Meeting notes (this is a wiki - please edit!)
existing cosima mom6 came from mom6 examples from om4 0.25° but with some changes
Adele: should we get the GFDL OM5 param? Steve says we should use their mixing parameters which they have carefully tuned (see MOM6 science testing · Issue #96 · COSIMA/access-om3 · GitHub) - Adele to follow up - GFDL OM5 parameters are here.
Adele: dttherm:dt ratio was only 2 for regional due to boundary issues but could presumably be increased for global
Dougie: WOMBAT has hard-coded 900s timestep for ecology - can be configurable in generic tracers
leave HFREEZE set to 10m (default) so we calculate freeze/melt potential
MINIMUM_DEPTH - changes land mask - set to zero so we control minimum depth via topog.nc
CHANNEL_CONFIG set to none
nk set to 75 at both 1° and 0.25° to match ACCESS-OM2-01 - use KDS75 z* - test hybrid adaptive coords later; initial target is something as robust as we can, eg for CMIP7
look in more detail at DTFREEZE_DP = -7.75E-08 - differs from default 0 and would be redundant if we used TFREEZE_FORM = “TEOS10” - should we move to TEOS10 for EOS as well?
do we want to save diagnostics on a lower resolution horizontal grid?
USE_MEKE: calculates evolution of mesoscale EKE budget and uses it to tune eddy parameterisation - something we should test; until we do that we should leave as-is or turn off - see 1deg_jra55do_ryf: Address MOM warnings · Issue #199 · COSIMA/access-om3 · GitHub
RESOLN_SCALED_KH, KHTH - leave on at both 1° and 0.25°
leave KHTH parameters as they are for mow, but needs testing - important to test if they came from NCAR as may not have been tested
TIDES - turn off, and all associated parameters
Lateral viscosity, smag etc - try using same values as ACCESS-OM2
mixed layer restratification - not very sensitive to these params so use the same as access-om2 for consistency
use KPP with existing for now but test EPBL for future configs (ask Wilton - he’s done some testing)
use pressure-dependent frazil
pen_sw_* - match access-om2 shortwave penetration if we can, and test/compare - may have bearing on WOMBAT - Dougie to do with Pearse - what about WOMBAT-mid shortwave attenuation? is that still planned? - might change
use neutral diffusion at both 1° and 0.25° - is this done the same way as in mom5? - to test
RESTART_CONTROL - try using default (1)
OCEAN_SURFACE_STAGGER = “A” - needed for NUOPC to avoid remapping in mediator
LATENT_HEAT_FUSION - make sure this matches CICE - add to list of cross-component tests
ADJUST_NET_SRESTORE_TO_ZERO = True - matches access-om2
ADJUST_NET_FRESH_WATER_TO_ZERO = True - matches access-om2
ADJUST_NET_FRESH_WATER_BY_SCALING - try this out
FLUXCONST = 0.11 matches access-om2 piston velocity
RESTORE_SALINITY - to investigate - can we have salinity restoring via salt flux in OM3? I guess we want a water flux in CM3?
GUST_CONST - leave as-is for now
MAX_ENT_IT and TOLERANCE_ENT - set to default - see MOM_parameter_doc_* incomplete · Issue #121 · COSIMA/access-om3 · GitHub
HMIX_MIN - use default (0.0) since we don’t use REGULARIZE_SURFACE_LAYERS in our configurations
WRITE_GEOM = 1
RESTART_CHECKSUMS_REQUIRED = True - needed for repro testing, leave as default
all OBC params irrelevant
BOUNDARY_EXTRAPOLATION_PRESSURE - delete - use default
PRANDTL_TURB = 1.0 (default)
SMAGORINSKY_AH - turn on
SMAGORINSKY_KH - turn off
Jump over EPBL for now
leave TIMEUNIT = 8.64E+04 (default)
MAXTRUNC = 1000
DIAG_COORDS - want both zstart and rho2
BAD_VAL_SS*_MAX/MIN - should work out why it crashes with default values Sudden crashes with excessive SSH · Issue #55 · COSIMA/access-om3 · GitHub
Summary of Tests we should run:
- Hybrid and/or Adaptive vertical coordinates (vs default z*)
- The Equation of State - switch to Roquet formulation??
- Can we output diagnostics at subsampled resolution??
- MEKE to tune eddy parameterisation.
- KHTH & neutral diffusion
- Smagorinsky lateral viscosity - move towards OM2 values?
- EPBL vs KPP
- SW penetration scheme - pen_sw_*
- RESTART_CONTROL
- Salinity restoring fluxconst
Note that some of these tests will require longer (10-20 year) simulations.