On the paradox of thriving cold-water coral reefs in the food-limited deep sea

The deep sea is amongst the most food-limited habitats on Earth, as only a small fraction ( < 4%) of the surface primary production is exported below 200 m water depth. Here, cold-water coral (CWC) reefs form oases of life: their biodiversity compares with tropical coral reefs, their biomass and metabolic activity exceed other deep-sea ecosystems by far. We critically assess the paradox of thriving CWC reefs in the food-limited deep sea, by reviewing the literature and open-access data on CWC habitats. This review shows ﬁ rstly that CWCs typically occur in areas where the food supply is not constantly low, but undergoes pronounced temporal variation. High currents, downwelling and/or vertically migrating zoo-plankton temporally boost the export of surface organic matter to the seabed, creating ‘ feast ’ conditions, interspersed with ‘ famine ’ periods during the non-productive season. Secondly, CWCs, particularly the most common reef-builder Desmophyllum pertusum (formerly known as Lophelia pertusa ), are well adapted to these ﬂ uctuations in food availability. Laboratory and in situ measurements revealed their dietary ﬂ exibility, tissue reserves, and temporal variation in growth and energy allocation. Thirdly, the high structural and functional diversity of CWC reefs increases resource retention: acting as giant ﬁ lters and sustaining complex food webs with diverse recycling pathways, the reefs optimise resource gains over losses. Anthropogenic pressures, including climate change and ocean acidi ﬁ cation, threaten this fragile equilibrium through decreased resource supply, increased energy costs


I. INTRODUCTION
The deep sea is the largest habitat on Earth, located below the continental shelf break from ca. 200 m water depth (Ramirez-Llodra et al., 2010, 2011. Here, far below the productive waters at the ocean surface, coldwater corals (CWCs, Fig. 1) form reefs of surprisingly high biodiversity, biomass, and metabolic activity (Fig. 2, Freiwald et al., 2004;Roberts, Wheeler & Freiwald, 2006;Cathalot et al., 2015;De Clippele et al., 2021a). CWCs encompass different taxonomic groups within the classes Hexacorallia, Octocorallia (sensu McFadden, Ofwegen & Quattrini, 2022), and Hydrozoa. In contrast to their well-known warm-water relatives, CWCs occur at relatively low temperatures (i.e. <14 C for reef-building CWCs; Freiwald et al., 2004;G omez et al., 2022), facilitating an almost global distribution in the deep sea (i.e. >200 m water depth, Fig. 3A). However, they can also occur at shallower depths, e.g. at 36 m depth in Norwegian fjords, when oceanographic conditions are suitable. Many CWCs create structurally complex and diverse 'marine animal forests' (reviewed by Rossi et al., 2017), including soft coral (Octocorallia) gardens (e.g. Long et al., 2020;Schejter et al., 2020), black coral (Antipatharia) gardens (e.g. Rakka et al., 2017Rakka et al., , 2020, lace coral (Stylasteridae) gardens (e.g. Di Camillo et al., 2017), and stony coral (Scleractinia) reefs. In this review, we focus on scleractinian CWCs that can form or contribute to large, long-lasting carbonate reefs in the deep sea (Table 1). Small coral polyps (<1 cm diameter; Filander et al., 2021) secrete an aragonite (calcium carbonate) skeleton and together can form 1.5 m-high CWC colonies ( Fig. 1; Wilson, 1979). Live polyps are restricted to the upper and outer parts of the colonies, as coral polyps in the inner parts become shaded from food-delivering currents and die ( Fig. 2; Wilson, 1979;Mortensen & Fosså, 2006;Hennige et al., 2021). The carbonate skeleton becomes exposed to physical and biological erosion (Beuck & Freiwald, 2005), which causes fragments to break off and develop into new colonies around the original colony (Fig. 2). The extending coral framework traps mobile (e.g. resuspended) sediment, leading to framework cementation and formation of elevated, kilometres-long CWC reefs (Dorschel et al., 2005;Roberts et al., 2006). Over time, CWCs develop large carbonate structures such as the CWC mounds at Rockall Bank in the North East Atlantic, which are 300 m high and thousands to millions of years old (Roberts et al., 2006).
The high biodiversity, biomass and metabolic activity of CWC reefs appear paradoxical, as the deep seafloor belongs to the most food-limited habitats on Earth (Ramirez-Llodra et al., 2010). Due to the absence of light, CWCs lack zooxanthellae (Freiwald et al., 2004), i.e. symbiotic dinoflagellates that contribute to the nutrition of most reef-building shallow-water corals (Goldberg, 2013). Like most deep-sea benthos, CWCs depend on organic matter produced in the sunlit surface waters, such as phytodetritus (remains of phytoplankton), zooplankton, and zooplankton remains (e.g. dead Reef-forming or structure-forming cold-water corals (CWCs) (see Table 1 zooplankton, faecal pellets) (Duineveld, Lavaleye & Berghuis, 2004;Duineveld et al., 2007;Naumann et al., 2015;van Oevelen et al., 2018). Phytodetritus and zooplankton faecal pellets aggregate and slowly sink from the photic zone to the deep sea as exported particulate organic matter (POM) (Turner, 2015). In addition, live zooplankton perform diurnal and/or seasonal vertical migrations between shallow and deep waters (Bandara et al., 2021). As passive suspension feeders, CWCs require currents to deliver food particles (Hamann & Blanke, 2022), which they capture with their tentacles or with mucus nets (Zetsche et al., 2016;Murray et al., 2019).
With increasing depth, the vertical flux of particulate organic carbon (POC) decreases exponentially (Fig. 4A), as zooplankton and bacterioplankton consume a substantial fraction of the sinking organic material (Suess, 1980;Lutz, Dunbar & Caldeira, 2002;Boyd & Trull, 2007). In the North Atlantic, 2.5 to 10 mol POC m −2 year −1 is exported from the photic zone in coastal areas, but only around 0.1 mol POC m −2 year −1 (1-4%) arrives in the bathyal deep sea (>200 m; Fig. 4A, see online Supporting Information, Table S1). On CWC reefs, 0.0003-3.5 mol POC m −2 year −1 are deposited, as measured by sediment traps deployed at 481-850 m water depth (Fig. 4A). At the same time, CWC reefs respire 4-45 mol oxygen m −2 year −1 , up to 20 times more than adjacent soft-sediment communities [respiration rates extrapolated from daily measurements of reef-community oxygen consumption on different reefs in the North Atlantic during spring/summer (Cathalot et al., 2015;De Froe et al., 2019)]. Assuming a respiratory quotient CO 2 :O 2 of 1 (as estimated for CWCs; Khripounoff et al., 2014), this leads to a C turnover of 4-45 mol C m −2 year −1 ( Fig. 4B; note that C refers to organic C throughout this review). To sustain this high C turnover, the reefs require an amount C that is orders of magnitude higher than the amount of POC deposited in sediment traps (Fig. 4B). The reefs even seem to require a substantial part of the entire primary production at the ocean surface [almost 100% of the local primary production above Mingulay Reef (De Clippele et al., 2021a); 5% of the primary production on the entire Norwegian shelf by all known Norwegian reefs combined (Cathalot et al., 2015)]. This poses the question: how can CWC reefs sustain their high biomass and metabolic activity in the food-limited deep sea, given this apparent mismatch in C supply (Fig. 4B)?
