Introduction

Coral reef communities have shown a remarkable persistence in taxonomic composition and diversity over at least the past 500,000 years (Pandolfi 2002). In the last 30 years, however, there has been an unprecedented decline in the distribution and abundance of coral communities (Pandolfi 2002; Gardner et al. 2003; Bruno and Selig 2007) with approximately 26% of the remaining coral reefs now considered to be under immediate or long-term threat (Wilkinson 2004). The decline of coral communities has been attributed to local factors such as over-fishing of key functional groups, declining water quality, and the physical degradation of coral reefs by human activities such as destructive fishing and unsustainable tourism. Moreover, rising atmospheric greenhouse gas concentrations pose a major threat to coral reefs principally through (1) global warming, which has dramatically increased incidences of mass coral bleaching and subsequent mortality events and (2) acidification of ocean waters, causing a decrease in coral growth and calcification rates (Kleypas et al. 1999; Hoegh-Guldberg et al. 2007; Cooper et al. 2008; Bak et al. 2009; De’ath et al. 2009). Current estimates of climatic changes dwarf even the fastest rates of change during the glacial cycles of the past 420,000 years, which are 2–3 orders of magnitude slower that that seen over the past 100 years (Hoegh-Guldberg et al. 2007).

The extent to which reefs are affected by environmental stressors is not uniform, and specific areas/zones may be able to (partially) escape certain disturbances. Such zones may constitute refugia (where species can survive during periods of adverse conditions elsewhere) and may play an important role in the recovery of impacted reef areas (Ridgway and Hoegh-Guldberg 2002; Hoegh-Guldberg et al. 2008). In particular, the observation that some areas are less affected by global thermal stress events, due to cooler or more stable conditions, has led to the idea that these areas may act as thermal refugia for corals (Vermeij 1986; Glynn 1996; Riegl and Piller 2003; Halfar et al. 2005). According to Glynn (1996), potential refugia include areas exposed to cool upwelling conditions, high-latitude communities, ocean banks or island shores, and moderate to deep reef ecosystems.

The ‘deep reef refugia’ hypothesis (DRRH) has gained in popularity in the recent literature (Hughes and Tanner 2000; Feingold 2001; Glynn et al. 2001; West and Salm 2003; Riegl and Piller 2003; Armstrong et al. 2006; Venn et al. 2009; Lesser et al. 2009a; Hinderstein et al. 2010). Glynn (1996), who first postulated this hypothesis, highlighted that deeper reefs are less affected by thermal stress events and thus have the potential to act as refugia. However, the refugia potential might not be limited to avoiding elevated seawater temperatures as the intensity of other stressors, such as storm-induced waves, is also negatively correlated with depth (Liddell and Ohlhorst 1988). With the recent evidence pointing to coral populations being largely self-seeding and more closed than previously thought (Ayre and Hughes 2000; Baums et al. 2005; Underwood et al. 2007; and references therein), the potential of deep reefs to act as a local recruitment source for the shallow has become an integrated part of the DRRH (Hughes and Tanner 2000; Lesser et al. 2009a).

Given the increasing pressure on coral reefs and the recent scientific findings on coral population connectivity, coupled with the recent interest in deeper coral communities, the purpose of this paper is twofold. First, it reviews the current understanding of the potential role of deeper reefs as (reproductive) refugia for scleractinian corals, and secondly, it provides a guide for future research directions. In doing so, the underlying assumptions of the DRRH are explored, which stipulate that deep reef areas (1) are protected or dampened from disturbances that affect shallow reef areas and (2) can provide a viable reproductive source for shallow reef areas following disturbance. Despite deeper tropical reef communities occurring in the Pacific and Indian Oceans (e.g., Fricke and Schuhmacher 1983; Maragos and Jokiel 1986; Kahng and Kelley 2007), the present study focuses specifically on the relatively well-documented deep reefs of the Caribbean in the Western Atlantic (e.g., Goreau and Goreau 1973; Bak 1977; Van den Hoek et al. 1978; Bak and Luckhurst 1980; Fricke and Meischner 1985; Bak and Nieuwland 1995; Bak et al. 2005; Jarrett et al. 2005; Armstrong et al. 2006; Culter et al. 2006) to evaluate these assumptions. Finally, critical information gaps are highlighted, and potential future research directions are presented.

