Previous bleaching gave no clear gain in later-year coral estimates in four Atlantic and Pacific samples, 2010-2019

We tested whether previous-year bleaching improved estimates of current coral bleaching beyond heat exposure, depth, season, calendar trend and site identity. We used Reef Check records from BCO-DMO 773466 version 2. Exact site and calendar-month matches in consecutive years gave 1,287 pairs. Earlier-year training gave 581 test predictions in 2010-2019. Adding previous bleaching changed mean absolute error by +0.098 and +0.192 percentage points in two Atlantic periods, and -0.093 and -0.281 points in two Pacific periods. All four adjusted regional bootstrap intervals included zero. Planned model checks also gave no clear repeated gain. A later context check found lower observed absolute error from always estimating zero in all four samples. Many zero values, few Atlantic records, selective return visits, coarse surface heat and few region groups limit interpretation. The results concern retrospective prediction in selected known sites. They do not identify adaptation, ecological memory or species loss. We provide source hashes, derived records, chronological model code, predictions, figures and numerical checks.

coral bleaching · marine heat stress · ecological memory · chronological validation · observation bias · Reef Check