The commit history of this repo tracks the changes made in developing namelists for ACCESS-OM2. Contact me if you want write access.

See namelist-check.pdf for an overview of namelists.

New namelists are in new/control

Run rsync_nml_from_raijin.sh to download all namelists from /g/data3/hh5/tmp/cosima/access-om2* to ./raijin

You may find nmltab useful for comparing namelists.

• cf a CICE nml from somebody we trust
• check CICE physical parameters are consistent with MOM - esp dynamics_nml
  • check that CICE and MOM are using the same definition of the freezing point - see David Bailey email 2017-11-29
• made changes discussed in Slack up to 2017-11-05: https://arccss.slack.com/archives/C6PP0GU9Y/p1508899952000034
• consider disabling max_delta_salinity_restore = 0.5 in ocean_sbc_nml: https://arccss.slack.com/archives/C6PP0GU9Y/p1511849053000081 and mom-ocean/MOM5#203
• make agreed nml changes
• cf. Griffies et al. (2015) p973
• cf. GFDL_ESM2M_input-cut.nml
• cf. MOM_SIS_TOPAZ_input.nml
• cf. MOM output file to show defaults for unspecified values
• Deleted this from 1deg: not in MOM5 code; no other configs other than old 1deg access use it

&bg_diff_lat_dependence_nml
    bg_diff_eq = 1e-06
    lat_low_bgdiff = 20.0
/

• check: do we want this at 0.25 and 0.1 deg?

&fms_nml
    domains_stack_size = 115200
/

• check: Russ said Matt C found neutral=.false. is better? False is default. or is this not used anyway? Looks like it's for atmosphere not ocean?

&monin_obukhov_nml
    neutral = .true.
/

• set ncar_boundary_scaling_read=.false. for first run only - see https://arccss.slack.com/archives/C6PP0GU9Y/p1508905387000065
• TEOS10: Russ says we're using eos_preteos10=.true. for performance. GFDL do this too. Bob: NEMO's polynomial TEOS-10 approx is equally accurate and much faster - see Fabian paper - ask Kial
• use do_bitwise_exact_sum in ocean_barotropic_nml (ESM2M does), ocean_blob_nml, ocean_density_nml, ocean_grids_nml (ESM2M does), ocean_rivermix_nml (ESM2M does), ocean_sbc_nml, ocean_tracer_diag_nml (GFDL don't), ocean_velocity_diag_nml ? Code doc says "Set true to do bitwise exact global sum. When it is false, the global ! sum will be non-bitwise_exact, but will significantly increase efficiency. ! Default do_bitwise_exact_sum=.false." but is this out of date? Bob says the exact order-independent summation has been put into latest FMS - so check whether it's in the version we use.
• use bitwise_reproduction in ocean_blob_nml?
• ESM2M uses smooth_eta_t_biharmonic=.true. but we have .false.. This is used only for depth-based coords (eg z*). We could have .true. but it may introduce noise and isn't recommended. Default is .false. so best to leave it .false.? But why do GFDL turn it on?
• smooth_eta_diag_laplacian=.true. in all cases and default, even though this is used only for pressure-based coords.
• ESM2M and old ACCESS use smooth_eta_t_laplacian=.false. but we have .true. which is the default. This is used only for depth-based coords (eg z*).
• ESM2M has smooth_pbot_t_biharmonic=.true. but we have .false. (the default). Not used in depth based coords anyway.
