Forest carbon growth-curve update, other-planted forest pool, and satellite forest ages

[flohump]

🐦 Description of this PR 🐦

This PR replaces MAgPIE's modelled forest growth curves with observation-based ones, adds an "other planted" forest pool matching the FRA 2025 taxonomy, and adds a satellite forest-age dataset — together with the mechanism change that lets a realistic carbon curve coexist with the FRA wood calibration.

1. Observation-based forest growth curves

MAgPIE's forest carbon growth curves are currently a process-model output (Braakhekke et al. 2019). This PR replaces them with observation-based curves, one per forest type — naturally regrowing forest (Robinson et al. 2025, global regrowth data), timber plantations (Bukoski et al. 2022, global plantation-plot database), and other planted forest (new curve, derived from the naturally-regrowing one). They grow faster and more realistically in the early decades, especially plantations and boreal forest; the mature-carbon level (asymptote) is unchanged — only the growth shape. (central figure below) The legacy Braakhekke curves remain selectable via c52_growth_par_source = braakhekke (other-planted forest, which those curves predate, then falls back to naturally regenerating forest). Note this applies the old curve shape inside the new framework — a sensitivity option, not a full revert to develop's behaviour.

2. Other-planted forest as its own pool

The FRA 2025 taxonomy distinguishes naturally regenerating → other planted → plantation forest. This other-planted forest — the planted forest present at model initialisation that is not an intensive plantation — already existed in MAgPIE, but it was not harvestable and was wrongly assigned to the NPI/NDC afforestation category (a policy-afforestation label, not an observed forest type). This PR promotes it to its own harvestable pool with its own growth curve, correctly separated from policy-driven afforestation. Planted forest is now split into Timber (plantations) and Other Planted across the land pools, reporting, and validation.

3. Satellite forest ages (GAMI)

The forest age-class distribution sets how much young vs. mature forest the model starts with — and therefore how much regrowth carbon it can credit. It currently comes from GFAD (Poulter et al. 2019). This PR adds the GAMI global forest-age dataset (Besnard et al. 2024, satellite-derived) as the new default age-class source (c28_ageclass_source); GFAD remains selectable (c28_ageclass_source = gfad). Globally the two carry the same total forest area but a different age shape: GFAD is strongly bimodal (most forest in the youngest 0–10 yr class or the >140 yr old-growth pool), while GAMI shifts area toward the middle age classes. (comparison figure below) Shipped via pik-piam/mrmagpie#66.

4. Separating the wood calibration from the carbon curve

One growth curve currently sets both a forest's carbon and its harvestable wood, and its rate was tuned (k-bisection) so the wood matched the FRA growing-stock (timber) target — bending the carbon curve to fit a wood target. This PR keeps one realistic carbon curve for all forest and meets the FRA wood target separately, with a per-region wood-only multiplier on the harvestable growing stock; carbon is never touched. Across the 12 regions this multiplier stays order-one — 0.5–3.3 for naturally regrowing forest and 0.3–1.6 for plantations — with a single arid outlier (MEA, which has almost no forest: 8.3 and 3.6). It reconciles the realistic-curve growing stock with the regional FRA timber target, not a large or arbitrary rescaling. That makes the old work-arounds obsolete — the k-bisection calibration, the "natural-origin" secondary-forest blend and harvest floor, and the young-secondary-forest harvest loophole.

5. Plantation rotation from an external table

Plantation harvest rotation used to be derived from the growth curve itself, via the current-annual-increment (CAI) criterion — cut at the age where the curve's annual increment peaks. That only worked while a single curve stood in for both a stand's carbon and its merchantable wood. Two reasons it no longer does: (i) the optimal age to cut a plantation is set by the growth of its merchantable wood volume, and total biomass carbon accumulates on a different shape and timescale than merchantable stem volume, so the carbon-increment peak is not the wood-increment peak; and (ii) once wood is calibrated separately from carbon (point 4), the carbon curve carries no wood signal left to optimise a rotation against. Rotation is therefore now an explicit input — a per-Köppen table of plantation rotation lengths (f32_plant_rotation.cs4; ~15 yr tropical, ~30 temperate, ~60 boreal) that sets both the cellular and regional rotation directly, consistent with the FAO-validated ranges reported by Mishra et al. 2021.

