Surface and column absorption benchmarks of extreme wildfire events — hypotheses, endpoints, and decision rules (Study D2, v1)

Pre-registration of Study D2 (v1, internally frozen 23 July 2026, Asia/Shanghai) before execution of the G2 full multi-event analysis: H1 (primary fingerprint — significantly positive within-station, within-season-window surface BrC absorption log ratios in >=2 independent events), H2 (co-occurring column-layer absorption anomalies at near-field/downwind AERONET sites), H3 (pre-registered negative prediction: no consistent European surface absorption enhancement during the 2023 trans-Atlantic smoke season, frozen window 1 June–31 October 2023), and H0 (publishable opposing outcome — a "surface invisibility" constraint). The document freezes primary/secondary endpoints, event windows, baselines, QC rules, the fixed-sequence gatekeeping chain H1→H2→H3, TOST equivalence bounds (SESOI ±log(1.5)), sign-consistency thresholds (>=70% same-sign, one-sided binomial p<0.05), the >=3/4 robustness-battery criterion, and the deviation policy. Appendix A carries SHA-256 fingerprints (first 16 hex digits) of the ten basis files; Appendix B records the AERONET AOD backfill executed simultaneously with this freeze. The accompanying backfill log is deposited with this record. Public deposit made 2026-08-14; the internal frozen timestamp precedes full-scale execution and is recorded inside the document.