Appendix: P29 prereg, data-source registry,

Update (2026-09-29). Which files are shipped and which are pre-registered. The code that ships with this volume is the engine: atl_bundle. Running validate_all.py gives 39/39 PASS, and every script, result and figure matches SHA256SUMS.txt. The file names given on this page for the prediction tests P1–P35 are pre-registration specifications. That covers the p15_*.py scripts, the *_gate.json results and the DataPack CSVs such as atl_opening_points.csv. Each of them fixes, in advance, what the test will read, how it will decide, and what would make it FAIL. They run once the data are in hand. Until then each prediction is HOLD, which means the data are pending. It is not a failed test and not lost code (see IRREPRODUCIBILITY_LEDGER.md, "Bounded HOLD items").

Appendix: P29 prereg, data-source registry, and HOLD roadmap — This appendix centralizes candidate “anchor datasets” used by modules such as P10/P19/P21/P31/P33/P35. In the formal DataPack, these are registered with the same IDs in config/data_sources.yml, and acquisition/processing must be reproducible under the codebook spec.

This appendix centralizes candidate “anchor datasets” used by modules such as P10/P19/P21/P31/P33/P35. In the formal DataPack, these are registered with the same IDs in config/data_sources.yml , and acquisition/processing must be reproducible under the codebook spec.

version: p29_joint_window_prereg_v0_1
enabled: false

data_sources:
  event_window_estimates: data/meta/event_window_estimates.csv

models:
  joint_window:
    method: "weighted_gaussian"
    width_ka: 2.0

metrics:
  p_value_unlock_max: 0.05
  sensitivity_required: true

outputs:
  json: results/p29_joint_window.json
  log: logs/TEST-P29.log

External data source registry (ERL template)

This appendix centralizes candidate “anchor datasets” used by modules such as P10/P19/P21/P31/P33/P35. In the formal DataPack, these are registered with the same IDs in config/data_sources.yml, and acquisition/processing must be reproducible under the codebook spec.

S1. Minimum principles

(S-PR1) Prefer primary sources (institutional releases / official catalogs).
(S-PR2) Commercial data may be used, but list an open alternative route; if irreproducible, downgrade the evidence rank (ERL).
(S-PR3) For each dataset, record mandatory metadata: (id, version, acquisition date, license, hash).

S2. Core anchor list (summary)

Module	Data anchor (ID)	Use / comment
P10	DS-SED-01	Global marine sediment-thickness grid (ridge→continental-margin profiles).
P1/P2	DS-SLAB-01 / DS-GCMT-01	Subduction-zone geometry (presence/absence) + global moment tensors (stress field).
P19	DS-SLB-01	Sea-level budget closure using altimetry (SSH) + GRACE (mass) + Argo (steric).
P21	DS-PLATE-01	Plate-motion model (long term) + GPS (short term) for deceleration testing.
P12	DS-VOLC-01	Holocene volcano catalog (construction/growth ages require literature linkage).
P31	DS-COAL-01	Coal rank / reflectance databases (global/national).
P33	DS-HF-01	Heat-flow / geothermal-gradient database for thermal covariate control.
P15/P16	DS-ICE-01 / DS-ICE-02	Ice-sheet reconstructions (meltwater scenarios / boundary fluxes).
P35	DS-EXPLORE-01	Exploration maturity (wells/seismic) and sediment-blanket covariates for discovery-bias quantification.

S3. Connection to FAIL conditions

If petroleum distributions/budgets are used as “evidence” in P22/P32, discovery bias must be blocked by P35 (H-DISC) in advance; if blocking fails, downgrade ERL.

HOLD DataPack lock roadmap (ARE→Ice Melt priority)

This appendix fixes a priority order for currently HOLD modules (P8, P9, P10, P12, P14, P17, P22–P23, P25–P28), ranked by direct contribution to the core causal claim ARE→Ice Melt. For each module it sketches required data types / minimum N / representative public repositories (DOI/endpoints) / and automated fetch/normalize scripts. Detailed instructions are synchronized with docs/hold_datapack_roadmap.md and the machine-readable plan config/hold_datapack_plan.yml.

