Add ACCESS-ESM3 by kieranricardo · Pull Request #12 · ACCESS-NRI/CMEPS

kieranricardo · 2026-02-02T23:41:15Z

Description of changes

CM3 related changes to CMEPS. This adds an "access" coupling mode and associate field exchange module. Also adds in a custom access coupling routine to "mep_phases_post_atm_mod.F90".

Specific notes

Contributors other than yourself, if any: @blimlim

Are changes expected to change answers? No, this shouldn't affect OM3.

anton-seaice

I didn't go through everything, but my main comment is - can we harmonise field names and code style with the rest of CMEPS

anton-seaice · 2026-02-03T04:37:54Z

+   nullify(is_local%wrap)
+   call ESMF_GridCompGetInternalState(gcomp, is_local, rc)
+
+   call ESMF_FieldBundleGet(is_local%wrap%FBImp(compice, compatm), fieldName='ia_aicen', field=ice_frac_cat, rc=rc)


Presumable we should be running the ChkErr error, most of the rest of the code does that

anton-seaice · 2026-02-03T04:54:21Z

+      ! to atm: from ocn
+      ! ---------------------------------------------------------------------
+      allocate(S_flds(3))
+      S_flds = (/'So_t', 'So_u', 'So_v'/) ! sea_surface_temperature


Suggested change

S_flds = (/'So_t', 'So_u', 'So_v'/) ! sea_surface_temperature

S_flds = (/'So_t', 'So_u', 'So_v'/)

There's lots of comments like this which are wrong or only partially complete

anton-seaice · 2026-02-03T04:56:46Z

+      ! ---------------------------------------------------------------------
+      allocate(S_flds(9))
+      S_flds = (/'Si_t', &
+                  'ia_aicen', &


Suggested change

'ia_aicen', &

'Si_ifrac_n', &

Should we try and reuse existing names for these ?

from

CMEPS/mediator/esmFldsExchange_cesm_mod.F90

Lines 1753 to 1754 in 63e4c6d

call addfld_from(compice, 'Si_ifrac_n')

call addfld_to(compocn, 'Si_ifrac_n')

anton-seaice · 2026-02-18T02:36:00Z

 project(cmeps Fortran)

-set(SRCFILES esmFldsExchange_cesm_mod.F90 med_fraction_mod.F90
+set(SRCFILES esmFldsExchange_access_mod.F90 esmFldsExchange_cesm_mod.F90 med_fraction_mod.F90


We don't use this file, so possibly this is confusing ([esmFldsExchange_hafs_mod.F90](https://github.com/ACCESS-NRI/CMEPS/blob/upstream-main/mediator/esmFldsExchange_hafs_mod.F90) isn't included, for example)

It'd be nice to be able to use the same version of CMEPS in CM3 and OM3, but this CMakeList.txt is overridden in access3-share anyway, so I could put the change there. What do you think? Then we have a mediator that can compute both CM3 and OM3 fluxes.

anton-seaice · 2026-02-18T02:41:32Z

+
+      do j = 1,lsize2
+         do i = 1,lsize1
+            if (ice_frac_cat_ptr(i, j) > 0.0) then


if ice_frac_cat_ptr is very small, won't there be a resulting overflow in ice_flux_cat_ptr ?

kieranricardo · 2026-02-24T05:15:05Z

@anton-seaice this edge case doesn't occur in CM3. CICE zaps small ice fractions and the UM considers ocean fractions <0.01 as land, which limits how small ice_frac_cat_ptr can be here.

anton-seaice · 2026-02-26T01:20:42Z

Changed base to cmeps1.1.35-x, looks like it will merge fine

anton-seaice

Thanks @kieranricardo

I made it as far through this as I had time for.

There are some features in esmFldsExchange_cesm which are missing here and are quite useful, including the configuration option for rof->ocn mapping files and the detection of which meshes are the same. There's also lots of duplication with the addfld/addmap/addmrg with those in cesm FldsExchange.

Would it work to call esmFldsExchange_cesm first, and then in this file, only add any extra fields needed? All the fields in esmFldsExchange_cesm are wrapped in if (fldchk(...) statements, so it should just pass over any fields which are not needed.

anton-seaice · 2026-03-13T00:12:32Z

+      !=====================================================================
+
+      call addfld_to(compatm, 'So_ofrac')
+      call addfld_to(compatm, 'Si_ifrac')


Needs an if comp_present ?

