Interactions between coral restoration and fish assemblages: implications for reef management

Corals create complex reef structures that provide both habitat and food for many fish species. Due to numerous natural and anthropogenic threats, many coral reefs are currently being degraded, endangering the fish assemblages they support. Coral reef restoration, an active ecological management tool, may help reverse some of the current trends in reef degradation through the transplantation of stony corals. While restoration techniques have been extensively reviewed in relation to coral survival, our understanding of the effects of adding live coral cover and complexity on fishes is in its infancy with a lack of scientifically-validated research. Here, we review the limited data on reef restoration and fish assemblages, and complement this with the more extensive understanding of complex interactions between natural reefs and fishes and how this might inform restoration efforts. We discuss which key fish species or functional groups may promote, facilitate or inhibit restoration efforts and, in turn, how restoration efforts can be optimised to enhance coral fish assemblages. By highlighting critical knowledge gaps in relation to fishes and restoration interactions, we aim to stimulate research into the role of reef fishes in restoration projects. A greater understanding of the functional roles of reef fishes would also help inform whether restoration projects can return fish assemblages to their natural compositions or whether alternative species compositions develop, and over what timeframe. While alleviation of local and global reef stressors remains a priority, reef restoration is an important tool; an increased understanding of the interactions between replanted corals and the fishes they support is critical for ensuring its success for people and nature. This article is protected by copyright. All rights reserved.


| INTRODUCTION
Coral reefs provide critical ecosystem services, including fisheries, coastal protection and tourist income, to millions of people (Barbier, 2017;De Groot et al., 2012;Woodhead et al., 2019). Despite their global importance, protection is currently inadequate (Mora et al., 2006;Pressey et al., 2015;Cox et al., 2017), and consequently key indicators of reef health, such as coral cover, are declining (Bruno & Selig, 2007;Gardner et al., 2003;Hughes et al., 2017;Pandolfi et al., 2003). This reef degradation is driven by anthropogenic impacts, including overfishing, global climate change, coral disease, sedimentation, extensive coastal development, introduction of invasive species and the release of pollutants (Hoegh-Guldberg et al., 2017;Hughes et al., 2003). The loss of coral cover and complexity caused by these stressors is affecting the ecosystem services provided by reefs (Cesar et al., 2003;Pratchett et al., 2014). Global threats require international action, but managing local threats is also critical (Kennedy et al., 2013).
Although establishing marine reserves is perhaps the most commonly used technique to address local reef degradation, it has been suggested that a wider range of methods are required to manage tropical coastal resources and to maintain reef processes (Allison et al., 1998;Anthony et al., 2015;Aswani et al., 2015;Rogers et al., 2014). Reef restoration is one of these potential tools to augment other management methods (Lirman & Schopmeyer, 2016).
The terms reef "restoration" and "rehabilitation" are often used interchangeably in the coral reef literature. Restoration is generally defined as bringing a degraded ecosystem back as close as possible to its original natural state, whereas rehabilitation refers to situations where the functional and structural properties of an ecosystem are replaced, not necessarily in the same manner as the original state (Edwards & Gomez, 2007). In most cases it is likely that although restoration may be desired, rehabilitation is the most achievable outcome, and a shift in management goals from a return to original species composition to the need to maintain ecological functions and ecosystem services of reefs has been suggested (Graham et al., 2014;Hughes et al., 2017). Effort may not always be focused on restoring original reefs, but sometimes creating new habitat for reef communities. There are many examples of entirely artificial reefs (e.g., "reef balls" or sunken ships), where the main focus is on the deployment of artificial structures, usually in areas where reefs did not previously exist or where they have been entirely degraded away, which can then be colonised by marine organisms (Baine, 2001). Although the authors draw on some of the artificial reef literature, the main focus of this review is specifically on how the restoration of existing coral reefs benefits, and is benefitted by, fishes, while acknowledging the potential addition of artificial structures being deployed as part of the process.
The most widespread method of coral reef restoration involves the introduction and distribution of nursery-reared or wild-collected coral fragments in areas previously affected by human actions or adverse environmental conditions (Johnson et al., 2011;McLeod et al., 2019a;Precht, 2006). Coral fragments are either directly transplanted to the substrate (Forrester et al., 2012;Ladd et al., 2019;Lohr et al., 2017) or may be attached to artificial structures which have proven successful in environments dominated by mobile substrata such as coral rubble (Clark & Edwards, 1999;Fadli et al., 2012;Williams et al., 2019). By outplanting corals, managers aim to enhance ecological processes and re-create self-sustaining naturally growing habitats due to the ability of corals to colonise and build complex structures (Edwards, 2010;Edwards & Gomez, 2007). In some instances, other techniques may be used such as the culture and release of coral larvae or juveniles (Chamberland et al., 2017;dela Cruz & Harrison, 2017;Heyward et al., 2002), the transplantation of entire mature coral colonies (Mbije et al., 2013;McLeod et al., 2019b;Schopmeyer & Lirman, 2015) or other organisms such as giant clams (Cabaitan et al., 2008), coral gardening including an intermediate coral nursery phase (Bongiorni et al., 2011;Frias-Torres et al., 2015;Horoszowski-Fridman et al., 2015), algal removal (McClanahan et al., 1999(McClanahan et al., , 2000(McClanahan et al., , 2001 or even the deployment of artificial structures alone to provide a stable substrate for future colonisation (Jayanthi et al., 2020;Ng et al., 2017;Thanner et al., 2006). A description of these coral reef restoration methods can be found in Boström-Einarsson et al. (2020), forming the basis for examining the relationship between these techniques and fish assemblages in this review.
