Common heat-response curvature did not remove the negative variability estimate in Pacific coral surveys from 2010 to 2019

We tested whether a curved common heat response accounts for a negative link between weekly temperature variability and the heat slope of coral bleaching. We used the same 4,625 repeated Reef Check site-month records as a prior study, with site, ocean-year, ocean-month and depth controls. In Pacific surveys from 2010 to 2019, the linear model gave -2.65 bleaching percentage points per four degree-heating weeks for each 0.5 C increase in variability. A model with slope changes at four and eight degree-heating weeks gave -3.02. Its six-test adjusted interval was -7.17 to +1.13, compared with -5.01 to -0.30 for the linear model. The fitted common curvature component shifted the linear estimate toward zero by 0.37 points. Thus this fitted component did not explain away the negative point estimate. All curved-model intervals included zero across six ocean-period samples. Limited heat support, selected surveys, few spatial clusters and unknown coral composition prevent causal or refuge claims. These results concern weekly satellite variability, not daily reef temperature. Code, source hashes, aggregate estimates, a figure and numerical checks are included. External records remain outside the bundle.

coral bleaching · temperature variability · nonlinear heat response · observation bias · model uncertainty · regional replication