I'm not 100% on this, but I think doing the fieldchecks in esmFldsExchange_accessesm_init is sufficient (which I've now added). We can advertise whatever fields we want (which is what this does), as long as we don't try to map between fields that don't exist.

Also I can't see any comp_present checks in esmFldsExchange_cesm, so I think this should work :)

anton-seaice · 2026-03-13T03:03:58Z

+      call addmrg_to(compatm, 'sstfrz', mrg_from=compice, mrg_fld='sstfrz', mrg_type='copy')
+
+      allocate(S_flds(7))
+      S_flds = (/'Si_ifrac_n', &


We should avoid redefining these lists twice and just reuse, them or set them as Parameters.

kieranricardo · 2026-03-16T02:08:48Z

@anton-seaice and @blimlim thanks for you reviews!

Re. using the esmFldsExchange_cesm, there's a few potential issues:

If any of the CESM fields use a different re-mapping to us this won't work (I think it will either crash, or the mappings will override each other without a guarentee of which mapping is actually used)
I'm fairly sure that this require's aofluxes to run (and crashes it is isn't run), which could potentially override the UM fluxes.
There's a potential additional maintainence burden - we'd have to keep track of which fields of ours they define, and what mappings are being used.

So I think we're better off just using our own self-contained module.

Re. the field names, most of these names were chosen to reuse existing fields in fd.yaml. Aside from rain, there aren’t many fields with the Faxa_ prefix there. Also, esmFldsExchange_cesm.F90 sometimes plays a bit fast and loose with the prefixes - e.g. Faxa_rain is exported directly to the ocean, whereas it would probably be more consistent for this to enter the mediator as Faxa_rain and then be mapped to Foxx_rain. Given that, I’m inclined to avoid introducing additional Faxa_... fields into fd.yaml, especially since the other esmFldsExchange... modules don’t fully adhere to that convention either. But I don’t feel strongly about it, so happy to go with whatever approach people think is best.

anton-seaice · 2026-03-25T03:17:00Z

+                  'Sa_u', &
+                  'Sa_v', &


'Sa_v', &

Sounds like these two are not needed? But otherwise should be mappatch_uv3d ?

Yep, I was able to delete this

anton-seaice

Shall we make an issue/issues for missing features, mostly:

-configurable rof2ocn_ice_rmap and rof2ocn_liq_rmap

I went through the remapping methods and added a few questions

anton-seaice · 2026-03-25T23:04:05Z

+    ! This is a mediator specific routine that determines ALL possible
+    ! fields exchanged between components and their associated routing,
+    ! mapping and merging
+    !---------------------------------------------------------------------


Shall we add a note here on the general approach to remapping, e.g. which method is used for which remapping ?

That might be a little complicated to capture here, but we should definitely document this.

I was just thinking, we use:

conservative remapping for states and volume fluxes (temp, area fractions, runoff etc)

patch remapping for stresses and wind
?

ah gotcha, sounds good!

conservative remapping for states

Prognostic states (temperature, humidity, etc) don't need to be conservative and would be better off being something smooth, e.g. bilinear for most of them, and patch for wind speed so derivatives are continuous. Only per-area quantities (e.g. fluxes and ice fraction) need to be conservative.

Are we remapping wind stress? I'm not sure what to do with that, as we want it to be smooth (e.g. so the curl is continuous, suggesting we use patch rather than bilinear), but stress is a momentum flux and patch is not conservative. Does it cause problems if the ocean-atm stress is different for the ocean and atmosphere? Or maybe it doesn't matter, as we also don't account for the momentum of the solid earth, so we're not conserving momentum for the whole earth system anyway?

anton-seaice · 2026-03-25T23:10:53Z

+             fldchk(is_local%wrap%FBImp(compocn, compocn), trim(fldname), rc=rc) &
+            ) then
+            call addmap_from(compocn, trim(fldname), compatm, mapconsf, 'ofrac', 'unset')
+            call addmrg_to(compatm, trim(fldname), mrg_from=compocn, mrg_fld=trim(fldname), mrg_type='copy')


I guess So_u and So_v should be mappatch_uv3d

I think we should add an issue for these, if mappatch_uv3d works better we should use this for the stresses as well. But I think we should properly evaluate this before implementing.