Although restored reefs remain susceptible to global influences such as climate change, disease and pollution, reef restoration may be the last resort for immediate reinforcement of critical ecological functions and services for reefs that have degraded significantly and may not have sufficient resilience to recover (Rinkevich, 2008), or where there is a desire to speed up recovery. For example, there is evidence that reef restoration methods can be used to manage tropical reefs damaged by destructive fishing practices (Fox et al., 2005;Raymundo et al., 2007); when sites have been degraded to the state of rubble fields, there is usually little chance for natural recovery without human intervention (Fox et al., 2003). This loss of benthic structure can have devastating consequences not only on natural fish populations but also on the livelihoods of coastal communities. Millions of people, mainly in developing countries, are dependent on tropical fisheries for income and protein needs (Cinner et al., 2009). Managing reefs through the implementation of restoration projects may help protect and enhance those ecosystem services for the benefit of coastal people (Mumby & Steneck, 2008;Rogers et al., 2015).
It is clear that reefs provide multiple benefits for fishes, the main ones being the provision of food, habitat and settlement substrate (Graham & Nash, 2013;Gratwicke & Speight, 2005;Luckhurst & Luckhurst, 1978). For example, some reef-associated fishes rely specifically on live coral, and many more species benefit from structural complexity provided by the reef environment (Coker et al., 2014).
Consequently, fishes are likely to benefit from restoration, and this is often an implicit or explicit reason for restoring habitat. The importance of fishes in benthic dynamics is also well documented, particularly herbivorous species aiding coral growth and survival by controlling macroalgal cover Hughes et al., 2007;Mumby et al., 2006a). Nonetheless, although the importance of the interactions between reef fishes and their habitat is well established, reef restoration research has focused almost exclusively on coral survival (Lirman & Schopmeyer, 2016;Young et al., 2012) with research into the effects of adding live coral cover and complexity on fishes in its infancy. This review has identified studies which have monitored effects of coral reef restoration on fishes and vice versa.
Following searches conducted with the keywords listed in Table S1 and excluding studies where the main aims did not concentrate on restoring coral reef ecosystems, 38 publications are summarised in Table 1. Although a few of these restoration publications have assessed fish populations directly, fishes were more commonly investigated as secondary qualitative observations. Throughout this review the authors consider the bidirectional interactions between fishes and restored reefs (Figure 1), and how this is governed by coral cover and reef complexity and the various functions of fishes on restored reefs. Furthermore, key research questions to help inform coral reef restoration are identified as restoration programmes intensify globally.

| THE ROLE OF HABITAT AND SEASCAPE COMPLEXITY
There is a general consensus that the availability of complex coral reef habitat is a prerequisite to abundant and diverse coral reef fish assemblages (Luckhurst & Luckhurst, 1978;Bell & Galzin, 1984;Gratwicke & Speight, 2005;Graham & Nash, 2013). Many reef fishes are dependent on complex corals for habitat, shelter from predators and water movement, foraging, spawning and nesting (Almany, 2004a;Caley & St John, 1996;Johansen et al., 2008;Robertson & Sheldon, 1979).
Consequently, the global decline of live corals and associated decrease in reef rugosity has affected resident fish populations, and coral reef fisheries (Alvarez-Filip et al., 2009;Jones et al., 2004;Pratchett et al., 2014;Sano et al., 1984). Restoration may increase coral cover and habitat complexity on a reef within a relatively short time period with the use of fast-growing coral species, which otherwise would take decades to re-establish naturally . For instance, the reported median length of restoration projects is 12 months, suggesting that coral reef restoration may have rapid effects on coral ecosystems (Boström-Einarsson et al., 2020). Nonetheless, although active management methods such as reef restoration have the potential to increase coral cover and fish stocks more quickly compared to some other management tools (Rinkevich, 2005(Rinkevich, , 2008, reported recovery time frames currently vary from months to decades and appear to be context-dependent (Table 1).
Coral reef restoration can provide shelter for fishes either between coral fragments and/or under constructed structures to which the coral fragments are attached (Clark & Edwards, 1999;Fadli et al., 2012). Provision of shelter allows reef fishes to avoid predation (Shulman, 1985), and shelter for herbivorous fishes may in turn help control algal overgrowth on restoration structures, as fish presence has been linked to reduced cleaning maintenance on introduced structures such as coral nurseries (Frias-Torres et al., 2015;Frias-Torres & van de Geer, 2015; see the section "The Role of Herbivorous Fishes").
In one of the few studies that have quantitatively investigated the effects of coral transplantation on fish colonisation, populations were surveyed at selected treatment and non-restored degraded control plots prior to transplantation of staghorn coral Acropora cervicornis fragments, and then again following restoration (Opel et al., 2017).
Fish numbers and diversity were significantly greater in restored plots when compared to control plots within a week of the transplantation, demonstrating the fast rate of fish recolonisation, with the benthic structure as the main predictor of change. By the end of the study, experimental sites had no resemblance to one another with regard to fish assemblages present, as each experimental site attracted unique and distinct populations. Therefore, although initial assemblages on restored reefs may reflect recruitment from adjacent reefs and original fish communities, restored sites may also attract new species and create novel fish assemblages. There is, however, mixed evidence that the artificial addition of live coral cover impacts coral fish populations.
