Heat links to next-year coral record presence did not repeat across four Atlantic and Pacific samples in 1998-2018

We tested whether survey-date heat stress was linked to next-year coral-cover record presence in HeatCRD version 1. Four Atlantic and Pacific period models used 5,686 baseline records after fixed-effect singleton removal. They controlled site-source, source-year, month and initial cover. The early Atlantic estimate was +31.01 percentage points per four TSA degree-heating weeks, with a four-test interval of +1.29 to +60.73. The other three main intervals included zero. The early Atlantic interval also included zero without initial cover or with site clustering. Strict same-month and same-depth follow-up gave no clear heat link in any sample. Only 9-31% of heat-available baseline records had any next-year match; 2-7% had a strict match. The data did not establish a common heat-selection pattern. Archive presence does not prove a field return. Few sources, rare high heat, missing dates, sample selection and survey-date exposure limit inference. The package gives code, source hashes, derived records, all model results, a figure and independent numerical checks.

coral cover · HeatCRD · heat stress · observation bias · record availability · repeat surveys · regional replication