Coral reefs only cover <0.1% of Earth's surface, yet house more than 30% of total marine biodiversity (Reaka-Kudla, 2005). Reefs are a key source of fisheries production (Moberg and Folke, 1999) and also provide shoreline protection for >100 million people living next to coastlines (Ferrario et al., 2014). However, the invaluable ecosystem services coral reefs provide are increasingly jeopardized as corals decline globally (Bruno and Selig, 2007; Jackson et al., 2014). In the Pacific Ocean, reefs have lost nearly half of their corals over the past four decades (Bruno and Selig, 2007). This alarming trend is even more pronounced in the Western Atlantic Ocean (henceforth, the Caribbean), where coral reefs have lost 50% of their coral cover since the mid 1970's (Gardner et al., 2003; Jackson et al., 2014). On many reefs, coral declines are accompanied by a loss of benthic diversity and increases in algae, weedy coral species, soft corals, and sponges (Burman et al., 2012; Ruzicka et al., 2013; Cardini et al., 2015). Such declines in coral cover and diversity often lead to the loss of structural complexity (Alvarez-Filip et al., 2009), diminished fish populations (Newman et al., 2015), and decreased coral recruitment (Dixson et al., 2014), jeopardizing the ecosystem function and economic value of reefs (Costanza et al., 2014).

To address these declines, coral restoration has gained increasing attention as a viable strategy to help degraded reefs recover, with large-scale restoration efforts now underway across the globe (Young et al., 2012). Current restoration efforts are primarily focused on restoring ecosystem engineers by outplanting nursery-raised corals to degraded reefs (Lirman et al., 2014; Cabaitan et al., 2015; Griffin et al., 2015). These projects have shown that coral size (Garrison and Ward, 2008), genotype (Lirman et al., 2014), and source location (Forrester et al., 2013) all influence the success of restored corals. While coral restoration is a rapidly progressing field, significant knowledge gaps remain regarding the drivers of restoration success.

One such gap is the optimal density and arrangement for outplanting restored corals. Restoration via transplantation of autogenic ecosystem engineers in systems ranging from tropical forests (Zahawi and Augspurger, 2012), to grasslands (Morgan and Scacco, 2006), to mangroves (Elster, 2000) suggests that the density and arrangement of organisms used for restoration can significantly influence their survival, growth (Li and Wilson, 1998) and recruitment (Mulligan et al., 2002). Further, the density of restored ecosystem engineers can mediate important ecological processes that drive community dynamics, such as herbivory and nutrient cycling (Holl et al., 2000).

Similarly, on coral reefs outplant density and arrangement will likely affect important responses such as growth rates, habitat production, disease dynamics, and, ultimately, coral survivorship. Indeed, Griffin et al. (2015) found that short-term growth rates of restored Acropora cervicornis over 3 months in the US Virgin Islands were inversely related to outplant density. In contrast, Shaish et al. (2010) found no differences in mortality or bleaching of restored Montipora digitata in high density, low density, or “patchy” arrangements of nursery-raised corals in the Philippines after 15 months.

Theoretical work suggests that corals outplanted in high densities and arranged with even spacing will maximize the development of topographic complexity on degraded reefs (Sleeman et al., 2005). The creation of habitat may aggregate important fishes such as schooling grunts, which can focus nutrient delivery from excretion and create nutrient hotspots that can increase the growth rates of restored corals as well as important processes such as herbivory (Shantz et al., 2015) and the removal of coral predators (Ladd and Shantz, 2016). However, there is a paucity of long-term studies that investigate the role of outplant density on coral restoration success. Given that coral restoration is an expensive and labor-intensive process, determining the most effective densities in which to outplant restored corals is an important step toward balancing the costs and benefits of coral restoration.