Whereas other studies have reported increases in fish densities and species richness due to coral transplantation (Cabaitan et al., 2008;Clark & Edwards, 1999;Hudson et al., 1989;Lecchini, 2003;Yap, 2009), a recent study by Ladd et al. (2019) investigating established restoration sites of varying coral transplant densities and maturity revealed little impact of the interventions on fish communities, with the exception of coral-associated damselfishes. As quantitative studies of the effect of coral restoration on fish assemblages are still scarce, with fishes rarely the main focus of restoration publications (Table 1) had more successful strikes, with prey mortality increasing when low complexity corals were transplanted (Beukers & Jones, 1998). Transplantation of high complexity corals provided greater refuge opportunity for the focal prey fish, juvenile lemon damsel (Pomacentrus moluccensis) Bleeker 1853. Nonetheless, the increased complexity associated with restoration may still be beneficial to predatory fishes in the longer term. If prey fishes survive capture by escaping into reef refuges, this enhances reef fish productivity, which in turn increases prey fish numbers (Rogers et al., 2014). As prey fish populations rise, some individuals are excluded from refuges through competition, thus exposing them to predators (Holbrook & Schmitt, 2002). The impact of habitat complexity on predators will also vary greatly depending on predatory strategies. Ambush predators may profit from increased structure compared to roaming predators (Almany, 2004b;Rogers et al., 2018). Restoration projects may provide prey shelters as well as predation opportunities by promoting complexity at different spatial scales. For example, creating different-sized holes within artificial reefs may benefit both prey and predators (Bohnsack, 1991). Future research should examine how changes in multi-scale coral complexity associated with coral transplantations affect predator-prey interactions (Table 2), and include both consumptive and non-consumptive ("fear") effects (Mitchell & Harborne, 2020 (23) Elkhorn coral fragments transplanted at a single site across 7 years. Three-spot damsel recruits increased over the last 3 years of the project. All recruits initially settled on transplanted elkhorn corals, but 57% of recruits settled on other microhabitats by end of study. Three-spot damselfish recruits positively selected for larger coral transplants. Recruits on elkhorn corals had higher survival than those on other substrata, but this was not statistically significant. Acropora intermedia and Acropora pulchra fragments transplanted in low and high densities on sandy and rubble substrate. Fish abundance increased by 50% on high density compared to low density and control plots. Fish species richness was higher at transplanted plots than at controls, but there was no difference between the two densities. Fish composition remained similar between all treatment plots. Fish biomass was significantly higher in the high-density plots than in the low-density and control plots.
*  (7) Before After Control Impact study design with nursery-reared transplanted Acropora cervicornis colonies. Mean fish abundance was higher in outplants than in controls despite a decrease in abundance 2 months post-outplanting. Species richness was higher in outplants than in controls. Little similarity in fish assemblage composition between outplants and controls shortly post-outplanting. Note: These studies were identified by systematic searches conducted in Jan/Feb 2020 in Web of Science. Keywords used in the searches are given in Table S1 (supporting information). This approach was complemented with searches in Google Scholar, hand-searching reference lists of included studies and by screening the recently compiled database of coral reef restoration studies in Boström-Einarsson et al. (2020). Artificial reef publications that (a) did not aim to mimic natural reefs (i.e., shipwrecks, jetties, breakwaters), (b) had the primary goal of increasing fishing productivity rather than ecosystem rehabilitation and (c) solely investigated ecological processes (i.e., succession, predation) without clear restoration aims were excluded from this review.
provision is considered (Bohnsack & Sutherland, 1985). In several studies deploying artificial reefs consisting of concrete blocks, the complexity of the blocks including the presence and size of holes within the structures had a significant effect on colonising fish assemblages (Hixon & Beets, 1989;Hixon & Beets, 1993;Sherman et al., 2002). Vertical topography has also been shown to have a significant effect on fish assemblages, with vertical jetty pillars experiencing much higher recruitment than low-relief natural reefs in the Red Sea (Rilov & Benayahu, 2000). Artificial reef design and the presence of holes and cracks providing protection for prey also seem to be particularly important in defining predatory fish assemblages (Da Rocha et al., 2015;Gregalis et al., 2009;Spieler et al., 2001). For example, in an experimental study where artificial reefs containing varying shelter sizes were deployed, smaller shelters were effective in excluding large predators, whereas the presence of larger holes increased the abundance of large piscivores and indirectly reduced prey numbers (Hixon & Beets, 1989). This work provides potential guidance for the spacing and coral growth forms that might most benefit fishes, along with a recognition that complexity occurs at multiple scales (Harborne et al., 2012b).
The provision of shade in addition to physical shelter provided by reef crevices and holes is also likely to be a contributing factor in attracting fish assemblages to restored reefs (Spieler et al., 2001).
Increasing complexity by reintroducing intricate and table-shaped corals may produce areas of shade in which juvenile and nocturnal coral fishes can take cover (Hair et al., 1994;Kerry & Bellwood, 2016;Sheppard, 1981;Stimson, 1985). Provision of shade may effectively conceal vulnerable fishes while allowing them to better spot predatory threats (Helfman, 1981), and shade may provide protection from the damaging effects of UV light (Kerry & Bellwood, 2015b  F I G U R E 1 Summary of interactions between fishes and their restored coral reef habitat. Benefits for fishes include the introduction of complexity for reef-associated fish species that provides shelter for reefassociated species either under artificial structures or within coral transplants (1), which is enhanced by providing transplant species with a range of morphologies, densities and shade-producing properties (2). Fishes will also benefit from increased food sources including coral (2) and other fishes (3). Through these trophic interactions, fishes play positive roles in restoration projects including herbivory to control algae growth (4) and provision of nutrients for coral growth (5), but may also have negative impacts through coral predation (6), and damselfish territories (7) often recruit to reefs post-disturbance and may be heavily impacted by subsequent increases in coral cover associated with restoration (Coker et al., 2012;Opel et al., 2017;Syms & Jones, 2000). Furthermore, artificial patch reefs with small-scale isolation have observed increased fish abundance, species richness and juvenile recruitment when compared to continuous reefs (Belmaker et al., 2011;Schroeder, 1987). Patch reef designs may, therefore, be preferable for reef restoration, providing a range of habitats and encouraging the recruitment of both coral and rubble-associated species. In addition, connectivity with adjacent ecosystems on a larger spatial scale is an important consideration as many coral fishes migrate between different habitats with the presence of nearby resources, such as nursery T A B L E 2 Summary of the interactions between natural reefs and fishes and how this information can be used to optimise the recovery of fish assemblages in reef restoration efforts

Introducing habitat complexity
High complexity corals provide shelter opportunities for prey items, e.g., juvenile fishes, cryptic fishes and invertebrates, reducing predatory success. Increased refuges may enhance prey fish numbers providing more opportunities for predators. Small holes provide shelter for prey; large holes increase predator abundance. Shade-producing corals offer shelter to juvenile and nocturnal fishes as well as protection from UV light. Tabular corals shape fish assemblages even when occupying a small proportion of total coral cover. Connectivity with other ecosystems will greatly affect fish abundance and biomass.
