Herbivory facilitates growth of a key reef‐building Caribbean coral

Abstract The decline of reef‐building corals in conjunction with shifts to short‐lived opportunistic species has prompted concerns that Caribbean reef framework‐building capacity has substantially diminished. Restoring herbivore populations may be a potential driver of coral recovery; however, the impact of herbivores on coral calcification has been little studied. We performed an exclusion experiment to evaluate the impact of herbivory on Orbicella faveolata coral growth over 14 months. The experiment consisted of three treatments: full exclusion cages; half cage procedural controls; and uncaged control plates, each with small O. faveolata colonies. We found that herbivorous fish exclusion had a substantial impact on both macroalgal cover and coral growth. Fleshy macroalgae reached 50% cover within some exclusion cages, but were almost absent from uncaged control plates. Critically, O. faveolata calcification rates were suppressed by almost half within exclusion cages, with monthly coral growth negatively related to overgrowth by fleshy macroalgae. These findings highlight the importance of herbivorous fishes for coral growth and the detrimental impact of macroalgal proliferation in the Caribbean. Policy makers and local managers should consider measures to protect herbivorous fishes and reduce macroalgal proliferation to enable coral communities to continue to grow and function.


| INTRODUCTION
On coral reefs, scleractinian corals perform an extensive ecosystem engineering role through the calcification of structurally complex physical structures (Graham & Nash, 2013;Jones, Lawton, & Shachak, 1994). Corals provide habitat for diverse fauna including herbivorous fishes, which are key organisms as they control the abundance and community structure of macroalgae which compete with corals (Hughes, 1994;Jackson, Donovan, Cramer, & Lam, 2014;Paddack et al., 2009). Although the impact of herbivory on algae has been well studied, facilitation of coral growth and calcification is poorly understood (Burkepile & Hay, 2006;Fong & Paul, 2011;McCook, Jompa, & Diaz-Pulido, 2001). Explaining the role of herbivores in facilitating coral ecosystem engineering is of fundamental ecological interest and can inform reef conservation strategies seeking to maintain ecosystem function.
Diminished grazing capacity on Caribbean reefs has prompted extensive evaluation of the role of herbivores in algal control by both observational and experimental studies. Observational studies cover a broad range of spatial and temporal scales and provide evidence for (Jackson et al., 2014;Newman, Paredes, Sala, & Jackson, 2006) and against (Cox, Valdivia, McField, & Bruno, 2017;Russ, Sarah-Lee, Rizzari, & Alcala, 2015;Suchley, McField, & Alvarez-Filip, 2016) the ability of grazers to control benthic algae. Herbivore exclusion has become a standard technique to experimentally assess the ability of both fishes and invertebrates to graze algae and has been adopted by many studies. These small-scale experimental studies generally involve the exclusion of herbivores via caging of experimental plots or artificial substrate units over time periods ranging from days to years.
In contrast to observational studies, experimental studies consistently tend to report a significant effect of herbivores on turf and macroalgal proliferation, with other factors such as nutrient levels and seasonality playing secondary roles (Burkepile & Hay, 2006Ferrari, Gonzalez-Rivero, Ortiz, & Mumby, 2012;Sotka & Hay, 2009).
Phase shifts from coral to algal dominance experienced by many Caribbean reefs have underlined the importance of the interaction between algae and scleractinian corals (Arias-González et al., 2017;Done, 1992;Hughes, 1989). However, despite the importance of the herbivore-algal-coral interaction, experimental evidence of the impact of herbivory on coral calcification is limited. In the Caribbean, Acropora spp. were major reef framework builders historically, but subsequent to a wide-spread epizootic event in the 1970/80s populations remain severely diminished (Alvarez-Filip, Dulvy, Côte, Watkinson, & Gill, 2011;Aronson & Precht, 2001;Gladfelter, 1982). Following the loss of Acropora, species of the genus Orbicella are the major framework builders on many of today's Caribbean reefs (McClanahan & Muthiga, 1998;Perry et al., 2012;Porto-Hannes et al., 2015). Although the effect of herbivory on Orbicella spp. has been considered (Foster, Box, & Mumby, 2008;Lirman, 2001;Vermeij et al., 2010), the impact of herbivory on coral calcification has seldom been assessed (Vu et al., 2009). To address this knowledge gap, we performed a 14-month herbivore exclusion experiment to determine the effect of fish herbivory on Orbicella faveolata calcification. Our hypothesis is that herbivorous fishes restrict algal growth and that algal interaction has a negative impact on coral growth and calcification.