Deep reefs in the Caribbean

To accurately evaluate the potential of deep reefs to act as refugia, it is important to first consider the abundance and particular physical characteristics that govern deep Caribbean coral communities. At the outset, it is imperative to define the somewhat relative (and subjective) term ‘deep reef’ in order to overcome potential confusion between terms such as ‘deep-water reef’ and ‘deep-sea reef’, which have been used in the past to refer to both cold-water and non-phototrophic systems as well as warm-water and light-dependant coral ecosystems. In this respect, the term ‘mesophotic’, which was traditionally used to describe the depth range of Halimeda bioherms (Pomar 2001), may be particularly useful and has recently been adopted to refer to deep but light-dependent coral ecosystems (Armstrong et al. 2006; Menza et al. 2007; Hinderstein et al. 2010). Although a technical definition for ‘mesophotic’ is still lacking in the coral literature, it is loosely defined as the depth range between 30 m and the depth at which light in the water column is too low to sustain the growth of corals that depend on their phototrophic obligate symbionts (dinoflagellates of the genus Symbiodinium). In the Caribbean, this lower depth limit is often around 80 m (Kahng et al. 2010), even though in some locations with exceptionally clear water, it may extend down to ~100 m (Reed 1985; Liddell and Ohlhorst 1988).

Even though information is scarce, it is possible to make some general observations about the distribution and occurrence of mesophotic reefs. The environmental conditions (e.g., temperature, salinity, nutrients, light availability, and aragonite saturation state) required for coral reefs to accrete net amounts of carbonate over time are outlined in Kleypas et al. (1999). Some of these parameters (principally temperature, light, and nutrients) can be depth dependent, and the geographic distribution of mesophotic coral communities does, therefore, not necessarily overlap with areas that sustain shallow coral reef communities. The majority of coral reefs occur in tropical seas at depths of less than 30 m (Huston 1985), and this is predominantly driven by the reduction in light availability with depth. The euphotic zone, defined as the depth range to which sufficient light is available to support photosynthetic activity, determines the lower depth limit of hermatypic corals. The actual depth range of the euphotic zone varies between locations in the Western Atlantic depending on local light attenuation coefficients (Kleypas et al. 1999; Kahng et al. 2010), which are determined by the concentrations of dissolved organic matter, phytoplankton, and sediment loads in the water. For example, the euphotic depth in Curaçao and southwest Florida ranges from 60 to 75 m (Van den Hoek et al. 1978; Jarrett et al. 2005), whereas the clearer, oceanic waters off Bermuda support a euphotic depth of >100 m (Fricke and Meischner 1985).

Scleractinian corals at mesophotic depths often exhibit flat, plate-like morphologies to maximize light capture (Titlyanov 1987; Merks et al. 2004) and may utilize different symbionts to cope with the environmental light field (intensity and spectral quality) of deeper reef zones (Iglesias-Prieto and Trench 1997; Iglesias-Prieto et al. 2004; Sampayo et al. 2007; Frade et al. 2008). Despite these adaptations to maximize light use, the total contribution of photosynthesis to calcification is reduced in mesophotic corals (McCloskey and Muscatine 1984). Although this deficit may be supplemented with heterotrophic feeding (McCloskey and Muscatine 1984; Mass et al. 2007; Alamaru et al. 2009; Lesser et al. 2009b), facilitated by the influx of deep nutrient-rich oceanic waters at greater depths (Leichter et al. 1996; Leichter and Genovese 2006), total reef accretion is negatively correlated with increasing depth (Grigg 2006).

The formation of mesophotic reefs is further influenced by the presence of thermoclines and the availability of suitable substrate. Strong shallow thermoclines (of several degrees Celsius) can prevent the development of mesophotic coral communities at locations where healthy shallow communities exist (e.g., Northwest Hawaiian Archipelago; Grigg 2006), but such thermal stratification of the water column is usually restricted to higher latitudes and has not yet been documented for the tropical reefs of the Caribbean. In contrast, the availability of hard substrate has been indicated as an important limiting factor in the formation of Caribbean deep reef communities (Bak 1977). The substrate supporting most mesophotic coral communities consists of limestone structures deposited during the late Pleistocene or early Holocene but also frequently consists of rhodolith banks (Fricke and Meischner 1985; Littler et al. 1991; García-Sais et al. 2008; Rivero-Calle et al. 2009), which provide suitable hard substrate when agglutinated into solid banks by red coralline algae (Fricke and Meischner 1985). Even though light and substrate conditions may favor the formation of mesophotic coral communities, the accumulation of sediment often prevents the growth of sessile organisms (Goreau and Goreau 1973) before light becomes limiting (Bak 1977). Substrate limitation due to sediment accumulation generally does not play a role on the vertical walls that are common in the Caribbean on most fore-reef slopes in excess of 50–60 m (Liddell and Ohlhorst 1988), with the exception of localized areas where sediment accumulates into sand channels or chutes (Hubbard 1989). However, near-vertical reef profiles receive lower irradiances relative to low-angle substrates (Brakel 1979) and only certain species can successfully grow on these steep, vertical walls (Liddell and Ohlhorst 1988).