Restoration must increase complexity, providing shelter to support fish communities. This can be incorporated through man-made structures and/or by the transplantation of intricate corals. Varying levels of complexity should be introduced; high complexity will provide shelter for prey fishes, but inclusion of gaps and moderate coral transplantation densities will ensure large-bodied predator success. The provision of shade needs to be included when designing restoration structures. Some shade-producing tabular corals should be introduced in addition to the more popular branching corals.
Where possible, reef restoration projects should be set up close to mangrove and seagrass habitats to enhance fish populations through provision of nursery and foraging areas.

Role of herbivorous fishes
Algae competes with corals for space and will opportunistically overgrow, shade or abrade coral colonies that are vulnerable or damaged.
Grazing may be enhanced with various management practices such as fishing reductions.
Territorial damselfishes can have deleterious effects on vulnerable coral colonies by biting coral polyps to promote algal growth. They are particularly attracted to fast-growing branching coral colonies.
Surveys to ensure sufficient herbivorous fishes are present are recommended prior to restoration. Removal of macroalgae may be necessary during initial stages while healthy grazing populations of fishes establish. Where possible, restoration projects should be located in marine reserves or at locations supporting a high biomass and diversity of grazers from various functional groups. Surveys of territorial damselfish and their known predators should be carried out prior to restoration to determine whether damselfish removal is required. A variety of coral morphologies should be transplanted to help minimise damselfish effects.
Nutrient provision Aggregating fishes supply nutrient-limited corals with added excretory products. Coral morphologies that promote low water flow between branches retain these nutrients more effectively. Fish farms may provide a source of natural enrichment.
In line with transplantation of varying coral morphologies, corals with closed morphologies should be included to enhance nutrient absorption and coral growth.
Consideration should be given to setting up coral nurseries near fish farms as nutrients may stimulate fragment growth. Nonetheless, this needs to be assessed alongside surveys of herbivorous fish populations, as algal overgrowth remains one of the main concerns on nutrient-enriched, coral-poor reef restoration sites.

Corallivory
Herbivores may induce coral recruit mortality through accidental grazing, and corallivores will target juvenile corals through predation. Corallivores may be selective in their coral preferences.
Nevertheless, the positive effect of the cropping of algae by herbivores appears to outweigh the negative effect of occasional predation by herbivorous and corallivorous fishes.
Where corallivory is a problem, rearing juvenile corals to larger sizes ex-situ prior to transplantation is recommended to decrease sizedependent mortality. Surveys to establish the presence of corallivorous fishes are recommended prior to restoration. Outplanting a range of coral species and morphologies could minimise the impact of corallivores.
Predatory fishes Predatory fishes have a vital role in maintaining healthy coral reef ecosystems and supporting fisheries.
Marine-protected areas and reserves, even of small size, can have significant positive impacts on predatory species through the prevention of fishing activity.
Attraction of predator assemblages should be a key aim of restoration projects. Although it may be difficult to identify specific mechanisms for this at the start of a restoration project, surveys of predatory fish populations over restoration time are recommended to inform this aim. Where possible, setting up coral restoration projects within established protected areas will increase their likelihood of success due to the protection of predatory species.
habitats, heavily impacting fish biomass (Mumby et al., 2004;Nagelkerken et al., 2000;Nagelkerken et al., 2002;Ogden & Quinn, 1984). For example, juvenile fishes may have entirely different refuge needs to adults and could greatly benefit from the presence of mangrove or seagrass nursery and foraging habitats near restoration sites. It is therefore essential that an integrated planning approach is taken at the design stage of restoration projects to maximise benefits to fish by considering small-to large-scale physical characteristics of the habitat and seascape (Table 2).

| Cryptic species
Although the need to provide shelter for ecologically and economically important fishes by using appropriate restoration design is well known, cryptic fishes are rarely considered in the context of reef restoration (Table 1). Though not primary targets of restoration, cryptobenthic fishes, or more commonly termed "cryptic" fishes, have an important role in reef assemblages and significantly contribute to fish abundance and diversity but are understudied (Ahmadia et al., 2012;Brandl et al., 2018;Depczynski & Bellwood, 2004;Harborne et al., 2012a). As they constitute common prey for piscivorous primary and secondary consumers and supply a substantial amount of energy to higher trophic levels (Brandl et al., 2018(Brandl et al., , 2019Depczynski & Bellwood, 2003), it is important that they are considered in the context of rebuilding food webs on restored sites. Cryptic species differ from more conspicuous species; they are small (<5 cm), have limited mobility and predominantly live in well-protected cavities formed within coral reef structures (Depczynski & Bellwood, 2003).