| MATERIALS AND METHODS
This study was conducted over a 14-month period from August 2015 to October 2016 in the "La Bocana" back reef located in the Parque Nacional Arrecife de Puerto Morelos, Mexico (20°52′26″N, 86°51′5″W). We reviewed national park monitoring data from the previous 3 years and preliminary surveys were performed at three sites within the park to evaluate the abundance of sea urchins and the biomass of fishes of the Scaridae and Acanthuridae families. Sea urchins were not observed at any site. Consequently, "La Bocana" back reef patch was selected as it displayed among the highest herbivorous fish biomass (average of 3,600 g/100 m 2 from national park monitoring and preliminary survey data). La Bocana reef patch is at a depth of 4 m and is characterized by large Orbicella faveolata colonies, Agaricia spp., dead Acropora palmata framework, gorgonians, and an algal community dominated by brown macroalgae, turf algae, green calcareous Halimeda spp., and red calcareous Amphiroa spp. Preliminary surveys and national park monitoring revealed reef patch coral cover of approximately 17% and fleshy macroalgal cover of 17%.
To test the effect of fish herbivory on coral growth, fishes were excluded from experimental units containing small O. faveolata colonies.
Half-ellipsoid O. faveolata colonies (approximately 5 cm maximum diameter) were collected prior to 2011 from the same site by removing fragments from large colonies and were left in the reef lagoon for later use. The experimental design constituted three treatments with replicate numbers limited by coral colony availability: (1) full cages excluded herbivorous fishes (n = 6); (2) half cages with caging at sides but no top panel acted as procedural controls testing for caging effects by allowing fish access (n = 6); (3) plates with no caging acted as full controls (n = 6). Cages were built upon 36-cm-square concrete plates, with 30-cm vertical steel reinforcing bars and thin 3 cm × 2 cm polyethylene mesh (Fig. S1). Concrete is commonly used as an artificial substrate in herbivore exclusion experiments (Burkepile & Hay, 2009Rasher et al., 2012). In late August 2015, coral colonies were retrieved and wet-weighed in the laboratory. Cages and plates were randomly located in the reef patch over an area of approximately 400 m 2 , with a minimum separation of 2 m, and coral colonies were affixed to plate centers using underwater epoxy plasticine (Fig. S1).
At the end of the observation period, coral colonies were removed from plates and reweighed in order to determine net calcification rates (in g cm −2 year −1 , based on average colony area over the observation period). Underwater Hobo data loggers tracked in situ water temperature at cage depth during the experiment, and the impact of caging on light levels was evaluated using three underwater Hobo data loggers in a cage, a half cage and on a control plate during a sunny day at the end of June 2016. Light levels were converted from Lux to μmol quanta m −2 s −1 by applying a conversion factor of 54 (Sager & McFarlane, 1997). Light levels were compared between treatments using a Friedman test for repeated measures due to non-normality.
Cages were monitored at the beginning of every month for 14 months and the mesh cleaned of algae and other fouling organisms with brushes every 1 to 2 weeks during this period. Every month topdown photographs of the plates were taken in order to assess algal cover. The software photoQuad v1.4 was used to determine percentage cover by area of a 30-cm quadrat contained within cages and on control plates by tracing algal mini-patch perimeters (Trygonis & Sini, 2012). Plate cover components were classified as short (<0.5 cm) or long (≥0.5 cm) filamentous turf algae; brown, green, or red fleshy macroalgae (to species or genus level); crustose coralline algae (CCA); calcareous algae; cyanobacteria; or gorgonian.
To identify which herbivorous fishes were browsing plate algae, in situ feeding was recorded by static GoPro video cameras.
Observations were performed over 1-hr intervals for a total of 3 hr for full cages, 3 hr for half cages, and 8 hr for uncaged control plates, between May and August 2016 between 11 a.m. and 4 p.m. For each 1-hr interval, feeding fish bite counts were recorded by species, total length (size categories estimated based on plate size), and bite target. Individual bites were defined as those involving a rapid movement of the head toward and away from the plate (Bellwood & Fulton, 2008). Fish biomass was determined using standard allometric length-weight conversions. Bite counts were subsequently converted to estimates of herbivory pressure by standardizing (multiplying) by herbivore biomass following Bellwood, Hughes, and Hoey (2006). Herbivory pressure was compared between treatments based on the first three hours of concurrent observation to standardize for temporal variation in fish activity. Herbivore identity evaluation was performed for control plates only (based on 8 hr of observation) due to low levels of consumption within exclusion cages.