Here, we address this information gap by investigating the influence of coral density on the growth, habitat production, and survivorship of restored corals. Over 13 months we monitored experimentally-established populations of A. cervicornis outplanted in a gradient of densities on a reef in the Florida Keys, USA. We tracked the growth, habitat production, tissue loss, and survivorship of restored corals as proxy for the success of coral restoration. We hypothesized that low-density treatments would demonstrate higher growth rates and per-coral habitat production compared to high-density treatments. Further, we predicted that per capita tissue loss and colony mortality would be greatest in high-density plots.

As coral reefs continue to decline globally (Bruno and Selig, 2007; Jackson et al., 2014), the need to develop effective strategies to restore degraded reefs becomes more urgent. Although coral restoration has gained increased attention as a viable strategy to restore reefs, it remains a labor- and cost-intensive strategy (Young et al., 2012; Rinkevich, 2014). Thus, there is a need to maximize ecological benefits while minimizing costs to continue scaling up coral restoration efforts. Here, we show that coral outplant density is a key factor to the success of coral restoration. We found that the survivorship of restored corals decreased with increasing density. Corals outplanted in moderate densities (12-colony treatments; 3 corals m⁻²) grew faster and lost less live tissue than high-density (24-colony and 12-clumped; 6 and 12 corals m⁻²) treatments. Further, corals in 12-colony treatments tripled in TLE during the course of the experiment and on average ended up with more live TLE at the plot level as compared to 24-colony treatments, though they started the experiment with half the number of corals. Importantly, our data suggest the presence of both positive and negative density dependent effects. Increasing density to 3 corals m⁻² resulted in more coral growth than expected. But, continuing to add more corals (6 or 12 corals m⁻²) resulted in negative density dependence and less coral growth than expected.

Despite substantial evidence of the impact of coral genotype on coral growth rates, survivorship, and disease prevalence (Vollmer and Kline, 2008; Lirman et al., 2014), we found no effect of genotype on any of our response variables. Instead, we show that density dependence plays a large, yet underappreciated, role in the success or failure of coral restoration efforts. Although we found stronger density effects than genotype effects, we only used three genotypes of A. cervicornis. Promoting genotypic diversity should still remain a restoration priority. Given the uncertain conditions reefs are likely to experience in the future (Pandolfi, 2015; Pendleton et al., 2016), including genotypes with a range of traits and environmental tolerances will likely be essential for successful coral reef restoration.

Density often influences the success of restoring ecosystem engineers in other ecosystems (Li and Wilson, 1998; Holl et al., 2000; Mulligan et al., 2002). Yet, limited research on the role of density exists for coral restoration, particularly in the Caribbean. Our findings suggest that outplanting A. cervicornis for restoration in moderate densities (3 corals m⁻² – our 12-colony treatment) maximizes growth rates and habitat production while minimizing tissue loss and coral mortality. To our knowledge, the only other study that manipulated the density of A. cervicornis colonies found a negative relationship between coral density and linear extension (Griffin et al., 2015). However, this study only tracked corals for a period of 3 months, had limited replication (n = 1 per density), and potentially confounded density effects with genotype effects. The longer (13 month) duration of our study allowed us to elucidate the effect of the density of restored corals over timescales more relevant to coral reef community recovery. Similar to Griffin et al. (2015), we found that growth rates in our 24-colony treatment were lower than the others at the end of the experiment. However, we show that density dependence can influence the success of A. cervicornis outplanted for restoration, and that the strength and direction of density dependence changes with coral density. These findings run counter to work done in the Philippines, which found no effect of outplant density or arrangement on M. digitata growth or mortality over a 15-month study, suggesting different species may display variable responses to outplant density (Shaish et al., 2010).

Density dependence has been heavily studied in terrestrial and intertidal systems, often with a particular focus on foundation species (Bertness and Callaway, 1994; Bruno et al., 2003). Work in marine systems, and specifically coral reef ecosystems, has largely focused on the effects of density dependence on the growth and survivorship of coral-associated organisms, such as coral reef fishes, rather than corals themselves (Hixon and Carr, 1997; but see Baird and Hughes, 2000; Hixon and Webster, 2002; Shantz et al., 2011; Marhaver et al., 2013). Here, we found that the densities of our densest treatments resulted in less habitat creation and live coral tissue than would be expected, suggesting negative density dependence.

