Underwater volcanic vents emit CO₂ (along with other trace gases, e.g., N₂, O₂, CH₄, Figure 2) into the adjacent seawater, reducing the Ω_arag and creating seawater conditions similar to those predicted to occur under ocean acidification. Local hydrology and variations in gas release can create large diel oscillations in pCO₂ at vent sites relative to adjacent control sites (Enochs et al., 2015). Even so, vent sites offer a unique natural environment to study ecosystem changes in response to frequent, low pH (Hall-Spencer et al., 2008). In the Mediterranean, the release of CO₂ at vent sites has been shown to alter a variety of ecosystems, including seagrass beds (Martin et al., 2008), rocky-reef systems (Hall-Spencer et al., 2008), and vermetid reefs (Milazzo et al., 2014). These temperate studies all report adverse effects of elevated pCO₂, including loss of diversity of marine calcifiers (Martin et al., 2008; Kroeker et al., 2011), reduced calcification (Milazzo et al., 2014), shifts from calcareous dominated systems to a state lacking scleractinian corals (Hall-Spencer et al., 2008) and reductions in trophic groups (Kroeker et al., 2011).

To date, there have only been three regions where vent systems have been described adjacent to coral reefs: Iwotorishima Island, Japan (Inoue et al., 2013), Papua New Guinea (PNG; Fabricius et al., 2011) and Maug, in the Commonwealth of the Northern Mariana Islands (Enochs et al., 2015). Findings from these vent sites have been variable, likely due to site-specific pH gradients, hydrology and conditions such as nutrient availability (Enochs et al., 2015). For example, in Iwotorishima, communities shifted from hard coral to soft coral as seawater acidified, with all corals absent from the highest pCO₂ zone (1,465 μatm, Inoue et al., 2013). In contrast, coral dominated sites at Maug (mean pH = 8.04 ± 0.016, min pH = 7.98) transitioned to macroalgal dominance (mean pH = 7.94 ± 0.051, min pH 7.72).

A common feature across the various vent sites are changes in community structure, where only a few acidification-tolerant species thrive under high pCO₂ (e.g., Hall-Spencer et al., 2008; Fabricius et al., 2011, 2015; Kroeker et al., 2011, 2013). There is also evidence that the structural integrity of any reef framework is compromised due to a loss of architectural complexity (Fabricius et al., 2011), increased rates of bioerosion (Enochs et al., 2016a,b), reductions in coral growth (Fabricius et al., 2011) and reduced skeletal density (Fantazzini et al., 2015; Strahl et al., 2015b).

The ability of corals around vent sites to maintain calcification under elevated pCO₂ appears to be species-specific (Rodolfo-Metalpa et al., 2011; Inoue et al., 2013). Both in laboratory work (Venn et al., 2011; McCulloch et al., 2012) and field investigations (Wall et al., 2016) species-specific abilities to buffer internal-pH have been recorded. There do appear to be defined physiological limits under which calcification can occur (McCulloch et al., 2012; Wall et al., 2016), and physiological plasticity appears important in supporting calcification at high pCO_2. For example, increased rates of calcification (Rodolfo-Metalpa et al., 2011), or the utilization of elevated external dissolved inorganic carbon (DIC) can fuel photosynthesis (Inoue et al., 2013; Strahl et al., 2015a) and offset the higher energetic cost of calcification under acidification. For soft corals (Inoue et al., 2013), sea anemones (Suggett et al., 2012a), and scleractinian corals (Strahl et al., 2015a), success at high pCO₂ seems to be driven by their capacity to enhance, or at least sustain, photosynthesis under elevated pCO₂. Other physiological processes such as higher cell protective capacities, total lipid content, tissue biomass (Strahl et al., 2015b) and changes in fatty acid metabolism (Kenkel et al., 2017) may also promote resistance to environmental stress at low pH.

Alongside physiological plasticity, live-tissue coverage (Rodolfo-Metalpa et al., 2011), and changes in microbial taxa (Morrow et al., 2014) can support species survival under elevated pCO₂. Variable changes in gene expression of Symbiodinium at the PNG vent system suggest population-specific responses to low pH (Kenkel et al., 2017). The collective response of the animal host, symbiotic algae and other microbial taxa demonstrate that fully considering the holobiont (i.e., the coral host, algal symbionts and microbiome) is important in determining success and survival under high pCO₂.

