Missing percentages permit both heat slope signs for severe coral bleaching in four Atlantic and Pacific samples in 2003-2019

We tested whether missing bleaching percentages can change the sign of the heat and severe-bleaching slope in archived Reef Check surveys. Four Atlantic and Pacific samples in 2003-2019 contain 7,103 surveys. Models control site, year, month and depth. With reported outcomes fixed, exact bounds permit both slope signs in all four samples. In the three samples with observed severe cases, a negative slope needs at least 10 of 57, 29 of 348, or 24 of 216 missing surveys to be severe, placed at the most influential records. A stricter site-month design retains sign ambiguity in three samples and gives a constant zero outcome in the fourth. At a 10% bleaching threshold, three main lower bounds are positive. These are attainable endpoints for selected archive records, not confidence intervals or causal effects. Surveys absent from the archive, report errors and restrictions across missing outcomes remain outside the test. The package includes code, source provenance, records, endpoint assignments, checks and a figure.

coral bleaching · thermal stress · missing outcomes · observation bias · partial identification · severe events · regional replication