To approach this 'CWC reef paradox', we review the literature and open-access data and discuss the following questions: are CWC reefs limited to locations with elevated food supply, and what mechanisms increase the food supply to the reefs (Section II); how are CWCs adapted to their food environment (Section III); how do CWC reefs function as ecosystem to maintain high metabolic rates (Section IV); and how do anthropogenic threats impact food supply, coral adaptations and reef functioning and jeopardise the existence of CWC reefs; from a trophodynamic perspective, how can we optimally protect CWC reefs? (Section V).    Table 1. Colonial, reef-building and/or structure-forming cold-water coral (CWC) species, their colony morphology, the degree to which they are reef-building, their distribution, depth range (mean ± standard deviation), and the number of publications mentioning each species (Google Scholar, 01 September 2022). All species are cnidarians of the order Scleractinia. Only species that occur at >200 m water depth were considered. References for presented information: Zibrowius (1980); Cairns (1982Cairns ( , 1995; Koslow et al.

II. FOOD SUPPLY TO COLD-WATER CORAL REEFS
(1) Are cold-water coral reefs limited to locations with elevated food supply?
CWCs are supplied with food through primary production at the ocean surface (measured as chlorophyll-a concentration or net primary productivity), export of primary production, flux of POM to the seafloor and/or currents (horizontal and/or vertical) that carry food particles (e.g. Duineveld et al., 2004;Kiriakoulakis et al., 2004Kiriakoulakis et al., , 2007Davies et al., 2009;De Froe et al., 2022). To evaluate whether CWC reefs are limited to locations with elevated food supply, we first reviewed modelling studies that predict habitat suitability for reef-building/structure-forming CWCs (Table 1), based on their environmental requirements including 'food supply' (primary production, export, POC flux and/or current velocity; Table 2). Currents were identified as an important predictor of CWC habitat suitability in nine out of 13 models that explicitly included currents ( Table 2). The importance of surface primary production and/or POM flux varied between the habitat suitability studies: 40% of the 30 relevant models (i.e. models that included the relevant parameters) describe the importance of primary production/POC flux as (relatively) high, 27% as moderate, and 33% as (relatively) low.
Secondly, we carried out a global analysis to test specifically whether surface primary productivity and/or benthic current velocity were above global average at those sites where either of the eight reef-building/structure-forming deep-sea CWC species (Table 1) occur. Publicly available data on the occurrence of these were obtained from OBIS (2022; data sets are available at https://doi.org/10. 5281/zenodo.7097065). Raster layers with values for (i) the annually averaged surface primary productivity; (ii) the annual range of surface primary productivity as an indicator of seasonality; and (iii) the annually averaged current velocity were extracted from Bio-ORACLE (Tyberghein et al., 2012;Assis et al., 2018;resolution 5 arcmin, i.e. ca. 9.2 km at equator). Bio-ORACLE compiles a global environmental data set for species distribution modelling, based on data sets provided by the E.U. Copernicus Marine Service Information (Assis et al., 2018). For a detailed description of the methodology, see online supporting information Appendix S1; the full R code for all analyses is available at https://doi.org/10. 5281/zenodo.7097065.
Our global analysis revealed that enhanced food supply through above-average surface productivity and currents drive the distribution of most reef-forming CWC species. Firstly, five out of eight species (i.e. Desmophyllum pertusum, Enallopsammia profunda, Goniocorella dumosa, Madrepora oculata, Solenosmilia variabilis) occur in locations with higher primary productivity than the global average (Figs 3B and 5A, Table 3, Appendix S2). Some species, however, occur in areas of non-enhanced (Enallopsammia pusilla, Enallopsammia rostrata) or even lower (Madrepora carolina) primary  Cold-water coral reefs productivity compared to global values (Figs 3B and 5A). A preference for sites with enhanced primary production does not correlate with species-specific differences in depth range (Table 1). Overall, the corals tolerate a broad range of annual primary productivity (Table 3). Secondly, six out of eight reef-forming CWC species occur in areas with above global-average current velocity (Figs 3D and 5C, Table 3). According to our analysis, reef-forming CWCs occur under current velocities of 0.11 ± 0.07 m s −1 (mean ± SD; Table 3). With their large, three-dimensional, branching skeletal framework, the corals locally reduce current velocities and create niches with optimal flow for their prey capture (Hennige et al., 2021;Sanna, Büscher & Freiwald, 2023), i.e. 0.05 m s −1 for phytoplankton (phytodetritus) capture and 0.02 m s −1 for zooplankton capture (Purser et al., 2010;Orejas et al., 2016). Currents further drive CWC distribution as they facilitate specific hydrodynamic processes, which spatially and temporally increase the food supply (see Section II.2).
In summary, food supply is an important driver of CWC reef distribution and the combination of enhanced current velocity and increased primary production (among other environmental drivers) can act as powerful predictors of CWC reef presence. However, food supply on and around CWC reefs varies at small spatial and temporal scales, as reviewed below. Accordingly, attempts to predict CWC reef distribution are more or less limited by the resolution of environmental data and the lack of true coral-absence data, due to the high logistic effort and associated costs of surveys (Davies et al., 2008;Georgian, Morgan & Wagner, 2021). In the future, increasing resolution of environmental data, especially of less-common parameters such as POM flux, high-resolution hydrodynamic models and an increasing number of benthic deep-sea surveys globally (Ramirez-Llodra et al., 2010) will likely improve our ability to predict CWC reef occurrences .
