Pre-Registration — The reservoir model at its own scale: long-solar-cycle phase → conflict onsets, 1400–1849

Registered: 2026-09-27, before any outcome data was joined to any solar phase. What was done first, and is disclosed here: the solar series was inspected alone to fix the phase method (§2), and the outcome catalog was inspected alone for coverage by century and region (§3). No conflict onset has been assigned a phase. Why this document exists: three registered tests found no eleven-year solar-phase signal in financial-crisis onsets (SOLAR_PHASE_CRISIS_TEST_RESULT_2026-09-17.md, ..._SOVEREIGN_DEFAULT_..._2026-09-17.md, ..._DEFAULTS_INDEPENDENT_SAMPLE_..._2026-09-27.md). The thesis those tests were a narrow proxy for is not eleven-year. It is the reservoir model: solar input builds through a long rising phase and its maximum, societies expand to fill what the peak provides, and the structures raised on peak conditions fail on the descending limb, when the input is withdrawn. The failure window is the descent, and the relevant cycle is the one a demographic mechanism can follow — the Gleissberg (~80–100-year) envelope of solar activity, not the Schwabe cycle. This registers the first direct test of that claim.

Status: RUN 2026-09-27 — H1 REFUTED at this scale on this catalog (rate ratio 0.91, permutation p = 0.75). Result: LONG_SOLAR_CYCLE_CONFLICT_TEST_RESULT_2026-09-27.md. Nothing below was revised. Revisions require a new dated document and both remain public.

---

1. Hypotheses, one direction each

H1 (phase asymmetry). Conflict onsets occur at a higher rate per year during the descending limbs of the Gleissberg envelope than during the ascending limbs. H0. Onset rate does not differ between limbs beyond what the limbs' durations predict.

H2 (reservoir amplitude, secondary and exploratory). Among descending limbs, the onset rate per year rises with the height of the preceding envelope peak: a larger build is a larger release. With three complete descending limbs in scope this is reported descriptively, with no significance claimed either way.

H3 (mediation, secondary). Global mean surface temperature is lower during descending limbs than ascending limbs, the first link of the proposed chain (solar input → climate → yield → strain). Tested on the climate series alone, independent of the outcome catalog.

2. Solar phases — fixed on the solar series alone

Usoskin, Solanki, Krivova et al. (2021), , A&A 649, A141; CDS J/A+A/649/A141, file : annual reconstructed sunspot number, 971–1899, from tree-ring radiocarbon. Independent of every outcome series below in source material and in the people who built it.

Reservoir model at Gleissberg scale — RESULT: refuted on the Brecke catalog, 1400–1849

Registration: PREREG_LONG_SOLAR_CYCLE_CONFLICT_TEST_2026-09-27.md, committed before the script existed, unchanged. Run: 2026-09-27. Script: scripts/research/prereg_long_cycle_conflict_test.py (re-runs from CDS, Georgia Tech and NOAA downloads in one command; raw output beside it). Verdict under §5: H1 REFUTED at this scale on this catalog. Conflict onsets per year are lower on the descending limbs of the Gleissberg envelope than on the ascending limbs (ratio 0.91, bootstrap 95 % [0.83, 1.00], permutation p = 0.75). H2's descriptive prediction fails in direction. H3 finds no temperature difference.

Solar envelope (51-yr mean, prominence ≥ 15): peaks 1124 (79), 1191 (65), 1359 (65), 1593 (41), 1764 (86), 1849 (82); troughs 1165, 1319, 1464, 1681, 1817. Limbs in scope: descending 1400–1464; ascending 1464–1593; descending 1593–1681; ascending 1681–1764; descending 1764–1817; ascending 1817–1849. Descending 205 yr, ascending 244 yr (descending share 0.457).

Events in scope: 2438 conflict onsets (Brecke), of which Europe 986, ≥1,000 fatalities 562.

Row	n	descending	ascending	rate desc /yr	rate asc /yr	ratio	bootstrap 95%	p permutation (PRIMARY)	p binomial	p reverse
All conflicts (PRIMARY)	2433	1053	1380	5.14	5.66	0.91	[0.83, 1.00]	0.752	0.991	0.010
Europe only (regions 3+4)	984	476	508	2.32	2.08	1.12	[0.95, 1.30]	0.373	0.047