What is reported
Paper parameters and prose aggregates
The canonical layer transcribes Table 4 parameters and exact or safely rounded values from the paper prose. Figure-only values remain null.
Reported versus reimplemented
Null stays null. Approximate stays approximate.
| Scenario | Independent duty cycle | Reported duty cycle | Independent collapses | Reported collapses | Comparison |
|---|---|---|---|---|---|
| S1 | 0.631 | not stated in prose | 9.73 | 10.00 | pass |
| S2 | 0.610 | not stated in prose | 5.95 | not stated in prose | pass |
| S3 | 1.000 | 1.000 | 0.00 | 0.00 | pass |
| S4 | 0.382 | 0.381 | 6.63 | 6.60 | pass |
| S5 | 0.990 | 0.990 | 0.50 | not stated in prose | pass |
| S6 | 0.794 | not stated in prose | 4.26 | 4.50 | pass |
| S7 | 0.580 | 0.570 | 7.37 | 7.50 | pass |
| S8 | 0.873 | 0.870 | 1.94 | 2.00 | pass |
| S9 | 0.911 | 0.910 | 1.47 | 1.50 | pass |
| S10 | 1.000 | 1.000 | 0.00 | 0.00 | pass |
What is independent
Equations, step semantics, seeded ensembles
The experiment implements Equations 1–3 from the publication. Hazard tie-breaking, recovery timing, uncertainty distributions, and extinction behavior are explicit independent choices covered by its ADR and tests.
Limits that remain
- This is an independent implementation from the CC BY paper equations, not official author code.
- Year-step ambiguities are fixed by the experiment ADR and remain implementation choices.
- Only prose-reported aggregates are validated; figure-only nulls are not guessed.
- The author-hosted code and aggregate CSV remain link-only because they declare no license.