Although TLE and live TLE increased for the 3-colony, 6-colony, and 12-colony plots, the densest treatments (12-clumped and 24-colony) saw little to no increase in these metrics over the 13-month study. Several mechanisms could be important for driving this negative density dependence. While small-scale alterations in water flow may benefit corals growing in close proximity, at some density threshold, such as in our 24-colony or 12-clumped treatments, coral branches could become so dense as to have a negative effect on water flow. At such high densities, reductions in water flow may reduce mass transfer of nutrients to the coral or efflux of oxygen as a byproduct of photosynthesis, contributing to declines in photosynthesis and the energy available for coral growth (Finelli et al., 2006). Additionally, crowding in high-density treatments could increase shading and intensify competition for light, effectively reducing photosynthesis (Chadwick and Morrow, 2011). Although A. cervicornis relies heavily on photosynthetic endosymbionts for energy, heterotrophic feeding is an important component of growth rates, particularly under stressful conditions (Houlbrèque and Ferrier-Pagès, 2009; Towle et al., 2015). Thus, corals in high-density plots may have experienced increased competition for food particles in the water column, contributing to lower growth rates in these treatments.

Additional mechanisms likely contributed to the negative density-dependent effects at the high-density treatments. High densities of corals may facilitate disease transmission among coral colonies. Some coral diseases can be vectored between A. cervicornis branches that are in direct contact (Williams and Miller, 2005). Thus, disease transmission may have been facilitated in 24-colony and 12-clumped treatments where branches of colonies were more likely to be in direct contact as compared to lower density treatments. The corallivorous snail C. abbreviata, and the fireworm H. carunculata can vector or act as a reservoir for coral diseases (Sussman et al., 2003; Williams and Miller, 2005; Gignoux-Wolfsohn et al., 2012). Consequently, these coral predators, which commonly feed on A. cervicornis (Miller, 1981), may cause initial disease infection or spread diseases within a plot of restored corals. Tightly clustered colonies in high-density treatments may have provided a physical escape from predation for small corallivores, increasing the probability of tissue loss and infection. Further, C. abbreviata prefers A. cervicornis colonies surrounded by conspecifics rather than solitary colonies or those surrounded by heterospecifics (Johnston and Miller, 2014), suggesting that high-density treatments may be preferred by corallivores.

While we did record predator density and disease presence as part of this study, our surveys were not frequent enough to track disease progression or link tissue loss to predator abundance. Although variable among replicates, we observed an average of 30% of corals showing signs of disease (rapid tissue loss) in 12-clumped treatments during our August 2013 survey, which coincided with significant increase in live tissue loss (Figure 4). Thus, the densest treatment appeared to facilitate disease, leading to dramatic loss of live coral. Partial colony mortality, highest in the 12-clumped and 24-coral treatments at the end of the experiment, likely depressed growth rates and contributed to the negative density-dependent effects we observed in the high-density treatments.

Farming damselfish (e.g., Stegastes planifrons) can rapidly colonize A. cervicornis colonies outplanted for restoration, cause significant amounts of partial mortality, and decrease growth rates (Schopmeyer and Lirman, 2015). While we did not specifically quantify damselfish abundance in our experiment, we also did not observe strong colonization by farming damselfish in our study. However, damselfish may selectively recruit to high-density plots that provide more shelter from predators (Almany, 2004) and substrate to create their algal lawns (Ceccarelli et al., 2001). Thus, at other restoration sites, the colonization of high-density coral treatments by territorial damselfish could be an additional mechanism contributing to negative density-dependence of corals used for restoration.