An accurate assessment of the distribution and abundance of mesophotic reef areas and their relative contribution to total coral reef surface area is currently restricted due to the lack of large-scale mapping studies that include deeper habitats. Nonetheless, mesophotic coral ecosystems have been reported throughout the Caribbean, including the Bahamas (Porter 1973), Barbados (Macintyre et al. 1991), Bermuda (Fricke and Meischner 1985), Curaçao (Bak 1977), Florida (Jarrett et al. 2005), Jamaica (Goreau and Goreau 1973), Puerto Rico (García-Sais et al. 2008) and the US Virgin Islands (Armstrong et al. 2006). Recent technological advances in multibeam sonar, remotely operated vehicles (ROVs), and autonomous underwater vehicles (AUVs) do, however, provide the means to map deep reefs on a large scale (with a limitation of the high costs involved), and initial mapping efforts have revealed extensive areas of mesophotic coral bank ecosystems on the Puerto Rican shelf (Armstrong 2007; Menza et al. 2007; Rivero-Calle et al. 2009), with an estimated area of 300 km2 for the mesophotic reef complex located south of the northern Virgin Islands (Smith et al. 2010). Similarly, large areas of mesophotic ecosystems are expected to occur elsewhere in the Caribbean, either as submerged banks (e.g., the Saba Bank; Van der Land 1977; Toller et al. 2008) or as part of the many fringing and barrier reef systems that occur directly adjacent to deep oceanic water.

Deep reefs as refugia for episodic disturbances

Disturbances and threats to coral reefs have received considerable attention over the past decade, but when considering the potential of mesophotic reefs to act as refugia, it is critical to determine the depth to which disturbances extend. Some episodic stressors may act indiscriminately over the entire euphotic depth range, whereas the effect of others may be limited to a certain depth range or diminish as depth increases. Despite a wealth of information in the scientific literature on the effects of disturbances on shallow-water coral reefs, only limited accounts can be found for deeper coral communities (Table 1). Processes that reduce the presence of coral reefs can be natural and anthropogenic, and their impacts range from local to global depending on scale and intensity.

Table 1 Reported disturbances in the Caribbean for which the effect on the deep reef (>30 m) has been assessed
Full size table

Natural disturbances

Over the past decades, storm events and disease outbreaks have strongly influenced coral reefs worldwide (Connell 1997). Storms are a common feature throughout the Caribbean, and the effects of storms on coral reef communities in relation to depth are relatively well documented (Banner 1961; Stoddart 1962; Glynn et al. 1964; Woodley et al. 1981; Rogers et al. 1982, 1983; Kjerfve et al. 1986; although few studies assess damage at depths >30 m, Table 1). These data indicate that mesophotic reef areas are largely sheltered from direct physical damage of storm-induced waves but are not spared from indirect effects (e.g., debris avalanches or sedimentation). Coral debris originating from storm impacts can be transported down the reef slope, causing damage to deeper coral assemblages (Dollar 1982), and this process is correlated with the angle of the reef slope. For example, in French Polynesia (Harmelin-Vivien and Laboute 1986), following the 1982/1983 hurricane season, large reef areas between 30 and 90 m with steep reef slope angles (>45°) suffered high mortality through debris avalanches, while reefs with low-angle slopes (<25°) in the same location were only damaged in the shallows. Besides damage from coral debris, storms can cause transport of fine sediments from the shallows into the mesophotic sections (Hubbard 1992), especially on fringing reefs bordering artificial sand beaches (Nagelkerken 2006). Fine sediments smother coral colonies and can cause significant mortality on deeper coral communities (Bak et al. 2005). Substrate and community structure of the deep reef is an additional consideration, because taxa with thin skeletal features such as Agaricia spp., which are often dominant on deep reef slopes, are somewhat more fragile and less adept at removing sediment compared to the hard plates formed by members of the Montastraea annularis species complex that dominate submerged banks (Smith et al. 2010). Finally, hurricane-related pressure on Caribbean reef systems may play a greater role over time given that a 0.5°C increase in sea surface temperature is predicted to increase hurricane intensity and frequency in the Caribbean (Webster et al. 2005; Saunders and Lea 2008). Despite this, the limited accounts on storm-related damage to mesophotic depths (Table 1) suggest that the direct effects of storm damage will generally always be greatest on shallower reef sections.

Disease forms part of the natural cycle of reef-building corals, but its abundance and prevalence within the marine environment can be amplified by anthropogenic influences. For example, stress may lower disease resistance, and the frequency of coral disease is projected to increase, as sea temperatures and/or local anthropogenic stresses increase (Harvell et al. 1999; Selig et al. 2006). As most coral diseases are host specific, their distribution over the reef slope is generally limited to the depth range of the coral host. For example, the outbreak of white-band disease on Caribbean Acropora in the 1980s and 1990s was limited to the shallows due to the restriction of Acropora to shallow water (Aronson and Precht 2001). Besides the outbreak of “intercostal mortality syndrome” in the US Virgin Islands (Smith et al. 2010), which was confined to deeper basin habitats (>35 m), other recent reports of coral disease along depth gradients (Calnan et al. 2008; Smith et al. 2008) do not show depth-specific disease prevalence in species with a broad depth distribution. Several coral diseases such as black band disease, dark spots disease, white syndrome, and yellow blotch (band) disease were first documented in shallower reefs but have since been reported on mesophotic reef sections (Table 1).