Every month close-up photographs of coral colonies (with a ruler for scale) were taken in order to evaluate coral planar area (substrate area occupied by colonies), perform polyp counts using the software ImageJ, and count occurrences of algal overgrowth (Rasband, 2014).
Photos were taken with consistent orientation to minimize variation in coral area and polyp counts due to camera angle. Corals were overgrown by Dictyota spp. and/or turf algae/sediment mix (also known as the "epilithic algal matrix"). Each month overgrowth was assessed and corals classified as overgrown if any partial overgrowth was evident.
As colony perimeters were obscured by algal overgrowth it was not possible to quantify interactions by coral area affected. Incidence of disease, coral bleaching, mortality, and algal overgrowth was recorded.
Mixed effect models were fitted to evaluate the influence of experimental treatment and water temperature on monthly total, fleshy macroalgal, and turf algal cover, with month and replicate included as random effects. Algal cover was evaluated at the beginning of each month, and consequently, the mean water temperature for the prior month was utilized. The algal cover time series were truncated to begin in December 2015 to account for initial algal growth after cage installation ( Figure 1). Linear model assumptions were assessed using residual diagnostic plots. All algal percent cover variables were arcsinsqrt transformed due to non-normality of residuals. Interaction terms between treatment and temperature were tested and subsequently removed if found not to be significant. ANOVAs were applied to the fitted models to determine variable-level significance and Tukey post hoc testing was performed to compare effects between treatments.
Comparison of coral calcification between treatments was performed using Welch's t tests or non-parametric Mann-Whitney U tests, based on an assessment of normality using Shapiro-Wilk tests.
Initial coral colony planar area, polyp count, and mass were compared between treatments using ANOVAs, with assumptions checked using diagnostic plots. To determine the drivers of coral growth over time, coral polyp count was preferred to coral planar area as it was observed to be a more consistent measure of individual colony growth. Mixed effect models were fitted to evaluate the drivers of month-on-month coral polyp count change. Experimental treatment, water temperature, plate fleshy macroalgal and turf algal cover, fleshy macroalgal, and turf algal overgrowth of coral colonies were included as fixed effects, with month and replicate included as random effects. One-month lagged versions of algal cover and overgrowth were also included as candidate predictor variables. The time series was truncated to begin in November 2015 as October 2015 was used as the baseline for polyp counts. Linear model assumptions were assessed using residual diagnostic plots. Multicollinearity among predictor variables was tested using variance inflation factors (VIFs), but no evidence of multicollinearity was observed (Graham, 2003). Least significant predictors were sequentially removed and models compared pairwise with partial F tests of significant differences in error sum-of-squares. The more parsimonious model was preferred until the partial F test revealed a significant difference and the prior model retained as the minimum adequate model. All statistical analyses were performed using R version 3.3.2 (R Core Team 2016) using packages including lme4 and nlme for mixed effect modeling, car for VIF assessment, multcomp for Tukey post hoc testing, and r2glmm for mixed effect model partial R-squared.

| Effect of herbivore exclusion on macroalgal cover
Experimental units were installed in late August 2015 and were subsequently rapidly colonized by turf algae and macroalgae (Figure 1).
Averaging across all monthly observation periods and plates, mean plate turf algal cover was 5.7%; fleshy macroalgae, 4.1%; cyanobacteria, 1.0%; CCA, 0.1%; and calcareous algae, 0.01%. Cyanobacteria were subject to short-term blooms such as in July 2016 which resulted in sudden increases in cover to over 20% in some cages. Fleshy macroalgal cover was dominated by Dictyota spp. (82% mean cover relative to total fleshy macroalgae) and showed substantial variation over time reaching above 50% total cover within some cages (Figure 1b), while turf algal cover appeared more stable over time (Figure 1c).
Total algal cover, fleshy macroalgal cover, and turf algal cover varied significantly by experimental treatment (Figure 1; Table 1). Algal cover was consistently significantly higher in full exclusion cages than on uncaged control plates (Figure 1d-f; Table 2). On average, macroalgal cover was 1731% relatively higher and turf algal cover was 202% higher in full exclusion cages than on uncaged control plates (Figure 1e,f). While there was no significant difference in fleshy macroalgal cover between full and half cages (Figure 1e; Table 2), turf algal cover was significantly (171%) higher in full cages than half cages ( Figure 1f; Table 2). Mean short (<0.5 cm) and long (≥0.5 cm) algal turf cover displayed similar relative levels between treatments (Fig. S2).