Underwater seeps (ojos) of the Yucatan Peninsula (Mexico) have physicochemical conditions entirely unique to the local area. Here, water that has associated with soil and limestone mixes with high CO₂ groundwater resulting in seawater that has a low pH (6.70–7.30), low Ω_arag (0.3–0.97 at the center of the seep), and high DIC, A_T, Ca²⁺ and nutrients (Crook et al., 2012). Although the combined physicochemical conditions differ from potential future ocean acidification scenarios and vent environments (e.g., low pH accompanied by high DIC, A_T, and Ca²⁺), the site provides insight into how coral diversity and growth are impacted when corals are exposed to low pH for an extended period of time [high CO₂ water has discharged for millennia, (Beddows et al., 2002)]. Crook et al. (2012) found that only three coral species (Porites astreoides, P. divaricata, and Siderastrea radians) persisted close to the ojos where the Ω_arag was <2.5. Further analysis of the growth rates of P. astreoides showed no change in linear extension rates, but a fall in skeletal density (ca. 40%) and calcification (ca. 30%) when Ω_arag was below 2.0, relative to an adjacent control site where Ω_arag was >3.5 (Crook et al., 2013). The ojos sites can experience changes in temperature, salinity and light that may in part explain the reduced calcification rates (Iglesias-Prieto et al., 2014), although the collective assessment from the laboratory and field experimentation suggests low pH is largely responsible for the reductions in calcification measured (Paytan et al., 2014). To date for this site, no molecular data or information on other life-history traits (e.g., metabolic properties) have been published, making it difficult to assess whether this reduction in density does indeed demonstrate a lack of acclimation ability for P. astreoides (Crook et al., 2013) or rather highlights a potential survival process to persist in adverse conditions (e.g., phenotypic or metabolic plasticity, see Pigliucci et al., 2006). Either way, these findings corroborate results from the vent sites that coral calcification will likely be compromised under low pH.

Elevated temperatures become particularly pronounced as reefs shallow and flushing becomes restricted. One key system that has been extensively studied since the early 2000s is the back reef tide pools on Ofu Island, American Samoa. The back reef pools harbor diverse coral communities (Craig et al., 2001) and feature distinct thermal regimes, governed by the thermal dynamics of pools of different sizes (Smith et al., 2007; Oliver and Palumbi, 2011b; Koweek et al., 2015); specifically, smaller high-variance (HV) pools where SST can briefly reach 35°C during strong noontime low tides (exceeding the local critical bleaching temperature of 30°C) and fluctuate by up to 6°C daily, vs. larger moderately variable (MV) pools, where SSTs rarely exceed 32°C and exhibit a more stable thermal environment.

Corals within the HV pool appear to have acquired putative heat tolerance through genetic, physiological, and ecological adaptations. For example, phenotypic plasticity of skeletal characteristics (Smith et al., 2007) that could potentially be linked to optimization of skeletal optical properties (Enríquez et al., 2017), a switch in symbiont type that is considered more heat-tolerant (Smith et al., 2008; Oliver and Palumbi, 2009, 2011a,b), enhanced thermal tolerance when exposed to heat stress (Oliver and Palumbi, 2011b; Palumbi et al., 2014), higher transcription of heat responsive genes than MV corals (Barshis et al., 2013) and a reservoir of alleles pre-adapted to high temperatures (Bay and Palumbi, 2014). Changes in the microbiome have also been demonstrated with bacterial communities changing in response to transplantation between the MV and HV pools (Ziegler et al., 2017). Higher heat tolerance of HV corals was found to be the result of both local acclimatization and fixed effects such as adaptation, with acclimatization in <2 years achieving the same heat tolerance that would be expected from strong natural selection over multiple generations (Palumbi et al., 2014). This unique system has thus provided critical insights into the mechanisms underlying coral heat tolerance, particularly of host-based contributions to holobiont heat tolerance. Nevertheless, it is currently unclear whether even naturally heat tolerant corals can continue to improve their heat tolerance fast enough to keep pace with global warming, with locations within American Samoa recently experiencing severe bleaching (Eakin et al., 2016).

