[jacobkasper]

Base: dev ← compare: bug/i101-catch-output-accounting

---

🌏

Closes #101. Closes #86. Closes #87.

This replaces #88 (fix for #86) and #91 (fix for #87), which I will close in favour of this PR, and adds the other output problems collected in #101. They were found while validating the detailed catch output in #32. That output turned out not to be needed, since OutCATCH.nc already gives catch by species, fishery and box, so #32 will be closed separately. The recreational catch outputs have the same kinds of problems but can't be tested here, so they are in their own issue, #99, and not changed in this PR.

Justification

Atlantis writes the same catch and discards to several outputs: OutCatch.txt, OutCatchPerFishery.txt, OutDiscard.txt, OutDiscardPerFishery.txt, OutCATCH.nc and OutTOTCATCH.nc. Summed over their extra dimensions they should agree, and anyone analysing catch by fishery, or discards from the NetCDF files, depends on that. They didn't. This PR fixes the causes, one commit each:

#	Problem (see #101 for details)	Fix
1	OutCatchPerFishery.txt about 0.15 % low every year: the year's last step is zeroed in bm->Catch before it reaches TotCumCatch (#86)	Add the step still in the buffer to the completed year before the annual report (from #88)
2	NetCDF discards (Tot_<sp>_Discards, <sp>_Discard_FC<n>) inflated, about 1.44 × in Icelantis and 4 × in SETas: box-level discards added once per layer (#87)	Add them once per box (from #91, without its rec-catch part, which moves to #99)
3	OutDiscardPerFishery.txt: each year's last step reported in the following year	Correct the written values; TotCumDiscards itself is unchanged
4	Last row of OutCatchPerFishery.txt repeats the previous year at the end of a run	Report the current year at the end-of-run write, including the step still in the buffer
5	Avg_Catch_Sze_* carry the effort variable's long name	Give them their own long name
6	Catch and discard variables in OutCATCH.nc / OutTOTCATCH.nc are named by group Code, unlike Out.nc and the age-class catch variables, which use the Name	Use the Name, e.g. Tot_FCD_Catch → Tot_Cod_Catch, FCD_Catch_FC7 → Cod_Catch_FC7

Results. dev (4fe7347) against this branch. Each figure is the output summed over boxes and fisheries, divided by OutCatch.txt or OutDiscard.txt:

Output	SETas, dev	SETas, branch	Icelantis, dev	Icelantis, branch
OutCatchPerFishery.txt	0.9986	1	0.9980–0.9988	1
…its last row (run ending mid-year)	repeats previous year	= OutCatch.txt	repeats previous year	= OutCatch.txt
NetCDF discards, water-column group	3.97–4.02	1	1.442–1.447	1
NetCDF discards, epibenthic group	1	1	–	–
OutDiscardPerFishery.txt	0.9986, then 0.9999–1.0001	1	0.9985, then 0.9995–1.0008	1
NetCDF catch	1	1	1	1

SETas: 1,100-day runs. Standard SETas has no discards, so the discard rows come from a run with discarding switched on for two groups (see reviewer instructions). Icelantis: 10-year runs. In Icelantis, Tot_<sp>_Catch also equals the sum of <sp>_Catch_FC<n> box by box (difference 0 t).

Everything else is unchanged. Comparing every output file between dev and the branch:

• Icelantis: 28 of 31 outputs are identical, including Out.nc, the biomass, mortality, SSB, catch, discard, TAC and effort outputs. The three that differ are the intended ones: OutCatchPerFishery.txt, OutDiscardPerFishery.txt and OutHarvestIndx.txt.
• SETas: the same; all NetCDF values are identical apart from the renamed and corrected catch variables.

Things reviewers should know:

• Fix 1 is not purely an output change. The added step also goes into TotCumCatch, CumCatch, TripCatch and TotCumBiCatch, which management, quota, assessment and economics code read. Biomass is unchanged in both test models, but in a model whose catch-based management reads these during the annual block, results could change slightly. That would be a correction: the step was simply lost before. OutHarvestIndx.txt changes for this reason (TotLanding up by about one timestep's catch per year).
• Fix 3 deliberately corrects only the written file. TotCumDiscards is also used by the quota and management code, so changing its year attribution would change management results. That is left for a separate discussion.
• Fix 6 is a breaking change for scripts that read these NetCDF variables by name. It broke my own check script, which now maps Name to Code using the groups CSV.

Types of changes

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Other change (if none of the other choices apply)

Items 1–5 are bug fixes. Item 6 renames output variables.

Further comments

• I've created an issue
• This is NA

AtlantisManual issue: #, to document the renamed variables in OutCATCH.nc and OutTOTCATCH.nc, and that the per-fishery outputs now match the totals.

Reviewer instructions:

Suggested reviewers: a maintainer, plus Andy, who reviewed #88 and #91.

1. Build dev and this branch, with your usual build (Docker or autotools):

git fetch origin
git worktree add ../at_dev origin/dev
git worktree add ../at_i101 origin/bug/i101-catch-output-accounting
for d in ../at_dev ../at_i101; do
  (cd $d/atlantis && aclocal && autoheader && autoconf && automake -a && ./configure && make)
done

2. Run SETas for 1,100 days with each build. Set this in VMPA_setas_run_fishing_F_Trunk.prm in a copy of example/:

tstop     1100 day
toutinc   365 day
toutfinc  365 day

Run with the command line in example/RunAtlantis.sh. Each run takes about a minute.

3. Optional, to test the discard fixes (2 and 3). Standard SETas has no discards. In a second copy of example/, edit VMPA_setas_harvest_F_Trunk.prm:

• FFCDR_FPS: set the 3rd value (midwcFP) to 0.1;
• FFCDR_BFS: set the 4th value (dredgeBFS) to 0.1;
• incidmort_FPS: set the 3rd value to 1.0;
• add an incidmort_BFS line (33 values, 1.0 in 4th place, 0.0 elsewhere).

FPS tests the water-column path and BFS the epibenthic path.

4. Compare with the attached R scripts:

• check_catch_nc.R compares every catch and discard output with OutCatch.txt and OutDiscard.txt by year. For the branch, set groups_file to SETasGroupsDem.csv, since the NetCDF variables now use the group Name.
• compare_opt_levels.R compares two runs file by file. With base_dir = the dev run and other_dirs = the branch run, only OutCatchPerFishery.txt, OutDiscardPerFishery.txt and OutHarvestIndx.txt should differ, plus the discard variables in the two catch NetCDF files.

Expected on the branch:

• all ratios are 1;
• the last row of OutCatchPerFishery.txt equals the final partial year of OutCatch.txt;
• all other outputs are identical to dev.

🤖 Generated with Claude Code

https://claude.ai/code/session_01CcJCYq31xJF2pEZfr66D3J