(2) Food pulses created by hydrodynamic processes Instead of a constantly low food supply, CWCs live in a dynamic environment, where hydrodynamic processes create periodic food pulses at different temporal scales, from a few hours for processes linked to internal tidal activity (e.g. Davies et al., 2009;Duineveld et al., 2012;De Froe et al., 2022), to seasonal cycles (e.g. Mienis et al., 2009;Navas et al., 2014;van der Kaaden et al., 2021) and multiyear cycles such as decadal oscillations (e.g. Guihen, White & Lundälv, 2012;Kazanidis et al., 2021b;Raddatz et al., 2022) and millennial-scale oscillations (Portilho-Ramos et al., 2022). Currents interact with the seafloor, especially with elevated seafloor structures such as oceanic banks [e.g. Galicia Bank and Rockall Bank in the North Atlantic (Duineveld et al., 2004White et al., 2005)], continental margins [e.g. the shelf edges of the Faroe islands and Norway (Frederiksen, Jensen & Westerberg, 1992;Thiem et al., 2006)], seamounts (globally, reviewed by White et al., 2007), and fjord sills [e.g. in Norway (Rüggeberg  Cyr et al., 2016;Soetaert et al., 2016;van der Kaaden et al., 2021). These current-topography interactions generate periodic hydrodynamic events like internal tides, trapped waves, and hydraulic jumps van Haren et al., 2014;Cyr et al., 2016). For instance, if an elevated structure partially blocks (tidal) currents, the isopycnals are depressed downstream of this structure, resulting in so-called 'hydraulic jumps' . Isopycnal depressions and hydraulic jumps accelerate the downward transport of organic matter, from typical particle sinking speeds of a few to hundreds of metres per day (Riley et al., 2012) to vertical transport at 10 cm s −1 , corresponding to >8.5 km day −1 (Davies et al., 2009;Juva et al., 2020). As a result, fresh organic matter is transported from surface waters to CWC reefs in less than 1 h [at the 140 m deep Mingulay reef (Davies et al., 2009;Findlay et al., 2013)]. As the tide reverses, this food pulse moves over the reef and supplies the entire reef community (Davies et al., 2009). Accordingly, fresh, lipid-rich (high-quality) suspended POM has been documented in the bottom water above several CWC reefs/mounds in the North Atlantic (Kiriakoulakis et al., 2007;Mienis et al., 2007;Davies et al., 2009;De Froe et al., 2022). Enhanced concentrations of fresh POM can occur in diurnal or semi-diurnal pulses, linked to the site-specific internal tidal cycle Mienis et al., 2007;Davies et al., 2009;De Froe et al., 2022). The downward transport of surface organic matter is most pronounced on or close to the reef crest or mound summit (Cyr et al., 2016) Jones, Lawton & Shachak, 1994). Next to accelerating vertical particle transport, tidal currents and internal waves resuspend deposited organic material into bottom or intermediate nepheloid layers, providing another temporal food source for the reefs (Frederiksen et al., 1992;White et al., 2005;Mienis et al., 2007). Besides tidally induced vertical transport, Ekman drainage was proposed as yet another mechanism D. pertusum  Cold-water coral reefs of food supply to CWC reefs at shelf break regions in the North Atlantic (White et al., 2005;Thiem et al., 2006;Simpson & McCandliss, 2013). Here, an along-slope surface current is deflected at depth, at an angle of 90 , due to the Coriolis force (Ekman, 1905). As a result, organic matter-rich bottom water from shallower areas above the continental shelf is transported downwards across the slope at velocities of ca. 2 cm s −1 (White et al., 2005;Thiem et al., 2006;Simpson & McCandliss, 2013). Due to lower velocities and longer distances cross-slope, particle transport from shallow to deep water by Ekman drainage is slower compared to tidally induced processes, but faster than passive settling (White et al., 2005;Davies et al., 2009;Riley et al., 2012;Juva et al., 2020).
In summary, depending on the prevailing hydrodynamic regime, food availability on CWC reefs can change drastically within a couple of hours, creating a 'feast-famine' environment. The relative contribution of different hydrodynamic regimes to food supply is likely reef/region-specific, but this remains to be studied.

(3) Seasonal variability of food pulses
Most reef-forming CWC species (six out of eight) occur in areas where primary productivity shows higher than average annual variation (Figs 3C and 5B, Table 3, Appendix S2). In temperate regions, such as the North Atlantic and the South Pacific, the seasonal cycle of light, temperature and nutrient replenishment gives rise to a pronounced spring phytoplankton bloom, followed by a peak in zooplankton abundance (Lalli & Parsons, 1997). In the Gulf of Mexico, seasonal upwelling, variations of mixed layer depth and riverine nutrient discharge cause a strong variation in primary productivity (Müller-Karger et al., 1991;Zavala-Hidalgo et al., 2006). Seasonal variations in upwelling are also responsible for primary productivity fluctuations on the Mauritanian continental margin (Eisele et al., 2011) and in the North East Pacific California Current System (Gruber et al., 2012;G omez et al., 2018). Periodic peaks of surface primary production create important food peaks for deep-sea benthos (Billett et al., 1983). During the seasonal phytoplankton bloom, POC flux, POC concentration and zooplankton abundance on Tisler reef (North East Skagerrak, depth <200 m) and Nakken reef (Norwegian fjord, depth 200 m) increase by a factor of two or more Maier et al., 2020a). Similarly, one/several annual peaks of fluorescence (ca. 30% increase) above the CWC reefs in the Cape Lookout area (NW Atlantic) and the Gulf of Mexico indicate seasonal pulses of fresh organic matter (Mienis et al., 2012. In between the seasonal food peaks, the availability of phytodetritus on the reefs is low (Duineveld et al., 2004Mienis et al., 2012Mienis et al., , 2014van Engeland et al., 2019). Alternative resources may then be available, e.g. resuspended, more degraded organic matter Maier et al., 2020a;van der Kaaden et al., 2021), bacterioplankton, and dissolved organic matter (DOM) (Wild et al., 2008).
Today we know that daily, seasonal, annual, and decadal cycles of primary production shape deep-sea ecosystems, just like in shallow waters. However, the study of 'deep-sea seasonality' remains difficult, especially in temperate and sub- Table 3. Primary productivity (annual average and annual seasonal range) and current velocity (annual average) at sites with cold-water corals (CWCs) compared to the global mean. Coordinates of CWC sites were obtained from OBIS (2022), environmental parameters at CWC sites were provided by Bio-ORACLE (Tyberghein et al., 2012;Assis et al., 2018), as described in Appendix S1. For CWCs, values are given separately for the indicated CWC species and pooled for all species (CWCs pooled). Temperatures in CWC habitats are additionally compared to the range of temperatures on tropical coral reefs (Freiwald et al., 2004) because temperature influences physiological processes (Dodds et al., 2007). All given values are model estimates, given at a resolution of 5 arcmin (ca. 9.2 km equatorial), not accounting for small-scale variability.