Corals from high-latitude environments are functionally different to tropical corals, having traits that may facilitate their survival under climate change (Beger et al., 2014). High-latitude, cooler water (sub-tropical) coral environments have received increasing attention as potential refugia from climate change and ocean warming for tropical coral reef organisms (Beger et al., 2014; Muir et al., 2015b). They currently exist at the environmental limits for growth, in terms of temperature, salinity and Ω_arag, where conditions are also highly seasonally variable (Kleypas et al., 1999; Guinotte et al., 2003). Well-studied locations include: Lord Howe Island and the Solitary Islands (30–32°S), Sydney Harbour (34°S) in eastern Australia, the Abrolhos Islands (28°S) and Rottnest Island (32°S) in Western Australia, Sodwana Bay in South Africa (28°S), several islands in temperate Japan (32–34°N), and Bermuda (32°N). Due to the marginal environmental conditions (particularly at more extreme high-latitudes), they can be considered natural extreme coral reef environments which differ from their tropical counterparts in many aspects. For example, the unique diversity of high-latitude reefs is due to the overlap of tropical and temperate species ranges, species that are rare or absent at lower latitudes and species endemic to these environments (Beger et al., 2014 and references therein). Species temporal turnover can be high depending on larval supply and recruitment from lower latitudes and fluctuations in environmental conditions. Whilst coral cover (Harriott et al., 1994; Thomson and Frisch, 2010; Denis et al., 2013) and even coral growth rates can be high (e.g., Ross et al., 2015), reef accretion and development are nevertheless often limited, with corals sometimes only forming living veneers on rocky substrate (Beger et al., 2014 and references therein). Furthermore, future reef accretion could become increasingly challenged with the percent changes in Ω_arag predicted to be greater at high-latitude reefs relative to their tropical counterparts (Hooidonk et al., 2014).

Greater fluctuations in environmental conditions at high-latitude coral reefs have been hypothesized to increase the environmental tolerance of high-latitude species as observed in tropical corals (Oliver and Palumbi, 2011b; Schoepf et al., 2015), thus potentially increasing their ability to cope with climate change. In high-latitude corals, phenotypic plasticity may support their survival, through diverse coral symbiont communities (Wicks et al., 2010b), enhanced symbiont tolerance to extreme (low) temperatures (Wicks et al., 2010a), enhanced heterotrophic plasticity (Bessell-Browne et al., 2014) and evidence of shifted thermal optima for calcification at cooler temperatures (Ross et al., 2015). High-latitude coral communities around the world have also been found to reproduce sexually and are therefore not solely dependent on recruitment from tropical region coral stocks (e.g., Babcock et al., 1994; van Woesik, 1995; Wilson and Harrison, 2003; Miller and Ayre, 2004; Madsen et al., 2014).

As predicted, corals have already extended their latitudinal range pole-wards in several locations such as Japan and eastern Australia (Booth et al., 2007; Yamano et al., 2011; Baird et al., 2012), although range expansion can be accompanied by reduced symbiont diversity (Grupstra et al., 2017). The ability and rate of this range expansion will be key in determining the ability of high-latitude locations to act as refugia. Furthermore, high-latitudinal corals themselves are increasingly threatened by climate change and ocean warming, as bleaching events have also occurred in many high-latitude locations in recent years (Celliers and Schleyer, 2002; Harrison et al., 2011; Thomson et al., 2011; Abdo et al., 2012; Smale and Wernberg, 2012). In addition, several other factors specific to high-latitude coral reefs, such as latitudinal light attenuation (Muir et al., 2015a), geographic isolation and small population sizes combined with greater predicted ocean warming and generally lower Ω_arag of cooler waters, could make high-latitude coral communities particularly susceptible to climate change (Beger et al., 2014). It is therefore currently unclear whether and to what extent high-latitude corals can in fact serve as refugia from continued climate change, despite evidence for some adaptations to their marginal, more extreme environments.