Few cases of invasive species have been documented in the Caribbean, and deleterious invasions such as those reported for temperate regions have yet to be reported (Coles and Eldredge 2002; Wilkinson 2004). Despite the fact that Caribbean reefs were hit somewhat harder hit by disease compared to their Indo-Pacific counterparts (Weil 2004), they have been less impacted by predatory species (e.g., Acanthaster planci and Drupella outbreaks; Colgan 1987; DeVantier and Deacon 1990; Turner 1994). The Indo-Pacific lionfish (Pterois spp.) and the ahermatypic coral Tubastrea coccinea are recent examples of invasive species that have successfully radiated throughout the Caribbean (Fenner and Banks 2004; Whitfield et al. 2007; Green and Cote 2009). T. coccinea occurs over a large depth range, but its distribution (and hence impact) is presently restricted to cryptic and artificial habitats (Vermeij 2005). The threat of the lionfish is currently considered to be restricted to fish communities (Albins and Hixon 2008; Green and Cote 2009) but could potentially have an indirect impact (e.g., increase in algal growth), particularly on the shallow coral community, through increased predation of juvenile herbivorous fish.

Local anthropogenic disturbances

The proximity of a coral reef to land-based disturbances is directly related to the level of exposure to nutrient enrichment, influx of toxins (such as herbicides and pesticides), and sedimentation (Menza et al. 2008; Smith et al. 2008). The differential effect of these stressors over depth has not been well described, although sedimentation is perhaps an exception in that it affects coral communities by reducing the amount of light and available substrate as well as interfering with photosynthesis and feeding. Sediments can smother coral colonies, even though corals may employ several physiological and behavioural avoidance strategies (Stafford-Smith and Ormond 1992). Colony and calyx morphologies play important roles in the passive removal of sediment, with dome-shaped colony morphology, and large polyps being particularly efficient at removing large particles (Hubbard and Pocock 1972). As a response to lower light levels, most mesophotic reef corals tend to have a flat plate-like morphology that traps sediment, and although this increased susceptibility to sedimentation is normally not problematic due to the relatively lower rates of sedimentation on the deeper reef (Bak and Engel 1979; Smith et al. 2008), increased sediment levels (due to storm activity) have been reported to result in large-scale mortality among mesophotic corals (Bak et al. 2005).

Nutrient enrichment adversely affects shallow reefs by fueling blooms of macroalgae and phytoplankton. Such blooms can be particularly catastrophic in combination with other stressors such as hurricanes and elimination of herbivores by overfishing (Hughes and Connell 1999). Increased macroalgal growth leads to an intensified competition for space and reduces survivorship of coral recruits (Bak and Engel 1979; Hunte and Wittenberg 1992; Vermeij 2006), while phytoplankton blooms decrease light availability for corals and other benthic organisms (Hallock and Schlager 1986). Isolated examples such as blooms of low-light-adapted macroalgae (e.g., Codium isthmocladum and Caulerpa spp.) have been reported on deep reefs (20–50 m) in Florida (Lapointe 1997; Lapointe et al. 2005; Lapointe and Yentsch pers. comm. Table 1). Additionally, in some instances, sewage outflow is discharged directly onto deep reef communities resulting in deep-water-restricted nutrient enrichment (Proni et al. 1994). Given that light is already limiting on mesophotic reefs, a further reduction in available irradiance through nutrient levels, macroalgae, and phytoplankton blooms has the potential to affect the viability of deeper light-dependent communities.

The overexploitation of key fish species is one of the most direct anthropogenic disturbances that impact coral reef ecosystems (Jackson et al. 2001; Hughes et al. 2007; Stallings 2009). Whether or not it has a depth-related component is unclear, although near-shore fishing intensities are usually more pronounced in the shallows (e.g., Polunin and Roberts 1993). Not only the intensity of fishing is important, but also the species that are targeted at the different sections of the reef. For example, herbivory diminishes with depth (Hay et al. 1983; Brokovich et al. 2010) and key grazers, such as parrotfish and damselfish species are usually restricted to the shallower reef (<30 m; Liddell and Avery 2000). Thus, the removal of herbivorous fishes might not have such deleterious effects on the deep reef but has significant effects on ecosystem functioning of shallow reefs (Jackson et al. 2001). This has also become apparent during the mass mortalities of the Caribbean sea urchin Diadema antillarum (Lessios 1988), which had major impacts only on shallower zones, due to D. antillarum’s predominance as an herbivore at these depths (Morrison 1988). Finally, a range of other local anthropogenic disturbances, such as diving and snorkeling-related activities, spear fishing, and ship groundings are largely restricted to shallow reef sections.

