Fesom2.6 recom tra diags and CI test by ogurses · Pull Request #802 · FESOM/fesom2

ogurses · 2025-10-29T14:24:22Z

No description provided.

Define new variables to track tracer changes due to advection and diffusion. We want to save for now diffusion and advection contribution to the tracer changes. Horizontal and vertical diffusion includes Redi parametrization (if it is set .true.).

Fill __ciso directive to ensure that carbon isotope code works. Medusa interface is added.

Use __coccos in src/CMakeLists.txt to include 2 more phytoplankton classes FESOM2.6-REcoM setup with 4 phytoplankton, 3 zooplankton and 2 detritus functional types. BREAKING CHANGE: FESOM2.6-REcoM setup with 4p3z2d

Correct temperature function for small phytoplankton and diatoms. Previous update for the fouth phytoplankton class and new temperature functions for small phytoplankton and diatoms resulted in very low Phaeocystis biomass. We recognised that this is related to missing code lines in the temperature function which we correct with this update.

…ostics Use namelist.io to activate diagnostics. Sinking velocity of particle (fast and slow) are calculated but not written into a file before. We fix the missing diagnostics for sinking velocity of detritus classes. Besides we Extracted the K0, solubility which is computed in gasx.F90.

…esom2.6_recom_tra_diags

We add __coccos directive to get the correct temperature function in silicate assimilation. In the coccos and Phaeocystis version, it should be dependent on the new temperature function, Temp_diatoms. Use the original Arrhenius function in simpler phytoplankton version. Besides, light limitation diagnostics is corrected. It does affect only the diatoms related output.

Correct reported salt flux bug in oce_ale_tracer.F90. #721

We commented out the linear model where degradation is directly proportional to light intensity and the chlorophyll degradation rate is capped at a maximum of 0.3. Reason for this is the unrealistic air2sea co2 fluxes in the 2p3z2d setup. The reactivated saturation model is more realistic. It accounts for light saturation effects. At low light, degradation increases roughly linearly with PAR At high light, the degradation rate approaches an asymptotic maximum

…tives with namelist parameters Replace preprocessor directives with namelist parameters: * `__3Zoo2Det` → `enable_3zoo2det` * `__coccos` → `enable_coccos` * Enhanced inline comments with ecological significance and biological context *Complete variable/parameter documentation with units and descriptions * Clear sectioning with visual separators and variable definitions * Equation explanations showing mathematical relationships and biological processes * Logic documentation for decision trees and conditional blocks * Comments now explain "why" not just "what" **Sequence Validation** * Generates expected tracer ID list based on configuration * Shows exact ID order and position-by-position comparison **Configuration-Specific Clash Detection** * Full Model (both flags): Don't use IDs 1023-1024 * Coccos-only: Use 1023-1028 (NOT 1029-1034) * 3Zoo2Det-only: Use 1029-1030 for microzoo (NOT 1035-1036) **Validation Checks** * Exact sequence matching against expected IDs * Duplicate ID detection *Forbidden ID detection per configuration BREAKING CHANGE: replace directives with namelist parameter for model configuration

ogurses · 2025-10-29T14:35:52Z

CI test for recom is on the way. Any help is welcome @suvarchal

JanStreffing · 2025-10-29T14:38:31Z

! O:G
! Save horizontal and vertical advective fluxes.
! We have the values on the nodes
! We do not know how much each edge contributes
! to the nodes it connects
! Notes from Patrick: del_ttf includes
! Low-order solution. But, del_ttf_advhoriz and
! del_ttf_advvert contain antidiffusive fluxes 
! from the FCT scheme

!if (.FALSE.) then
! O:G - tra_diag
!#if defined (__recom)
!        if (tracers%data(tr_num)%ltra_diag) then
!           do n=1, myDim_nod2D+eDim_nod2D
!              nu1 = ulevels_nod2D(n)
!              nl1 = nlevels_nod2D(n)
!              do nz = nu1, nl1-1
                 ! Horizontal advection part
!                 tracers%work%tra_advhoriz(nz,n,tr_num) = tracers%work%del_ttf_advhoriz(nz,n)
                 ! Vertical advection part
!                 tracers%work%tra_advvert (nz,n,tr_num) = tracers%work%del_ttf_advvert(nz,n)
!              end do
!           end do
!        end if
!#endif
!endif

Still needed @ogurses?

