv:initalize): Error reported from marbl_io_open
MARBL ERROR (marbl_compute_cols_drv:test): Error reported from initialize

  2. Kristen is happy with her update to make her coccolithophore branch CESM 2.1 compatible (and that's been tagged), but we haven't found time to bring her up to date with the latest MARBL development: that's holding up #330

  3. PGI latest community edition (i.e. free version) is available - PGI 19.4 includes macos mojave support (18.10, the last community edition, runs on macos 10.13 but not 10.14). No issues building out of the box with it.

POP Software Updates

  1. Keith and I talked about what moving POP to git means for MARBL development -- will abandon the idea of a marbl_dev branch and just push our changes directly to master. This is in line with the typical git workflow, and I think MARBL is stable enough that we've outgrown the need for marbl_dev anyway.

No Meeting

  • April 23, 2019 (Mike home with a sick kid)

April 9, 2019

General discussion

MARBL Software Updates

  1. Merged #328, which adds Arrhenius temperature scaling option and finer-grained grazing diagnostics (Jessica's SPECTRA code should run out-of-the-box now)
    • Answer-changing on some compilers (including intel), so only available in CESM 2.2 or later (not in 2.1)
  2. Upcoming work:
    • After cesm2_2_beta02 tag is made, Kristen and I will work on getting her explicit calcifier work onto development (#330)
    • Sounds like #146 (adding latitude and other variables to marbl_domain_type) is a priority now as well

POP Software Updates

  1. cesm_pop_2_1_20190322 has changes necessary to run SPECTRA (larger max tavg count, different initial condition file when running SPECTRA)

  2. Working on updates to both CESM 2.1 release branch and CESM 2.2 development to run with poorly-balanced PE layouts

    • Run when nblocks_clinic < max_blocks_clinic on some tasks
    • Run in the special case of above when nblocks_clinic = 0 on some tasks

    Note that this issue affect the base POP code as well as the MARBL driver... so Keith fixed the base code and I'm getting the C1850ECO tests to pass

No Meetings

  • March 26, 2019 (everyone out of town for local school spring break)
  • March 12, 2019 (goodbye party for J. Luo)

February 26, 2019

General discussion

  1. Diagnostics
    • Current work -- (a) & (b) done a few minutes ago, I'd like to review; (c) not yet started:
      1. Restructure YAML configuration files for better extensibility
      2. Update existing annual climatology plots
        • add contour lines
        • better colormaps (rainbow for tracer values (?), definitely need improved scaling for non-uniform scales; bwr, centered at 0, for difference plots)
      3. Additional diagnostic plot options
        • E.g. fields Keith looked at in comparing JRA & CORE: nitrogen fixation and other fields under the heading "Ecosystem: Maps" in CESM 1.3 BGC diags
    • I talked with Alice, have a much better grasp on how marbl-diags will interact with CESM_postprocessing. In the meantime, jinja2 in marbl-diags to generate stand-alone HTML?
  2. MARBL-adjacent projects
    • Kristen is looking for a way to smooth SSH before computing some correlations, Matt asked me to investigate Loess filter as option. I found an existing python package, but have some concerns:
      1. I'm not sure I can shoehorn in global data (was thinking about applying filter region-by-region, using POP's region mask)
      2. It's an expensive method (filter weights at a grid point are based on distance to nearest neighbors)
      3. I need to read up a little more on the parameters controlling the filter (first attempt at filtering a global dataset didn't come out looking so great)

MARBL Software Updates

  1. New PR: #330 has Kristen's explicit calcifier updates to allow adding coccolithophores
  2. Single column testing update: mostly stalled as I focus on other work
    • Decent notes should allow me to pick it back up again when I have the time
    • I'd like to get the explicit calcifiers and size-resolved plankton settings onto development before coming back to the single-column tests (requires code review and more thorough CESM test suite; I think both PRs can be done in a bit-for-bit manner but that might require moving Fortran parameters into MARBL's input settings file)

POP Software Updates

January 29, 2019

General discussion

MARBL Software Updates

  1. Stand-alone driver:
    • Adding a regression test that calls surface_flux_compute() and interior_tendency_compute() is top priority
      • Planning on pulling initial conditions / forcing data from a single column of POP, as I am not aware of any applicable analytic test case
      • up for debate: driver can read a netcdf file, or I can hard-code values in Fortran (downside to netcdf is the need to make it available for anyone running the test)
      • I like the idea of hard-coding values from a shallow-ish (5 - 10 level?) column if no
    • Will need to update documentation to refer back to this test
    • This test will be a good launching point for a true stand-alone driver that includes time-stepping
      • No need for that driver to be Fortran -- maybe take advantage of python ODE solvers?
      • Another advantage to python-based driver: jupyter could incorporate a simple test case AND generate plots
  2. Diagnostics:
    • Did a little more tweaking to plot generation, see PR #11
    • Currently on back-burner waiting for Alice B to come back and answer some questions (I have local copy of text document, no need to get into details of what the questions are)
    • When I come back to this, will work closely with Matt to ensure we're on the same page

POP Software Updates

January 15, 2019

General discussion

  1. Update on marbl-diags and CESM_postprocessing:
    • Working on a branch with a goal of recreating ecosys sections of CESM 1.3 diags
    • First step: Ecosystem: Maps at Depth section, goal is plots similar to NO3 an 0m and Fe at 0m
    • Second step: I don't think the current workflow in marbl-diags will play nicely with the way CESM_postprocessing parallelizes tasks, so it might be worth rethinking workflow. It shouldn't be too hard to introduce a new interface that lets us loop over variables to plot at the driver layer, which I think is the goal.
    • Still unclear of how to generate HTML (Alice B has been working remotely and I've been focusing on the plot generation, but pretty soon I'll be ready to make the web pages)
    • Planning to compare to WOA 2013 -- I originally wanted to also include WOA2005 to make sure statistics are being computed correctly, but /glade/p/cesm/bgcwg/obgc_diag/data/obs_data/WOA2005/WOA05_ann_avg_gx1v6.nc doesn't match the min / max from the links above and I don't know if it's worthwhile to track down the dataset used in those original plots.

MARBL Software Updates

  1. DOIs: after setting up a DOI for Kristen, went ahead and got DOIs for MARBL in CESM 2.x releases:
  2. Documentation: I think this should get bumped up in priority -- our github.io page is missing the CESM 2.1 documentation (and there have been several interface changes so some pages are out of date)

POP Software Updates

No Meeting

  • January 1, 2019 (NCAR closed for New Year's Day)
  1. Design Document
  2. How to Submit Code Modifications
  3. Dev meeting agenda / rough notes
  4. Processes Still Being Developed
Clone this wiki locally