Consequently they typically have high site fidelity and are affected by a range of physical characteristics, including habitat complexity and shelter quality (Depczynski & Bellwood, 2004;Kobluk, 1988;Prochazka, 1998;Syms, 1995;Willis & Anderson, 2003). Cryptic species specialising in living within coral habitats are likely to be positively affected by the increase in structural complexity and live coral cover through the transplantation of stony corals, and the introduction of structures to which they are attached (Jaap, 2000). Indeed, the introduction of artificial reefs has previously increased the abundance of small fishes such as cardinalfishes and gobies (Clark & Edwards, 1999;Thanner et al., 2006). Without a fuller understanding of the impact of reef degradation on cryptic species, any potential positive effects of restoration on this group of fish species will be difficult to manage or monitor. Further research is critical to explore cryptic fish population structures across different restoration designs, and how they may aid the recolonisation of higher trophic species.

| THE ROLE OF HERBIVOROUS FISHES
Herbivorous fishes are vital for reef restoration programmes; sufficient grazing is a necessity to prevent algae from smothering coral fragments and outcompeting coral transplants, particularly during the early stages of restoration (Edwards, 2010;Edwards & Gomez, 2007).
Macroalgae rarely overgrow thriving coral colonies. Nonetheless, when coral colonies are already damaged or dead, algae can colonise in an opportunistic manner (McCook et al., 2001), overgrow coral recruits (Box & Mumby, 2007), and algae may also act as disease vectors (Nugues et al., 2004), so that the control of macroalgal cover by grazing fishes is vital. Small coral fragments are particularly susceptible to sub-lethal effects from contact with macroalgae (Ferrari et al., 2012).
This susceptibility is particularly relevant for restored coral colonies where coral fragments may be small and are already stressed due to the transplantation process. Research conducted at several restoration sites within the Florida reef tract found high cover of macroalgae to be a major threat to the survival of A. cervicornis fragments (van Woesik et al., 2018). Indeed, reef restoration is not recommended in areas where grazing populations of fishes and/or invertebrates are scarce as this would prevent restored corals from recruiting in the future, therefore rendering the exercise futile (Edwards, 2010). Surveys of existing fish populations at proposed sites are, therefore, essential (Edwards & Gomez, 2007), and the active removal of macroalgae has been suggested on reefs with reduced herbivory in association with coral reef restoration efforts to improve chances of coral survival while coral fragments establish (Ceccarelli et al., 2018).
Most reef restoration projects are expensive and require extensive time spent cleaning algae from introduced structures such as coral nurseries (Frias-Torres & van de Geer, 2015) and artificial reef modules , often due to the lack of healthy herbivorous fish populations. With their fused beaks, parrotfishes are particularly efficient at removing algae, consequently freeing up space for coral recruits and reducing coral-algal interactions (Abelson et al., 2016b;Bellwood et al., 2004;Ogden & Lobel, 1978), either through targeting algae directly (Adam et al., 2018) or indirectly while feeding as microphages (Clements et al., 2017). Although seemingly less clear in the Pacific (Russ et al., 2015), in the Caribbean there are evident relationships between parrotfish biomass and the abundance of large-sized individuals with macroalgal cover (Shantz et al., 2020;Williams & Polunin, 2001), and consequently restoring parrotfish populations is often the focus of conservation initiatives in the western Atlantic (Jackson et al., 2014). Nonetheless, it is necessary to rebuild the entire herbivorous fish guild, including macroalgae-eating browsers to keep algae communities in an early successional stage (Adam et al., 2015;Burkepile & Hay, 2010;Cheal et al., 2010). For example, the reversal of an experimentally induced algal phase shift was attributed to the batfish Platax pinnatus (Linnaeus 1758), a species not previously classified as a conventional grazer, whereas grazing from parrotfishes and other key herbivorous species had little impact on direct removal of macroalgae (Bellwood et al., 2006). Additional fishes within the herbivorous guild continue to be identified in both the Pacific and Caribbean faunas (Tebbett et al., 2020). Non-fish species such as urchins are also functionally important grazers on many reefs (Edmunds & Carpenter, 2001); thus, a diversity of fish and invertebrate grazers is advocated to promote restoration success.
In an marine protected area (MPA) at Cousin Island in the Seychelles, coral fragments were set up to grow at a coral nursery site located near a healthy local reef, aiming to reduce cleaning costs during the first phase of coral gardening prior to coral transplantation.
The presence of reef fishes removing the nursery of biofouling organisms, such as algae and invertebrates, reduced the usual cleaning time by 60% (Frias-Torres et al., 2015). This trophic facilitation has significant implications for coral reef restoration projects in terms of costeffectiveness (Toh et al., 2013). During a different study at the same MPA, restoration structures were filmed to investigate the importance of grazers and to test a novel cleaning station technique (Frias-Torres & van de Geer, 2015). Within 48 h of nursery rope structures being placed at the experimental site, all biofouling reef algae had been removed by herbivores, therefore eliminating the need for maintenance-cleaning and the risk of coral shading by macroalgae. The benefits herbivorous fish species provide by reducing algal competition are thought to outweigh any damage to juvenile coral recruits and coral fragments caused by intense grazing activities (see also the section "Corallivory"), at least on natural reefs (Mumby, 2009). Moreover, grazing opens new settlement space for coral larvae to colonise (Doropoulos et al., 2016), thus facilitating natural ecological recovery processes.
Although the reduction in fishing pressure to protect herbivorous fish stocks is often a key management step to increase reef resilience, the enhancement of grazers has also been proposed as a complementary method to reef restoration (Abelson, 2006). Typically this is achieved through marine reserves, but region-wide fishing bans on herbivores are increasingly being utilised O'Farrell et al., 2015). Although the recovery of parrotfishes can be rapid (<5 years, O'Farrell et al., 2015), the re-introduction of grazers by releasing fish larvae on restored, but recruitment-limited, reefs has been suggested as a useful management technique in accelerating stock recovery and increasing herbivory (Abelson et al., 2016b). In a modelling study, different simulated scenarios of fish stock enhancement predicted that fish restocking could substantially increase the success of coral reef restoration projects. Restocking was shown to lead to enhanced coral cover and grazing fish density while reducing macroalgal cover in a significantly shorter amount of time when compared to restoration without restocking interventions (Obolski et al., 2016). Nonetheless, restocking remains a logistically challenging management option, and field tests are lacking. For example, post-settlement mortality of fish larvae needs to be addressed before attempting restocking activities, as restored reefs with limited food or shelter from predators may not be adequate for supporting juvenile communities (Almany & Webster, 2006;Booth & Hixon, 1999;Forrester, 1990;Juanes, 2007).