JanStreffing · 2025-10-29T14:55:31Z

src/oce_ale_tracer.F90

+!#if defined (__recom)           ! not necessarily should belong to recom case
+!    tracers%work%tra_advhoriz = 0.0 ! O:G - tra_diag
+!    tracers%work%tra_advvert  = 0.0
+    ttf_rhs_bak = 0.0


Has no impact on CI test, so probably ok. Just flagging to be checked by @patrickscholz

suvarchal · 2025-10-29T14:58:35Z

CI test for recom is on the way. Any help is welcome @suvarchal

last-minute working on compute time reporting and proposal, can take a look tomorrow

Problem: Fortran NetCDF library and CDO/NCO tools use different fill value conventions, causing NaN values in tracer initialization. Solution: Make fill value detection more robust: 1. Try _FillValue attribute (Fortran NetCDF standard) 2. Fall back to missing_value attribute (used by CDO/NCO) 3. Use NetCDF default fill value (9.969e+36) if neither exists 4. Broadcast fill value to all MPI ranks This fixes the 'REcoM_DIC: NaN' error during initialization by properly handling land/missing values in input NetCDF files created with different tools. Fixes issue reported by colleague about incompatible fill value conventions between different NetCDF tools.

Added diagnostic output to track where DIC becomes NaN: - Check state(idic) and sms(idic) before DIC calculation - Check DIC after calculation - Print node, depth, lon/lat location when NaN detected - Stop execution immediately to get clean traceback This will help identify whether: 1. The state already contains NaN (initialization issue) 2. The SMS contains NaN (biogeochemistry issue) 3. The addition creates NaN (numerical issue) The error occurs at lon=-43.88, lat=14.55 consistently, so this should pinpoint the exact cause.

Added diagnostic checks at the point where DIC is read from state array and converted for CO2 flux calculation. Since the SMS debug didn't trigger, the NaN must appear between SMS calculation and forcing. This will catch: 1. If state(one,idic) is already NaN when forcing starts 2. If the conversion to REcoM_DIC creates NaN Will print node number, coordinates, and related state values to identify the exact source of NaN.

JanStreffing · 2025-11-12T22:31:10Z

A couple of interesting changes so far. Some of the recom code was missing allocations, that triggered GNU errors. I am left with NANs in one init field that I failed to resolve.

@ogurses, @suvarchal: Can you check this out?

When ballasting is used, vertical sinking is adjusted if the detritus carbon biomass exceeds a treshold value (0.001). This results in large air-sea co2flux uptake. Weremove this constraint.

JanStreffing · 2026-02-25T08:30:30Z

@ogurses, @suvarchal, @patrickscholz
I think we should try to resolve the last issues on the CI test and merge this branch. I'm at AWI to join next week Monday to Wednesday. After that, I'm away for some weeks.

patrickscholz · 2026-02-25T10:06:38Z

src/oce_ale_tracer.F90

+    if (tracers%data(tr_num)%ltra_diag) then
+          do n=1, myDim_nod2D+eDim_nod2D
+          nu1 = ulevels_nod2D(n)
+          nl1 = nlevels_nod2D(n)
+          do nz = nu1, nl1-1
+             ttf_rhs_bak(nz,n) = del_ttf(nz,n)
+          end do
+       end do
+    end if


it would be nice if we could encapsulate this in a subroutine not to make our code messy and unreadable! Especially that this pieces repeat itself

patrickscholz · 2026-02-25T10:07:51Z

src/oce_ale_tracer.F90

+    if (tracers%data(tr_num)%ltra_diag) then
+       do n=1, myDim_nod2D+eDim_nod2D
+          nu1 = ulevels_nod2D(n)
+          nl1 = nlevels_nod2D(n)
+          do nz = nu1, nl1-1
+             ! horizontal diffusion (w/out Redi)           
+             tracers%work%tra_diff_part_hor_redi(nz,n,tr_num) = (del_ttf(nz,n) - ttf_rhs_bak(nz,n)) / hnode_new(nz,n) ! Unit [Conc]
+             !if (mype==0)  print *, tracers%work%tra_diff_part_hor_redi(nz,n,tr_num)
+          end do
+       end do
+    end if