Parameter
Mean ± SD at locations with CWCs (4) The role of live zooplankton Vertically migrating zooplankton export substantial amounts of C, especially lipids, from the ocean surface to the deep sea (J onasd ottir et al., 2015;Kiko et al., 2017;Bandara et al., 2021). Most zooplankton spend the night in surface waters, grazing on phytoplankton, and descend to deeper waters (>300 m) at dawn to escape visual predators (Zaret & Suffern, 1976 Zooplankton are typically also abundant on and above seamounts, sustaining high CWC abundance (Rogers, 1994;Duineveld et al., 2004;Rowden et al., 2010). By contrast, the large CWC reefs on the slope of Rockall Bank (North-East Atlantic, 800 m depth) showed a low zooplankton abundance De Froe et al., 2022). Site-specific differences in zooplankton abundance in the deep sea relate to patterns of primary productivity (Hern andez-Le on et al., 2020). Furthermore, local zooplankton populations have specific depth ranges, varying from 300 to 600 m, e.g. at the Bermuda-Atlantic-Time-Series station (Sargasso Sea; Steinberg et al., 2000), to >1000 m depth, e.g. in the Gulf of Mexico (Ochoa et al., 2013;Ursella et al., 2021). Finally, zooplankton overwinter at depth in specific areas like the Norwegian Sea at ca. 600 m depth, whereas other areas like the deep Rockall Trough or shallower Norwegian fjords show low zooplankton abundance in winter (Heath et al., 2000;Campbell & Dower, 2003;Maier et al., 2020a). Corresponding to variations in abundance, the importance of live zooplankton in the CWC diet varies among reefs (Duineveld et al., 2004Carlier et al., 2009;van Oevelen et al., 2018) and seasons (Maier et al., 2020a). However, most information on zooplankton food for CWCs originates from the North Atlantic, and future research is required to reveal global patterns.
In conclusion, CWC reefs occur in a highly dynamic 'feast-famine' environment, where food availability changes substantially, depending on the season, the prevailing hydrodynamic regime, and presence of vertically migrating zooplankton. During feast conditions, food supply likely sustains reef C demand. For instance, during the productive season, POC fluxes of 3-67 mmol C m −2 day −1 were measured by sediment traps (Duineveld et al., 2004;Lavaleye et al., 2009;Mienis et al., 2012;Khripounoff et al., 2014), which is in the same order of magnitude as the reef C turnover of 11-123 mmol C m −2 day −1 (4-45 mol C m −2 year −1 , Fig. 4B). It should be noted that sediment traps tend to underestimate POC fluxes, especially when they tilt under high currents (Khripounoff et al., 2014). Furthermore, additional input of zooplankton and/or dissolved organic carbon (DOC), which are not measured by sediment traps, may fill the remaining gap. More precise reef C budgets, however, require C flux data at higher temporal and spatial resolution and the inclusion of C fluxes through zooplankton and DOC. The comparatively low food supply to the reefs in 'famine' periods (Duineveld et al., 2004Mienis et al., 2012Mienis et al., , 2014van Engeland et al., 2019) is reflected in the large annual C mismatch (Fig. 4B) and suggests that other physiological and ecological mechanisms are at play, as outlined in the following sections.

III. ADAPTATIONS OF COLD-WATER CORALS TO THE FEAST-FAMINE ENVIRONMENT
As established in Section II, CWCs live in a feast-famine environment with strong temporal fluctuations in resource availability and hydrodynamic conditions. At the same time, CWCs have a high C demand, as illustrated by the following simplified annual C budget for Desmophyllum pertusum (Fig. 6), the best-studied scleractinian CWC species. Respiration rates of D. pertusum, measured by ex situ and in situ incubations, range between 1.1 and 2.8 mmol C (g coral dry mass) year −1 , if CO 2 :O 2 = 1 (see Section I; Larsson et al., 2013b;Larsson, Lundälv & van Oevelen, 2013a;Khripounoff et al., 2014;Maier et al., 2019Maier et al., , 2020aBaussant et al., 2022). C flux ('loss') associated with mucus release has been assessed in fewer studies by ex situ incubations (Maier et al., 2011(Maier et al., , 2016Naumann, Orejas & Ferrier-Pagès, 2014;Maier et al., 2019) and amounts to 0.2 mmol POC (g coral dry mass) −1 year −1 ; other studies used different units, see Appendix S3). Together, the C demand of D.
pertusum amounts to 1.3-3.0 mmol C (g dry mass) −1 year −1 (Fig. 6, Appendix S3); similar C budgets for other CWC species remain to be investigated. Additional energetic costs incur for reproduction, but these are less well established: it is known that D. pertusum, E. rostrata, G. dumosa, and S. variabilis release a large number of small eggs and sperm (i.e. 'organic C') once per year (broadcast spawning), while M. oculata produces several smaller cohorts of larger eggs and sperm (Burgess & Babcock, 2005;Waller & Tyler, 2005). The related C expenditure is difficult to assess, but D. pertusum showed a 50% lower organic C content after than before the spawning season (January-March in Norway; Brooke & Järnegren, 2013). This difference of 1.5 mmol C (g dry mass) −1 may correspond to the amount of spawned C and suggests a relatively high reproductive

Cold-water coral reefs
cost (Maier et al., 2020a). In total, the annual C budget of D. pertusum (Fig. 6)  CWCs seem to meet large parts of their annual C demand by efficiently exploiting large food pulses. As passive suspension feeders, CWCs rely on water flow for their food supply (Gili & Coma, 1998). The corals are optimally adapted to site-specific hydrodynamic conditions and food availability, through variable colony morphology among and within species (De Clippele et al., 2017b;Vad et al., 2017;Hennige et al., 2021;Sanna & Freiwald, 2021;Sanna et al., 2023). For example, fan-shaped colonies (M. carolina, E. rostrata) often grow perpendicular to unidirectional currents, optimising prey capture (Fricke & Meischner, 1985). D. pertusum adapts its morphology based on the prevailing hydrodynamic conditions (Sanna et al., 2023), forming bush-like colonies with more compact upstream branches under higher, unidirectional flow velocities and symmetrical cauliflower-shaped colonies with thinner, ramified branches under more sheltered conditions with lower, multidirectional flow (De Clippele et al., 2017b). These adaptations create efficient suspension feeders. At concentrations of 100 copepods l −1 , D. pertusum can catch around 12-23 copepods polyp −1 h −1 , depending on the flow speed (Orejas et al., 2016; highest capture at 2 cm s −1 , lowest at 10 cm s −1 ). This corresponds to a C uptake of 11-150 μmol C polyp −1 h −1 (using the copepod C content range mentioned above; Appendix S3). Prey capture rates increase further with increasing zooplankton concentration (Purser et al., 2010). The smaller polyps of M. oculata show lower capture rates of large zooplankton compared to D. pertusum, but on the scale of a coral colony, this effect is outweighed by the higher polyp density (Tsounis et al., 2010). At high phytoplankton concentrations, as may occur during rapid downwelling events in the productive season (Davies et al., 2009), D. pertusum is able to retain up to 6 × 10 4 phytoplankton cells polyp −1 h −1 , i.e. ca. 200 μmol C polyp −1 h −1 (Orejas et al., 2016). Hence, during such zooplankton or phytoplankton food pulses, the corals might be able to sustain 1-17% of their annual C demand in only 1 hour (for calculation, see Appendix S3). In situ data confirm the potential of CWCs to exploit food pulses and their energetic value: on the Norwegian shelf, D. pertusum changes its polyp (feeding) activity in accordance with diurnal changes in current speed and direction, which likely cause periodic changes in food availability (Buhl- Mortensen, Tenningen & Tysseland, 2015;Osterloff et al., 2019). On Nakken reef (Norwegian fjord), D. pertusum doubled its organic C content or biomass during the annual spring bloom (Maier et al., 2020a), showing the importance of seasonal food pulses for the annual C budget of these corals.