Most fish and coral restoration interactions are considered to be beneficial; nonetheless, certain fish species are known to have deleterious effects on restoration success and create considerable challenges for reef managers (Forrester et al., 2012). Herbivorous damselfishes are well known for their effects on coral colonies, and are often among the first fish groups to colonise restored reefs (Schopmeyer & Lirman, 2015). Within their territories, damselfishes may intentionally bite and damage live coral polyps to promote the growth of the algae they consume on the coral skeleton (Ogden & Lobel, 1978), which becomes a major issue on restoration projects where coral fragments are already fragile (Ladd et al., 2018;Williams et al., 2019; Table 1). For example, Isopora palifera colonies transplanted within territories of the white damsel Dischistodus perspicillatus (Cuvier 1830) eventually died due to the metabolic cost of combating algal smothering (Potts, 1977). When coral fragments were transplanted inside and outside Australian gregory (Stegastes apicalis) (De Vis 1885) and dusky farmerfish (Stegastes nigricans) (Lacepède 1802) territories, transplant mortality was higher inside the territories than in control areas (Casey et al., 2015). Schopmeyer and Lirman (2015) studied the effects of territorial damselfish on a coral reef restoration project in Florida. Immediately following, and even during, the outplanting of nursery-reared A. cervicornis colonies, damselfishes colonised the restored sites and established algal mats within the first 6 months with large coral colonies experiencing up to 45% colony mortality. Williams et al. (2019) found that after the first few weeks of coral transplantation, it was critical for coral survival that the large D. perspicillatus and Cross's damsel Neoglyphididon crossi Allen 1991 were actively managed to prevent algal overgrowth.
Although algal-farming by damselfishes is a natural ecological process on coral reefs, locating restoration programmes in areas where predators of damselfishes are present in higher densities (e.g., MPAs) may mitigate the negative effects of algal farms through predation and indirectly reduce the incidence of coral disease . The removal of territorial damselfishes (Casey et al., 2015;Schopmeyer & Lirman, 2015) may also help to ensure their presence does not compromise restoration success (Table 2). Transplanting a diversity of coral species is likely to be beneficial and may additionally minimise the impact of damselfishes. A prevalence of fast-growing branching corals may attract damselfishes away from slower-growing corals that may be less able to compete with algal growth stimulated by damselfish gardening. It is, however, important to note that impacts will vary depending on geographic location and damselfish species. Although territorial damselfishes are a significant challenge to coral reef restoration efforts, particularly in the Caribbean, their effects are likely to be context-dependent (Ladd et al., 2018). For instance on Indo-Pacific reefs, territorial damselfishes can exclude corallivores from their territories (Gochfeld, 2010;White & O'Donnell, 2010), resulting in increased coral growth, diversity (Glynn & Colgan, 1988) and recruitment (Gleason, 1996) Burkepile et al., 2013). Coral reefs are primarily nutrient-limited, and yet they are some of the most productive ecosystems on the planet (Davis et al., 2009;Szmant-Froelich, 1983). Fish communities store and supply substantial quantities of nitrogen and phosphorus in the form of excretion and egestion (Allgeier et al., 2017). This nutrient source is crucial in supporting coral reef productivity (Allgeier et al., 2014;Holmlund & Hammer, 1999), and will increase with increasing fish abundances on restored reefs. Nutrients may be translocated from other sources such as seagrass beds as coral fishes forage on adjacent habitats during the night but return to shelter in coral colonies during the day, thus creating nutrient hotspots (Meyer & Schultz, 1983). For example, the nutrient-rich gill excretions and phosphorus-rich faeces of grunts were found to increase the growth of Acropora and Porites coral colonies (Meyer & Schultz, 1985), and high nutrient delivery has been associated with increased herbivorous activity and reduced algal cover on outplanted coral colonies (Shantz et al., 2015). Holbrook et al. (2008) (Table 2).
To date, only a handful of studies have tested the benefit of fishderived nutrients on restored coral fragments and colonies (Table 1). Bongiorni et al. (2003) compared growth and gonad development in coral fragments suspended near a fish farm and in an oligotrophic control site. Despite nutrients potentially being deleterious to corals by enhancing algal growth and increasing water turbidity, Bongiorni et al. (2003) found that proximity to the fish farm greatly enhanced growth and reproductive activity of Acropora eurystoma and Stylophora pistillata. Coral fragment growth rates were 3 to 4 times higher at the nutrient-enriched site, and oocyte numbers were significantly higher, compared to the fragments located at the reference site. Shafir et al. (2006) also suggested that placing their suspended coral nursery 10 m from a large fish cage containing gilthead seabream Sparus aurata Linnaeus 1758 was instrumental in its success and recommended placing coral nurseries within nutrient-rich environments to enhance coral growth, shorten nursery incubation time and reduce costs and threats of predation and competition. Such benefits may extend to fishes repopulating restored reefs.