We use patch for tau, so surely this should be tau too ?

the _uv3d part does seem contentious, so how about just mappatch

sounds good to me

Agreed - if by 3d you mean what was discussed here, then I think it's best avoided. ACCESS-NRI/access-om3-configs#1045 (comment)

anton-seaice · 2026-03-25T23:20:06Z

+                  'Sa_u', &
+                  'Sa_v', &
+                  'Sa_shum', &
+                  'Sa_tbot', &


Sa_tbot from atm uses mapbilnr and So_t from ocn to atm uses consf - is that correct ?

I was able to remove Sa_tbot, so now all our state fields are using conservative remapping. Conservative remapping is useful for ice->atm, bc it gives us a good option of weighting by ice fraction (weight by ice fraction, conservatively remap, divide by ice fraction). Also, because the atmosphere is a lot coarser that the ocean/ice grid it doesn't matter too much which re-mamping we use.

kieranricardo · 2026-04-21T02:15:50Z

@anton-seaice @blimlim I've changed the coupling names around. Some choices still seem ambiguous to me, but at least we're no longer labelling fluxes from the atmosphere with Foxx!

blimlim

Thanks @kieranricardo, this is looking good to me! The naming choices seem difficult given the restrictions from the MOM and CICE caps, and so the current decisions seem sensible. I've just added a couple points on some of the comments and arrangement of the fields.

blimlim · 2026-05-01T05:27:44Z

+      call addfld_to(compice, 'Faii_taux')
+      call addfld_to(compice, 'Faii_tauy')


Would Faia be better here as they're being calculated in the atmosphere?

It would be good for clarity to move them out of the ! precipitation section too

nice catch! I'll change this

kieranricardo · 2026-05-04T05:54:17Z

@blimlim and @anton-seaice thanks for your comments! Things are ready for another review

anton-seaice · 2026-05-04T06:14:57Z

Could you update the access3-share build please ? ACCESS-NRI/access3-share#28

Their were some gcc build failures that might still need addressing ?

blimlim · 2026-05-04T10:10:50Z

+    ! mapping and merging
+
+    ! In general, first order conservative remapping is used for flux fields
+    ! and bilinear remapping is used for flux fields. Wind stress is a special case


and bilinear remapping is used for flux fields

Should this be for state fields? I guess it's a bit complicated as there's a mix of bilinear, conservative, and patch. Is something like In general, first order conservative remapping is used for flux fields and a mix of bilinear, patch, and conservative remapping are used for state fields depending on the field and mapping direction too vague?

hmm, how about:

! In general, first order conservative remapping is used for flux fields ! while bilinear remapping is applied to state fields. ! Velocity and stress fields are remapped using a higher order patch method. ! Some sea-ice related fields are conservatively remapped with weighting by ice fraction. !---------------------------------------------------------------------

Thanks, that reads well to me :)

kieranricardo · 2026-05-05T00:25:26Z

@anton-seaice here's an access3-share PR with the branch I've been working on: ACCESS-NRI/access3-share#42

kieranricardo · 2026-05-05T02:26:33Z

@anton-seaice gcc issues are fixed now!

anton-seaice

Thanks Kieran

Please write a clear commit description and use a squash commit

blimlim

Thanks @kieranricardo, this is looking good to me

Co-authored-by: Anton Steketee <79179784+anton-seaice@users.noreply.github.com>

Co-authored-by: Spencer Wong <88933912+blimlim@users.noreply.github.com>

…c-cmeps1.1.31-x # Conflicts: # mediator/med_field_info_mod.F90 # mediator/med_methods_mod.F90

kieranricardo requested a review from a team February 2, 2026 23:43

anton-seaice reviewed Feb 3, 2026

View reviewed changes

kieranricardo requested a review from anton-seaice February 16, 2026 03:38

kieranricardo mentioned this pull request Feb 17, 2026

Update to CMEPS with CM3 changes ACCESS-NRI/access3-share#28

Closed

anton-seaice reviewed Feb 18, 2026

View reviewed changes

anton-seaice changed the base branch from rc-cmeps1.1.31-x to cmeps1.1.35-x February 26, 2026 01:20

kieranricardo requested a review from anton-seaice March 11, 2026 04:48

anton-seaice requested changes Mar 13, 2026

View reviewed changes

blimlim reviewed Mar 16, 2026

View reviewed changes

Comment thread mediator/esmFldsExchange_access_mod.F90 Outdated

anton-seaice reviewed Mar 25, 2026

View reviewed changes

Comment thread mediator/esmFldsExchange_accessesm_mod.F90

anton-seaice reviewed Mar 25, 2026

View reviewed changes

kieranricardo requested review from anton-seaice and blimlim April 21, 2026 02:16