CWCs store excess assimilated resources from food pulses or periods of high food availability as tissue reserves . To store reserves, CWCs contain a substantial amount of neutral lipids, i.e. triacylglycerols for short-term storage and wax esters for long-term storage (Dodds et al., 2009;Larsson et al., 2013a;Galand et al., 2020). The build-up of lipid reserves depends on the CWC species and diet. Experiments showed that D. pertusum prefers a zooplankton diet to build-up storage lipids . In situ (Norway), this species formed lipid reserves after the zooplankton bloom in summer, while the earlier spring phytoplankton bloom (phytodetritus peak) was mostly invested in proteins (Maier et al., 2020a). By contrast, M. oculata benefits more from a phytodetritus or mixed phytodetrituszooplankton diet to build up lipid stores .
The original hypothesis that CWCs draw on their lipid reserves in food-limited periods was, however, not confirmed for D. pertusum. When experimentally starved, the corals showed either no decline in storage lipids (Baussant et al., 2017), or a decline in storage lipids that was independent of their feeding status (high versus low food; Larsson et al., 2013a). In temperate areas, the lipid content of D. pertusum did not decrease steadily over winter (Dodds et al., 2009). Nevertheless, the species did show a seasonal cycle of build-up and decline in lipids closely related to its gametogenesis and spawning. Lipid reserves were formed during summer, in the period of highest oocyte growth, maintained until late winter (December), following which >50% were released between December and February, presumably during a mass-spawning event (Brooke & Järnegren, 2013;Maier et al., 2020a). Hence, CWCs seem to use their lipid reserves rather to sustain their high reproductive costs than to overcome low-food periods. It appears likely that food availability governs the reproductive timing of broadcast-spawning CWCs for two reasons. Firstly, synchronising cost-intensive spawning with the approaching spring bloom allows the resource-depleted adult coral colonies to restock (Maier et al., 2020a). Secondly, the coral larvae start feeding at 3 weeks of age and might profit from abundant phytoplankton (Strömberg & Larsson, 2017). Increased food availability may also trigger periodic spawners (e.g. M. oculata) to produce and spawn gametes several times a year (Waller & Tyler, 2005).
Conservation of tissue reserves, in spite of low food availability (e.g. in winter), could be facilitated by a switch to alternative resources. For example in the Rockall Bank area, increased internal wave activity in winter resuspends sediment and resupplies the CWC reefs with more degraded organic particles van der Kaaden et al., 2021), which could, in turn, release DOM and attract bacterioplankton. Feeding experiments demonstrated that CWCs are able to consume these alternative resources, i.e. bacterioplankton, detritus (D. pertusum; Mueller et al., 2014) and DOM [D. pertusum, M. oculata Mueller et al., 2014)]. Their close association with a species-specific microbiome (Hansson et al., 2009;Schöttner et al., 2012) further allows D. pertusum to feed on inorganic resources, such as inorganic C (coupled to nitrification), dinitrogen and ammonium. However, these chemoautotrophic pathways only contributed 2% to their respiratory C demand (Middelburg et al., 2015). Two studies underline consumption of alternative resources in situ. Firstly, the occurrence of E. rostrata and M. oculata in the New Zealand region is driven by DOM concentration (Tracey et al., 2011). Secondly, D. pertusum in a Norwegian fjord switched from a zooplankton-phytodetritus diet to more degraded material/bacteria in winter, indicated by decreased δ 13 C and increased bacteria fatty acid trophic markers (Maier et al., 2020a). Other studies, however, indicate that M. oculata is a less-opportunistic feeder than D. pertusum  and therefore less adapted to variations in resource supply (Chapron et al., 2020). Future research should address variability in resource flexibility among CWC species.
Finally, CWCs may also conserve tissue reserves and energy through low metabolic activity and growth. Facilitated by lower temperatures (Table 3), the respiration rate of CWCs is almost 60% lower compared to their tropical zooxanthellate relatives (Naumann et al., 2011). Correspondingly, CWCs grow about 10 times slower compared to their tropical, shallow-water relatives, i.e. at rates between 0.02% mass increase per day (D. pertusum) and 0.2% per day (M. oculata) (Orejas et al., 2011a,b). Nevertheless, growth rates vary considerably depending on the local environmental conditions and can reach up to 4 cm year −1 , rates comparable to some shallow-water corals (Chapron et al., 2020). While low metabolism and growth may represent general adaptations of CWCs, their reaction to seasonal food shortage differs among species. Under long-term (7-month) experimental food deprivation, D. pertusum slowly reduced its metabolic rate by 40-50% in total, but maintained skeletal growth rates (Larsson et al., 2013a;Baussant et al., 2017). By contrast, its close relative Desmophyllum dianthus (a non-reefbuilding, solitary CWC) immediately and strongly reduced skeletal growth and metabolic rate in response to short-term experimental food deprivation (Naumann et al., 2011). In situ, D. pertusum maintained skeletal growth and budding in seasons with reduced food supply (Lartaud et al., 2014;Maier et al., 2020a). However, increased metabolic rates and decreased skeletal growth during the period of highest oocyte growth suggested an energetic trade-off between reproduction and skeletal growth (Maier et al., 2020a). M. oculata (Mediterranean) showed more pronounced seasonal differences in budding (reduced in summer) and skeletal growth (reduced in winter/spring) compared to D. pertusum, possibly related to its periodic reproduction or a lower dietary flexibility (Lartaud et al., 2014). M. oculata may be more sensitive to varying food availability than D. pertusum (Lartaud et al., 2014;Chapron et al., 2020). Species-specific differences in CWC C budgets and adaptations to low-food periods may explain differences in distribution and resilience; hence, future research efforts should aim at including a broader range of different CWC species.