Global anthropogenic disturbances

Sea surface temperature increases, or anomalies cause mass coral bleaching events that have been responsible for the significant mortality of reefs worldwide over the past two decades (Hoegh-Guldberg 1999; Hoegh-Guldberg et al. 2007). Coral bleaching is mainly caused by elevated temperatures but is exacerbated by high irradiance levels (Jokiel and Coles 1977). Coral bleaching has been reported down to 30–60 m of depth (Bak et al. 2005; Bunkley-Williams et al. 1991; Lang et al. 1988), but the effects of warm-water bleaching are generally considered to be more pronounced in shallow water (Fisk and Done 1985; Wilkinson and Souter 2008).

The bathymetric effect of bleaching susceptibility can be explained through the relatively homogenous and temporally stable irradiances beyond 30–40-m depth (Vermeij and Bak 2002). Also, the thermal regimes between shallow and deep reefs can be markedly different, due to the pulsed delivery of oceanic sub-thermocline water to the deeper reef sections. This phenomenon is common on reefs located adjacent to deep oceanic water (such as the many fringing reef slopes of the Caribbean) (Leichter et al. 1996; Bak et al. 2005; Leichter and Genovese 2006; Lesser et al. 2009a), and although these influxes may cause higher temperature variability in deep water, long-term averages and maximum temperatures remain lower compared to the shallows (Frade et al. 2008). Examples of this differential include a 1°C difference in average summer temperature between the shallow (5–7 m) and deep (35–40 m) deep reef in the Florida Keys (Leichter et al. 1996) and Curaçao (Frade et al. 2008) but can be substantially greater (i.e., >1°C difference) over broader depth ranges (Bak et al. 2005; Lesser et al. 2009a). Given that exposures to temperatures 1–2°C higher than the long-term monthly average have repeatedly led to mass coral bleaching events (Hoegh-Guldberg 1999), the regular cold-water influxes on deeper reefs may indeed offer an escape from thermal stress. Nonetheless, bleaching susceptibility varies between species and across environmental gradients due to acclimation/adaptation to different thermal regimes (Coles and Brown 2003), and shallow corals may exhibit a broader thermal tolerance (Birkeland 1997). Additionally, extensive cooling of deeper water can lead to so-called coldwater bleaching as observed in Bonaire (Kobluk and Lysenko 1994) and the US Virgin Islands (Menza et al. 2007). Thus, although there are indeed more studies that report bleaching at shallow (rather than deep) reef sections, the general view that bleaching is usually restricted to shallow water should be interpreted with caution due to the lack of quantitative bleaching studies over large depth ranges.

Finally, global warming is driving increases in sea level as a result of the thermal expansion of seawater and the melting of ice trapped in landlocked glaciers and ice sheets. The Intergovernmental Panel on Climate Change (IPCC) consequently has predicted sea level rises of 18–59 cm during the twenty-first century (IPCC 2007). Recent changes in the arctic ice sheet plus growing evidence from other sources have shown that IPCC projections for sea level rise are conservative, and that sea levels will increase by at least 1 m by 2100 (Rahmstorf et al. 2007). Even though vertical reef accretion rates have kept up with rapidly rising sea levels in the geological past, coral reefs have also ‘drowned’, vanishing into the aphotic zone when accretion rates fall behind rising sea level rates (Grigg and Epp 1989). Projected sea level changes, therefore, have the potential to push deep reefs below the euphotic zone (Brown 1997), especially when increased thermal stress and ocean acidity simultaneously reduce coral growth rates (Cooper et al. 2008).

Deep reefs as a source of propagules for shallow areas

The ability of deeper sections of coral reefs to supply recruits to shallow reef areas has become a central assumption within the DRRH. Recovery from disturbances depends largely on the recruitment of larvae from either local or distant sources, and the availability of such larval sources is therefore crucial (Hughes and Tanner 2000). Recent genetic evidence demonstrates that many coral populations are largely self-seeding (see Ayre and Hughes 2000; Baums et al. 2005; and references therein), and larval dispersal can be as low as 100 m in some species (Underwood et al. 2007), which challenges the idea that reef systems may be rapidly repopulated from external larval sources after adult populations have declined. Deep reefs can only act as a (local) reproductive source for the shallow if (1) sufficient overlap exists in community structure between the shallow and deep and (2) sufficient larval exchange occurs from deep to shallow communities.

