Cice config #17

anton-seaice · 2023-10-27T05:25:58Z

Copy OM2 config across from https://github.com/COSIMA/1deg_jra55_ryf/blob/master/ice/cice_in.nml

Noting

initial conditions are set per CICE6 initial condition COSIMA/access-om3#50 , and to move to using initial ice conditions based on SST and Latitude would need (very minor) code changes
nprocs and the block sizes haven't been tuned (check in OM2 technical report)
chunking in output files would need code changes (Parallel IO in CICE COSIMA/access-om3#81)
It looks like dalb_mlt was set in the namelist in OM2, but still hardcoded to the default

anton-seaice · 2023-10-29T23:18:49Z

@aekiss Can you review?

aekiss · 2023-10-31T06:37:58Z

Great, thanks @anton-seaice I'll hopefully have a chance to look at this on Thursday.

aekiss · 2023-10-31T06:42:04Z

Here's a table comparing this PR to ACCESS-OM2 and the current MOM6-CICE6 settings. Differences are in bold.
This table shows only the differences.

anton-seaice · 2023-10-31T22:10:00Z

Here's a table comparing this PR to ACCESS-OM2 and the current MOM6-CICE6 settings. Differences are in bold. This table shows only the differences.

Thanks - at some point we might want to figure out how to add a notes column to the table? (And ideally the default values too!)

dougiesquire · 2023-11-01T01:20:12Z

Here's a table comparing this PR to ACCESS-OM2 and the current MOM6-CICE6 settings. Differences are in bold.
This table shows only the differences.

@anton-seaice, how come there are there are lots of variables in your namelist that are different to both ACCESS-OM2 and the current MOM6-CICE6?

anton-seaice · 2023-11-01T02:36:23Z

@anton-seaice, how come there are there are lots of variables in your namelist that are different to both ACCESS-OM2 and the current MOM6-CICE6?

Noting I deleted all the lines where we use the default value, other differences are:

sometimes CICE5 compiler options are now CICE6 config options (e.g. nblyr)
sometimes the directories/paths/filenames are controlled by NUOPC now instead of CICE
I need to double check albedo_type

Did you have specific examples?

dougiesquire · 2023-11-01T02:59:55Z

Did you have specific examples?

Possibly most of what I was looking at is due to default stuff being removed. But yes, the albedo_type and also write_ic I suppose

anton-seaice · 2023-11-07T01:09:40Z

There are a few items defined in both ice_in and nuopc.runconfig which is confusing at best:

I deleted nprocs from ice_in becuase the one in runconfig is the one that is used
ice_ncat, ice_nx, ice_ny in nuopc.runconfig appear to not be used (they are set also in ice_in ) but I haven't deleted them.
I didn't dig into the ICE_modelio section in nuopc.runconfig but suspect most of those fields don't matter.

I will turn off the writing initial conditions again, and set albedo_type to default in ice_in.

I thought / assumed all the output fields were not saved by default. By there is a whole list here: https://github.com/CICE-Consortium/CICE/blob/4450a3e8c64bc07d1173eb3e341cd8dea91d5068/cicecore/cicedyn/analysis/ice_history_shared.F90#L202 which are saved monthly by default, which I will need to work through to set to match the OM2 config.

anton-seaice · 2023-11-07T04:56:46Z

I have gone through and fixed this to correct for the icefields which are on by default. Here are the CICE log extracts which list the output fields - based on the latest commit (76b2c60).

om2_ryf_example.log
cice_config_extract.log

The only substantive difference is fsensn_ai_m which looks like if was off by mistake in OM2 (we were saving the closely related flatn_ai_m)

anton-seaice · 2023-11-21T05:47:58Z

@aekiss Have you been able to have a look?

aekiss · 2023-11-29T01:57:39Z

ice_in

  block_size_x = 20
  block_size_y = 16
-  distribution_type = "sectrobin"


I'd prefer to keep these to make it explicit, as we'll be tuning them

distribution_type = "sectrobin" distribution_wght = "blockall"

ice_in

aekiss · 2023-11-29T02:28:07Z

ice_in

  ns_boundary_type = "tripole"
  nx_global = 360
  ny_global = 300
-  processor_shape = "square-ice"
+  processor_shape = "slenderX2"


Suggested change

processor_shape = "slenderX2"

processor_shape = "slenderX1"

to match access-om2

With the block size and nx/ny settings here, we have 18 blocks in the x direction and 19 in the y direction.

With slenderX1, this gives 19 blocks per processor but only 18 processors would be used. With slenderX2 we get 9/10 blocks per processor and 36 processors used.