6. Removed the minimum carbon-price floor

MAgPIE applied a small floor to the carbon price in every timestep (s56_minimum_cprice, 3.67 USD/tC) to dampen sudden year-to-year jumps in carbon-stock change. As a side effect, that floor suppressed historical deforestation. This PR sets it to 0, giving more realistic historical land-use change. The jump-damping the floor provided is now handled in reporting instead, by the legacy land-clearing reframe in magpie4 (which spreads land-clearing emissions over a decay tail — see the companion PR below). ⚠ This changes calibrated historical land-use change — flagged for review.

Input data: ships as preprocessing rev4.133 (FRA2025 forest data + the new growth curves via additional_data; GAMI ages via mrmagpie).

Validation: compiles clean (codeCheck); 4 test runs (develop vs. this PR × NPi + PkBudg650) solve to 2100. The realistic curves roughly triple the mid-century land-carbon sink under mitigation (PkBudg650, 2050: −5.4 vs −1.7 GtCO2/yr).

Companion PRs:

• Reporting — Forestry reporting + legacy land-clearing emissions reframe pik-piam/magpie4#143
• Validation — Fix and align FRA2025 forest-type and LUC-CO2 validation series pik-piam/mrvalidation#40 (merged)
• GAMI age-class dataset — Add GAMI v2.1 forest age-class dataset option pik-piam/mrmagpie#66

Suggested label: High risk (changes a model default and a core mechanic).

Figures

Comparison, develop vs. this PR:

🔧 Checklist for PR creator 🔧

If a point is not applicable, check the checkbox anyway and write "non-applicable" next to the checkbox.

• Label pull request from the label list.

  • Low risk: Simple bugfixes (missing files, updated documentation, typos) or changes in start or output scripts
  • Medium risk: Uncritical changes in the model core (e.g. moderate modifications in non-default realizations)
  • High risk: Critical changes in model core or default settings (e.g. changing a model default or adjusting a core mechanic in the model)

• Self-review own code

  • No hard coded numbers and cluster/country/region names.
  • The new code doesn't contain declared but unused parameters or variables.
  • magpie4 R library has been updated accordingly and backwards compatible where necessary.
  • scenario_config.csv has been updated accordingly (important if default.cfg has been updated)

• Document changes

  • Add changes to CHANGELOG.md
  • Where relevant, put In-code documentation comments
  • Properly address updates in interfaces in the module documentations
  • run goxygen::goxygen() and verify the modified code is properly documented

• Perform test runs

  • Low risk:
    • Run a compilation check via Rscript start.R --> "compilation check"
  • Medium risk:
    • Run default run via Rscript start.R --> "default"
    • Check logs for errors/warnings
    • Fill in performance section below
  • High risk:
    • Run test runs via Rscript start.R --> "test runs"
    • Check logs for errors/warnings
    • Default run from the PR target branch for comparison
    • Provide relevant comparison plots (land-use, emissions, food prices, land-use intensity,...)
    • Fill in performance section below

• Reporting produces no errors and no new warnings
  3 warnings regarding the piaminterfaces mapping will be fixed after approval and before merging the PR

• Get two approving reviews (at least one from RSE)

📉 Performance 📈

• This PR's default : ~33 mins (NPi / PkBudg650 test runs, coup2100)
• For comparison: runtimes of weekly test

🚨 Checklist for reviewer 🚨

• PR is labeled correctly
• Code changes look reasonable
  • No hard coded numbers and cluster/country/region names.
  • No unnecessary increase in module interfaces
  • model behavior/performance is satisfactory.
• Changes are properly documented
  • CHANGELOG is updated correctly
  • Updates in interfaces have been properly addressed in the module documentations
  • In-code documentation looks appropriate