blimlim reviewed May 3, 2026

View reviewed changes

kieranricardo requested a review from blimlim May 4, 2026 05:53

blimlim reviewed May 4, 2026

View reviewed changes

kieranricardo mentioned this pull request May 5, 2026

Update to CMEPS with CM3 changes ACCESS-NRI/access3-share#42

Merged

anton-seaice changed the title ~~CM3 change rc cmeps1.1.31 x~~ Add ACCESS-ESM3 May 5, 2026

kieranricardo requested a review from blimlim May 5, 2026 03:35

anton-seaice approved these changes May 5, 2026

View reviewed changes

blimlim approved these changes May 5, 2026

View reviewed changes

kieranricardo and others added 26 commits May 5, 2026 14:53

revert CMakeLists.txt and Makefile

6fd0f89

check ice_frac_cat_ptr > 1e-11 (puny) before division

9c59a4b

update coupled names in post_atm tti

47af1a5

update coupling mode check

3675fa1

Indent changes

56e1a59

Co-authored-by: Anton Steketee <79179784+anton-seaice@users.noreply.github.com>

Delete duplicate rainc and snowc

3a813ae

Co-authored-by: Anton Steketee <79179784+anton-seaice@users.noreply.github.com>

fix comments

50da986

fix indentation

92c5486

rename access->accessesm

583384a

delete fieldcheck

82cc47c

delete unecessary fields going into CICE

3734336

add fieldchecks around re-mapping

4a2c9be

fix syntax errors

684e9e0

fix import error

3a2c243

fix import error

9107e8d

map wind stress to sea-ice and remove atm->ice state fields

1a1a029

rename fields with CMEPs prefixes

d98dbe2

revert Sa_duu10n to So_duu10n

769738b

update comments, and ocn->atm mapping of SST and ocn currents

1944cd7

Update mediator/esmFldsExchange_accessesm_mod.F90

991cf2a

Co-authored-by: Spencer Wong <88933912+blimlim@users.noreply.github.com>

Update mediator/esmFldsExchange_accessesm_mod.F90

0a04469

Co-authored-by: Spencer Wong <88933912+blimlim@users.noreply.github.com>

Update mediator/esmFldsExchange_accessesm_mod.F90

75bdb49

Co-authored-by: Spencer Wong <88933912+blimlim@users.noreply.github.com>

Update mediator/esmFldsExchange_accessesm_mod.F90

103ea3f

Co-authored-by: Spencer Wong <88933912+blimlim@users.noreply.github.com>

comment on mapping methods

cadadfb

comment on mapping methods

81d165b

specify character length in array constructor

5a222d0

kieranricardo force-pushed the cm3-change-rc-cmeps1.1.31-x branch from 246796a to 5a222d0 Compare May 5, 2026 04:54

Merge remote-tracking branch 'origin/cmeps1.1.35-x' into cm3-change-r…

bcccb0a

…c-cmeps1.1.31-x # Conflicts: # mediator/med_field_info_mod.F90 # mediator/med_methods_mod.F90

kieranricardo merged commit e9e1f2f into cmeps1.1.35-x May 5, 2026

kieranricardo deleted the cm3-change-rc-cmeps1.1.31-x branch May 5, 2026 05:04

	S_flds = (/'So_t', 'So_u', 'So_v'/) ! sea_surface_temperature
	S_flds = (/'So_t', 'So_u', 'So_v'/)

	call addfld_from(compice, 'Si_ifrac_n')
	call addfld_to(compocn, 'Si_ifrac_n')

		call addfld_to(compice, 'Faii_taux')
		call addfld_to(compice, 'Faii_tauy')

Conversation

kieranricardo commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description of changes

Specific notes

Uh oh!

anton-seaice left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

anton-seaice Feb 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kieranricardo commented Feb 24, 2026

Uh oh!

anton-seaice commented Feb 26, 2026

Uh oh!

anton-seaice left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

kieranricardo commented Mar 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

anton-seaice left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

kieranricardo commented Feb 2, 2026 •

edited

Loading

anton-seaice Feb 3, 2026 •

edited

Loading

kieranricardo commented Mar 16, 2026 •

edited

Loading

kieranricardo May 4, 2026 •

edited

Loading