Now that I write this, maybe we should change the block_size_x to 16 and block_size_y to 15 with slenderX2 as this gives still 10 blocks per processor but they are slightly smaller and all 48 processors are actually used.

Oh. processor_shape only matters if we set distribution_type="cartesian" so I set

Distribution type = cartesian Distribution weight = latitude

to match OM2

ok it doesn't matter a lot at this stage as the core count will probably change

ice_in

aekiss · 2023-11-29T02:39:24Z

ice_in

-  coriolis = "latitude"
-  cstar = 20.0d0
-  e_plasticpot = 2.0
-  e_yieldcurve = 2.0
  k1 = 8.0d0


Suggested change

k1 = 8.0d0

suggest using default landfast parameterisation here until we have a reason to change

ice_in

aekiss · 2023-11-29T04:50:37Z

ice_in

-  rotate_wind = .true.
-  saltflux_option = "prognostic"
-  tfrz_option = "mushy"
+  tfrz_option = "linear_salt"
  update_ocn_f = .true.
  ustar_min = 0.005


Let's use

Suggested change

ustar_min = 0.005

ustar_min = 0.0005

for now to match access-om2. This is also used in https://github.com/CICE-Consortium/CICE/blob/b14cedfaed8b81500fc5422cfc44b6d80e5893ef/configuration/scripts/ice_in#L253. Also see https://bb.cgd.ucar.edu/cesm/threads/ustar_min-default-documentation.9059/

ice_in

aekiss · 2023-11-29T05:47:40Z

ice_in

  albsnowv = 0.98
-  dt_mlt = 1.50
+  dt_mlt = 1.0


Suggested change

dt_mlt = 1.0

dt_mlt = 1.0

dalb_mlt = -0.02

to match access-om2 since this isn't the default

If would be nice if we can figure out the history / justification for these. It looks like in earlier "AusCOM" definitions dt_mlt meant

change in temp to give dalb_mlt albedo change" (which is hardcoded to 1 now)

but now it means :

change in temp for non-melt to melt snow grain radius change (C)

And looks like its only used for shortwave=dedd (we use default shortwave=ccsm3)

I guess dalb_mlt could have been set to reflect Antarctic vs Arctic differences? (although possibly surprising to not also set dalb_mltv dalb_mlti)

It looks like dalb_mlt was added to the namelist for AusCOM and is not configurable by default. (https://github.com/COSIMA/cice5/blob/edcfa6f9c76ed05b63196ce4b5355fa5a8f4fe3a/source/ice_init.F90#L178)

looks to me like AusCOM used dalb_mlt, but subsequently the same name was used in an update to cice to mean something else

ok let's omit both dt_mlt and dalb_mlt

aekiss · 2023-11-29T06:05:31Z

ice_in

  sw_dtemp = 0.01d0
  sw_frac = 1.0d0
  sw_redist = .true.


these three seem to be for a new feature in cice6 - so should we set

sw_redist = .false.

for now to match access-om2?

actually, that's the default, so all 3 could be deleted, right?

Ah - I get this warning with it set to false:

(input_data) WARNING: ktherm = 1 and sw_redist = F (input_data) WARNING: For consistency, set sw_redist = .true.

It looks like it used to be on by default for BL99, so sw_redist=.true. would be correct to be consistent with OM2

ref: CICE-Consortium/Icepack#326 (review)

ok how about we set

sw_redist = .true.

and leave the others as default values

aekiss · 2023-11-29T06:34:36Z

OK I've looked through everything, apart from diagnostics (which we'll adjust anyway) and BGC (which we can think about later, since CICE6 has significantly improved BGC capabilities).

I've noted just a few things to change and then it will be ready to merge.

aekiss · 2023-11-29T06:35:42Z

I've also updated the comparison table

anton-seaice · 2023-11-30T01:22:27Z

Thanks @aekiss . I responded to all the comments I think

aekiss

thanks @anton-seaice, happy for you to merge

anton-seaice · 2023-11-30T22:29:31Z

@aekiss - I don't have write access to this repo!

micaeljtoliveira · 2023-12-03T23:44:05Z

@aekiss - I don't have write access to this repo!

Should be fixed now.

* Copy OM2 CICE Config * updates for icefields which are on by default * address review comments * review comments * review comments

Copy OM2 CICE Config

f915402

anton-seaice force-pushed the cice_config branch from 3d0971d to f915402 Compare October 29, 2023 22:24

anton-seaice marked this pull request as ready for review October 29, 2023 23:18

anton-seaice mentioned this pull request Nov 3, 2023

Parallel IO in CICE COSIMA/access-om3#81

Open

updates for icefield which are on by default

76b2c60