Coral community structure over depth

The mesophotic reefs of the Caribbean exist as extensions of the shallow reef slope or as isolated coral dominated communities on deep submerged plateaus or ridges (e.g., Flower Garden Banks (Gulf of Mexico) and the Puerto Rican Shelf). When forming part of a slope, the deeper reef typically shows a gradual depth zonation, progressing from moderate coral cover to a few isolated colonies in deeper water. In terms of coral biodiversity, the general trend is increasing species richness from the surface to an intermediate depth, followed by a continuous decrease with depth into the mesophotic zone (Bak 1977; Sheppard 1982; Huston 1985). Caribbean mesophotic reefs are usually dominated by members of the genera Agaricia (mainly A. grahamae and A. lamarcki), followed by Montastraea (mainly M. cavernosa) and Madracis (mainly M. formosa and M. pharensis) (Bak 1977; Van den Hoek et al. 1978; Bak et al. 2005; Jarrett et al. 2005; Culter et al. 2006; Venn et al. 2009), with the exception of submerged banks, where members of the Montastraea annularis species complex (which form thick plates at these depths) are usually the dominant members of the coral community (García-Sais et al. 2008; Rivero-Calle et al. 2009; Smith et al. 2010). The decrease in species richness over depth extends to the aphotic zone (<1% of surface irradiance) where light-dependent corals gradually disappear and are replaced by azooxanthellate corals, stylasterids, sponges, and coralline-, turf- and macro-algae (mainly Halimeda, Lobophora and Dictyota spp.) (Van den Hoek et al. 1978).

Knowledge of the distribution of coral species down the reef slope is useful to assess which species are the most likely candidates in the deep to replenish shallow reef zones. So-called ‘shallow-specialist’ species obviously do not occur on the mesophotic reef, whereas ‘deep-specialist’ species do not appear to be successfully recruiting to shallow water (Bak and Engel 1979). In contrast, ‘depth-generalist’ species exhibit wide depth ranges, and reef populations of these species are able to at least partially escape depth-related stressors. The fact that these species occur on both shallow and mesophotic reefs (although sometimes with different substrate preferences over depth; Vermeij and Bak 2003) suggests that they may form a single metapopulation over the reef slope and as such would be likely candidates to supply offspring up the reef slope. A detailed compilation of data on the vertical distribution of zooxanthellate coral species in the Caribbean and Bermuda (Table 2) indicates that ~25% of the species (total n = 53) occurs over large depth ranges (i.e., are ‘depth-generalists’) encompassing both the shallow reef and upper mesophotic zone (30–60 m). Several of these species (e.g., Madracis pharensis, Montastrea cavernosa, and Stephanocoenia intersepta) can be considered ‘extreme depth-generalists’, as their distribution extends into the lower mesophotic zone (>60 m); however, the majority of species does not occur deeper than the upper mesophotic zone (30–60 m) (Reed 1985; Rezak et al. 1990; Phillips et al. 1990; Jarrett et al. 2005; Culter et al. 2006). Only a few species (13%, Table 2; e.g., Agaricia grahamae, Scolymia cubensis, and Madracis formosa) are observed exclusively on mesophotic reefs (i.e., ‘deep-specialists’), and a relatively large number of species are limited to the shallow reef (55%, Table 2; <30 m; i.e., ‘shallow-specialists’). These numbers correspond roughly (although the numbers of ‘depth-generalist’ species are somewhat higher) with depth-distribution patterns described for species by Goreau and Wells (1967) in Jamaica (‘depth-generalist’ spp. 36%; ‘shallow-specialist’ spp. 53%; and ‘deep-specialist’ spp. 11%) and by Bak (1977) at two reef localities in Curaçao (‘depth-generalist’ spp. 41%; ‘shallow-specialist’ spp. 44%; and ‘deep-specialist’ spp. 18%). Interestingly, these percentages are also similar to those observed for highly diverse coral communities (total n = 152) in the Indian Ocean (Chagos Archipelago), with ~25% of the species occurring over large depth ranges, and 57 and 18% of the species representing ‘shallow- and deep-specialists’, respectively (Sheppard 1981). Even though proportions will vary geographically between regions and locations (as will the presence/absence on shallow and deep reefs of the species in Table 2), roughly a quarter of the scleractinian coral species in the Caribbean exhibit distributions that encompass both the shallow reef and upper mesophotic zone, and therefore based purely on community composition and species distribution ranges, represent candidate species that have the potential to provide propagules for recruitment in shallower zones.

Table 2 List of zooxanthellate coral species of the Caribbean and Bermuda with information on their bathymetric distribution, symbiont zonation, and reproductive mode
Full size table

Reproduction and recruitment as a function of depth

Although ‘depth-generalist’ coral species on the mesophotic reef are likely candidates to provide propagules for the shallow reef, there is currently no direct evidence that larval exchange between shallow and deep populations actually occurs (i.e., that these species form panmictic populations over the reef slope). Both coral life histories and symbiont associations have the potential to strongly influence recruitment patterns over depth, and it is, therefore, pertinent to explore these to evaluate whether species from the deep reef can act as a reproductive source for the shallow.