Interannual growth patterns also relate to fluctuating food supply. In the NW Mediterranean, D. pertusum and M. oculata grew faster (polyp budding and/or linear growth) in years with higher seasonal downwelling intensity caused by episodic dense shelf water events (Chapron et al., 2020). Furthermore, in years with higher sedimentation, the CWCs grew slower, related to less-efficient feeding and energetic costs of Biological Reviews 98 (2023)  Cold-water coral reefs sediment cleaning; in years with particularly strong currents, the corals formed thicker colonies and allocated energy to polyp budding over linear growth (Chapron et al., 2020).
In summary, D. pertusum is well adapted to the feast-famine environment through its high physiological flexibility: (i) phytodetritus and zooplankton food pulses are effectively exploited, but when absent, the corals switch to alternative resources; (ii) growth rates are low, but can be boosted during high food availability; (iii) build-up and expenditure of tissue reserves, mostly for reproduction, are synchronised with the seasonal changes in resource supply. Additional work is needed on the physiological flexibility of other CWC species. By contrast, the accumulated knowledge on D. pertusum provides the chance to move from simple C budgets (Fig. 6) to more complex dynamic energy budget (DEB) models (Kooijman, 2000;van der Meer, 2006). While rarely applied to deep-sea species due to limited data availability, DEB theory provides a useful model framework to understand, for example, how species can grow (in size, tissue reserves) and reproduce under varying food availability in the deep sea (Gaudron, Lefebvre & Marques, 2021).

IV. HOW DO COLD-WATER CORAL REEFS SUSTAIN THEIR HIGH FOOD DEMAND?
Reef-building CWCs may be well adapted to their 'feastfamine' environment, yet, the entire CWC reef community, i.e. the corals and reef-associated fauna and microbes require more food to sustain their high biomass and metabolic activity than they receive in terms of deposited particulate organic matter (see Section I). Based on the literature reviewed here, we suggest two interdependent mechanisms that may explain this mismatch in the reef (organic) C budget: (i) the reef acts as a mechanical and biological filter for phytodetritus (POM); and (ii) recycling (re-use) of C and N within the reef community limits material loss.
The branched, porous reef framework brings abundant suspension-feeding epifauna into close contact with detritivores (Mortensen et al., 1995;Henry & Roberts, 2007) and a diverse microbial community (van Bleijswijk et al., 2015), giving rise to a complex food web (van Oevelen et al., 2009). Particles mechanically intercepted by the reef framework serve as a food source for detritivores such as echiuran worms (Kiriakoulakis et al., 2004) or ophiuroids . The tube-building polychaete Eunice norvegica (Linnaeus, 1767) forms a symbiosis with D. pertusum: the polychaete benefits from stealing food particles from the coral and at the same time it keeps its host clean from accumulating detritus and stabilises the reef framework by building tubes between the framework branches (Mortensen, 2001;Roberts, 2005;Mueller et al., 2013).
Moreover, suspension feeders can directly consume and recycle DOM (Fig. 7, Table 4). Deep-sea sponges are wellknown DOM consumers (Table 4), just like their shallowwater counterparts (reviewed by De Goeij, Lesser & Pawlik, 2017). Initially, symbiotic microbes were considered responsible for the high DOM uptake of sponges (Reiswig, 1981;Yahel et al., 2003;Ribes et al., 2012). Specifically in high-microbial-abundance (HMA) sponges, microorganisms contribute 20-35% to the total sponge biomass (Reiswig, 1981;Hentschel, Usher & Taylor, 2006). Recent research, however, challenged this paradigm: Firstly, lowmicrobial-abundance (LMA) sponges consume and assimilate DOM at high rates (Table 4; De Goeij et al., 2017), but host microorganisms at concentrations only equivalent to those in the surrounding sea water (Hentschel et al., 2006). Secondly, stable isotope tracer experiments demonstrated direct uptake of DOM by sponge cells (Rix et al., 2020) and incorporation of DOM-derived C into de novo-synthesised and/or sponge-specific fatty acids (Rix et al., 2016; Cold-water coral reefs Bart et al., 2020). Thirdly, several other suspension-feeding taxa on CWC reefs consume and assimilate DOM, including several scleractinian CWC species Mueller et al., 2014), hydrozoans, stylasterid corals, and bivalves (Maier et al., 2020b. The faunal and microbial community growing on and inside the reef framework meets 30% of its respiratory C demand by uptake of natural DOM , demonstrating the quantitative importance of this recycling pathway. In a recycling pathway termed the 'sponge loop', sponges recycle substantial amounts of assimilated DOM to particulate 'sponge detritus' (POM; Fig. 7), which is consumed by reef detritivores (De Goeij et al., 2013). Originally described for tropical coral reefs (De Goeij et al., 2013), the sponge loop finds an equivalent on CWC reefs (Rix et al., 2016;Bart et al., 2021a). Other suspension feeders, such as bivalves, likewise recycle substantial amounts of DOM to detrital POM, in this case bivalve (pseudo-)faeces, which can be consumed by reef detritivores (Maier et al., 2020b). The prevalence of DOM consumption by reef invertebrates (Table 4) suggests that these 'suspension feeder' loops (Fig. 7;Maier et al., 2020b;Bart et al., 2021a) are ubiquitous within the CWC reef community and future research is likely to reveal more.
