Equal country weights raised the fitted coral heat slope in the Pacific sample from 2003 to 2010

We tested whether country survey weights change coral heat slopes. Four Atlantic and Pacific samples in 2003-2019 used 5,336 monthly units at repeated sites. Models controlled site, year, calendar month and depth. Equal country weights raised the early Pacific slope from 3.97 to 8.04 bleaching points per four TSA degree-heating weeks. The paired change was +4.07 points. Four-test intervals were +1.54 to +6.60 with country clusters and +1.82 to +6.32 with year clusters. The other three changes had intervals that included zero. Equal country then equal site weights gave an early Pacific change of +3.09, with a country-cluster interval of -0.07 to +6.26. Thus the primary result depended on the visit weight rule and did not repeat across all samples. Uneven coverage, sparse countries, missing reports, few clusters and one reporting system limit inference. The weights do not correct unknown reef selection probabilities. This is a survey-weight audit, not evidence of causal climate damage or species loss. Code, aggregate results, source hashes, checks and a figure are included.

coral bleaching · heat stress · observation bias · survey weights · weighted regression · regional replication