Three-year endpoints missed 19 of 101 windows with at least 30% annual coral-cover loss in selected Great Barrier Reef surveys, 1992–2025

We tested how removing intermediate surveys changes large coral-cover loss flags. We used AIMS LTMP photo-transect data at 9 m. The primary sample had 587 four-visit windows at 62 Great Barrier Reef reefs in 1992–2025. Each window had the same three sites and adjacent date gaps of 0.6–1.4 years. A qualifying annual loss was a relative cover fall of at least 30%, from cover of at least 10%. Annual comparisons flagged 101 windows. Three-year endpoint comparisons missed 19, or 18.8%. The nominal 95% sector-bootstrap interval was 11.1%–25.8%. Resampling sectors and shared three-year blocks gave 6.5%–34.7%. The endpoints also flagged nine windows with no qualifying annual loss. Their total of 91 flags thus concealed disagreement about window identity. The 101 overlapping annual-loss windows contained 45 distinct loss intervals. Two-year endpoints missed seven of 77 annual-loss windows. Planned checks gave three-year missed fractions of 12.7%–24.8%. This audit measures information loss on selected recorded paths. Most windows started before 2003. It does not estimate whole-reef mortality or an optimal survey schedule. The package has code, source hashes, source restoration, all aggregate results, bootstrap draws, and a figure. An independent calculation passed 12,479 checks. An exact repeat matched all 11 outputs. AIMS source values and reef-level derived tables remain in separate retained storage under their original non-commercial terms.

coral cover · survey interval · temporal sampling · Great Barrier Reef · ecological monitoring · reproducibility