Recent coral-cover loss gave no clear next-year record link in four Atlantic and Pacific samples in 1999-2018

We tested whether recent coral-cover loss predicted a next-year archive record after controls for current cover. Four Atlantic and Pacific Reef Check samples in HeatCRD covered 1999-2018. Models used 2,543 records and controlled site, country-year, survey month, survey heat and changes in survey month and depth. Slopes ranged from -8.41 to +2.24 percentage points of record probability per ten cover points lost. All four corrected intervals included zero. Both later point estimates were negative, but strict matching changed the late Pacific sign. Only 8-12 countries remained per main model. Selected recurrent sites, uncertain survey identity and few clusters limit inference. Archive absence is not known field absence. The results do not establish random missingness, a causal damage response or a climate-change effect. Code, aggregate evidence, exact source hashes and external-data regeneration instructions are included.

coral cover · observation bias · HeatCRD · marine heatwaves · record availability · regional replication · null result