The benefits of enhanced nutrient supply by fishes may be context-dependent, as high levels of fish excretions can trigger shifts to algal-dominated states on coral-depauperate reefs as opposed to coral-dominant reefs (Burkepile et al., 2013). Furthermore, whereas natural enrichment tends to enhance coral growth, the addition of nutrients by means other than fish-associated processes is not recommended as clear negative associations between anthropogenic nutrient enrichment and coral reef health have been reported (D'Angelo & Wiedenmann, 2014). The differential effects of natural vs. anthropogenic nutrients on corals are attributed to a range of distinctions including nutrient identity (ammonium and phosphorus vs. nitrate) and concentrations (discrete pulses vs. heavy discharge) (Shantz & Burkepile, 2014). Anthropogenic nutrification increases turf algae competition over corals (Vermeij et al., 2010) and affects susceptibility of corals to bleaching (Wiedenmann et al., 2013). In a field experiment, Zaneveld et al. (2016) demonstrated that nutrient pollution can increase coral disease, which was exacerbated at high temperatures, and aggravate the impact of corallivory on coral survival. In their study, although parrotfish predation had a negligible impact on Porites coral survival in control plots, coral survival was significantly impacted in nutrient-enriched plots with 92% of Porites losing tissue through predation resulting in 62% mortality (Zaneveld et al., 2016).
Overall, studies on the effects of added nutrients on reefs are conflicting (Koop et al., 2001;Lapointe, 1997), and effects are likely to be context-dependent (Mumby et al., 2006b;Sotka & Hay, 2009). Thus, further work is required to quantify benefits of fish excretions for coral growth in restoration projects. This could be of particular importance to restoration managers as enhanced fragment growth may reduce the high costs and setbacks associated with coral gardening.
Nonetheless, although increased nutrients may stimulate coral growth in some cases, algal overgrowth remains one of the main concerns on coral-poor reef restoration sites (Bowden-Kerby, 2001;Yap, 2004;Young et al., 2012).

| CORALLIVORY
Corallivorous fishes such as butterflyfishes (Chaetodontidae) can be positively affected by the addition of live coral cover due to increasing food availability (Cole & Pratchett, 2014;Hourigan et al., 1988). Taira et al. (2017 reported that coral nurseries were adequate habitats for juvenile Chaetodon octofasciatus, where their densities were higher than at nearby natural reefs. Predation on coral, while providing an important food source for reef fishes, is however a concern in reef restoration projects, where new coral transplants are particularly vulnerable to native predators and other disturbances (Edwards & Gomez, 2007;Jayewardene et al., 2009;Omori, 2005).
Consumers of coral tissue differ in their feeding strategies and effects on coral fitness; butterflyfishes remove single coral polyps without affecting the underlying skeleton. In contrast, parrotfishes, pufferfishes, triggerfishes, filefishes and wrasses also remove part of the underlying skeleton, with a few species acting as bioeroders by actively consuming the dead coral matrix (Rotjan & Lewis, 2008).
Therefore, corallivory by reef fishes may adversely affect restoration success. For example, intense corallivory by Scaridae and Chaetodontidae caused major tissue loss and coral detachment in transplanted Stylophora coral fragments (Horoszowski-Fridman et al., 2015), and high Acropora formosa fragment mortalities at a coral nursery site were attributed to severe predation by fish and other corallivores (Xin et al., 2016). The coral-feeding butterflyfish Chaetodon capistratus Linnaeus, 1758 was also reported to increase the spread of black-band disease to coral fragments (Aeby & Santavy, 2006).
Although the transplantation of large coral fragments and mature colonies remains the most commonly used method of coral reef restoration, the need for sexually propagated corals has been increasingly recognised (Chamberland et al., 2015;Villanueva et al., 2012). Outplanting coral juveniles raised from sexually derived larvae, as opposed to using more cost-effective clonal fragments, may help conserve the genetic diversity of restored coral populations (Baums, 2008). Nonetheless, translocating juvenile corals remains challenging as they are particularly at risk of damage from corallivorous fishes (Page et al., 2018). For instance, in a study investigating the susceptibility of coral recruits to predation by using settlement plates, parrotfish abundance was correlated with coral recruit mortality, attributed to accidental grazing, whereas butterflyfish abundance was correlated with juvenile coral mortality, attributed to predation (Penin et al., 2010). Nonetheless, both survival and growth rates of juvenile corals increase with transplant size and time spent at a nursery prior to transplantation (Guest et al., 2014;Ligson et al., 2020).
Augmenting the size of juvenile corals ex situ, thus decreasing sizedependent mortality due to predation, may be preferable when considering optimal transplant size, despite the added maintenance cost (Raymundo & Maypa, 2004;Toh et al., 2014). Baria et al. (2010)

| PREDATORY FISHES
There has been a sharp decline in transient apex predator abundance in most reef ecosystems (Baum et al., 2003;Essington et al., 2006;Jackson et al., 2001;. Predatory fishes are typically highly valued by fishers, and their densities have been significantly reduced on many reefs, potentially leading to top-down effects (Baum & Worm, 2009;Heithaus et al., 2008;Myers et al., 2007). For example, predators are important in preventing prey species such as territorial damselfishes from proliferating (Schopmeyer & Lirman, 2015). Corallivorous invertebrates such as the crown-ofthorns starfish (Acanthaster planci), the gastropod Coralliophila abbreviata and the bearded fireworm (Hermodice carunculate) can also negatively affect corals species used in restoration projects (Dulvy  Table 1). This is despite the rebuilding of fisheries being an aim of restoration projects, either explicitly or implicitly. Regardless of local species richness, functional roles on reefs may be performed by only a few species (Hughes et al., 2017), thus restoration of functional roles may be a more important goal for the return of top predators than restoration of species diversity per se. As the number of restoration projects increases globally, there is an urgent need to assess the impact of different restoration methods on the behaviour of marine predators as some designs, such as biorock-associated electric fields, can deter and reduce their feeding rates (Uchoa et al., 2017). Predatory fishes have a vital role in maintaining the ecological balance on reef ecosystems and structuring coral fish assemblages; therefore, the replenishment of their populations and the maintenance of natural behaviour patterns should remain a priority for restoration projects (Ritchie et al., 2012).