• ESM2M has smooth_pbot_t_laplacian=.false. but we have .true. (the default). Not used in depth based coords anyway.
• We have vel_micom_lap_diag=0.2 (the default) but ESM2M uses 1.0
• We use cdbot=0.001; ESM2M uses 0.002 - that's ok - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192485000002?thread_ts=1509185894.000029&cid=C6PP0GU9Y
• We use cdbot_hi=0.007; ESM2M uses default (0.003) - that's ok - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192485000002?thread_ts=1509185894.000029&cid=C6PP0GU9Y
• We use cdbot_roughness_uamp=.true.; ESM2M uses .false. (default). "This approach is more relevant for coarse models than the constant roughness length used in the cdbot_law_of_wall option" - that's ok - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192485000002?thread_ts=1509185894.000029&cid=C6PP0GU9Y
• We have use_geothermal_heating=.false. (default); ESM2M has .true. - leave this .false. for now - heating data is not well constrained - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192485000002?thread_ts=1509185894.000029&cid=C6PP0GU9Y
• ESM2M uses bottom_5point in &ocean_bihgen_friction_nml; don't turn this on unless we need it? It's on at 1 deg but not higher. our vel_micom_bottom also differs from ESM2M at 0.25, 0.1deg (matches at 1deg) but is presumably ignored.- see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192485000002?thread_ts=1509185894.000029&cid=C6PP0GU9Y
• our vel_micom_iso and visc_crit_scale in &ocean_bihgen_friction_nml matches ESM2M at 1 deg but differs at 0.25, 0.1 deg - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192485000002?thread_ts=1509185894.000029&cid=C6PP0GU9Y
• ESM2M radiation requires a chl file - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509185901000006
• in &ocean_density_nml our neutralrho_max, neutralrho_min differ from ESM2M & ACCESS. They look like they were mistakenly copied from potrho_max, potrho_min - FIXED
• FIXED THESE in &ocean_lapgen_friction_nml
  • removed ncar_only_equatorial=.true. at 1deg to match ESM2M (default is false) - differs from old ACCESS https://arccss.slack.com/archives/C6PP0GU9Y/p1509192207000072?thread_ts=1509185883.000090&cid=C6PP0GU9Y
  • changed viscosity_ncar to .false. at 1deg to match ESM2M & old ACCESS have .false. (the default). see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192207000072?thread_ts=1509185883.000090&cid=C6PP0GU9Y
  • vconst1,3,4,6 differed from default in 1deg; set to default to match ESM2M - see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192207000072?thread_ts=1509185883.000090&cid=C6PP0GU9Y
  • at 1 deg left viscosity_ncar_2000, viscosity_ncar_2007 unspecified to match ESM2M. This changes viscosity_ncar_2000 from .false. to .true. (default) and changes viscosity_ncar_2007 from .true. to .false. (default). see https://arccss.slack.com/archives/C6PP0GU9Y/p1509192207000072?thread_ts=1509185883.000090&cid=C6PP0GU9Y
• ESM2M uses a mask in &ocean_mixdownslope_nml; we don't.
• ESM2M has frazil_only_in_surface, freezing_temp_simple true; we have false.
• is fms_io_layout overridden by io_layout and therefore redundant?
• check: sigma_2 or sigma_0 for

