Fisheries Restrictions and Their Cascading Effects on Herbivore Abundance and Macroalgae Removal at Kenyan Coral Reefs

The increase of macroalgae at degraded reefs impedes ecosystem services and calls for effective methods to facilitate a return to coral dominance. Removal of macroalgae (i.e. browsing) is typically realized by �sh, but the role and identity of browsers at the heavily-shed East African Coast is still largely unknown. This study investigated how browsing pressure at Kenyan reefs (-4.700, 39.396) related to �sheries management and herbivore abundance. From October 2018 to January 2019, consumption during 24-h buffet assays using the brown macroalgae Sargassum and Padina was determined and video recorded at six sites: two in �shed zones, two in marine reserves (traditional �shing allowed) and two in no-take zones. Herbivorous �sh composition and biomass and urchin abundance were also determined. Consumption of both Sargassum and Padina was signi�cantly lower in the �shed zones (25% and 27% of macroalgal biomass consumed, respectively) compared to the no-take zones (66% and 81%), with intermediate consumption in the marine reserves (49% and 73%). Biomasses of herbivorous �sh (browsers, grazers and scrapers) were between 3 to 30 times higher in the protected zones, whereas sea urchins and territorial damsel�sh were more abundant in �shed zones. Macroalgae consumption correlated positively with the biomass of browsing and grazing �sh and negatively with sea urchin abundance. Fish from various functional groups were identi�ed as dominant browsers and these differed strongly across sites. These results indicate that �shing restrictions are likely to support reef resilience by increasing herbivorous �sh biomass of key species and thereby promote macroalgae removal.


Introduction
Important ecosystem services such as coastal protection and sustainable sheries provision become reduced when coral reefs degrade into seascapes dominated by macroalgae (Pratchett et al. 2014; Rogers et al. 2018).Herbivores, especially sh, play a crucial role in promoting coral over macroalgal dominance either by preventing the establishment of macroalgal recruits through continuous removal of algal turf (i.e.grazing) or by reversing macroalgal dominance through selective removal of mature macroalgae (i.e.browsing).Grazing of algal turf creates favourable conditions for coral growth and settlement and thereby supports coral dominance (Lefcheck et al. 2019) and consequently reef biodiversity and resilience (Nyström et al. 2008).However, with climate change-induced temperature stress weakening the competitive potential of corals (Sully et al. 2019) in combination with over shing (McCauley et al. 2015) and eutrophication (Norström et al. 2009), the impaired grazer community becomes increasingly unable to prevent macroalgae establishment.First observed in the Caribbean (Hughes 1994), coral to macroalgae phase shifts are now repeatedly reported from Indo-Paci c reefs as well (Ledlie et al. 2007).Once established, macroalgae prevent a return to a coral-dominated state by supressing the survival, fecundity and recruitment of corals (Hughes et al. 2007;Schmitt et al. 2019).It is only through the active removal of macroalgae by browsers that reversal of such phase shifts can likely be achieved (Puk et al. 2016).Therefore, a good understanding on the distribution of key browsing species and the factors in uencing their potential to control macroalgae is important.
The use of functional groups has proven helpful to understand the principles of ecological processes and coral reef resilience (Heenan and Williams 2013).Even though herbivory is a well-studied process on coral reefs, research focussed on browsers speci cally has been challenging and full of surprises.
Identi cation of key species responsible for macroalgae removal based on their abundance only has remained precarious (Puk et al. 2016) and therefore browser biomass estimates alone remain inadequate as an indicator for reef resilience (Nyström et al. 2008;Brandl and Bellwood 2014).Due to the cautious nature of browsers, biomass estimates from visual surveys tend to underestimate browser diversity and biomass (Hoey and Bellwood 2010;Michael et al. 2013) and the use of video-recorded macroalgae buffet assays have provided a wealth of additional information on browsers (Bennett and Bellwood 2011).
Browsing on macroalgae seems to be commonly done by a small subset of the diverse browser community due to spatial, temporal and behavioural variation (Bennett and Bellwood 2011; Lefèvre and Bellwood 2011; Puk et al. 2016), and is sometimes dominated by species traditionally not considered as browsers (Bellwood et al. 2006; Chong-Seng et al. 2014; Tebbett et al. 2020).In addition to these various groups of sh, sea urchins can also contribute to the control of macroalgae (McClanahan et al. 1994) and this has been especially important for Caribbean reefs (Francis et al. 2019).At certain over shed Indo-Paci c reefs increasing numbers of sea urchins have become more dominant in macroalgal control (Humphries et al. 2014), but also contribute heavily to reef erosion through their scraping feeding method (Norström et al. 2009).In short, the identity and impact of browser communities appear variable across broad ranges but also between reefs with apparent plasticity in functional roles, making it di cult to predict browsing function from biomass estimates alone.
Mismatches between abundance and ecological relevance of browser species have important management implications, where protection of certain species might not result in the desired coral reef resilience (McClanahan 2008).A good understanding on which species contribute most to macroalgae removal can enable managers to increase reef resilience and the likelihood of phase shift reversal when restoration towards coral dominance is desired (Ladd et al. 2018).Although several studies on the wellprotected Great Barrier Reef have proven invaluable to identify dominant browsers and potential drivers for the regional variability in browsing pressure (Hoey and Bellwood 2009;Bennett and Bellwood 2011;Michael et al. 2013;Streit et al. 2015), the application of this knowledge to other geographical areas and sites with higher shing pressure may be limited.Indeed, superimposed on geographic differences (Heenan et al. 2016) is the divergent shing pressure among coastal populations, in which large-bodied shes such as browsers are preferred targets (Plass-Johnson et al. 2015).The high susceptibility of browsers to shing pressure and their important yet complex role in the coral-algae balance call for a better understanding of these dynamics in general and variability among geographical areas in particular.
This study aimed to further improve our understanding on browsing by expanding the geographic scope and including the impact of sheries management within the study domain.We provide an identi cation of key browsers and quanti cation of their browsing capacity at six Kenyan reefs, which form part of a mostly intensely-utilized and understudied fringing reef in the Western Indian Ocean.Here, the enforcement of three distinct sheries management zones (open access shed zone, marine reserve for traditional shing only and a well-enforced no-take zone) are well suited to investigate the impact of management on the browser community and their in uence on reef resilience.We hypothesise that increasing levels of sheries restrictions will increase herbivorous sh biomass and result in more effective control of macroalgae.Through video-recorded macroalgae buffet assays we identi ed key browsing species and their browsing pressure and compared this with biomass estimates from stationary sh surveys and sea urchin counts.Lastly, we used benthic surveys to investigate how two commonlyused reef condition indicators (coral cover and macroalgal cover) relate to the browser community and their consumption of macroalgae to allow for a perspective on reef resilience.

Study sites
The study was performed around Wasini Island in southern Kenya from October 2018 till January 2019 during the dry northeast monsoon.Tidal differences in the area are signi cant, reaching over four meters during spring tide.Six study sites were chosen (Fig. 1

Benthic and sh surveys
A 20-m point intercept transect with 0.5 m interval was used to map benthic cover in broad categories (hard coral, soft coral, turf algae (< 1 cm), eshy macroalgae (> 1 cm) and a pooled category 'other' including mainly sand, rubble and uncommon invertebrates such as sponges and tunicates (Hill and Wilkinson 2004).Within a 2 x 20 m 2 belt transect sea urchins were identi ed to species level and counted to determine their abundance (Hill and Wilkinson 2004).A stationary sh census (radius of 7.5 m) was used to quantify the composition and abundance of all diurnal, non-cryptic sh (Bohnsack and Bannerot 1986).Fish sizes (fork length) were estimated in classes of 5 cm for shes smaller than 20 cm, and in 10cm size classes for larger individuals.Per study site, 10 replicate benthic surveys and 11 -15 replicate sh surveys were performed.

Macroalgae buffet assay
At each study site, browsing pressure was determined by recording consumption from macroalgae buffet assays over 24 h (Fig. 2).Two brown macroalgae were used: Sargassum ilicifolium and Padina boergesenii (henceforth referred to by genus name only).These brown algae were chosen as they are the dominant macroalgae in the area and are typical representatives of algal climax communities (Humphries et al. 2014).One day before use, the macroalgae were collected from the shallow reef at at study site 1 and stored in seawater basins.Before and after deployment, the drip-dry wet weight (shaken 10 times to remove excess water) of each macroalgae was determined.The macroalgae were kept in their natural growth form, resulting in the following average starting weights (mean ± SD): Sargassum (38.3 ± 4.6 g) and Padina (18.4 ± 2.9 g).The buffet assay also included the seagrass Thalassia hemprichii to allow for comparisons with an older buffet assay study from the Kenyan coast (McClanahan 2008), but these results are discussed separately in the supplementary materials (Fig. S1) as the focus of this report is on macroalgae.For deployment, the three different macrophyte species were strung equidistant and in random order on a 1-m shing line.The line was weighted with three 10-cm metal pins to enable secure placement on the reef substratum and to provide access to both vertebrate and invertebrate, bottom-dwelling browsers.The macrophytes were transported in basins with seawater and deployed at the reef slope of study sites with a fringing reef or on patch reefs at the other sites.One replicate buffet assay consisted of 10 lines, each line approximately separated 2 meters from the next.
Per replicate assay, one additional control line was placed inside a weighted plastic cage of 30 x 30 x 50 cm 3 with 1.3 x 1.3 cm 2 mesh size to exclude all macro-browsers and was used to check for weight loss due to handling, following Seah et al. (2021) amongst others.For each study site, the assays were repeated on ve non-consecutive days throughout the 4-month study period.

Remote underwater video
To identify the species responsible for the reduction in macroalgae biomass and to quantify their browsing activity with minimum disturbance, the rst 75 mins of deployment of each buffet assay was recorded on remote underwater video (RUV).A Canon 600D DSLR camera in a Neewer 40M case was placed on a weighted tripod, approximately 2 meters from one randomly chosen line.The camera was programmed to take 10-min clips, with a starting delay and interval of 5 min, resulting in a total recording time of 45 min per assay.Recording took place between 1000 h and 1400 h, which coincides with the peak in foraging activity of most roving (i.e.mobile) herbivorous sh (Hoey and Bellwood 2009).In total, 30 recordings were made.

Data processing and analysis
Data on benthic cover, sh composition, sh biomass and grazing sea urchin density is presented descriptively.Grazing sea urchins include all sea urchin species except two burrowing species (Echinometra mathaei and Echinostrephus molaris).Data from stationary visual surveys was used to estimate sh biomass using the midpoint of each size class and published length-weight relationships (Froese and Pauly 2015).Average herbivorous sh biomass was subdivided into the following functional groups: browsers, grazers, scrapers and excavators, based on reported species' functional traits following Green & Bellwood (2009).Two additions were made: Platax spp.were also considered browsers (Bellwood et al. 2006) and territorial damsel sh were considered as a separate functional group, including the common genera Amblyglyphidodon, Plectroglyphidodon and Stegastes.
The fraction consumption of buffet assays was calculated following Bennett & Bellwood (2011): where M 0 is the initial macroalgal mass, M 1 the remaining mass after 24 h and C the mean fraction of biomass loss at the control treatment (calculated separately for each site and each macroalgae).The mean of 10 lines was taken for each replicate buffet assay and data analysis was performed in R (R Core Team 2020).A beta regression model with logit link was used to account for the proportional nature of the consumption data (Douma and Weedon 2019) using the glmmTMB package (Brooks et al. 2017).A mixed-effects model was built to determine the xed effects of Protection (Fished, Reserve and No-take) and Species (Sargassum and Padina) on consumption.Study site and Assay were included as nested random factors to account for the non-independence of repeated measurements at each study site within protection zone and the non-independence of macrophyte species on the same line.The Wald Chi-Squared Test from the car package (Fox and Weisberg 2018) was used to determine the signi cance of the xed factors and their interaction.Model assumptions were validated by visual inspection of DHARMa diagnostic plots for mixed regression models (Hartig 2021).Within-level differences between Protection for each Species were examined using pair-wise means comparisons with Tukey adjustment using the emmeans package (Lenth 2020).
All RUV recordings were viewed and for each bite the targeted macroalgae and involved sh species were noted.In addition, sh's fork length was estimated (using the buffet line as reference), transformed to weight using published length-weight relationships (Froese and Pauly 2015) and multiplied by the number of bites taken to calculate mass-scaled bites (ms-bites) following Hoey and Bellwood (2009).Sums of ms-bites were standardized to hour to correct for slight variations in RUV recording length.Bite impact estimated from RUV is thus expressed as ms-bites in kg h -1 .
Average macroalgae consumption per study site was correlated against the study site's average herbivorous sh biomass (browsers, grazers, scrapers, territorial damsel sh and total), grazing sea urchin density, hard coral cover and macroalgal cover using Pearson's test.Macroalgae consumption was also speci cally correlated against the biomass of unicorn sh (Naso spp.) as these sh were identi ed as abundant and important browsers.Excavators were not included in these correlations because of their low numbers.Patterns were similar for both Sargassum and Padina and it was decided to pool the two macroalgae to aid visualisation and increase the low statistical power, but it should be noted that the two macroalgae were not independent of each other (presented on the same assay line).The correlations are thus mainly intended to explore broad patterns and conclusions drawn from them will be moderated accordingly.

Benthos and sh
Average hard coral cover was relatively high across study sites with values ranging between 25 -47%, except at site 3 where only 6% hard coral cover was found (Fig. 1).In contrast, macroalgal cover was low for all sites (< 8%) except at site 3 where half of the substrate was covered by macroalgae (mainly Sargassum spp.).The density of grazing sea urchins (Fig. 1) was highest in the shed zones (0.8 and 1.1 sea urchins m -2 at sites 1 and 2, respectively) and lower in the marine reserves and no-take zones (< 0.5 sea urchins m 2 at sites 3 -6).Total sh biomass was low in the shed zones and marine reserves, with values ranging between 150 -285 kg ha -1 for sites 1 -4 (Fig. 1).In the no-take zones, total sh biomass was much higher: 898 kg ha -1 for site 5 and 1667 kg ha -1 for site 6.In line with total sh biomass, the biomass of herbivorous sh was highest in the no-take zones, but also site 4 in the marine reserve hosted a considerable biomass of herbivorous sh (Fig. 3).These higher herbivore biomasses were not only attributable to more herbivores being present, but also due to the presence of larger (> 30 cm) individuals, which were completely absent from sites 1 -3 (Fig. S2).
The composition of functional groups within the herbivorous sh community also differed between sites.Browsers were rarely seen at shed sites 1 and 2 or at site 3 in the marine reserve but made up between a third to more than half of the herbivorous sh community at site 4 in the marine reserve and in both notake zones (Fig. 3).Grazers made up about a third of the herbivorous sh community at most sites, except at site 2 where over two-thirds of the herbivores were grazers and site 6 where only a fth were grazers.Scrapers and excavators were only regularly encountered in the no-take zones and made up slightly more than a third of the herbivorous sh community there.Territorial damsel sh made up only a small proportion of the herbivorous sh biomass for all sites, except site 1 where these small sh contributed a striking 70% to the herbivore biomass.Macroalgae buffet assay A signi cant interaction was found between Protection and Species (X 2 = 10.917,df = 4, p = 0.0275), see Table S1 for detailed model output.For Sargassum, the consumption was more than two-fold higher in no-take zones (66 ± 9%) compared to the shed zones (25 ± 7%; p = 0.0064), with intermediate results for the marine reserves (49 ± 10%) that were not signi cantly different from the other two management types (Fig. 4).For Padina, consumption was comparably low in the shed zones (27 ± 7%) and consumption signi cantly increased over two fold in the marine reserves (73 ± 8%; p = 0.0002) and three-fold in the notake zones (81 ± 7%; p < 0.0001); the marine reserves and no-take zones were not signi cantly different from each other (see Table S2 for details on Tukey's post hoc).Across all three protection zones, consumption of Padina was higher than for Sargassum (Fig. 4), and this was most evident at sites 3 and 6 (see also Fig. 6).

Recorded bites
Mass-scaled bites as recorded on RUV were dominated by a few sh species (Fig. 5).Only three species were recorded taking substantial ms-bites of Sargassum (Hipposcarus harid, Naso elegans and Zebrasoma desjardinii).Bites on Padina were predominantly taken by a small group of scarids (Hipposcarus harid, Scarus tricolor and Calotomus carolinus) and the unicorn sh Naso elegans.The majority of the ms-bites were recorded at sites 4 -6 (Table S3), whereas only very few ms-bites were recorded at sites 1 -3 and those were mainly taken by a small territorial damsel sh (Plectroglyphidodon lacrymatus).At sites 4 -6, a different species dominated at each site, with Naso elegans taking most ms-bites at site 4, Scarus tricolor dominating ms-bites at site 5 and Hipposcarus harid taking most msbites at site 6 (Table S3).

Correlations
The fraction of consumed macroalgae correlated positively with the total biomass of herbivorous sh, though seemed to level off above an herbivorous sh biomass of around 200 kg ha -1 (Fig. 6a).Looking at the separate functional groups, macroalgae consumption correlated positively with the biomass of browsing herbivorous sh (Fig. 6b) and this correlation appeared mainly driven by browsers of the genus Naso (Fig. 6c).A strong positive correlation was found for grazing herbivorous sh and macroalgae consumption (Fig. 6d).The fraction of consumed macroalgae did neither correlate signi cantly with the biomass of scrapers (Fig. 6e) nor territorial damsel sh (Fig. 6f).The abundance of grazing urchins correlated negatively with the fraction of consumed macroalgae (Fig. 6g).No statistically signi cant correlation was found between the fraction of consumed macroalgae and the cover of macroalgae (Fig. 6h) or hard coral cover (Fig. 6i).

Discussion
A major threat to coral reefs is the phase shift from corals towards macroalgae, promoted by eutrophication and warmer waters and exacerbated by the removal of herbivorous sh by over shing (Hughes et al. 2007; Ledlie et al. 2007;Pratchett et al. 2014).Key browsing sh species, their impact on reefs and the relationships to sheries management are geographically variable and still largely unknown from the East African Coast.We characterized the herbivore community and quanti ed their browsing pressure at six Kenyan reefs within three distinct sheries management zones, and related realized browsing pressure to herbivore biomasses and other common reef indicators.Browsing pressure on the presented macroalgae was two-fold greater in areas with partial shing restrictions and up to three times higher at fully protected reefs, and this higher browsing pressure correlated positively with the higher biomasses of herbivorous sh present at these protected reefs.In line with previous studies, only a select few dominant browsers were identi ed (Puk et al. 2016), the key species varied strongly across reefs (Cvitanovic and Bellwood 2009) and also included herbivores not speci cally classi ed as browsers (Chong-Seng et al. 2014).Browsing pressure correlated poorly with coral cover suggesting that coraldominated reefs can persist even under low levels of browsing (Cernohorsky et al. 2015;Holbrook et al. 2016).The implication of limited browsing activity, however, is that these reefs might be impaired in their ability to absorb disturbances that stimulate macroalgal growth (Holbrook et al. 2016).Overall, our results a rm that shing restrictions can have a strong positive in uence on herbivorous sh biomass and highlight how this can be expected to increase reef resilience by supporting higher rates of browsing.
Shifts in the herbivore community and strong reductions in key functional groups including browsing, scraping and excavating herbivores were found at the shed study sites, con rming results found at the northern Kenyan coast (Humphries et al. 2015), and potentially undermining the resilience of these reefs (Nyström et al. 2008;Holbrook et al. 2016).Herbivorous sh biomass in the no-take zones and marine reserve (except site 3) was comparable with worldwide averages from protected reefs (Edwards et al. 2013) and this biomass was around 10 times lower at the sites without shing restrictions.An exception was the macroalgae-dominated study site 3 in the marine reserve, which had an equally low sh biomass as the shed reefs.At this site and at the shed reefs, no large herbivores (> 30 cm) were recorded, indicative of severe over shing (McClanahan et al. 2008), habitat degradation (Rogers et al. 2018) or both.The observed reductions in herbivore biomass were most striking for large-bodied and functionally important sh such as browsers, scrapers and excavators.This impact of high shing pressure on key functional groups has been observed worldwide (Edwards et  It is therefore promising that the small and recently-established community managed no-take zone of Wasini (study site 6) has been able to sidestep this trend and now boasts the highest sh biomass of all the sites studied here, despite its nearshore location (Johansson et al. 2013).Unlike other young community managed reserves in Kenya where only grazers recovered (Humphries et al. 2015), also browsers and scrapers are abundant at Wasini.Our data suggest that of all herbivorous guilds, grazers are least impacted by high shing pressure, with 'only' a three-fold reduction of their biomass at shed sites compared to the no-take zones, conform ndings of Heenan et al. (2016) in American Samoa.Sea urchins and territorial damsel sh showed highest abundances in shed zones and it is likely that they bene t from reduced competition as well as reduced predation by larger sh (Ceccarelli et al. 2005;McClanahan 2008).
In line with the higher biomasses of roving herbivorous sh, the removal of macroalgae was up to 3-fold greater in the no-take zones and more than two-fold greater in marine reserves compared to the shed zones.Consumption of Sargassum in the no-take zones and marine reserves was higher than found in a previously studied community managed area in northern Kenya where only 20% was consumed in 24 h (Humphries et al. 2015), but somewhat lower than consumption rates (81 -92% in 24 h) found at the Great Barrier Reef (Hoey and Bellwood 2010).Padina consumption fell broadly within the ranges previously found (Humphries et al. 2015;Plass-Johnson et al. 2015).It seems that despite widely varying species compositions across broad geographic scales, realized browsing pressure at un shed reefs is quite comparable (Tebbett et al. 2020), highlighting the role local drivers can play in determining browsing pressure.Interestingly, consumption at the sh-depauperate and macroalgae-dominated study site 3 was also relatively high.This result contrasts with previous studies where higher densities of macroalgae were associated with lower browsing rates, supposably through feeding dilution (Chong-Seng et al. 2014) or predator avoidance (Hoey and Bellwood 2011).The combination of both low sh and urchin biomass, the absence of browsing recorded on RUV, but relatively high macroalgae consumption at this structurally-eroded and macroalgae-dominated site is indeed surprising.It should be noted, however, that the high consumption was mainly driven by removal of Padina, the macroalgae which appeared overall more palatable in this and other experiments (Humphries et al. 2015), compared to Sargassum, the macroalgae which dominated this reef and is most often associated with phase shifts (Hughes et al. 2007).Still, the organism responsible for the high removal of Padina at this site remains unidenti ed and could possibly include overlooked species such as nocturnal crabs (Francis et al. 2019).
At the two shed study sites, consumption was higher compared to reports of other over shed or macroalgae-dominated reefs.For example, Sargassum sp.removal rates of only 2% in 4.5 h were found on macroalgae-dominated reefs in the Seychelles (Chong-Seng et al. 2014).At the heavily shed zones of this study, despite the low biomass of browsing sh, macroalgae removal is possibly still realized by small-bodied grazers and sea urchins.
The possibility that small-bodied grazers can endure high shing pressure and control macroalgal establishment could be seen as hopeful (Cernohorsky et al. 2015;Knoester et al. 2019;Müller et al. 2021), yet there are several reasons to be cautious.First, small herbivorous sh are likely to be targeting leaves or epiphytes only, without removing the holdfasts of macroalgae (Streit et al. 2015).Second, small herbivorous sh appear more vulnerable to bleaching events and the ensuing habitat loss of branching coral (Nash et al. 2016).In addition, when bleaching events open up large areas of space, macroalgal settlement and growth is likely to overwhelm the grazing capacity of small herbivores, increasing chances of a phase shift (Williams et al. 2001).We suppose this could have happened at study site 3 during the  (Humphries et al. 2020).In addition, as illustrated by the negative correlation between sea urchin abundance and macroalgae consumption in this study, browsing pressure by urchins (at over shed reefs) appears to be relatively small compared to the browsing pressure by herbivorous sh (at protected reefs).Hence, at over shed reefs, small-bodied sh and sea urchins may partially take over the role of larger herbivorous sh in controlling macroalgal growth, but such a change in control is likely to undermine reef resilience.
The apparent limited functional redundancy of browsers at the studied protected reefs may also have implications for reef resilience, as the loss of key species can have large detrimental impacts on ecosystem functioning (Cheal et al. 2013;Nash et al. 2016).In accordance with reports from numerous preceding studies using macroalgae buffet assays, browsing in this study was dominated by a few species only (Puk et al. 2016) and with marked variation in dominant species across sites (Cvitanovic and Bellwood 2009).Naso elegans was among the dominant browsers in this study and closely-related species have been identi ed as dominant browsers across the Indo-Paci c (Hoey & Bellwood, 2009;Plass-Johnson et al., 2015;Welsh & Bellwood, 2015), highlighting the importance of this genus in macroalgal control across broad geographic scales.Browsing was not only performed by those classi ed as browsers and this supports several studies that suggest plasticity in functional roles exists (Bellwood et  Interestingly, Siganus spp.and Kyphosus spp., species frequently identi ed as dominant browsers at the Central Indo-Paci c and Great Barrier Reef (Michael et al. 2013;Puk et al. 2016) were not recorded biting in this study despite their presence in nearby seagrass habitat and at the reef crest.Instead, Siganus spp. in the Western Indian Ocean have been shown to target more delicate and turf algae (Humphries et al. 2015;Ebrahim et al. 2020).Kyphosus spp., situated higher in water column, might have been feeding on drifting algae instead (Carpenter 1986).Altogether, these results align with other browsing studies in the Indo-Paci c in that they identify only a select and sometimes surprising group of species responsible for macroalgal removal from the diverse assemblage of potential browsers present.This variability may in part explain the nding that the reef with the highest herbivore biomass in this study did not bolster the highest browsing pressure, and especially consumption of Sargassum was relatively low here.Thus, in addition to the biomass of herbivores present, realized browsing pressure is likely also depending on many more factors such as spatial restrictions (Puk et  and coral dominance are still realized with herbivorous sh biomasses as low as 80 kg ha -1 .Nevertheless, reefs like these might be pushed to macroalgae dominance through an external disturbance such as bleaching (Williams et al. 2001).The minimum herbivore biomass needed to absorb such increasingly common disturbances remains unknown for the Indo-Paci c (Roff and Mumby 2012), but could possibly lie around 200 kg ha -1 at our study sites as above this threshold browsing pressure was found to level off.
In the marine reserve, a coral-dominated and a macroalgae-dominated reef co-exist under roughly equal browsing pressure.This co-existence could be indicative of alternative steady states (Holbrook et al. 2016) and illustrates that shifts to alternative stable states can be di cult to reverse even when ambient browsing pressure is relatively high (Schmitt et al. 2019).In such cases, the protection of herbivorous sh could be reinforced by removing macroalgae manually in an effort to push the ecosystem back to coral dominance (Ceccarelli et al. 2018;Williams et al. 2019).More effective, however, would be to keep herbivore levels well above phase-shift thresholds and prevent macroalgal dominance in the rst place (Anthony et al. 2015).Our results indicate that sheries management through marine reserves and notake zones in particular, even small-scale and community-managed, have the potential to safeguard the diversity and biomass of functionally important herbivorous sh.Following effective management, a high level of macroalgal control is realized as large-bodied browsing and scraping sh seem to bene t markedly from sheries protection.Although gains in browsing pressure appear to level off with increasing herbivore biomass and reasonable levels of browsing were still realized at shed study sites, the long-term resilience of these shed reefs is uncertain given the eroding nature of urchin browsing (Carreiro-Silva and McClanahan 2001), the high susceptibility of small-bodied herbivorous sh to coral loss (Nash et al. 2016) and their limited capacity to control sudden increases in macroalgae (Williams et al. 2001;Streit et al. 2015).Therefore, we recommend to continue the establishment of a network of community managed no-take zones to allow for the recovery of herbivorous sh biomass well above potential phase-shift thresholds, increase ecosystem resilience, promote local stewardship and move towards sustainable use of coral reefs (Topor et al. 2019).Such local management could help restore and maintain coral dominance and provide heightened resilience against large-scale disturbances during the Anthropocene.Fraction of macroalgal biomass consumed in 24 h (F) for both Sargassum ilicifolium and Padina boergesenii, split between three types of sheries management.Bars present mean ± standard error (n = 10).Letters above indicate signi cant differences (p < 0.05) between sheries management for each macroalgae Figure 5 Recorded mass-scaled bites (kg h -1 ) on presented macroalgae by sh species recorded on remote underwater video, summed across all study sites Figure 6 Correlations between the fraction of macroalgal biomass consumed and a the total biomass of herbivorous sh b the biomass of browsing herbivorous sh and c browsers of the genus Naso speci cally, d grazing herbivorous sh biomass, e scraping herbivorous sh biomass, f territorial damsel sh biomass, g urchin abundance, h macroalgal cover and i hard coral cover.Linear trend lines are added with their associated Pearson correlation coe cient (R) and signi cance indicated (*p < 0.05, **p
strong 1998 El Niño (McClanahan et al. 2001), despite the implemented partial shing restrictions (Williams et al. 2019).Lastly, while increasing numbers of sea urchins can partially compensate for the loss of herbivorous sh (McClanahan 2014), the intensity of their scraping feeding method can undermine long-term reef development through bioerosion (Carreiro-Silva and McClanahan 2001) and hinder coral settlement al. 2006; Chong-Seng et al. 2014; Tebbett et al. 2020).Indeed, scraping parrot sh were recorded taking substantial amounts of bites as has been found in previous studies (McClanahan et al. 1994), but are more likely to have been targeting epiphytes (Lefèvre and Bellwood 2011; Clements et al. 2017).
al. 2016), behavioural variation (Bennett and Bellwood 2011) as well as temporal variation (Lefèvre and Bellwood 2011; Seah et al. 2021).The variation in browsing pressure found can be indicative of divergent resilience between the studied reefs and their sheries management (Nyström et al. 2008).Holbrook et al. (2016) experimentally identi ed a coral to macroalgae tipping point when herbivore biomass dropped below about 50 kg ha -1 in Moorea.This experimental biomass threshold corroborates eld observations from reefs in Indonesia, where a lack of macroalgal removal was observed at herbivorous sh biomass levels below 50 kg ha -1 (Plass-Johnson et al. 2015).Three of the six Kenyan reefs included in the current study harboured an herbivore biomass that was just above this threshold.On two of these reefs, coral is still dominant over macroalgae.This nding is in agreement with data from Singaporean (Bauman et al. 2017), Indian (Cernohorsky et al. 2015) and Indonesian (Plass-Johnson et al. 2015) reefs, where macroalgae removal

Figure 1 Map
Figure 1