F I G U R E 1 Total algal, fleshy macroalgal and turf algal plate mean cover and temporal trends. (a-c) Mean (±SEM) monthly total plate algal cover (a), fleshy macroalgal cover (b), and turf algal cover (c) for full exclusion cages (n = 6), half cage controls (n = 6), and uncaged control plates (n = 6). Algal cover monitoring was performed at the beginning of each month. Mean water temperature for the prior month measured by Hobo data loggers is also shown. (d-f) Mean (±SEM) total plate algal cover (d), fleshy macroalgal cover (e), and turf algal cover (f) over the observation period for full cages (n = 6), half cages (n = 6), and uncaged control plates (n = 6). Numbers represent significant differences indicated by mixed model Tukey post hoc testing ( The interaction of water temperature and cage type had a significant effect on fleshy macroalgal cover (Figure 1b; Table 1).

| Herbivorous fish identity
In situ herbivore feeding observations revealed grazing on plates by herbivorous fishes. Uncaged control plates experienced an average of 87 bites/hr, and both full and half cages were subject to 63 bites/hr, with the majority (87%) directed at turf algae. Fishes up to 10 cm total length (TL) were observed to enter full cages through the 3 cm × 2 cm mesh. Acanthurus spp. larger than 10 cm TL entered half cages from above, although these did not feed on half cage plate algae. Consequently, less herbivory pressure was observed within full and half cages (64 and 520 bites g hr −1 plate −1 respectively) compared with control plates (6,874 bites g hr −1 plate −1 ; Fig. S3 Table 3). The interaction of cage type and Dictyota overgrowth was tested and found not to be significant. Further interaction term testing was limited by model degrees of freedom.

| Effect of herbivore exclusion on coral growth and calcification
Over the study period, no coral mortality (except for colony perimeter tissue loss) occurred and few disease symptoms were observed.
Coral bleaching in late summer 2016 was limited to two polyps on one Orbicella faveolata colony. Caging was observed to have an effect on light intensity (Friedman test, chi-squared = 722, df = 2, p < .001).
T A B L E 2 Effect of experimental treatment on monthly algal cover. Plate monthly total algal cover, fleshy macroalgal cover, and turf algal cover compared between experimental treatments by applying Tukey post hoc tests to fitted mixed models (Table 1). Experimental treatments are full exclusion cages (n = 6), half cage controls (n = 6), and uncaged control plates (n = 6). Asterisks denote level of significance (* denotes p < .05 and ** denotes p < .01) On a sunny day at the end of June 2016, mean daylight incident at uncaged control plates (408.6 μmol quanta m −2 s −1 ) was slightly (4.7%) higher than within half cage procedural controls (390.3 μmol quanta m −2 s −1 ), which in turn was (10.9%) greater than within full exclusion cages (352.0 μmol quanta m −2 s −1 ).

| DISCUSSION
The substantial impact of reduced fish herbivory on algal cover mediates growth rates of the important reef-building coral, Orbicella faveolata. Over our 14-month study, macroalgal and turf algal cover were greater within herbivore exclusion cages than on uncaged control plates (Figure 1; Table 2). In situ observations revealed negligible herbivory pressure within exclusion cages (Fig. S3). Fleshy macroalgae were almost always absent from uncaged control plates while cover reached over 50% within some herbivore exclusion cages. Coral-algal interactions were common within cages and caged coral colonies experienced significantly lower growth than those on control plates (Figures 2 and 3; Table 3). Net annual calcification rates on uncaged control plates were comparable with benchmark values, while in both full and half cages calcification was reduced by over 40% (Figure 2 in order to conserve community calcification in the mid and long term (Kuffner & Toth, 2016).
Ecosystem processes are often scale-dependent (Hewitt, Thrush, & Lundquist, 2010;Levin, 1992). External drivers and internal feed- ). On small spatial and temporal scales, herbivorous fishes may control algal growth while at larger and longer scales nutrient input, herbivore depletion below critical thresholds and other ecological processes may promote algal proliferation (Mumby, Hastings, & Edwards, 2007;Paddack et al., 2009).
In this study, in situ herbivore feeding observations suggest that turf grazing was primarily performed by the surgeonfish Acanthurus bahianus and juvenile Scarus iseri parrotfishes. Acanthurus spp. have a mixed diet Burkepile & Hay, 2011) and may also have been responsible for (unobservable) macroalgal propagule removal from uncaged control plates; however, longer observation periods would be required to detect this effect. Scarus parrotfishes are primarily turf grazers (Adam, Kelley, et al., 2015;Burkepile & Hay, 2010) and small-bodied S. iseri have been shown to exert substantial grazing pressure on certain Caribbean reefs (Kuempel & Altieri, 2017).