Finally, DIC and DIN are recycled by reef microbes that grow on and inside the porous reef framework and as symbionts in invertebrates (Fig. 7). Nitrifying bacteria and archaea associated with the reef framework (van Bleijswijk et al., 2015) subsist on (faunal) ammonium, which they transform to nitrate (nitrification; Maier et al., 2021); thereby gaining energy for chemoautotrophic DIC fixation. Furthermore, HMA sponges and their diverse microbiome (Hentschel et al., 2006) perform a variety of nutrient fluxes together, e.g. aerobic and anaerobic respiration, nitrification of sponge-derived ammonium and coupled DIC fixation, denitrification of nitrate to dinitrogen gas, and anaerobic ammonium oxidation (anammox) (Hoffmann et al., 2009;De Kluijver et al., 2021). Microbially fixed C is transferred to the sponge hosts (van Duyl et al., 2020), e.g. via the consumption of microbes via phagocytosis (Leys et al., 2018). Mediating diverse internal and external recycling pathways, sponges act on and connect several trophic levels and play an ubiquitous role in deep-sea ecosystems (Hanz et al., 2022). CWCs also host a microbiome, enabling them to perform similar nutrient fluxes, but these play a minor quantitative role in the total CWC metabolism (Middelburg et al., 2015). Altogether, the microbial community contributes substantially to the total organic matter mineralisation of CWC reefs (van Oevelen et al., 2009;Maier et al., 2021).
In conclusion, the high ecological efficiency of CWC reefs may explain the mismatch in their organic C budget (Fig. 4B). While the 'reef filter' ensures optimal retention of zooplankton, phytodetritus, and bacterioplankton, sediment traps only measure the deposition of phytodetritus (POM). Hence, sediment traps underestimate the actual food retention by CWC reefs, especially under high current velocities (Gardner, Biscaye & Richardson, 1997;Mienis et al., 2009;van Oevelen et al., 2009). Moreover, diverse trophic interactions and material recycling facilitate optimal utilisation of the retained food and exploitation of additional resources beyond deposited POM, such as dissolved organic and inorganic matter. Nevertheless, the quantitative importance of recycling pathways and their link in the reef food web remain vague. Modern mapping attempts of reef biomass and metabolic activity detail the importance of CWC reefs in the regional C cycle (De Clippele et al., 2021a,b). In a logical next step, reef food web (C cycling) models (e.g. van Oevelen et al., 2009) could be updated with recycling pathways to estimate how much of the retained and/or metabolised material is recycled, how filtration and recycling contribute to reef (biomass) growth, and how much metabolic 'waste' material is lost from the reef ecosystem. Another essential question in this context is the importance of (functional) biodiversity for the filtration and recycling capacity of the reef, and hence its biomass, metabolic activity and resilience to changing conditions, which we discuss in the following section.

V. PERSPECTIVES: COLD-WATER CORAL REEFS IN THE ANTHROPOCENE
In the Anthropocene, each of the mechanisms discussed above that sustain CWC reefs in the food-limited deep sea have become threatened. The ocean, including the deep sea, is becoming warmer, less well mixed and more acidic, with pollution, fisheries and mining aggravating global change (Gruber, 2011;Roberts & Cairns, 2014;Sweetman et al., 2017). Anthropogenic environmental change disbalances the energy budget of reef-building corals (Fig. 8A) and destabilises the reef framework, resulting in decreased reef biodiversity and functioning (Fig. 8B).
(1) Disbalanced energy budget of cold-water corals A disbalanced energy budget of CWCs is the result of decreasing food supply on the one hand, and increasing energetic costs on the other (Fig. 8A). As the ocean surface is warming more rapidly than the rest of the water column, the water column becomes more stratified (Bopp et al., 2001;Gruber, 2011;Capotondi et al., 2012). Enhanced stratification decreases the intensity of the hydrodynamic mixing processes (Bopp et al., 2001;Li et al., 2020) that supply CWC reefs with important food pulses (see Section II.2). In addition, reduced upwelling of nutrient-rich bottom water limits diatom growth, further reducing POM export .
At the same time, bottom-water temperature in the bathyal is projected to increase by 3-4 C by 2100 (Mora et al., 2013;Sweetman et al., 2017). Increasing temperatures spur the corals' respiratory activity and metabolic costs (Dodds et al., 2009;Dorey et al., 2020;G omez et al., 2022). Initially, CWCs may benefit from higher temperatures, through enhanced polyp activity, higher prey capture rates (Chapron et al., 2021) and growth (Büscher, Form & Riebesell, 2017;Büscher et al., 2022). Nevertheless, beyond a certain temperature threshold (+4 C), enhanced food intake is no longer sufficient to offset the metabolic energy costs (Chapron et al., 2021) or prey capture decreases, leading to decreased growth, tissue reserves, and eventually death (G omez et al., 2022). Furthermore, even small changes in temperature affect the CWC microbiome, with potential consequences for microbially assisted nutrient acquisition and immune responses (Chapron et al., 2021).
In situ, CWCs already show signs of energetic shortfalls. In seasonal periods of enhanced metabolic activity, possibly due to reproductive tissue modifications, D. pertusum showed on average ca. 70% lower linear skeletal extension rates compared to other seasons (Maier et al., 2020a). Likewise, in years of low downwelling/water-column mixing intensity, the species showed on average ca. 80% lower linear skeletal extension rates compared to high-mixing years (Chapron et al., 2020). Furthermore, corals at heavily trawled sites are non-reproductive, possibly because their energy reserves (and colony size) are too low to afford sexual reproduction (Waller & Tyler, 2005). Altogether, global change is predicted to reduce the habitat suitable for CWCs by 79%  and CWC reef biomass by 38% (Jones et al., 2014) by 2100.
(2) Reduced reef functioning on 'crumbling' reefs On the ecosystem level, decreased coral growth on the one hand, and increased erosion of the reef framework on the other hand (Fig. 8B), threaten the reef carbonate budget (Perry et al., 2013;Büscher et al., 2019). Under low pH, CWC skeletons are more porous and form a less stable reef framework (Hennige et al., 2015). Furthermore, ocean acidification accelerates chemical dissolution and bioerosion of the calcium carbonate reef framework (Wisshak et al., 2012(Wisshak et al., , 2014Hennige et al., 2015). Altogether, this 'coralporosis' produces instable, 'crumbling' reefs of reduced structural complexity, and in case of aragonite undersaturation, CWC 'reefs' consisting primarily of live coral colonies without a dead framework foundation (Hennige et al., 2020).