Reef systems that extend across broad latitudes and temperature gradients have provided evidence of juvenile (Woolsey et al., 2015) and adult corals with enhanced thermal tolerance (Ulstrup et al., 2006; van Oppen et al., 2011; Howells et al., 2012, 2013). On the Great Barrier Reef (GBR), mean summer maximum temperatures have been reported to have nearly 2°C difference at reefs 250 km apart (South Molle Island to Magnetic Island; Howells et al., 2013). Different thermal histories on the GBR have been found to influence bleaching susceptibility (Ulstrup et al., 2006; Ainsworth et al., 2016) and the physiological performance of a generalist coral symbiont (Howells et al., 2012). Warmer latitudes have also been shown to enhance heritable thermal tolerance, up to 10-fold in Acropora millepora from the GBR (Dixon et al., 2015). Such latitudinal gradients thus provide important information over the spatial scales through which thermal tolerance can be achieved, and opens discussion on the ability for more “tolerant” species to provide “genetic rescue” through the exchange of more tolerant larvae across latitudes (Ingvarsson, 2001). That said, studies of the coral Oculina patagonica in the Mediterranean across locations with different thermal temperature regimes demonstrated limited capacity to acclimate to conditions comparable to future warming (Rodolfo-Metalpa et al., 2014). Thus, the ability for corals to inherit thermal tolerance is likely to be species-specific, and highly-dependent on the dynamics of local temperature regimes. Costs may also be associated with greater thermal tolerance, for example, a thermo-tolerant symbiont, Symbiodinium trenchii (D1a), has been shown to reduce host mortality rates by 30% under thermal stress; however, that is accompanied by a reduction in growth rates of 50–60% (Little et al., 2004; Smith and Iglesias-Prieto, 2010; Pettay et al., 2015). Energetic trade-offs need to be considered alongside the benefit of tolerance to assess whether coral reef form and function can be maintained.

Inshore turbid environments (Figure 2) often have a suite of extreme conditions, such as low light and elevated water temperatures, and thus have historically been overlooked as important coral environments. Corals employ physiological plasticity to persist in turbid waters, including flexible photosynthetic properties, which may include a change of algal endosymbiont genetic type, and/or enhanced host heterotrophy (Anthony, 2000; Anthony and Fabricius, 2000; Hennige et al., 2008, 2010; Suggett et al., 2012b). Morgan et al. (2016) recently described over 21 genera of coral surviving in muddy waters on the inshore GBR suggesting that turbid systems may afford important range (niche) extension beyond that currently considered for many key “blue water” tropical oligotrophic reef-forming coral taxa.

High irradiance exacerbates susceptibility to coral bleaching, and thus turbid reef environments where irradiance is reduced could act as potential refugia during bleaching events. Cacciapaglia and van Woesik (2015) predicted that 9% of coral environments previously considered uninhabitable under ocean warming, due to the combination of elevated temperatures and high irradiance, would be protected by the mitigating effect of shading in high turbid sites. Studies in Palau (van Woesik et al., 2012), Florida (van Woesik and McCaffrey, 2017), and on the GBR (Morgan et al., 2017) have all provided compelling evidence that coral bleaching can be reduced at high-turbidity sites. Consequently, as the oceans warm and bleaching events become more frequent, there could be a direct selection of corals able to survive in inshore turbid environments (van Woesik and McCaffrey, 2017). Potts and Jacobs (2000) propose that turbid inshore habitats have provided ecological and evolutionary continuity and refugia for scleractinian corals during non-reefal periods when environmental conditions have been too extreme for reef growth.

High-anthropogenic stress experienced by shallow reefs has seen deeper reefs proposed as potential refugia (Glynn and Morales, 1997; Riegl and Piller, 2003; Armstrong et al., 2006). Under this hypothesis, deeper reefs are able to: (i) protect or reduce the disturbances faced on shallower reefs, and (ii) provide a genetic stock of corals that can re-seed shallower reefs after a disturbance. Supporting this theory are studies evidencing reduced bleaching on deeper coral reefs (e.g., Riegl and Piller, 2003; Bridge et al., 2014). However, this bleaching pattern is not always observed; for example, with deeper corals in the Seychelles bleaching first, and more severely, than shallow corals during the 1997–1998 bleaching event (Spencer et al., 2000). Smith et al. (2016) demonstrated that deep (30–75 m) coral reefs have lower bleaching threshold temperatures than shallow-reefs, and that any increase in temperature above the local mean can result in stress and bleaching. Furthermore, deepwater generally has lower pH and Ω_arag than shallow-water, complicating the ability of these systems to act as refugia.

Capacity for deep reefs to act as potential refugia remains debated and depends in part on the connectivity of shallow and deep coral populations (Bongaerts et al., 2011). van Oppen et al. (2011) found evidence of deep-to-shallow coral recruitment for some reefs (Scott's Reef) on the GBR, but not others (Yonge Reef). Similarly, Serrano et al. (2014) found geographic differences in vertical connectivity due to local geophysical conditions. Species-specific differences in population connectivity with depth have also been reported in the Caribbean (Bongaerts et al., 2017).