Recent studies suggest that abrasion may have a limited impact as a mechanism for coral-algal interaction (Diaz-Pulido, Harii, McCook, & Hoegh-Guldberg, 2010;Rasher & Hay, 2010). We observed that coral growth was unrelated to overall plate algal cover, and consequently, direct contact may be necessary for other coral-algal interaction mechanisms such as allelopathy, microbial transmission, and local hypoxia to take effect (Barott et al., 2012;Nugues, Smith, van Hooidonk, Seabra, & Bak, 2004;Rasher & Hay, 2010;Wolf, Wild, & Nugues, 2012). The negative effects of turf algal overgrowth were negligible in this study potentially due to the low severity of interactions or the inconsistent nature of hypoxia generated at the coral-algal interface (Wangpraseurt et al., 2012).
While some herbivore exclusion studies have found evidence of coral growth suppression, other studies report no impact of reduced herbivory levels on coral growth. This is likely a result of the spe-  (Fong & Paul, 2011;McCook et al., 2001). For example, canopy-forming algae (e.g., Sargassum spp.), foliose macroalgae (e.g., Dictyota spp. or Lobophora spp.), or filamentous turf algae, generate distinct physical coral interactions due to their varying morphologies (Box & Mumby, 2007;River & Edmunds, 2001;Titlyanov, Yakovleva, & Titlyanova, 2007). Exclusion studies typically employ small, shortlived, "weedy" species such as Caribbean Porites spp. (Burkepile & Hay, 2010;McClanahan et al., 2003;Miller & Hay, 1998;Sotka & Hay, 2009) which may be more susceptible to algal overgrowth and contribute little to reef framework (Alvarez-Filip et al., 2013;Tanner, 1995 , 2001). Accordingly, these studies often include procedural controls in an attempt to control for experimental manipulation. In this study, we adopted the common procedural control of half exclusion cages to allow fish access from above, but retain the manipulation and barrier and shading effects of side panels. However, we observed that half cage procedural  however, insufficient coral colonies were available for these.
Few studies have evaluated the effect of caging on the enclosed environment. Water movement has been assessed using different techniques and found not to be significantly affected within exclusion cages (Castro-Sanguino et al., 2016;Lewis, 1986;Vermeij et al., 2010).
Studies report differing effects of caging on incident light levels, potentially due to cage construction, ambient light levels, other environmental conditions, or measurement procedure (Castro-Sanguino et al., 2016;Ferrari et al., 2012;Vermeij et al., 2010). In this study, we used Hobo data loggers to observe that mean light intensity incident at uncaged control plates on a particular day was slightly (4.7%) higher than within half cages, which in turn was (10.9%) greater than within full cages. However, coral calcification rates are relatively insensitive to such minor reductions in light intensity and consequently cage effects could only account for a small fraction of the differences in calcification between herbivore treatments (Venti, Andersson, & Langdon, 2014).
Macroalgae compete with corals, reducing fitness and suppressing growth (Chadwick & Morrow, 2011;Fong & Paul, 2011;McCook et al., 2001). In this study, we have shown that fish herbivory combats macroalgal growth and facilitates Orbicella coral calcification, a major contributor to Caribbean reef framework. However, in recent decades, Caribbean reefs have lost both overall coral cover and suffered shifts to more short-lived coral species which contribute less to reef framework. Consequently, herbivorous fishes and macroalgal control are increasingly important for today's Caribbean reefs. Policy makers and local managers should consider measures to protect herbivorous fishes and reduce macroalgal proliferation in order for reefs to continue to grow, function, and survive anthropogenic sea-level rise. Fisheries management strategies including quotas, size limits, and gear restrictions can serve to protect fish populations. Spatial management such as the implementation of marine protected areas, and no-take zones are also useful tools if well enforced. However, these measures must be combined with watershed management strategies to address uncontrolled coastal development and inadequate wastewater treatment which have caused large-scale nutrient enrichment of coastal Caribbean waters (Risk, 2014).

ACKNOWLEDGMENTS
We are grateful to R. Iglesias-Prieto who provided coral colonies and insightful comments during manuscript preparation. E. Perez-Cervantes, J. González-Barrios, B. Gubri, E. Navarro-Espinoza, A.
E. Escalante-Mancera provided technical assistance with water temperature and light measurement. Our manuscript was significantly improved by insightful comments from two anonymous reviewers.
The study was carried out with the permission and support of the Parque Nacional Arrecife de Puerto Morelos authorities. We thank the Instituto de Ciencias del Mar y Limnología, UNAM, for paying the publication fees for this article.