Reduced structural complexity will likely diminish CWC reef biodiversity and ecosystem functioning (Fig. 8B), similar to tropical coral reefs (Nelson, Kuempel & Altieri, 2016;Sunday et al., 2016;Doo, Edmunds & Carpenter, 2019;Dove et al., 2020). A flat reef structure does not induce downward transport of POM-rich surface water, leading to (further) diminished food supply and retention (White et al., 2005;Mienis et al., 2007;Soetaert et al., 2016). Reef sessile suspension feeders cannot attach to strongly degraded reef framework and coral rubble (Mortensen & Fosså, 2006;Maier et al., 2021). Furthermore, suspension feeders appear particularly sensitive to temperature increase, indicated by their reduced abundance under episodic, interannual temperature highs at the Mingulay Reef (Kazanidis et al., 2021b). Loss of structural complexity and suspension feeders on CWC reefs will likely impair the mechanical and biological 'reef filter' and concomitantly food particle retention ( Fig. 8B; see Section IV). Reduced food availability and decreased biodiversity may restrain recycling pathways. For instance, lower abundance of CWCs could result in lower production of mucoid DOM, less recycling of DOM by the reef fauna, and less production of detrital waste for detritivoresi.e. an attenuated 'suspension feeder (sponge) loop' (Fig. 8B). Reduced resource availability might lead to a further decrease of biodiversity (Fig. 8B), creating a detrimental feedback loop that jeopardises the stability of the reef community (Worm & Duffy, 2003). In conclusion, the Anthropocene climate may substantially damage the functioning of CWC reef ecosystems, but unlike for tropical coral reefs (Hughes et al., 2010), virtually nothing is known about their resilience and potential phase shifts. In a similar way, global change will likely affect other complex deep-sea ecosystems, such as coral gardens and sponge grounds .
(3) Conservation of cold-water coral reefs in the Anthropocene The vulnerability of CWC reefs and their importance as ecosystem service providers (see Section I) has been recognised by the United Nations (UN) Food and Agriculture Organisation (FAO, 2009), declaring CWC reefs as vulnerable marine ecosystems (VMEs), according to the United Nations General Assembly (UNGA) resolution 61/105 (UNGA, 2007). VMEs require special protection, e.g. through marine spatial planning with the designation of marine protected areas (MPAs; United Nations, 2017). Yet, effective marine spatial planning remains difficult, due to limited scientific knowledge on global CWC distribution, lack of historical baseline data (Dur an Muñoz & Sayago-Gil, 2011; Kazanidis et al., 2020;Lim, Wheeler & Conti, 2021), and scarce data on MPA effectiveness (Huvenne et al., 2016). Some stressors are not constrained by protected area boundaries, including oil, other pollutants and impacts from global change; however protected areas can provide resilience to global stressors by maintaining ecosystem function. Continued mapping and characterisation of CWC habitats is critical, but we argue that our improved understanding of the 'cold-water coral reef paradox' should also be incorporated into ecosystem assessment and conservation efforts. Live coral cover has been used as a proxy for CWC reef health (Flögel et al., 2014;Juva et al., 2020), yet differences in CWC cover may be natural and do not necessarily provide information on whether a reef is new or on the verge of disappearing (Hughes et al., 2010). An integrative approach to assess the environmental status of CWC reefs was presented by Kazanidis et al. (2020), including biodiversity indices, coral cover, fish biomass, signs of anthropogenic impacts, etc., as proxies for reef status. Based on Sections II-IV, we suggest complementing these ecosystem descriptors by including (i) organic and inorganic C budgets for (a) CWCs and (b) the reef ecosystem, to evaluate reef growth versus erosion; and (ii) reef functional diversity and food-web complexity, to judge ecosystem functioning and resilience to changing oceanographic conditions. Future research should create a framework to facilitate the assessment of these ecosystem descriptors. For 1a, the energetic status (energy budget) of CWCs at different reefs should be regularly measured, e.g. their metabolic activity and tissue stores (somatic and reproductive). In addition, these measures could be introduced into DEB models, to assess coral Biological Reviews 98 (2023)  Cold-water coral reefs growth under local and potentially changing environmental conditions (e.g. temperature, food supply). Regarding 1b, video transect annotations of CWC reefs can be combined with predictive modelling to create reef-scale biomass, and organic and inorganic C maps; newly developed machinelearning algorithms will accelerate video annotations in the future (De Clippele et al., 2021a). Regarding 2, future research could develop a functional traits database for reef-associated organisms (e.g. their feeding guild) plus open-source code for simple implementation of food-web models. This would allow evaluation of the trophodynamic resilience of different CWC reefs without strong programming skills. Based on these approaches, future research may help to identify CWC reef 'refugia' that are likely to persist during future global change . Spatial measures protecting networks of these refugia from, e.g. bottom-trawling fisheries, appear to be our best chance to preserve CWC reefs and their function as diversityproductivity hotspots in the deep sea.

VI. CONCLUSIONS
(1) Recent major advances allow us to approach the paradox of how CWC reefs sustain high biodiversity, biomass and metabolic activity in the food-limited deep sea. Suggested answers to this paradox reveal key drivers of CWC reef distribution that are required to achieve effective conservation measures.
(2) Most reef-building CWC species occur in areas of enhanced primary production with high seasonal fluctuations and under elevated current velocity (relative to global averages), indicating that food production and supply are important drivers of CWC reef distribution.
(3) Food supply on CWC reefs is not constantly low, but highly dynamic. Within a couple of hours, food availability can change from very low to very high, depending on the season, the prevailing hydrodynamic regime, and presence of vertically migrating zooplankton.
(4) The best-studied reef-forming CWC species D. pertusum is well adapted to these extreme temporal fluctuations in food availability, by (i) high capture rates of phytodetritus and zooplankton, (ii) high resource flexibility (DOM, bacterioplankton, inorganic resources), (iii) investing in large tissue reserves for reproduction, and (iv) synchronising activity with (seasonal) fluctuations in food availability. (5) On the ecosystem level, CWC reefs sustain high metabolic activity and biomass. They achieve this firstly by efficient retention of phytodetritus, zooplankton and bacterioplankton in the 'reef filter', a combination of a mechanical filter provided by the structurally complex, porous reef framework and a biological filter consisting of diverse, abundant suspension-feeding epifauna. Secondly, diverse trophic interactions and material recycling facilitate optimal resource utilisation and exploitation of additional food sources, such as dissolved organic and inorganic matter.
(6) Climate change, ocean acidification, fisheries, mining, and pollution impact reef functioning in various ways, e.g. by reducing organic matter supply, increasing the animals' energy demands, and dissolving the carbonate reef framework, thereby decreasing structural and biological diversity and ecosystem functioning. Research has only started to reveal the vast complexity, drivers and functioning of CWC reefs, but it is crucial to continue this path to facilitate knowledge-based habitat management for sustaining these diversity-productivity hotspots in the future ocean.