COSIMA TWG Meeting Minutes 2024

COSIMA TWG
17

Summary from TWG meeting today - please correct and elaborate as needed.

Date: 2024-09-18

Attendees:

AM: CLIVAR OMDP meeting (on the side of CLIVAR/COMMODORE)

  • OMIP papers being prepared, eg
    • Fox-Kemper on diagnostics and forcing strategy and protocol
    • Gokhan: ocean spinup
    • ocean-sea ice data request
  • First phase to use JRA55-do 1.5
  • Gokhan leading ERA5-based forcing (CFORCE) for the next phase of OMIP
    • person doing work only just started
    • testing by NCAR/GFDL/ Los Alamos. We’ll do 2nd stage testing (with others)
    • Plan to update to ERA6 when avail
  • Bill Large leading revision of bulk formulae
  • Discussion on spinup protocol
    • Moving away from 4x61-year cycles to avoid warm → cold jump from 2018 → 1958.
    • Maybe repeat decade for several cycles, then IAF at end. TBC.
      • Advantage of repeat decade is inclusion of interannual variability in spinup, and closer to climatology than a single-year RYF.
      • But could make validation awkward if repeated decade is (say) 1958-1968, since both forcing and obs are poor. But later decades include more climate change signal.
  • Relative or absolute winds - in discussion. Eric Chassignet in favour of mixed (70% of ocean vel). But depends on how ERA5-based forcing is developed
  • Whether to use SSS restoring
    • previously problematic, with differing parameters and issues extending into future.
    • Harrison et al 2022 switch off salinity restoring and have freshwater transport between latitude bands based on CMIP models, and land storage in each band to have delays. Fixed many biases eg AMOC and T drift in spinup. Suggest using this instead of SSS in OMIP, with atm and land bulk reservoirs. Other groups unsure.
    • Multiple groups to test this, including us. Details to come (Adcroft/Harrison). Invite Adcroft/Harrison to present on this at a COSIMA meeting? [NB: this has now been scheduled for the COSIMA meeting on 14 Nov 2024]
  • We should follow OMIP protocol to facilitate comparison with other OMIP studies.

CB: what’s the plan for spinup, given we won’t have finalised OMIP?
AH: go with something consistent with ACCESS-OM2 for comparison
AM: why match with OM2? why not match best with obs?
AH: two objectives: 1: good enough for coupled modellers (benchmark=OM2); 2: perfect the OM3 model, eg using repeat decade etc from OMIP

AH: COMMODORE meeting updates

  • quite technical - 5-6 engineers there, including some talks
  • ACCESS-OM2 bottom water was news to many attendees, and some were unaware the AABW shouldn’t be from open-ocean polynyas
  • emphasis on Gulf Stream separation to justify high res

AM:

  • lots on machine learning, eg for parameterisations; danger of us falling behind?
  • a lot of work on climate model emulators

AK:

AM: what are our options for icebergs? Can CICE6 do it?
AS: unclear what is being done. Was work ~10yr ago with icebergs in CICE but not coupled to ocean

AM: What about distributing iceberg flux in OM3? We should spread according to a prescribed pattern rather than at coast. See issue.

  • Martin/Siobhan/Chris: at the moment we have HADCM3 seasonal climatology from coupled run
  • Chris: has used interannually-varying iceberg distribution
  • Adele: any spread is better than nothing
  • Andes : about 50% at the coast, 50% spead by distribution
  • Anton: Who to take this on?

AS: @minghangli’s tracer timestep investigation:

  • seem to be settling on 3hr irrespective of resolution
  • to be tested at 0.25°

AS: @ezhilsabareesh8 progressing well with new grid

  • error of too many CFL truncations
  • log files not working?

NCMAS

  • NCMAS applications are likely open soon, although they appear delayed. We need numbers scaling performance for both OM2 and OM3. Its hard to provide solid OM3 numbers at this point.
  • AM: need good numbers to include OM3 in proposal. Last year’s application was criticised for lack of OM3 detail
  • few proposals in this round are likely to need OM3

Andy: OM3 evaluation

  • had a lot of test cases lined up
  • timing, timestepping etc needed to be done first - now basically done
  • revisit test plans, or proceed with plans?
  • strategy of something decent for CM3, then refine
  • run and analyse all, or involve community in analysis and/or runs?

Anton: Best case is with tracer time-step and new grids, OM3 will be stable and efficient enough for CM3 without further investigation. C-grid cice testing to be done still at 0.25° - expect to be ok.

Andrew: for refinements after an intial version of CM3, we want to be able to keep those in sync

Martin: @kieranricardo has custom mediator and different build process, using OM3 object file as library. But OM3-CM3 upgrade path fairly frictionless so can adopt OM3 changes easily.

Adele: for model evaulation we need to get COSIMA to agree on a priority list of metrics and acceptable values (eg T bias, AABW formation rate, etc), and ideally to write assessment scripts

  • Andy: Cookbook or ESMvaltool?
  • Anton: going with Cookbook style - lots of overhead with ESMvaltool
  • Andrew: want to avoid additional barriers to engagement - already have MOM6, etc
  • Andy: need a plan and timeline once we have Ezhil’s version stable and Minghang’s timings

COSIMA leadership meeting at Australian Antarctic Research Conference, Hobart 19-22 Nov

  • intro movie
  • ethics
  • attribution/credit

AK: For a later discussion: post-CMIP7 priorities: high resolution? BGC? WW3? DOCN? @lgbennetts is keen on CICE-WW3-DOCN but needs excessive wave attenuation to be resolved (perhaps with this bugfix? need to wait until available in ESCOMP before we test it in OM3)

2 Likes