The variation in bathymetric distribution ranges of different coral species seems to be determined by pre-settlement rather than post-settlement processes (Mundy and Babcock 2000), as depth distributions of juveniles mirror those of adult colonies (Bak and Engel 1979), and larvae preferentially select parental habitat substratum (Baird et al. 2003) by differentiating between light intensity and spectral composition (Mundy and Babcock 1998). Therefore, even though recruitment processes on bare substrata can differ from those occurring in mature communities (Grigg and Maragos 1974; Tomasik et al. 1996; Vermeij 2006), it appears unlikely that ‘deep-specialist’ species will colonize bare substratum on the shallow reef. ‘Depth-generalist’ species are, therefore, the most likely candidates to facilitate recruitment to the shallow reef. However, pre-settlement (in addition to post-settlement) processes could potentially vary between shallow and deep populations of ‘depth-generalist’ species, resulting in (or reinforcing) intra-specific genetic structuring and limiting larval connectivity up the reef slope.

Genetic structuring could originate through local adaptation of coral populations (reviewed in Baums 2008) to the unique environmental conditions in shallow versus deep habitats. However, genetic assessments that specifically address genetic structuring of coral host populations over depth are lacking to date. Broadcasting species are less likely to form genetically distinct populations in the shallow and deep, because mass spawning in certain species of deep corals is observed to be synchronized with their shallow-water counterparts (e.g., M. cavernosa, M. franksi, and Diploria strigosa; Vize 2006), and mixing of sperm and eggs from the entire depth range is expected to occur at the surface (Willis et al. 2006). Most coral species in the Caribbean (and all species exclusive to the mesophotic zone) exhibit a brooding reproductive mode, and these species are more likely to exhibit small-scale genetic structuring (e.g., Ayre and Hughes 2000).

As scleractinian corals live in a mutualistic symbiosis with Symbiodinium, the combination of the physiological tolerances of each partner determines the ability of the holobiont (host plus endosymbiont) to occupy, compete, and thrive within its environment (Iglesias-Prieto and Trench 1997). In terms of the endosymbionts, the presence of distinct varieties has been shown to influence the survival and competitive ability of juvenile corals (Rodriguez-Lanetty et al. 2004; Little et al. 2004; Gómez-Cabrera et al. 2007). Comparing endosymbiont community composition of parental populations and offspring reveals that, similar to selection of parental habitat, juveniles adopt Symbiodinium types similar to that of the parent (Coffroth et al. 2001; Weis et al. 2001; Rodriguez-Lanetty et al. 2004). ‘Depth-generalist’ coral species can either harbor a single symbiont over their entire range or show a cladal or subcladal shift of symbionts with depth (e.g., Rowan and Knowlton 1995; Warner et al. 2006; Sampayo et al. 2007; Frade et al. 2008). However, nine out of ten ‘depth-generalist’ Caribbean species studied to date exhibit a zonation of Symbiodinium over depth with the exception of Siderastrea siderea (Table 2). In addition, processes of symbiont transfer and specialization of the symbionts to particular environments may impose limitations to the colonization and survival of coral offspring within certain habitats. For example, coral species with vertical transmission (maternal acquisition of symbionts) may be limited in their ability to settle outside the direct parental range if depth-specific symbionts are transferred to the offspring (e.g., Meandrina meandrites and Porites astreoides; Table 2). Such limitations are not expected for corals with horizontal transmission strategies (i.e., that acquire their Symbiodinium from the water column), or those that harbor a single Symbiodinium type throughout their entire distribution range. The prevalence of Symbiodinium zonation and a vertical symbiont acquisition mode in ‘depth-generalist’ species further increases the likelihood of genetic differentiation (and reduced gene flow) between shallow and deep coral populations.

Evaluation of the ‘deep reef refugia’ hypothesis

This paper set out to explore the potential role of mesophotic coral communities to act as ‘deep reef refugia’ (Glynn 1996; Riegl and Piller 2003; Armstrong et al. 2006). Given the rapid changes coral reefs currently face, the potential of deep reef sections to function as refugia has generated growing scientific and management interest in these communities (e.g., this issue and http://www.mesophotic.org). In this study, the current knowledge of deep reefs in the Caribbean was used to assess whether mesophotic reef areas (1) are protected or dampened from disturbances that affect shallow reef areas and (2) can provide a viable reproductive source for shallow reef areas following disturbance.

Some of the disturbances on coral reefs have the potential to act indiscriminately over the entire depth range (e.g., sedimentation, nutrient enrichment, and influx of toxins). However, the case history of the Caribbean provides clear examples of major disturbances that only affected the shallow sections of the reef (Table 1) such as the outbreak of white-band disease, the mass mortality of Diadema, and several hurricanes, and thermal bleaching events, leaving deeper reef sections relatively unaffected. This said, mesophotic communities face their own set of occasional stressors, including catastrophic sedimentation, deep-water macroalgal blooms, and cold-water bleaching (Table 1) and consequently are not immune to disturbances. Additionally, the slow growth (Hughes and Jackson 1985), fragile skeletons, and plate-like morphology of mesophotic corals make these communities more susceptible to damage through breakage and smothering. Nonetheless, mesophotic communities have on several occasions provided an escape to the effects of storm-induced waves and anomalies in sea surface temperature (Table 1); two acute threats that are predicted to become more severe in the coming decades. Even though the ability to escape the impact of storm-induced waves has limitations (depending on bathymetry and levels of sedimentation), and the ability of deep reefs to provide a thermal escape is poorly understood (i.e., the roles of light/temperature, local acclimation/adaptation, and species-specific differences), it does provide some support for the validity of the DRRH.