&ocean_density_nml
    eos_linear = .false.
    eos_preteos10 = .true.
    layer_nk = 80
    neutralrho_max = 1038.0
    neutralrho_min = 1028.0
    potrho_max = 1038.0
    potrho_min = 1028.0
/

• checked ocean_frazil_nml is used by ACCESS
• should frazil_factor=0.5 ?
• viscosity_scale_by_rossby_power: Bob said GFDL uses 100 but GFDL namelist actually uses 4 and default is 2. So set to 4
• many nml groups (eg &ocean_lap_friction_nml) don't include use_this_module - how many have use_this_module = .false. by default and are therefore disabled? - checked - all ok
• ocean_albedo_option = 2 looks dodgy? or is this not used anyway? (handled by coupler/CICE?) ESM2M uses ocean_albedo_option = 5 - but Steve says "Also note that "ocean_albedo" is set for a coupled model, and it is different for ocean/ice simulations. That is a major "gotcha" that Spence can share with you if interested.'"
• changed clock_grain from 'LOOP' to 'COMPONENT' (as used by ESM2M) - may give better performance?
• delete ocean_polar_filter and ocean_vert_kpp_iow? apparently not in MOM5 code
• try barotropic_split > 80 to get dt>1800 at 0.25deg (2160s is very nearly OK) - would need to be >84 to pass startup check - but none of the models use >80 - why not? - see COSIMA-TODO.md and run-025deg_jra55_ryf_broadwell_test.md and Ch11 of Griffies2012a-mom-elements-5-updated.pdf
• Bob suggests paying close attention to tuning ocean_nphysics*. Also check for consistency across resolutions.
• Russ: set river_insertion_thickness=10.0 in 0.1deg? Bob: check code - may need to set to 10m for all resolutions if we refine vertical resolution
• Nic: is 600 OK for dt_cpld at 0.1deg? was 150.
• check ocean_submesoscale_nml at 1deg - don't use submeso_diffusion_biharmonic?
• check ocean_vert_tidal_nml with Steve - use Hogg 1deg settings for background_diffusivity, decay_scale and shelf_depth_cutoff for all resolutions? (nb: shelf_depth_cutoff=-1000 turns this off)
• check xgrid_nml settings with Marshall
• should we be more precise than grav = 9.8?
• bbc namelist changes now require a roughness_cdbot file - revert this?
• s_min_limit=2.0psu - might we get less than this in estuaries? Minimum salinity below which will employ upwind advection instead of quicker, and horizontal diffusion instead of neutral physics.
• enable these?
  • ocean_overexchange_nml (Griffies et al 2015: "None of the CM2-O ocean configurations make use of an overflow parameterization.")
  • ocean_overflow_nml
  • ocean_overflow_ofp_nml
  • ocean_mixdownslope_nml
  • ocean_sigma_transport_nml (GFDL use it)
  • remove do_alltoall (already default) and do_alltoallv (NOT default)? what do they do?
  • make_exchange_reproduce (doc says "This option has a considerable performance impact" but Bob says it won't slow it down in recent versions of MOM)
  • we should go with massless ice for all cases
  • SSS restoring is quite strong (15days) at 1 deg. Should be 60 days
  • the tracer advection scheme sweby was changed by the multi-dimensional piecewise (MDPPM) method since recommendations from Steve to the ACCESS team back to 2010 (access_ocean_notes attached). - check MDPPM is enabled (is it default?)
  • Also following Steve’s suggestion from last year, the truncate_velocity in ocean_velocity_nml should be set to FALSE.

Aidan's namelist recommendations (email 2017-10-26) have all been made:

1. You've got debugging turned on for the ocean_tracer_advect module, which is fine if you want to debug tracer advection, but otherwise just takes extra time

&ocean_tracer_advect_nml debug_this_module=.false. advect_sweby_all=.false. /

2. The minimum salt is negative, but the model will blow up with salinity < 0. Probably not important, but I see this a lot in MOM namelists and often wonder why it is

&ocean_tempsalt_nml s_min = -1.0 /

3. You have barotropic smoothing instead of laplacian. This is not recommended explicitly in the code

&ocean_barotropic_nml smooth_eta_t_laplacian=.false. smooth_eta_t_biharmonic=.true. /

4. Russ advised me that kbl_standard_method and smooth_blmc should always be set to false

&ocean_vert_kpp_nml smooth_blmc=.true. kbl_standard_method=.true. /

5. All your diag_step variables are set to 120. For your 1800s time step this is every 60 hours. If you're in production mode and don't need to monitor these then the GFDL people typically set them to a month, or even the entire runtime (1 year?).

• ncar_boundary_scaling_read = .true.
• vel_micom_bottom = 0.1
• matched &ocean_bihgen_friction_nml to ESM2M following https://arccss.slack.com/archives/C6PP0GU9Y/p1509185883000090 - but now these differ from higher resolution

• viscosity_ncar_2000 = .true. by default and in ESM2M

• red sea fix - make generic and also apply to bering str? or just fix both with topg changes and set redsea_gulfbay_fix=.false.? But Andy found topog changes alone did not fix it, so needed to turn fix on as well: https://arccss.slack.com/archives/C6PP0GU9Y/p1509922922000020 so should topog changes be reverted?

##CICE5p1

see AK email to Petra 2017-11-15 and highlights in HunkeLipscombTurnerJefferyElliott2015a-CICE5p1.pdf

• how many ice classes do we want? any implications for load balancing? any implications for coupler & mom?
• cice_in.nml: set ktherm = 2 (mushy ice) as recommended by Bob; see COSIMA/access-om2#56 ktherm=2 should hopefully fix cice message:

cat work/ice/ice_diag.d  
 --------------------------------
   CICE model diagnostic output  
 --------------------------------
  
 WARNING: ktherm = 1 and tfrz_option = mushy
 WARNING: For consistency, set tfrz_option = linear_salt

• adjust diagfreq in cice_in.nml to get sensible output frequency? (NB: dt overridden by value from config.yaml?)
• enable melt ponds? at present tr_pond_*=false in cice_in.nml

• make MATM namelist consistent