Here that same this, this structure repeats itself and could be a re-callable subroutine

patrickscholz · 2026-02-25T10:18:27Z

I talked a while a ago with Sergey about this, and we should try to encapsulate as much of RECOM as we can, To achieve that as much of RECOM lives within the recom_...F90 code files and is only called by function and subroutines (e.g boundary conditions, initialisation, ...). We have been a bit sloppy about this with all the iceberg and transient tracer stuff in the past, which made large part of the FESOM code especially the initialisation, tracer and boundary condition part very crowded and messy and more difficult to maintain.

JanStreffing · 2026-02-25T10:25:36Z

Perfect task for the AI. I had it search for more cases:

Similar repeated-structure cases in REcoM code (candidates for future refactor)

src/int_recom/recom_main.F90
- Pattern: In the main node loop (do n=1,myDim_nod2D) there is a recurring
  "backup state -> do work -> compute diagnostic delta" idiom.
- Example: ttf_rhs_bak is filled for all tracers with ltra_diag, and later
  tracers%work%tra_recom_sms(...) is computed as values - ttf_rhs_bak.
- Why it’s similar: conceptually the same as the backup/delta calculation
  refactored in src/oce_ale_tracer.F90 (just organized differently).
src/int_recom/recom_sinking.F90
- Pattern: long sequences of if(tracer ID == ...) blocks repeat nearly the same
  update structure for different element pools:
  - update Benthos(n,k)
  - optionally update SinkFlx_tr(n,k,tr_num) when use_MEDUSA
  - optionally update Benthos_tr(n,k,tr_num)
    repeated for N/C/Si/CaCO3 and also isotope variants.
- Why it’s similar: repetitive boilerplate where only the tracer-ID set and the
  target index (k) changes.
src/int_recom/recom_init.F90
- Pattern: repeated min/max scanning over nodes for multiple REcoM tracers,
  followed by repeated MPI_Allreduce calls with identical structure.
- Why it’s similar: repeated blocks of the same operation on different tracer IDs.
src/int_recom/recom_sms.F90 (benthic remineralization section)
- Pattern: repeated update equations with the same shape:
  decayBenthos(i) = rate(i) * LocBenthos(i)
  LocBenthos(i) = LocBenthos(i) - decayBenthos(i) * dt_b
  repeated across elements and (optionally) isotopes.
- Why it’s similar: identical structure differing only by indices/parameters.

Recom refactoring dedup

JanStreffing · 2026-03-03T13:53:15Z

I tried my hands on this, but I was not able to figure out what is wrong with the DIC_PI file. Could your try this out @ogurses ?

ogurses and others added 17 commits October 15, 2024 15:08

add recom_sms diag

3e84a88

feat(recom): add ciso and medusa components

434f0e6

Fill __ciso directive to ensure that carbon isotope code works. Medusa interface is added.

input to bc_surface from medusa added and Sed_input deleted

0864237

fix(recom) Add missing directives

eafe8e9

two small fixes

4408f0e

feat(recom)!: FESOM2.6-REcoM setup with 4p3z2d

db00bd1

Use __coccos in src/CMakeLists.txt to include 2 more phytoplankton classes FESOM2.6-REcoM setup with 4 phytoplankton, 3 zooplankton and 2 detritus functional types. BREAKING CHANGE: FESOM2.6-REcoM setup with 4p3z2d

Merge remote-tracking branch 'origin/fesom2.6_recom_tra_diags' into f…

2836492

…esom2.6_recom_tra_diags

fix(recom) clean merge conflicts

66a5aeb

fix(recom): Correct Light limitation diagnostics(continued)

4258874

fix(recom): Correct namelist.recom files in config folder

d82dbcb

fix(fesom): Bugfix: sign of real salt flux

502d4d6

Correct reported salt flux bug in oce_ale_tracer.F90. #721

JanStreffing requested review from JanStreffing and patrickscholz October 29, 2025 14:25