The evidence to support the idea that mesophotic reef areas will act as a viable reproductive source for shallow reef areas is, however, limited. Much of our knowledge of recruitment and dispersal in corals is limited to data on established or mature communities and does not include metrics on the possibilities when new habitat becomes available (e.g., Grigg and Maragos 1974; Tomasik et al. 1996; Vermeij 2006). Based on patterns of the distribution across depth alone, the upper mesophotic zone (30–60 m) yields the greatest potential for larval linkages between deep and shallow communities, because it contains a large number of ‘depth-generalist’ species. It must be noted that some ‘depth-generalist’ species occur in cryptic locations on the shallow reef (e.g., Madracis pharensis; Vermeij and Bak 2002), which limits their ability to colonize bare substrate on the shallow reef. Despite the overlap in coral community structure between the shallow reef and upper mesophotic zone, our understanding of the genetic structure and recruitment biology of such coral species is insufficient to be able to distinguish panmixis versus genetic structuring over depth. The predominance of corals with a brooding reproductive mode and zonation of symbionts over depth suggests that the capacity for genetic exchange between the shallow and deep reef may be limited to a subset of coral species on Caribbean reef slopes.

Recommendations

The considerable information available for shallow-water coral populations contrasts the relatively small amount present for coral communities occurring over 30 m (Bak et al. 2005; Menza et al. 2008). Given their potential importance with respect to global issues such as climate change, greater efforts should be exerted to understand the ecology and structural heterogeneity of mesophotic coral communities, and the potential linkages that exist with shallow-water counterparts. In formulating this review, a number of critical research areas become apparent not only for the Caribbean but for reef ecosystems globally.

The first area concerns the distribution and abundance of mesophotic coral ecosystems (in relation to shallow coral reefs), and the role of marine protected areas (MPAs) in providing protection for these important and often biologically diverse ecosystems. So far, only a small number of mesophotic communities have been described (Kahng et al. 2010). Thus, targeted mapping explorations should be undertaken at locations where the requirements for deep reef formation are likely to be met (i.e., oligotrophic conditions and clear waters, availability of hard substrate, and absence of a strong shallow thermocline). Given the logistical complexity of research in the mesophotic realm, such studies would greatly benefit from international collaborations whereby the cost-intensive facilities can be shared among different research groups and funding agencies. The second area involves improving our understanding the relationship between physical, chemical, and biological stress factors and depth. This is especially important in terms of understanding the resilience of reefs to future disturbances arising from climatic change. In particular, the notion that temperature-induced bleaching is generally restricted to shallow depths urgently needs to be evaluated by pre- and post-bleaching surveys over broader depth ranges (i.e., encompassing the mesophotic zone), coupled with long-term temperature monitoring over depth. It is crucial to determine the causal mechanisms behind the bathymetric patterns of coral bleaching, such as the role of cold-water influxes in providing thermal relief and the occurrence of local adaptation/acclimation in coral communities to the distinct thermal patterns occurring on the shallow and deep reef.

Finally, perhaps the most urgent research question concerns whether mesophotic coral communities are able to contribute recruits to shallow-water habitats. High-resolution genetic studies should be able to assess the connectivity between shallow and deep coral populations in a variety of species covering both brooding and broadcasting strategies (and both vertical and horizontal symbiont acquisition modes). Ideally, these studies should assess both the coral host and their associated Symbiodinium and should be conducted in conjunction with regional geographic population genetic assessments to allow comparison between intra- and inter-reef genetic structuring. Even though genetic studies will only provide an indirect measure of reef connectivity, the information will indicate whether ‘depth-generalist’ species are indeed suitable candidates to aid in the rapid recovery of shallow reefs. Additionally, reciprocal depth-transplantations of offspring from shallow and deep coral colonies will inform us about the respective roles of genetic adaptation versus phenotypic plasticity in the opportunistic success of ‘depth-generalist’ coral species and the ability of offspring to survive outside their parental depth range.

Conclusion

Reef communities represent a dynamic equilibrium between processes that reestablish coral reefs (‘recovery’) and those that lead to the deterioration of reef communities (‘disturbance’). If anything, the upper mesophotic (30–60 m) zone holds the greatest potential to aid in reef recovery following disturbance due to the species overlap with the shallow reef and the ability to (partially) escape certain disturbances. Although there is clearly much to learn about mesophotic coral ecosystems and restrictions seem to apply, their potential to act as refugia and subsequent reproductive sources for shallow reef areas remains a hopeful aspect of the biology of coral reefs as they enter a century of unprecedented human disturbance.