Initial-cover limits gave no clear heat link to annual losses of ten coral-cover points in four ocean-period samples in 1998-2019

We tested whether survey-date heat was linked to annual losses of at least ten coral-cover percentage points. HeatCRD version 1 gave 1,115 exact annual survey pairs in the Atlantic and Pacific in 1998-2019. Initial cover was below ten percent in 11-29% of pairs. These pairs could not have a ten-point loss. Four models kept only eligible initial cover and controlled initial heat, initial cover, year, month and source. Heat slopes were +1.42 to +3.43 probability percentage points per four TSA degree-heating weeks. All four corrected intervals included zero. Quadratic initial-cover control, country clusters and loss thresholds of five and twenty points also gave uncertain slopes. Exact follow-up was available for only 3.5-7.8% of baseline cells. Sparse high heat, selected visits, source imbalance and heat measured only near survey dates limit inference. The results do not show that heat causes no coral loss. We test coral habitat cover, not species richness or air pollution. Code, aggregate results, source hashes and checks are included; record tables are rebuilt outside the publication.

coral cover · marine heat · observation bias · initial cover · absolute loss · null result · HeatCRD