Low-light conditions of deeper reefs are considered sub-optimal, and may select against some corals that do not have the physiological plasticity to survive. For example, Smith et al. (2017b) identified the importance of photoconvertible red fluorescent proteins (pcRFPs) to support coral adaptation/acclimation to deeper waters, while Cooper et al. (2011) showed that both the host and symbiont needed to have photobiological flexibility to survive across light regimes. Similarly, Lesser et al. (2010) found for the coral Montastrea cavernosa in the Bahamas that both physiological and morphological adaptation was required to survive in depths down to 91 m. Changes in the endosymbiont community (e.g., Frade et al., 2008; Lesser et al., 2010; Bongaerts et al., 2015) and density (Frade et al., 2008) have also been reported with depth. Specialization of deeper coral communities suggests plasticity will be required for species to survive across depths and ultimately utilize these systems as potential refugia.

Mangroves are widespread reef-associated environments that have recently received attention for their potential services to corals threatened by climate change (e.g., Yates et al., 2014; Camp et al., 2016a). Mangrove systems have dynamic temperature, pH and oxygen profiles, and as such, expose resident corals to very different physicochemical conditions than neighboring reefs (Camp et al., 2017). Mangroves are highly heterogeneous in nature, with freshwater input, organic content, system size and consequently water residency time, along with other benthic components (e.g., seagrass composition) influencing the local physicochemical conditions. As such, mangrove systems appear to offer a range of potential services to corals, dependent on local spatial and temporal biogeochemical conditions. For example, work in the Caribbean (Yates et al., 2014) suggested that mangroves could act as a potential refuge for corals, through physical shading and the elevation of downstream A_T as a result of carbonate-sediment dissolution (i.e., buffering potential similar to seagrass habitats), as well as frequent exposure to more variable temperatures. In other locations [e.g., the Seychelles, Indonesia (Camp et al., 2016a) and New Caledonia (Camp et al., 2017)] mangrove systems have exposed corals to simultaneously low pH, low oxygen and warmer waters relative to adjacent reefs. As such, these mangrove systems may precondition corals to future climates (Camp et al., 2016a) or house corals more resilient to future climate conditions (Figure 2, Camp et al., 2017). In New Caledonia, over 20 coral species were regularly exposed to pH as low as 7.3, temperature 1–2°C higher than adjacent reefs, and oxygen levels below 3 mg L⁻¹ (Camp et al., 2017). Initial research at this site suggests that physiological plasticity (e.g., enhanced respiration indicative of increased heterotrophy) is facilitating survival of diverse coral communities, which include numerous species typically thought to be more sensitive to stress, e.g., Acropora spp.

Physicochemical conditions of shallow reef systems are dynamic and highly influenced by their coastal interface, ecosystem structures and local metabolic activity. Across global locations, there are reports of shallow in-shore reef systems being regularly exposed (albeit for different durations) to conditions predicted for the open-ocean under climate change by the end of the century (e.g., Hofmann et al., 2011; Price et al., 2012; Albright et al., 2013; Guadayol et al., 2014; Kline et al., 2015). The shallow water depth and reduced residency time of inshore waters can create an inshore to offshore carbonate chemistry gradient (Manzello et al., 2012; Shamberger et al., 2014; Camp et al., 2016a). Nutrient run-off and freshwater input can influence A_T, while the metabolic activity of local benthos can change pCO₂ levels. Sites in Palau (Shamberger et al., 2014; Barkley et al., 2015), Puerto Rico (Gray et al., 2012), Florida (Millero et al., 2001; Yates et al., 2007), and the GBR (Santos et al., 2011; Shaw et al., 2012, 2015; Kline et al., 2015) have all been characterized as having extreme in-shore low pH reef systems. In the in-shore low pH bays of Palau, coral cover and richness remained similar to control areas, but community composition shifted, with replacement of Acropora, Montipora, and Pocillopora with Porites (Barkley et al., 2015). For the abundant Porites sp. skeletal extension and density remained unchanged but macrobioerosion increased at the lower Ω_arag sites (Barkley et al., 2015). Interestingly, one of the low pH Palau communities was not dominated by Porites but instead, Leptastrea, Platygyra, Favites, and Dipsastraea, contributing to the emerging evidence that a range of species are tolerant to low pH conditions. These findings suggest that acidification conditions are not deterministic, and that other local factors are also important in shaping coral community composition under low pH and low Ω_arag (Barkley et al., 2015).

