Missing year-five records permit minority and majority coral recovery in four Atlantic and Pacific loss cohorts from 1998 to 2013

We tested whether missing follow-up records limit estimates of five-year coral cover recovery. HeatCRD version 1 gave 348 distinct Atlantic and Pacific sites with an annual cover loss of at least 10 percentage points in 1998-2013. Only 116 had year-five cover. Under a 0-100% range for missing cover, all four ocean and period groups permit both a minority and a majority to reach 90% of prior cover. The later-minus-early recovery difference permits either sign in each ocean. Records after the missing target year occur at 56 of 232 sites with missing year-five cover; source endpoints alone cannot explain all gaps. Outside Reef Check, two early subgroups imply minority recovery. These are exact finite-cohort bounds, not confidence intervals or heat effects. Survey timing, changing methods, selected losses and few high-heat events limit inference. Code, aggregate evidence, provenance, a figure and an independent reconstruction are included. External records remain outside the bundle.

coral recovery · observation bias · missing follow-up · partial identification · marine heat · HeatCRD