T0. r11 DataPack updates (summary)

P10 (sediment thickness). The GlobSed v2 thickness grid (data/external/globsed/GlobSed_v2.nc) is included in the bundle, and sediment-thickness profiles along 12 fixed transects (data/sediment/atl_sed_thickness_transects.csv) are extracted. However, the crust-age/spreading-rate grid (DS-AGE-01; EarthByte agegrid) cannot be auto-downloaded in the current sandbox environment, so the age–thickness regression gate is not locked. Therefore P10 remains HOLD.
P14 (heat flow). The IHFC Global Heat Flow Database (GHFDB-R21; DOI 10.5880/fidgeo.2021.014) is included in the bundle, and an Atlantic-only extraction data/thermal/heatflow_points.csv is generated. But because kinematic constraints and prereg parameters are not fully fixed/executed, P14 remains HOLD.

T1. Priority (fixed)

Tier 1 (ARE self-constraints; kinematics/geometry/physics): P10 → P14 → P8 → P13 → P12 → P9
Tier 2 (ARE→ice response bridge; indirect): P24 → P25 → P26 → P23
Tier 3 (low directness / controversial / non-core): P17, P22, P27, P28

T2. Per-module minimum requirements (summary)

{

Module	Core data type	Minimum N	Representative repository (DOI/endpoint) + script
Module	Core data type	Minimum N	Representative repository (DOI/endpoint) + script
4r(continued on next page)
P10	sediment-thickness grid + crust-age/spreading-rate grid + transects	transects≥12, drill≥10	`doi:10.25921/xs0z-z845` (sed grid), EarthByte agegrid(`10.1029/2020GC009214`)
			`python code/fetch_hold_datapacks.py –module P10 –download –normalize`
P14	heat-flow points + kinematic constraints	HF≥200, vel≥5	IHFC heat-flow DB + EarthByte rate grid
			`python code/fetch_hold_datapacks.py –module P14 –normalize`
P8	age/rate grids (magnetics-based) + (optional) magnetic profiles	profiles≥20, seg≥3	EarthByte agegrid(`10.1029/2020GC009214`)
			`python code/fetch_hold_datapacks.py –module P8 –download –normalize`
P13	friction/pore-pressure priors (literature)	refs≥5	Byerlee-type priors (DS-FRIC-01)
			`python code/fetch_hold_datapacks.py –module P13 –normalize`
P12	volcano catalog + (partial) construction-age/stage fields	volc≥50	GVP VOTW(`doi:10.5479/si.GVP.VOTW5-2022.5.0`)
			`python code/fetch_hold_datapacks.py –module P12 –download –normalize`
P9	exhumation/high-metamorphism casebook (age/depth)	cases≥10	Example: PANGAEA `doi:10.1594/PANGAEA.969013`
			`python code/fetch_hold_datapacks.py –module P9 –download –normalize`
P24	global endorheic-lake status/time-slice	basins≥100	Oxford LLDB `doi:10.25921/mxae-9596`
			`python code/fetch_hold_datapacks.py –module P24 –download –normalize`
P25	bathymetry-based shelf/canyon proxies	profiles≥30	GEBCO grid(`https://www.gebco.net/...`)
			`python code/fetch_hold_datapacks.py –module P25 –download –normalize`
P26	unconformity/ hiatus catalog (API-based)	sites≥50	Macrostrat API(`doi:10.1029/2018GC007467`)
			`python code/fetch_hold_datapacks.py –module P26 –download –normalize`
P23	refugia catalog + heat-flow covariate	refugia≥20	IHFC heat-flow DB
			`python code/fetch_hold_datapacks.py –module P23 –normalize`

}

T3. Locking execution (recommended)

# (1) Lock from Tier 1
python code/fetch_hold_datapacks.py --module P10 --download --normalize
python code/run_hold_gates.py --module P10

python code/fetch_hold_datapacks.py --module P8  --download --normalize
python code/fetch_hold_datapacks.py --module P14 --normalize

# (2) Bridge modules
python code/fetch_hold_datapacks.py --module P24 --download --normalize
python code/run_hold_gates.py --all

# (3) Final roll-up
python code/generate_pass_hold_fail.py