Deeper coral reefs experience reduced temperatures and light and are often shielded from localized anthropogenic stressors such as pollution and fishing. The deep reef refugia hypothesis posits that light-dependent stony coral species at deeper depths are buffered from thermal stress and will avoid bleaching-related mass mortalities caused by increasing sea surface temperatures under climate change. This hypothesis has not been tested because data collection on deeper coral reefs is difficult. Here we show that deeper (mesophotic) reefs, 30-75 m depth, in the Caribbean are not refugia because they have lower bleaching threshold temperatures than shallow reefs. Over two thermal stress events, mesophotic reef bleaching was driven by a bleaching threshold that declines 0.26 °C every +10 m depth. Thus, the main premise of the deep reef refugia hypothesis that cooler environments are protective is incorrect; any increase in temperatures above the local mean warmest conditions can lead to thermal stress and bleaching. Thus, relatively cooler temperatures can no longer be considered a de facto refugium for corals and it is likely that many deeper coral reefs are as vulnerable to climate change as shallow water reefs.

The persistence of natural metapopulations may depend on subpopulations that exist at the edges of species ranges, removed from anthropogenic stress. Mesophotic coral ecosystems (30–150 m) are buffered from disturbance by depth and distance, and are potentially massive reservoirs of coral diversity and fecundity; yet we know little about the reproductive capabilities of their constituent species and the potential for these marginal environments to influence patterns of coral reef persistence. We investigated the reproductive performance of the threatened depth-generalist coral Orbicella faveolata over the extent of its vertical range to assess mesophotic contributions to regional larval pools. Over equal habitat area, mesophotic coral populations were found to produce over an order of magnitude more eggs than nearby shallow populations. Positive changes with depth in both population abundance and polyp fecundity contributed to this discrepancy. Relative larval pool contributions of deeper living corals will likely increase as shallow habitats further degrade due to climate change and local habitat degradation. This is a compelling example of the potential for marginal habitat to be critical to metapopulation persistence as reproductive refugia.

Species intolerant of changing climate might avoid extinction within refugia buffered from extreme conditions. Refugia have been observed in the fossil record but are not well documented or understood on ecological time scales. Using a 37-year record from the eastern Pacific across the two most severe El Niño events on record (1982-1983 and 1997 1998) we show how an exceptionally thermally sensitive reef-building hydrocoral, Millepora intricata, twice survived catastrophic bleaching in a deeper-water refuge (> 11 m depth). During both events, M. intricata was extirpated across its range in shallow water, but showed recovery within several years, while two other hydrocorals without deep-water populations were driven to regional extinction. Evidence from the subfossil record in the same area showed shallow-water persistence of abundant M. intricata populations from 5000 years ago, through severe El Niño-Southern Oscillation cycles, suggesting a potential depth refugium on a millennial timescale. Our data confirm the deep refuge hypothesis for corals under thermal stress.