JanStreffing assigned ogurses Oct 29, 2025

JanStreffing added the RECOM label Oct 29, 2025

JanStreffing added this to the FESOM 2.7 milestone Oct 29, 2025

JanStreffing changed the title ~~Fesom2.6 recom tra diags~~ Fesom2.6 recom tra diags and CI test Oct 29, 2025

JanStreffing linked an issue Oct 29, 2025 that may be closed by this pull request

Create REcoM CI test #620

Open

ogurses requested a review from suvarchal October 29, 2025 14:34

Merge branch 'main' into fesom2.6_recom_tra_diags

80bbbe7

JanStreffing reviewed Oct 29, 2025

View reviewed changes

JanStreffing added 3 commits November 12, 2025 19:25

JanStreffing force-pushed the fesom2.6_recom_tra_diags branch from 619ee7c to c6c21e0 Compare November 12, 2025 22:28

JanStreffing modified the milestones: FESOM 2.7, FESOM 2.7.1 Nov 17, 2025

ogurses and others added 3 commits December 1, 2025 10:49

fix(recom): Bugfixes related to ballasting

2a1ed34

When ballasting is used, vertical sinking is adjusted if the detritus carbon biomass exceeds a treshold value (0.001). This results in large air-sea co2flux uptake. Weremove this constraint.

Merge branch 'main' into fesom2.6_recom_tra_diags

8f94ed3

overwrite Sweeney path for recom ci test

075f709

JanStreffing force-pushed the fesom2.6_recom_tra_diags branch from e5e0fe2 to 075f709 Compare January 26, 2026 16:00

JanStreffing modified the milestones: FESOM 2.7.1, FESOM 2.8 Feb 18, 2026

Merge branch 'main' into fesom2.6_recom_tra_diags

e296ae8

Merge branch 'main' into fesom2.6_recom_tra_diags

87fbce9

patrickscholz reviewed Feb 25, 2026

View reviewed changes

Refactor tracer diag loops

f56ee08

JanStreffing added 4 commits February 25, 2026 11:30

REcoM: deduplicate ltra_diag backup/delta

b1e09a8

REcoM: deduplicate benthos accumulation

cd2928f

REcoM: deduplicate init sanity-check reductions

f179d3c

REcoM: deduplicate benthos decay updates

c9287bc

This was referenced Feb 25, 2026

Recom refactoring dedup #863

Merged

Fesom2.7 recom3.1 #864

Draft

patrickscholz approved these changes Feb 27, 2026

View reviewed changes

Merge pull request #863 from FESOM/recom_refactoring_dedup

6056444

Recom refactoring dedup

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fesom2.6 recom tra diags and CI test#802

Fesom2.6 recom tra diags and CI test#802
ogurses wants to merge 68 commits intomainfrom
fesom2.6_recom_tra_diags

ogurses commented Oct 29, 2025

Uh oh!

ogurses commented Oct 29, 2025

Uh oh!

JanStreffing commented Oct 29, 2025

Uh oh!

JanStreffing Oct 29, 2025

Uh oh!

suvarchal commented Oct 29, 2025

Uh oh!

JanStreffing commented Nov 12, 2025

Uh oh!

JanStreffing commented Feb 25, 2026

Uh oh!

patrickscholz Feb 25, 2026

Uh oh!

patrickscholz Feb 25, 2026

Uh oh!

patrickscholz commented Feb 25, 2026

Uh oh!

JanStreffing commented Feb 25, 2026

Uh oh!

JanStreffing commented Mar 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

ogurses commented Oct 29, 2025

Uh oh!

ogurses commented Oct 29, 2025

Uh oh!

JanStreffing commented Oct 29, 2025

Uh oh!

JanStreffing Oct 29, 2025

Choose a reason for hiding this comment

Uh oh!

suvarchal commented Oct 29, 2025

Uh oh!

JanStreffing commented Nov 12, 2025

Uh oh!

JanStreffing commented Feb 25, 2026

Uh oh!

patrickscholz Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

patrickscholz Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

patrickscholz commented Feb 25, 2026

Uh oh!

JanStreffing commented Feb 25, 2026

Uh oh!

JanStreffing commented Mar 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants