Biomass of key grazing fishes is an important determinant of coral growth and fouling control in coral nurseries

Restoration is an emerging tool for coral reef conservation, yet despite small‐scale successes there are concerns about high costs and ecological setbacks. Integration between reef ecology and restoration could help address such concerns. A prime example is the use of grazing by herbivores to reduce coral nursery cleaning costs. However, the relation between herbivore communities and cleaning benefits remains unquantified. This study aimed to measure links between herbivorous fish communities, grazing intensity and coral nursery performance. Six reefs were selected in southern Kenya, equally divided across three levels of fisheries management (fished < reserve < no‐take). Fish surveys determined herbivorous fish biomass and composition, and remote underwater videos recorded grazing intensity on coral nurseries. Accumulated fouling and coral growth were measured at the end of the 4‐month study. Grazing intensity was sixfold lower and fouling density fourfold higher in the fished areas compared to protected zones. Higher fouling strongly correlated with lower coral growth: exponential growth constants in fished areas were respectively twice and three times as low compared to marine reserves and no‐take zones. Across study sites grazing was dominated by bristletooth tangs (Ctenochaetus spp.), except where these were outcompeted by territorial damselfish. Thus, better coral nursery performance in protected areas can be partially linked to higher grazing intensity, which in turn is determined by both fish biomass and local species composition. We recommend protecting herbivorous fish species and placing coral nurseries in areas with high biomass of key grazers to improve coral nursery performance and reduce maintenance costs.


Introduction
Restoration is establishing itself as an additional coral reef conservation tool (Rinkevich 2019), complementing traditional measures and climate action to maintain coral reef ecosystem functions and services under present-day stressors (Anthony et al. 2017;Mcleod et al. 2019).The appropriate use of coral gardening, a commonly-used restoration technique where corals are first grown in nursery structures before being outplanted on the reef (Rinkevich 1995(Rinkevich , 2005)), has been outlined in several science-based guidelines (Edwards et al. 2010;Johnson et al. 2010;Shaver et al. 2020).This two-step coral gardening technique has realized substantial increases in hard coral cover locally (Hein et al. 2020) and the adaptability of this low-tech technique allows for easy implementation in emergent regions such as East Africa (Mbije et al. 2010(Mbije et al. , 2013)).However, the majority of coral reef restoration projects around the world are still small and costly, featuring a median restored area of just 100 m 2 (Boström-Einarsson et al. 2020) at a median project cost of roughly 400,000 US$ ha À1 (Bayraktarov et al. 2019).If reef restoration aspires to have a positive impact on socioecological Author contributions: EGK, RO, AJM conceptualized the research; ER, EGK performed the research; EGK, ER analyzed the data; EGK wrote the manuscript; ER, AJM, RO edited the manuscript.
scales, further improvements in the cost-effectiveness and scalability of techniques are imperative.
A considerable cost for numerous reef restoration projects using coral gardening is the maintenance of coral nurseries.Bio-fouling such as macroalgae and tunicates can settle onto nursery structures and impair the growth and survival of nursery-grown corals (e.g.Dehnert et al. 2022) due to shading, abrasion, overgrowth and allelopathy (McCook et al. 2001).To reduce these competitive interactions, frequent and time-consuming cleaning of nurseries is common practice (Precht 2006;Edwards et al. 2010;Johnson et al. 2010;Ferse et al. 2021).Fortunately, reef restoration and reef ecology are becoming increasingly integrated (Ladd et al. 2018).The importance of herbivores such as surgeonfishes consuming and thereby controlling algae on natural reefs has been long established (Carpenter 1986;Hay 1997), but only more recently has herbivory gained appreciation in the process of reef restoration as well.For example, temporary (Frias-Torres et al. 2015) or permanent (Knoester et al. 2019) placement of coral nurseries in proximity to fish communities of nearby natural reefs or the coculturing of grazing gastropods (Toh et al. 2013) have all demonstrated the benefits of fouling control by herbivores.Still, only just a fraction of restoration publications examine such facilitative opportunities (Abelson et al. 2020;Ladd & Shantz 2020), despite the great potential such ecological integration could have to improve the cost effectiveness of reef restoration efforts (Ladd et al. 2018(Ladd et al. , 2019)).
The current neglect of ecological knowledge could stall restoration progress and, even though awareness on the importance of herbivory is growing (Rinkevich 2019;Seraphim et al. 2020), there are no tangible guidelines yet.For example, the positive interactions between herbivorous fish and coral nurseries have been demonstrated (Frias-Torres & Van de Geer 2015; Knoester et al. 2019), yet site-specific criteria such as a minimum recommended fish biomass or presence of certain key species for effective herbivory are lacking.To establish such guidelines, the quantification of site-specific herbivorous fish communities and their effect on coral nursery performance (i.e.coral growth and condition) are needed.This study aimed to determine the links between herbivorous fish communities, their grazing intensity and the performance of coral in mid-water (floating) nurseries.Herbivorous fish biomass and species composition were determined at six study sites in southern Kenya, two per level of fisheries management (fished < reserve < no-take).At each study site, remote underwater videos were used to record grazing intensity exerted on coral nurseries, and accumulated fouling density and coral growth were also measured.We hypothesized that placing nurseries in areas with higher fish biomass would result in higher grazing intensity by a diverse assemblage of herbivorous species, less fouling and better coral performance (i.e. higher coral growth and higher percentage live coral tissue).The identification and quantification of such ecological links can be used to develop a more ecologically integrated and cost-effective reef restoration approach (Abelson et al. 2020;Boström-Einarsson et al. 2020).

Study Sites
The six study sites were located around Wasini Island in southern Kenya (Fig. 1) and were equally distributed over three levels of fisheries management.Study sites 1 and 2 were located in a fished zone, where artisanal fishing was both intense and unselective.Sites 3 and 4 were located in the Mpunguti Marine Reserve (11 km 2 established in 1973), where only nondestructive, traditional fishing methods were officially allowed by the Kenyan Wildlife Service.The remaining two sites were situated inside well-enforced no-take zones: study site 5 in Kisite Marine National Park (28 km 2 established in 1973 and enforced by the Kenya Wildlife Service) and study site 6 in the Wasini Community Managed Area (0.31 km 2 established in 2008 and enforced locally by the Wasini Beach Management Unit).Study sites 1, 2 and 6 were situated in a sea strait between Wasini Island and the Kenyan mainland and experience relatively turbid water conditions (average visibility of $7 m) and therefore exhibit a shallow (up to 8 m depth below Mean Lower Low Water) and patchy reef development.An overview of the benthic composition of the various study sites is given in Fig. 1, as the different structural complexity provided by for example hard corals and macroalgae can influence the type of fish community present at each site (Rogers et al. 2014;Heenan et al. 2016).The sea strait is directly bordered by three fishing villages, home to roughly 8,000 people in total (Kenya National Bureau of Statistics 2019).Sites 3, 4 and 5 were situated around small, uninhabited coral islands further offshore (3-7 km from the nearest village) and featured an average visibility of $15 m and fringing reef development up to around 16 m depth (see Table S1 for details).Tidal differences across all study sites were significant and reached over four meters during spring tide, resulting in moderate to strong currents at all study sites.Water temperature was similar across study sites and increased from 27 C in November 2017 to 29 C in March 2018.

Nursery and Experimental Setup
From November 2017 to March 2018, largely coinciding with the dry northeast monsoon, eight replicate mid-water (i.e.positioned floating in the water column) coral nurseries were placed and monitored at each study site.This study looked at the first 4 months of coral performance in nurseries, as the initial small fragments are deemed most responsive to the effects of fouling and grazing by fish.The structures (Fig. 2) closely resembled the design by Knoester et al. (2019): a plastic (PPR) cross holding eight coral fragments in monofilament loops, kept afloat 1 m from the sea bottom by a 2-l glass bottle and anchored by a 10-kg concrete sinker.This design is adapted from the commonly used coral tree nursery design (Nedimyer et al. 2011) and prevents access by bottom-dwelling herbivorous and corallivorous invertebrates, which were therefore not relevant for this study.Coral nurseries were placed on sand or rubble approximately 1 m from either a coral patch or a fringing reef (see Fig. S1).Depth varied per study site and depended on the extent of reef slope development and light availability: nursery placement was shallower in the turbid sea strait compared to the offshore islands (Table S1), so that irradiance levels were estimated to be roughly similar among study sites.Replicate nurseries were separated two meters from each other and placed parallel to relatively homogenous stretches of reef.Following Knoester et al. (2019), all nurseries were filled with healthy, thumb-sized (range: 3.9-4.5 cm) clonal coral fragments of Acropora verweyi, which were harvested from a large coral nursery at study site 2. Transportation of fragments to other study sites was realized in shaded seawater bins during boat rides of 5-50 min during which sea water was replaced roughly every 10 min.No visible signs of stress were noticed on the coral fragments (e.g.no excess mucus production or bleaching) and in the first month after deployment no fragment mortality was observed.No evident positive or negative effects were observed between coral performance at study site 2 and the other study sites that could indicate the site of coral collection influenced the results.During the experiment, nursery structures were not manually cleaned by divers.At the end of the 4-month experiment, fouling was collected from the PPR pipes, monofilament loops and coral fragments.

Stationary Fish Surveys
The species composition and biomass of all diurnally-active, noncryptic fishes were determined by a stationary underwater census with a 7.5-m radius and 5-min initial time slot, following Bohnsack and Bannerot (1986).Fork length was estimated for each fish in 5-cm classes for fishes smaller than 20 cm and in 10-cm size classes thereafter.Estimation of lengths underwater was practiced before commencing the surveys.Between November 2017 and March 2018, 11-15 nonoverlapping, replicate surveys were performed around the coral nurseries of each study site, covering a stretch of roughly 200 m per site.Fish abundance was transformed to biomass using the midpoint of each size class and published length-weight relationships of species (Froese & Pauly 2015).Herbivorous fish biomass was further subdivided into the following functional groups: grazers (targeting turf algae <1 cm), browsers (targeting macroalgae >1 cm), scrapers, excavators and territorial damselfish, based on reported species' functional traits following Green and Bellwood (2009).

Videos Recording Grazing
Remote underwater video (RUV) was used to identify key grazers and their grazing intensity with minimal diver disturbance.A Canon 600D DSLR camera with Neewer 40M case was positioned approximately two meters from a coral nursery on a weighted tripod (Fig. 2C).The camera started recording after a 12-min delay and took five 10-min recordings with 12-min breaks, thus the 2-hour deployment resulted in a total recording time of 50 min per replicate video.After installing the camera, divers would either leave the water or move at least 50 m away.Recording usually took place between 10:00 hours and 14:00 hours, coinciding with the peak in foraging activity of most roving herbivorous fishes (Hoey & Bellwood 2009).Each coral nursery was recorded once at a randomly chosen moment throughout the study period, resulting in eight replicate recordings per study site.All RUV recordings were viewed and each bite targeting the PPR frame, monofilament loops or coral fragments was counted and the fish species noted.Bites of all fish species were noted down and over 97% of the bites were taken by nominal herbivorous fish.In addition, each fish's fork length was estimated (using the nursery structure as size reference), transformed to weight and multiplied by the number of bites taken to calculate mass-scaled bites (ms-bites), following (Hoey & Bellwood 2009).Sums of ms-bites were standardized per hour to correct for slight variations in RUV recording length.Grazing intensity on nurseries is thus expressed as ms-bites in bites Â kg h À1 .

Fouling and Coral Performance
Fouling collected from the nursery structures at the end of the study was categorized in the following broad functional groups: turf algae (<1 cm), macroalgae (>1 cm), crustose coralline algae (CCA), shelled animals (including both molluscs and barnacles) or others (consisting mainly of tunicates and sponges).Fouling was sundried and the dry weight was standardized to fouling density by dividing through the nursery surface area (0.16 m 2 ).
Measurements on coral performance were taken right at the start and end of the experiment.Using scaled photographs, the ecological volume (Shafir et al. 2006) was determined using ImageJ and the live coral tissue was quantified visually as a percentage.The specific growth rate of healthy fragments (live coral tissue ≥80%) was determined using the same formulas as Knoester et al. (2019) and references therein.

Analyses
All analyses were performed in R (R Core Team 2020) and data presented as means AE standard error.To compare average herbivorous fish biomass and grazing intensity between study sites, a generalized linear model with Gamma distribution and loglink from the stats package (R Core Team 2020) was used.To determine the effect of study site on the various types of fouling as well as the specific growth rate of coral fragments, simple linear models were fit using the nlme package (DebRoy 2006).Nursery structure was included here as a random factor to account for the nonindependence of multiple coral fragments in the same nursery.To determine the effect of study site on percentage live coral tissue, a beta regression model with logit link was used using the glmmTMB package (Brooks et al. 2017), as this accounts for the proportional nature of the live coral tissue data (Douma & Weedon 2019).All model assumptions were validated by visual inspection of residual plots, using DHARMa diagnostic plots in the case of generalized linear models (Hartig 2021).Wald chi-squared tests from the car package (Fox & Weisberg 2018) were used to determine the significance of the fixed factor study site for all models.Pairwise comparisons with Tukey adjustments were made with the emmeans package (Lenth 2020).Lastly, Pearson correlation analyses were performed to explore potential links between herbivorous fish biomass, grazing intensity, fouling development and coral performance.As depth varied across study sites, nursery depth was correlated against all above variables to check for any effects of depth on these key processes (Table S2).

Results
The total biomass of herbivorous fishes differed significantly between study sites (χ 2 = 34.355,df = 5, p < 0.001) and mostly increased with stricter levels of fisheries management (Fig. 3).Herbivorous fish biomass was low in the fished zone at site 1 (58 AE 11 kg ha À1 ) and site 2 (34 AE 8 kg ha À1 ), low to moderate in the marine reserve at site 3 (41 AE 16 kg ha À1 ) and site 4 (162 AE 70 kg ha À1 ) and clearly higher in the no-take zones at site 5 (248 AE 105 kg ha À1 ) and site 6 (391 AE 154 kg ha À1 ).Appreciable numbers of browsers, scrapers and excavators were only observed in the protected areas, whereas grazers and damselfishes were relatively more abundant in fished zones (Fig. 3).Overall, grazers (predominantly represented by the genera Acanthurus and Ctenochaetus) were common across study sites, comprising two-thirds of the herbivorous fish biomass at site 2, one-third at sites 3-5 and around a fifth at sites 1 and 6.Grazing intensity on nursery structures differed significantly between study sites (χ 2 = 23.538,df = 5, p < 0.001) and was significantly higher at sites 2-6 compared to site 1 (Fig. 4), where grazing was practically absent (ms-bite rate of 0.6 AE 0.3 bites Â kg h À1 ).The grazing intensity experienced at site 2 (8 AE 3 bites Â kg h À1 ) and site 6 (10 AE 7 bites Â kg h À1 ) was less than half that of sites 3, 4 and 5 (24 AE 16, 34 AE 14 and 28 AE 11 bites Â kg h À1 , respectively), though these differences were not significant likely owing to the high variation in grazing intensity within study sites.The dominant grazers at site 2 (Centropyge multispinis) and site 6 (Amblyglyphidodon indicus) differed from sites 3 to 5, where Ctenochaetus spp.were dominant (see Fig. 4 for genera and Table S3 for species details).
Taken together over all study sites, Ctenochaetus spp.(mainly Ctenochaetus striatus and Ctenochaetus binotatus) were the dominant grazers on nursery structures, accounting for 73% of all recorded ms-bites, followed by the genera Amblyglyphidodon, Centropyge and Scarus, which were each contributing 6% to the total of ms-bites.The remaining 9% of ms-bites were recorded from a diverse group of 48 other species (Table S3).Observed grazing intensities were neither found to correlate with the total herbivorous fish biomass (Fig. 5A), nor with the biomass of grazing herbivorous fishes specifically (Fig. 5B).Grazing intensity did correlate with Ctenochaetus spp.biomass, the identified key grazer on RUV (r = 0.86, p = 0.029; Fig. 5C).S3).Study sites are grouped according to level of fisheries protection, as indicated on top.Study sites not sharing any lower-case letters experienced significantly different ( p < 0.05) average number of mass-scaled bites.Total fouling density accumulated on the nursery structures over the 4-month study (Fig. 6) differed significantly between sites (χ 2 = 97.304,df = 5, p < 0.001) and was clearly higher at site 1 (289 AE 53 g m À2 , mean AE SE) and site 2 (130 AE 32 g m À2 ) compared to sites 3-6 (all below 60 g m À2 ).In addition to an accumulation of molluscs, barnacles and turf algae at the sites in the fished zone, even macroalgae (predominantly brown algae of the genera Padina and Dictyota) became established on the nurseries at study site 1 (see also Fig. 2B).In stark contrast, the little fouling that accumulated at study sites 3-6 consisted primarily of crustose coralline algae (see also Fig. 2A).For details on significant differences between study sites for all fouling groups, see Table S4.Herbivorous fish biomass and grazing intensity were negatively correlated with fouling density, though these correlations were not statistically significant (Fig. 5D & 5E).
Live coral tissue at the end of the study was high at all study sites Figure 5.A compilation of correlations to depict the links between herbivorous fish biomass around coral nurseries and the grazing intensity, fouling accumulation and coral growth in coral nurseries.Correlations are depicted between study site averages of the biomass of (A) all herbivorous fish, (B) grazing herbivorous fish specifically and (C) Ctenochaetus spp.against the rate of mass-scaled bites (i.e.grazing intensity) on coral nursery structures.The (log10-transformed) accumulated fouling density present on nursery structures is correlated against (D) the biomass of herbivorous fish and (E) grazing intensity.Coral growth rate (SGR: specific growth rate constant) is correlated against (F) the biomass of all herbivorous fish, (G) grazing intensity and (H) fouling density.Linear trend lines are added with their associated Pearson correlation coefficient (r) and significance indicated (*p < 0.05, **p < 0.01; nonsignificant correlations are drawn as dotted lines).Each study site is indicated by its number and colored according to fisheries protection level.SEs are shown for each study site to depict the variation in fish surveys (n = 11-15) and bite rate, fouling and coral growth measurements (all n = 8, except for study site 5 which has n = 4 due to lost structures).Correlations were performed on site averages to avoid pseudo-replication.(>90% live coral tissue), except for study site 3 where live coral tissue was 83% (Fig. S2).This lower average at study site 3 can be attributed to a relatively large share of fragments that died (i.e. had 0% live coral tissue): out of all 384 coral fragments at the start of the study, 14 did not make it to the end and 8 of those were at site 3.A different pattern was seen for coral growth (Fig. 7), which increased stepwise from very poor at site 1 (mean specific growth rate of 0.001 day À1 ) to superb at site 5 (0.012 day À1 ) and 6 (0.014 day À1 ), with intermediate growth values for sites 2, 3 and 4 (0.006, 0.008 and 0.010 day À1 , respectively).A strong positive correlation between herbivorous fish biomass and coral growth was found (r = 0.83 p = 0.040; Fig. 5F), though the positive correlation between grazing intensity and coral growth was not significant (Fig. 5G).Percentage live coral tissue did not correlate with fouling density.A very strong negative and significant correlation (r = À0.91,p = 0.0073) was found between fouling density and coral growth (Fig. 5H).It is worthwhile to point out here that, despite comparably low fouling densities in both the marine reserve and no-take zone (Fig. 6), substantially  S1 and S4 for the statistical results for each fouling type separately across study sites.Grazing fishes and coral nurseries 1526100x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/rec.13982by Schweizerische Akademie Der, Wiley Online Library on [16/09/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License higher coral growth rates were found for study sites in the no-take zone (Figs.5H & 7).

Discussion
This study aimed to determine the links between herbivorous fish communities, grazing intensity and the performance of early-stage coral fragments in mid-water nurseries.As hypothesized, higher coral growth rates were found at study sites with higher fish biomass, which were situated inside protected areas.As expected, higher coral growth rates strongly coincided with lower fouling accumulation, but, interestingly, growth rates were even higher in no-take zones compared to marine reserves despite equally low fouling densities.This suggests there might be additional benefits of a healthy fish stock besides grazing, such as nutrient recycling (Shantz et al. 2015).Contrary to our expectations, the presence of key species was a better indicator of grazing intensity than total fish biomass.The majority of ms-bites were taken by Ctenochaetus spp.and overall grazing intensity correlated strongly with the biomass of this key species.A high abundance of Ctenochaetus spp.could as such be indicative for locations where nurseries need little humanassisted cleaning (e.g.sites 3, 4 and 5).Also in contrast to our expectations, the link between grazing intensity and fouling was not clearcut, and may have been confounded by the presence of territorial damselfish (site 6).These feisty fish prevented other grazers access to nurseries within their territory, yet grazing by certain damselfish species also resulted in low fouling accumulation and good coral performance.A relatively low herbivorous fish biomass appeared sufficient to keep coral nurseries free of fouling (e.g.sites 2 and 3), though nurseries placed in areas with low fish biomass that also lacked key grazers were quickly overgrown with macroalgae (site 1).Overall, these results indicate that fish communities can improve coral nursery performance in various ways and that, especially at low overall herbivorous fish biomass densities, successful facilitation likely depends on the presence of site-specific key species.
Biomasses of functionally important herbivorous fishes such as scrapers, browsers and large-bodied grazers were low outside protected areas, affirming the importance of fisheries management (Edwards et al. 2014;Heenan et al. 2016;Knoester et al. 2023).Notably, even inside the relatively small protected area at Wasini (site 6), large-bodied herbivores were abundant and this is encouraging for the approach of community managed areas (Kawaka et al. 2017).The biomass of small-bodied grazing fishes such as Ctenochaetus spp.appeared unaffected by the level of fisheries management, supporting previous studies that found these detritivorous surgeonfishes to be more affected by bottom-up processes such as resource availability (Miller et al. 2012;Robinson et al. 2020).Indeed, unlike the majority of reef fishes, detritivorous surgeonfishes can benefit from large-scale disturbances that reduce live coral cover (Russ et al. 2018) and thrive on overexploited reefs as long as sufficient structural complexity remains (Nash et al. 2016;Obura et al. 2017).Also the high biomass of damselfishes in one of the fished areas conforms to global trends, and likely relates to both reduced predation and competition (Edwards et al. 2014;Seraphim et al. 2020).
Total herbivorous fish biomass was not correlated with fish grazing intensity on nursery structures, indicating a more refined approach is advisable that discriminates the functional diversity (e.g.grazers, browsers, scrapers) within the broad group of herbivores (Heenan & Williams 2013).However, even the biomass of grazing herbivores specifically did not correlate with grazing intensity.Instead, in accordance with the findings of similar study nearby (Knoester et al. 2019), grazing intensity was driven by a select group of key species, principally Ctenochaetus spp.The limited impact of fishing pressure on these species in combination with high grazing rates found for small-bodied and juvenile herbivores (Cernohorsky et al. 2015;Robinson et al. 2020) could explain the unexpected high grazing intensity found on nursery structures located in study sites with overall low fish biomass (e.g.site 3).While other differences in biological or physical properties between study sites might impact grazing intensity, the difference in abundance of this key species is likely the main driver.As Ctenochaetus spp.are among the most abundant surgeonfishes throughout the Indo-Pacific (Marshell & Mumby 2015), feature a high bite rate (Marshell & Mumby 2012) and target early-successional stages of turf (Hamilton et al. 2014;Kelly et al. 2016;Tebbett et al. 2017), the importance of these fishes for bio-assisted cleaning of nursery structures can conceivably be more widespread.Interestingly, another identified key grazer was a territorial damselfish species, Amblyglyphidodon indicus, which was also able to prevent the build-up of macroalgae, indicating that other species might take up grazing when Ctenochaetus spp. is uncommon (i.e.functional redundancy).The potential role of macroalgae control by damselfishes where grazing herbivores are rare has been identified at the Great Barrier Reef (Ceccarelli et al. 2011) and contrasts the negative effects of damselfishes more often found at Caribbean restoration efforts (Ladd & Shantz 2020).
The high grazing intensity by key herbivorous fish species even at depauperate fish communities can be crucial for restoration projects as these generally target degraded and overexploited reefs.Nonetheless, caution is warranted.A system reliant on key species, especially in the absence of functional redundancy, can be vulnerable to variation in grazing intensity due to for example seasonal changes (Paddack et al. 2006;Lefèvre & Bellwood 2011;Seah et al. 2021) or community shifts following acute disturbances such as coral bleaching (Cheal et al. 2010;Nash et al. 2016).Furthermore, coral gardening is a two-step process and grazing intensity on nursery structures is unlikely to be directly translated to grazing intensity around outplanted corals.The higher diversity and complexity of natural reef substrates likely requires complementary feeding by a broader assemblage of herbivorous species (Burkepile & Hay 2008;Fox & Bellwood 2013;Brandl & Bellwood 2014;Brandl & Bellwood 2016;Kelly et al. 2016;Lefcheck et al. 2019).Furthermore, the grazer community might vary depending on the exact habitat being restored (Bonaldo & Bellwood 2010;Hoey & Bellwood 2010;Roff et al. 2019) and will also include invertebrates such as sea urchins that especially graze nearby the seafloor (Carreiro-Silva & McClanahan 2001).Altogether, this means that for successful reef restoration a more diverse herbivore community is likely needed than was found sufficient for grazing of mid-water suspended coral nurseries in this study.Indeed, a study in central Kenya using plates positioned on the seafloor found macroalgae accumulation within a few months despite an herbivorous fish biomass of 180 kg ha À1 (Humphries et al. 2014).
Accumulation of macroalgae on the studied nurseries here only happened at the reef (site 1) with the lowest biomass of roving herbivorous fish (16 kg ha À1 ), of which grazers constituted 14 kg ha À1 .This fits within the grazing fish biomass range of 10-20 kg ha À1 which Robinson et al. (2018) identified as critical threshold below which Pacific coral reefs become (turf) algae dominated.A corresponding threshold for grazing intensity as determined in this study might lay around 10 ms-bites h À1 , with higher grazing intensities effectively keeping coral nurseries free of fouling.Interestingly, the natural reef at this studied site with a grazing intensity below this threshold was still coral dominated, suggesting a fragile state in which newly opened substrate can quickly become colonized by algae, a threat also described for other exploited Kenyan reefs (Humphries et al. 2014).Invertebrate and nocturnal herbivores such as sea urchins can partially substitute the grazing role of diurnal herbivorous fish (Humphries et al. 2014), but were found to have limited capacity for macroalgae control at this specific site (Knoester et al. 2023) and furthermore can contribute to reef erosion (Carreiro-Silva & McClanahan 2001).The effective fouling control on other nurseries in this study at a roving herbivorous fish biomass as low as 31 kg ha À1 might relate again to the simple grazable surface of mid-water nursery structures as opposed to natural reef substrate.The accumulation of bivalves and barnacles on nursery structures is not expected to influence coral performance as explained by Knoester et al. (2019), and would only hinder coral gardening efforts when settling directly onto coral fragments (Frias-Torres & Van de Geer 2015).Another interesting finding was that grazing intensity at a macroalgae-dominated reef (site 3) was sufficiently high to prevent the accumulation of algae on nursery structures.Besides factors discussed above, this might also be explained by a past disturbance such as the 1998 bleaching event opening up a large area for algal settlement (McClanahan et al. 2001) and subsequent maturation of an unpalatable algae community that is difficult to reverse (Humphries et al. 2014;Schmitt et al. 2019).
In line with previous studies, coral growth was negatively associated with the accumulation of fouling, especially macroalgae (Hughes et al. 2007;Knoester et al. 2019).The observed lower coral growth rates at site 1 can be explained by direct macroalgal competition (McCook et al. 2001), while reduced live coral tissue in absence of direct macroalgal contact (at study site 3) could possibly be inflicted through allelopathy by the dense Sargassum community there (Smith et al. 2006).Besides through the control of algal competitors, the fish community can also facilitate reef restoration in other ways (Seraphim et al. 2020).Indeed, while featuring equally low levels of biofouling, coral growth rates were higher in no-take zones compared to the reserves.An explanation could be the delivery of beneficial nutrient pulses by the abundant fish community around nursery structures in the no-take zones (Burkepile et al. 2013;Shantz & Burkepile 2014;Shantz et al. 2015).In light of these benefits, any suggested potential negative effects of a diverse and abundant fish community on coral nursery performance (e.g.coral predation) appear negligible (Knoester et al. 2019).Additional variables that might have influenced coral performance differently across study sites seem limited, but remain to be studied.Furthermore, the response of different coral species is worth exploring.Substantial differences in algal production due to bottom-up processes such as nutrient availability are not expected based on previous research along the Kenyan coast (Humphries et al. 2020) and are generally considerably less important than grazing intensity (Belliveau & Paul 2002;Burkepile & Hay 2006).Potential differences in light availability were assumed to be minimal as deeper sites were chosen because these locations featured better visibility than shallower sites.Indeed no correlations were found between water depth and coral performance (Table S2).
This study aligns with earlier work that found effective fouling control on coral nurseries by herbivorous fishes (Frias-Torres et al. 2015;Frias-Torres & Van de Geer 2015) and corroborates an earlier study that related the grazer-induced reduction in fouling to improved coral performance (Knoester et al. 2019).Such free bio-assisted cleaning bypasses costly human cleaning and thereby facilitates upscaling of restoration activities (Bayraktarov et al. 2016;Shaver & Silliman 2017;Abelson et al. 2020;Ladd & Shantz 2020).Interestingly, a relatively low grazing pressure appeared sufficient to keep nurseries clean of fouling.In this study area, a minimum grazing herbivorous fish biomass of around just 20 kg ha À1 was sufficient to keep coral nurseries macroalgae free, though potential functional redundancy among grazing species might quickly erode below that threshold.Given the apparent greater importance of key species compared to overall herbivore biomass found here and elsewhere (Humphries et al. 2015;Plass-Johnson et al. 2015;Ruttenberg et al. 2019;Knoester et al. 2023), this threshold will likely vary depending on the region and the local species community and therefore limits the formulation of simple site selection guidelines regarding a minimum recommended herbivorous fish biomass.Instead, the identification of suitable nursery sites with sufficient grazing pressure by local herbivorous key species is recommended through small pilot studies.Lastly, while a marginalized fish community could suffice to keep coral nurseries free of fouling, the protection of the full fish community or, potentially, restocking of key species (Abelson et al. 2016) will likely provide additional benefits for reef restoration including nutrient recycling, increased functional redundancy and more effective grazing pressure around outplanted corals.Therefore, we consider the integration of restoration and protection the most effective way forward to rehabilitate and preserve functional coral reefs locally, while global stressors are being addressed simultaneously (Knowlton et al. 2021).
Grazing fishes and coral nurseries 1526100x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/rec.13982by Schweizerische Akademie Der, Wiley Online Library on [16/09/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License like to acknowledge the Kenya Wildlife Service and their Strategic Adaptive Management through which collaborative data collection in Kisite-Mpunguti Marine National Park was possible.We thank Hamadi Mwamlavya for assisting with the fish surveys.A special thanks to Cindy Chorongo, Bulisa Masiga, Judy Nduta and Mercy Zawadi for counting countless fish bites.We thank the reviewers for their time taken to provide valuable input.This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.The data and code used for analysis is available as an archived Github repository on: https://github.com/ewoutknoester/HerbivoryGrazing.

Figure 1 .
Figure 1.Map of Kenya showing study area (insert) and detailed map showing the six study sites.Three different fisheries management zones can be identified: unrestricted fished zone (unshaded) including sites 1 (Firefly House Reef) and 2 (Pilli Pipa Restaurant), the Mpunguti Marine Reserve where traditional fishing is allowed (shaded orange) encompassing sites 3 (Lower Mpunguti) and 4 (Dolphin Point) and no-take zones (shaded red) covering sites 5 (Kisite Marine National Park) and 6 (Wasini Community Managed Area).Boxes shows additional information for each study site on benthic cover, total fish biomass and sea urchin density.Bars represent means AE standard error (n = 10 for benthic surveys and n = 11-15 for fish surveys).Figure reused from Knoester et al. (2023).

Figure 2 .
Figure 2. Schematic representation of the experimental setup with inset photos showing (A) coral nursery structure, (B) a close-up of a nursery structure with some fouling and (C) a diver preparing a remote underwater video recording.Artwork assisted by: Vrijlanser, photo sources (A)-(C): EGK.

Figure 3 .
Figure 3. Herbivorous fish biomass (kg ha À1 ) per study site and type of fisheries management.Average biomass (n = 11-15 fish surveys) is stacked by genus.Colors indicate functional groups: grazers (blue), browsers (green), scrapers (red), excavators (purple) and territorial damselfish (yellow), and shades further identify each genus.The group Other includes a mixture of uncommon herbivorous fish.Error bars denote + SE and lower-case letters denote significant differences between study sites ( p < 0.05).

Figure 4 .
Figure 4. Average + SE numbers of mass-scaled bites per hour (bites Â kg h À1 ) on coral nursery structures (n = 8) per study site, split by fish genus and functional group.Colors of the genera match those in Figure 3.The group "Other" contains 30 species of nonherbivorous fish, which were recorded taking infrequent bites (TableS3).Study sites are grouped according to level of fisheries protection, as indicated on top.Study sites not sharing any lower-case letters experienced significantly different ( p < 0.05) average number of mass-scaled bites.

Figure 6 .
Figure 6.Average + SE density of total fouling (g m À2 ) on coral nursery structures (n = 8, except for study site 5 which has n = 4 due to damaged structures) per study site, split by type of fouling.Turf algae are categorized as a multi-species assemblage of benthic algae that are smaller than 1 cm in height, Macroalgae include brown, red and green fleshy algae, CCA are crustose coralline algae and Shelled animals include both bivalves and barnacles.The group 'Other' contains remaining uncommon benthic groups consisting predominantly of sponges and tunicates.Study sites are grouped according to level of fisheries protection, as indicated on top.Study sites not sharing any lower-case letters have significantly different ( p < 0.05) average density of total fouling.See TablesS1 and S4for the statistical results for each fouling type separately across study sites.

Figure 7 .
Figure 7. Average growth rates per study site of the coral Acropora verweyi in nursery structures (n = 8, except for study site 5 which has n = 4 due to lost structures) during the 4-month study period.Growth rates are expressed as a constant growth factor (SGR: specific growth rate in day À1 ) of the exponential increase in coral volume over time.Study sites are grouped according to level of fisheries protection, as indicated on top.Study sites not sharing any lower-case letters have significantly different ( p < 0.05) average growth rates.Error bars denote AE SE.