Although our analyses showed only marginally significant effects for positive density dependence, the patterns were suggestive of positive effects of increasing density to moderate levels. This pattern was most obvious going from the 6-colony to the 12-colony densities where TLE and live TLE increased 2.5–3 times. These data suggest the existence of positive feedback mechanisms for coral growth and habitat creation under moderate increases in density. For corals, more live tissue affords increased opportunity for growth, and therefore increasing structural complexity, particularly for branching corals such as A. cervicornis. Positive density dependence could also be expected for corals through the improvement of microclimatic conditions, as observed for plants in terrestrial systems (Bruno et al., 2003). For example, increased coral density could decrease laminar flow and increase mixing, reducing the boundary layer and enhancing delivery of nutrients and dissolved oxygen to nearby corals (Atkinson and Bilger, 1992; Lesser et al., 1994). For species that rely heavily on asexual fragmentation, such as A. cervicornis, high densities can function to trap and stabilize asexual fragments, contributing to a positive density dependent feedback (Tunnicliffe, 1981). Over time branches can fuse together to form dense thickets that increase resistance to physical disturbance and further promote fragment retention.

Reef fishes, often limited by habitat availability as both juveniles and adults, selectively recruit to live coral, where they grow faster than fishes recruiting to non-living structure (Holbrook et al., 2000; Feary et al., 2009; Kerry and Bellwood, 2015). The topographic complexity provided by corals can aggregate fishes, concentrating fish-derived nutrients that can increase coral growth (Holbrook et al., 2008; Shantz and Burkepile, 2014). These fish-derived nutrient hotspots also increase grazing by herbivorous fishes and decrease algal abundance, both of which likely help facilitate coral growth and survivorship (Shantz et al., 2015). Further, many of the fishes that aggregate around structurally complex corals are invertivores, such as white grunts (Haemulon plumierii), possibly promoting top-down control on coral predators (Lirman, 1999; Ladd and Shantz, 2016).

Our findings have important implications for how we approach coral restoration. Ultimately, the goal of coral restoration is to promote ecological processes and positive feedbacks that foster self-sustaining coral reef communities. While the goal of coral restoration is not focused solely on corals, these ecosystem engineers are the foundation upon which other essential species and ecological processes depend (Mumby and Steneck, 2008; Newman et al., 2015). For example, coral and fish larvae are able to track the smell of corals to use as positive settlement cues (Dixson et al., 2014). Larger corals also have higher reproductive potential (Szmant, 1986) and therefore are more likely to contribute to sexual reproduction, a key component of coral reef recovery. Thus, restoration efforts that maximize coral growth, coral survivorship, and habitat creation will likely promote these important positive feedbacks and more quickly foster the recovery of coral reef communities.

Here, we show that the direction and intensity of density dependence on the success of corals used for restoration is context-dependent. These findings highlight the need for restoration practitioners to consider the density of corals when planning restoration efforts. For A. cervicornis, the primary species used for coral restoration in the Caribbean, our data suggests that outplanting in densities of three corals m⁻² can take advantage of positive density dependent processes that maximize habitat production and reduce mortality (Figure 9A). We posit that by capitalizing on positive density-dependent processes, restoration practitioners can maximize the benefits of coral restoration (Figure 9B). Further, this would avoid overloading areas with corals that could be used to restore other areas. Importantly, we found that increasing the density of coral to 6 or 12 corals m⁻² can actually induce negative density dependent processes that increase coral mortality and slow coral growth, working against restoration goals. These results demonstrate the need to evaluate the influence of density on the success of other coral species used for restoration, which likely display density-dependent relationships that could be exploited to facilitate coral restoration. Further work is needed to determine the effect density has on important factors such as disease transmission and predator attraction as well as ecosystem processes such as herbivory and nutrient recycling that will likely also influence coral restoration success. Long-term studies investigating how the density of corals influences the development of coral reef communities and the ecological processes that maintain healthy reefs will advance our ability to effectively restore coral reef communities.

ML, DB designed the study. ML, AS conducted fieldwork. ML conducted the analyses and wrote the manuscript with assistance from AS, KN, and DB.