Along with environments that have persistently low pH (e.g., Palau), there are highly variable pH habitats that routinely, through tidal and diurnal cycles, experience low pH beyond what is predicted for coastal systems by 2100 (Duarte et al., 2013). For example, the Heron Island reef flat can experience a seasonal range in temperature of 8.1°C (17.9–26°C) and pH of 1.04 (7.57–8.61; Kline et al., 2015). Daytime photosynthesis, and night-time respiration in the absence of photosynthesis, create large diel cycles in pH (0.75 units, Santos et al., 2011) and Ω_arag (5.4 units, Shaw et al., 2012). At these highly variable in-shore sites, carbonate chemistry conditions can be optimal (elevated pH and Ω_arag) or sub-optimal (low pH and Ω_arag), and the presence of diverse and abundant coral communities at these sites indicates that corals can tolerate, at least periodic, short-term exposure to low pH. It remains unclear whether exposure to periodic low pH and low Ω_arag enhances resistance to future ocean acidification. In the Cayman Islands, corals from a naturally variable inshore pH environment experienced no enhanced tolerance when exposed to conditions predicted for 2100 (Camp et al., 2016b). Similarly, three corals species in Florida Bay regularly exposed to seasonal and diel variance in pCO₂ and Ω_arag showed no reduced sensitivity to low Ω_arag, with calcification rates equivalent to corals from a more stable physicochemical environment (Okazaki et al., 2013). However, the natural oscillations of pCO₂ in these in-shore environments may reduce the negative effects of elevated pCO₂, by ensuring at least periodic conditions for optimal calcification. For example, Comeau et al. (2014) observed calcification rates for Acropora hyacinthus that were increased by 21% under 400 to 2,000 μatm oscillatory pCO₂ vs. 1,000 μatm constant pCO₂ treatment. Such heterogeneity of in-shore reef systems, means that some locations dominated by photoautotrophs could in fact see an elevation in mean pH relative to offshore, due to the daytime consumption of CO₂ elevating pH and local Ω_arag, as observed for the back reefs of the Cayman Islands (Camp et al., 2016a) and in seagrass dominated environments.

Coral reefs subjected to large tidal ranges are similarly exposed to extreme environmental conditions. The most extreme macrotidal reef site examined to date is the Kimberley region in northwest Australia, which features the world's largest tropical tides (up to 12 m). Corals on the reef flat of the Kimberley region have to cope with extended aerial exposure during spring low tides (up to several hours, Figure 2) and periods of stagnant water alternating with strong tidal currents (>10 knots). These physical conditions create extreme fluctuations in environmental parameters, such as temperature (e.g., the well-studied Shell Island where temperature can fluctuate by up to 7°C daily, Dandan et al., 2015; Schoepf et al., 2015), pH (e.g., Tallon Island where pH can fluctuate from 7.6 to 8.8 units over a spring low tide, Pedersen et al., 2016), dissolved oxygen (Pedersen et al., 2016; Gruber et al., 2017), unusually turbid waters, and monthly SST exceeding 30°C for several months a year. Despite these extreme conditions, highly diverse coral reefs exist throughout the Kimberley of up to 225 species (Rosser and Veron, 2011; Richards et al., 2015), which are comparable to inshore reefs in the central GBR ~2 decades ago (Richards et al., 2015).

Research is just beginning to uncover the biological properties that allow corals to thrive in this extreme macrotidal environment. Firstly, algal endosymbionts appear to be locally adapted with clade C dominating northern Kimberley Acropora assemblages (Thomas et al., 2014) and southern Acropora and Dipsastraea corals (Schoepf et al., 2015); however, further research is required to determine how Symbiodinium diversity within clade C is linked to environmental gradients in this region. Furthermore, thermal history has been shown to influence bleaching susceptibility of corals harboring the same symbiont type, with more variable temperature environments enhancing heat resistance (Schoepf et al., 2015). During the first documented regional-scale bleaching event in the Kimberley in the austral summer of 2016 (Hughes et al., 2017), corals were more severely bleached (Le Nohaïc et al., 2017) and experienced much greater mortality (V. Schoepf, unpublished data) at the less-variable subtidal site on Shell Island.

