Pre-registered predictions P1 to P4

Pre-registered predictions P1 to P4 — This section is designed for adjudication, not narrative. Each prediction module explicitly contains (A) observable signature, (B) data, (C) test, and (D) falsifier (LOCK → Derive → Gate). Module IDs, file paths, and PASS/HOLD/FAIL semantics follow the Korean final (r16).

This section is designed for adjudication , not narrative. Each prediction module explicitly contains (A) observable signature , (B) data , (C) test , and (D) falsifier (LOCK → Derive → Gate).

This section is designed for adjudication, not narrative. Each prediction module explicitly contains (A) observable signature, (B) data, (C) test, and (D) falsifier. Module IDs, file paths, and PASS/HOLD/FAIL semantics follow the Korean final (r16).

P1: Atlantic-rim boundary character (subduction vs passive margins)

Bundle verdict (2025-12-27): PASS. (results/p1_plate_boundaries_summary.json)

Observable signature. Around the Atlantic rim, “passive margins + ridge” signatures must dominate over “persistent subduction” signatures.

Data (minimum). (e.g.) trench/subduction distributions, margin structural maps, ridge/transform patterns, and sedimentary-basin development patterns.

Test (example). Sample the perimeter at fixed spacing and estimate the “subduction-dominant segment fraction.” Pre-registered criterion: e.g., the “subduction-dominant fraction ” must be satisfied to UNLOCK.

FALSIFIER. If subduction is widespread and persistently dominant around the Atlantic rim, P1 is FAIL.

Recommended quantitative metrics.

Pre-registered thresholds.

Note: numerical values may be modified only before release; after release they are fixed.

Hard-gate (TEST-P1) outputs (recommended).

r17 computed result. Using literature-derived segment lengths, the only active subduction along the Atlantic rim is the Lesser Antilles arc (~850 km; Bouysse & Westercamp 1990; Christeson et al. 2003) and the South Sandwich trench (965 km; Five Deeps multibeam 2019), totalling ~1,815 km — only ~3.6% of the global subduction length (~51,000 km; Bird 2003). Against an Atlantic-rim perimeter of ~32,000 km (two passive margins, each ≳ the ~16,000 km Mid-Atlantic Ridge), R_sub≈0.057; even a deliberately worst-case choice (maximal subduction including a contested Gibraltar segment, minimal perimeter) gives R_sub≈0.13. Both are well below the UNLOCK threshold 0.25, so P1 = UNLOCK (PASS), robustly: the Atlantic is a passive-margin ocean with subduction confined to two short arcs.

P2: Spatial concentration of deformation/acceleration (near the void edge)

Observable signature. Maximum deformation must be concentrated near the void edge (continental shelf / major fault zones).

Data (minimum). Deformation indicators (faults, shear zones, fracture patterns) and maps of deformation localization (seismology/gravity/structure).

Test (example). Quantify whether deformation indicators decrease with distance r from the void edge (e.g., an r-regression). If there is no spatial correlation or the sign is opposite, FAIL.

Recommended quantitative metrics. Define a deformation intensity D(r) as a function of distance r from the void edge (e.g., fault density, shear-zone frequency, cumulative seismic moment).

Pre-registered thresholds.

Hard-gate (TEST-P2) outputs (recommended).

P3: RSL asymmetry (Pacific oscillation vs Atlantic filling)

Observable signature. Even after corrections, the Pacific should retain an overshoot/oscillatory component, whereas the Atlantic/Mediterranean should be dominated by a relatively monotonic-rise component.

Data (minimum). RSL time series (multiple sites per basin) and correction metadata (GIA/crustal motion).

Test (spec).

Recommended control expansion: include the Indian Ocean (INDIAN). If P3's power depends only on a Pacific vs Atlantic/Mediterranean dichotomy, it is hard to judge reproducibility in other basins. Thus we recommend labeling the Indian Ocean separately as basin_type=INDIAN (mixed/rupture-type hypothesis), and adjudicating whether its pattern is closer to PACIFIC or ATLANTIC by the same feature-distance / separability metrics. For implementation, add basin_type to the DataPack schema (Appendix F). Recommended reporting format: (Group A) PACIFIC vs (Group B) ATLANTIC+INDIAN and (Group C) INDIAN alone.

FALSIFIER. If corrected records converge to the same shape in the two basins, P3 is FAIL or HOLD.

Additional recommendation: lock the “phase/relaxation metric” down to file format. RSL adjudication must be numeric, not “graph impression.” Therefore define features so the code computes them directly. (e.g.) t₅: time after the event until RSL “stabilizes” to the 5% level of the initial response (or another prereg reference), as a phase/relaxation metric. If the t₅ definition changes at release, treat it as a version change.

Example prereg (a form that can be borrowed from internal templates).

Note: values above are examples; the principle is pre-registration lock itself.

P4: Reduced effective stress / overpressured fluids / lubrication (low-friction) signatures

Bundle verdict (2025-12-27): PASS. (results/p4_les_summary.json)

Observable signature. In this model, “lubrication” is not just “a thin water film” but the structural/mineral/microstructural signatures of reduced effective stress (hydraulic jacking) + fluid involvement. Accordingly, P4 may appear in one of two branches (or a mixture):

Data (minimum). Petrography (thin section), mineral composition/alteration indicators, shear-zone maps plus fracture/injection-vein network maps, fluid inclusions/porosity observations, brecciation distributions.

Test (example). Define a checklist of “lubrication/overpressure signature indicators”, and pre-register the explanatory comparison versus alternatives (dry friction) and other low-friction mechanisms (dynamic separation, grain-fluidization, etc.).

FALSIFIER. If (1) there is no overpressure/fluid-involvement signature (negating P4-a), (2) no melting signature either (negating P4-b), and (3) dry friction signatures suffice as a simpler/better explanation, then P4 is FAIL. That is, to prevent P4 from remaining HOLD forever under “the signature might be missing,” we specify required signatures (L3 or L5 below).

Recommended quantitative metric: LES+ (Lubrication/Overpressure Evidence Score). To avoid reducing P4 to a binary “exists/doesn't exist,” score the following five items as 0/1 (LES⁺∈[0,5]):

Pre-registered thresholds.

Hard-gate (TEST-P4) outputs (recommended).