With some similarities to coral reef restoration, MPAs and marine reserves aim to maintain ecosystem functions and increase marine habitat quality (Gell & Roberts, 2003;Halpern, 2003;Lester et al., 2009), and may provide insight into the effects of restoration on predatory fishes. Coral reef restoration programmes set up within protected areas may reap their combined benefits on fish recovery (Abelson et al., 2016a). Several studies have demonstrated that protected areas can have positive effects on fish predators, even in small reserves (Clemente et al., 2011;Pilyugin et al., 2016;Russ & Alcala, 2004). Therefore, restored and protected coral reefs could provide visitation sites to large transient predatory species (e.g., jacks) and territories for resident predators (e.g., groupers and eels). It is important to note, however, that reserve benefits to transient species are more likely to be related to protection from fishing and prey availability rather than habitat structure (Roberts & Hawkins, 1997). Nonetheless, some restored reefs also experience a certain degree of protection due to the addition of artificial structures to which the coral fragments are attached, obstructing net-based fishing and thus discouraging certain practices, such as trawling (Edwards & Gomez, 2007). The effect of such protection on stand-alone restoration projects may be limited for wide-ranging species due to the small-scale nature of most reef restoration projects relative to the home ranges of many of the large, high-value predatory fishes . A reduction in fishing pressure on large fish species, alongside coral transplantation efforts, is obviously recommended to aid restoration of functional ecosystem food webs.
Although increased habitat complexity influences the abundance of small-bodied resident predators, and to some extent the abundance of transient predators (see the section "The Role of Habitat and Seascape Complexity"), it is more likely that the increase in prey abundance will be the main attractant to piscivorous fishes with large home ranges (Grossman et al., 1997;Newman et al., 2006;Wickham et al., 1973). If reef fish and invertebrate densities benefit from restoration, foraging opportunities for predators will increase in the long term. Habitat complexity may also affect the hunting efficiency of reef predators, but this is largely unexplored. Several recent studies investigating the effect of reef degradation on predators have highlighted a higher abundance and diversity of reef piscivores on recovering reefs compared to degraded reefs due to the availability of higher-quality prey (Hempson et al., 2018a(Hempson et al., , 2018b. On degraded reefs predators feed lower down the food chain, potentially leading to lower nutrition, survival, fecundity and growth (Hempson et al., 2017). Reef restoration may, therefore, be able to reverse the effects of trophic downgrading by increasing prey availability and improving predator diet. As most restoration projects ultimately aim to restore top-down trophic interactions and positively affect species of commercial importance, research is urgently needed to understand the factors which will influence the return of large piscivorous fishes. A comparison of foraging success by predators with differing prey pursuit or ambush behaviours on restored reefs would be worth examining, particularly in relation to reef design.

| CONCLUSION
This review considers fish-benthic interactions in the growing field of reef restoration research, which has received much less attention than research on effective outplanting of corals, and a number of immediate questions for future research are highlighted (Table 3). In expanding these research questions for coral reef restoration, it is suggested that there is an initial requirement to first understand whether restoration projects can return fish assemblages to their original species composition, or whether restored reefs are likely to support altered or novel fish assemblages. Such altered fish assemblages may function in T A B L E 3 Future key research questions surrounding the recovery of fish assemblages in coral reef restoration efforts

Introducing habitat complexity
Through what mechanisms does the level of habitat complexity generated by coral transplantation affect both consumptive and non-consumptive predator-prey interactions during the process of rebuilding fish assemblages? How important is shade provision in facilitating the return of fish assemblages on restored reefs? How does reef restoration design influence the abundance and diversity of cryptic species?

Role of herbivorous fishes
As a key ecological process, how does herbivory of different species (specifically grazing intensity) change on a restored reef over time? How can the diversity of coral transplant morphologies be manipulated to reduce the detrimental impact of territorial damselfishes? Nutrient provision How can the known benefits of fish excretions on coral growth be utilised to benefit reef restoration projects, in particular coral gardening?

Corallivory
How does food preference of existing corallivore fishes affect restoration success?
Predatory fishes How can reef design (e.g., spacing and a variety of coral transplants) be altered to encourage the return of large predatory fishes? What is the relationship between specific reef designs and the different predatory behaviours they facilitate, e.g., ambush, pursuit?
Over-arching questions Which key factors influence whether restoration projects return fish assemblages to their original species composition or generate conditions likely to support novel fish assemblages? How do patterns of fish recovery vary with biophysical gradients and across biogeographic regions? different ways to natural reefs (e.g., different predator-prey interactions). The timeframe over which these changes occur also warrants attention and therefore should be reflected in the monitoring of restored reefs. On healthy reefs, specific guilds of fishes can have both positive and negative effects on corals, and this is also true for restored reefs. Understanding these interactions is likely to be critical in large-scale efforts to increase coral cover through transplants.
With the global challenges currently facing coral reefs, the requirement to explore their restoration has never been greater. Furthermore, it may not be possible to fully restore reefs to pristine conditions, creating a pressing need to understand the novel reef ecosystems that arise through restoration. Irrespective of whether original species composition and diversity are attainable, understanding the interactions between coral restoration and fish assemblages will be vital to ensure that anthropogenically manipulated reef ecosystems still function and provide ecosystem services. Much work has been focused on the benthic component of reef restoration, although very little is known concerning the impact of restoration on fish assemblages in the short and long-term, a clear omission given the integrated relationship between fishes and their reef habitat. As a greater understanding of the interactions between reef restoration and fishes is gained, and as fish-focused research is integrated into the core of restoration efforts, the effectiveness of this important management tool will increase significantly.