Changing reports below 1% kept fitted heat slopes positive in three of four coral samples in 2003-2019

We tested whether uncertain trace reports can reverse fitted heat links to the presence of coral bleaching. Four Atlantic and Pacific samples in 2003-2019 used 5,336 monthly units. We allowed every report above zero but below 1% bleaching to count as presence or absence. Exact fixed-sample slope bounds stayed positive in three samples. The later Atlantic sample permitted both signs, and one report change could reverse its small positive slope. All four primary country-cluster uncertainty intervals for reported presence included zero. With a 2% trace cutoff all four classification bounds included negative slopes. Same-calendar-month checks kept the three positive 1% bounds, with one near zero. These are classification bounds, not population confidence intervals or causal effects. Missing visits, changing detection, selected complete reports, and few clusters limit inference. The package gives code, aggregate results, a figure, source hashes, and a retrieval recipe. External records stay outside the bundle.

coral bleaching · heat stress · observation bias · classification bounds · trace reports · regional replication