Reef flat communities of the Kimberley appear to have adapted their physiology to maintain metabolic balance of photosynthesis-to-respiration (P:R) despite the extreme physicochemical conditions of the reef-flat (Gruber et al., 2017). Furthermore, a seasonal calcification study revealed high resilience of coral calcification to both intertidal and subtidal temperature conditions as branching and massive corals calcified at rates that were comparable to those of similar species at a more typical coral reef 1,200 km to the southwest (Ningaloo Reef; Dandan et al., 2015). Given that rising sea levels are expected to reduce extreme temperature variation on macrotidal reef flats such as Tallon Island (Lowe et al., 2016), reef flat communities might have more time than other reef types to acclimatize and/or adapt to ongoing ocean warming. However, further research is needed on these systems to disentangle the complex interactions of acclimation/adaptation processes that support coral survival under such extreme physicochemical conditions.

Dispersed across the world's oceans are sites where deep, nutrient-rich, cooler ocean water wells-up to the surface through local water circulation. Such waters have been proposed as potential localized coral refugia under times of thermal stress (Riegl and Piller, 2003); however, the cool water is also high in CO₂ and low in Ω_arag; consequently, upwelling environments have the potential to be ocean acidification hot-spots (Feely et al., 2008; Manzello, 2010). Upwelling sites clearly comprise a complex mix of abiotic stressors, with any refuge potentially limited by upwelling time (Chollett et al., 2010). In regions such as the Eastern Tropical Pacific (ETP) and Arabian Sea, marginal coral reef formation has been attributed to the cool upwelled waters creating a “pseudo-high-latitude effect” (Sheppard and Salm, 1988) by reducing reef-building processes (Benzoni et al., 2003). Here, corals do not form a true reef structure, but instead form low-relief coral carpets (often monospecific) or remain as isolated coral colonies that are generally small in size. These locations can illustrate the changes in coral community composition when reef formation ceases, as is predicted to occur in a high CO₂ world.

A unique upwelling system exists in the ETP where high CO₂ water upwells to the surface, resulting in coral reef and coral communities off the coast of the Galápagos Islands having some of the lowest Ω_arag documented for modern coral reefs (<2.5, (Manzello et al., 2008)). Corals of the Galápagos Islands and neighboring Panamá reefs also contend with frequent exposure to the El Niño-Southern Oscillation (ENSO, + 3–4°C during ENSO events for >2 months), creating a unique opportunity to study the long-term effects of low pH and thermal stress events on corals (Manzello et al., 2014). Only one coral reef exists within the low Ω_arag waters of the Galápagos archipelago and it is dominated by P. lobata (Manzello et al., 2014). High bioerosion rates (>25 kg CaCO₃ m⁻² yr⁻¹) and low reef cementation are characteristic of sites within the Galápagos region (Manzello et al., 2008).

Upwelled waters of the ETP are also high in nutrients, which may in part have enhanced coral growth rates for four coral species of the ETP (Jiménez and Cortés, 2003). Elevated nutrient levels have the potential to alleviate some of the negative impacts of decreased coral calcification associated with the low Ω_arag of the upwelled waters, by potentially enhancing heterotrophy (Langdon and Atkinson, 2005; Cohen and Holcomb, 2009), or increasing inorganic nutrients that can enhance cell growth and proliferate abundance of the algal symbiont (see review by Fabricius, 2005). However, enhanced heterotrophy may not always be an advantage for corals around cooler upwelling sites, with zooplankton feeding rates of three coral species reduced under cooler upwelling conditions in Panamá (Palardy et al., 2005). High nutrients, in particular phosphates, can also reduce skeletal density (Dunn et al., 2012), and may explain why corals around the Galápagos Islands maintain their linear extension rates, but experience reduced skeletal density (Manzello et al., 2014). Increased exposure to nitrogen levels has also been documented to enhance bleaching susceptibility when exposed to heat and light stress (Wiedenmann et al., 2013) and could add to the potential bleaching susceptibility of corals at upwelling sites (D'Croz and Maté, 2004). Clearly the net-gain or cost of elevated nutrients for corals around upwelling sites remains debated and is an area for further investigation. An important consideration with upwelling sites is that the severity or even presence of upwelling is often seasonal (as occurs for sites of the ETP), meaning the physicochemical conditions representative of future scenarios may be limited in duration. Timing of assessment at these sites is therefore important, with seasonal differences having large impacts on the biodiversity recorded (e.g., Diaz-Pulido and Garzón-Ferreira, 2002; Witman and Smith, 2003).