Sedimentation and overfishing drive changes in early succession and coral recruitment

Abstract

Sedimentation and overfishing are important local stressors on coral reefs that can independently result in declines in coral recruitment and shifts to algal-dominated states. However, the role of herbivory in driving recovery across environmental gradients is often unclear. Here we investigate early successional benthic communities and coral recruitment across a sediment gradient in Palau, Micronesia over a 12-month period. Total sedimentation rates measured by ‘TurfPods’ varied from 0.03 ± 0.1 SE mg cm⁻² d⁻¹ at offshore sites to 1.32 ± 0.2 mg cm⁻² d⁻¹ at inshore sites. To assess benthic succession, three-dimensional settlement tiles were deployed at sites with experimental cages used to exclude tile access to larger herbivorous fish. Benthic assemblages exhibited rapid transitions across the sediment gradient within three months of deployment. At low levels of sedimentation (less than 0.6 mg cm⁻² d⁻¹), herbivory resulted in communities dominated by coral recruitment inducers (short turf algae and crustose coralline algae), whereas exclusion of herbivores resulted in the overgrowth of coral inhibitors (encrusting and upright foliose macroalgae). An ‘inducer threshold’ was found under increasing levels of sedimentation (greater than 0.6 mg cm⁻² d⁻¹), with coral inducers having limited to no presence in communities, and herbivore access to tiles resulted in sediment-laden turf algal assemblages, while exclusion of herbivores resulted in invertebrates (sponges, ascidians) and terrestrial sediment accumulation. A ‘coral recruitment threshold’ was found at 0.8 mg cm⁻² d⁻¹, below which net coral recruitment was reduced by 50% in the absence of herbivores, while recruitment was minimal above the threshold. Our results highlight nonlinear trajectories of benthic succession across sediment gradients and identify strong interactions between sediment and herbivory that have cascading effects on coral recruitment. Local management strategies that aim to reduce sedimentation and turbidity and manage herbivore fisheries can have measurable effects on benthic community succession and coral recruitment, enhancing reef resilience and driving coral recovery.

1. Introduction

Coral reefs are subject to a wide range of anthropogenic stressors. At global scales, increases in sea surface temperatures over the past decades have resulted in mass coral bleaching events [1,2], causing widespread declines in coral cover [3] and coral recruitment [4]. At local scales, widespread catchment clearing, dredging and coastal development has resulted in increasing sediment loading in near-shore coral reefs [5–7], and overfishing of herbivorous fish can result in shifts from coral to macroalgal-dominated states [8]. While mitigation of rising sea surface temperatures impacting coral reefs requires global commitments [9], local stressors can be ameliorated or reversed through applied management, stemming the decline and ultimately driving recovery of coral reefs [10]. However, the impacts of sediment and herbivory on succession and recovery of coral reefs are not well understood, and thresholds for management targets remain elusive.

The recovery of coral reefs following disturbance depends on the successful settlement, metamorphosis and survival of coral individuals into a benthic population [11,12]. Sedimentation and overfishing can alter benthic succession from assemblages dominated by recruitment ‘inducers’ including crustose coralline algae, short algal turfs and epilithic algal matrix [13–15] to assemblages dominated by recruitment ‘inhibitors’ including macroalgae, heterotrophs (e.g. sponges, byrozoans, ascidians) and long algal turfs [16–18]. Shifts in benthic assemblages have been documented in the Pacific, both through experimental herbivore exclusions associated with a fourfold decrease in coral recruitment [13] and monitoring following natural disturbance, whereby macroalgal proliferation inhibited coral settlement at reef scales [15].

Catchment clearing and run-off can alter ecological trajectories of coral reef benthos through light attenuation [19], smothering [20] and eutrophication [21]. Suspended sediments can also directly impact coral recovery by negatively affecting coral fertilization success [22] and settlement [23], while causing a reduction in growth and survival of newly settled recruits [24] and juveniles [25]. Increasing evidence suggests that sedimentation can interact with herbivory to impact negatively on coral recovery. Increased suspended sediments and sediment deposition on coral reef benthos can deter herbivorous fish from grazing [26] or may displace them entirely [27], resulting in proliferation of sediment-laden long algal turfs [28].

With widespread development and increasing population growth throughout the Pacific region [29], understanding the roles of sediment and overfishing of herbivores on ecological trajectories and coral recovery is of increasing importance. Here, we take advantage of the natural environmental gradient of sedimentation in Palau, to quantify how herbivory and sedimentation interact to drive trajectories of benthic succession on coral reefs. Through a 12-month herbivore exclusion experiment, we document patterns of benthic succession and net coral recruitment across 15 sites from clear lagoonal waters to highly turbid marginal reef environments.

2. Methods

The study was conducted on inshore reefs of Palau archipelago (Micronesia). Recent population growth on the main island (Babeldaob, figure 1) has led to increased development and catchment clearing within many of its watersheds [30]. Study sites were located adjacent to three locations on the west coast of the Babeldaob (figure 1a): Ngardmau (North), Ngeremlengui (Central) and Ngermeduu (South) following a previous study on sediment dynamics [31]. Within each location, five sites were chosen with increasing distance from river mouths, where site 1 was the closest to the river mouth, and site 5 was located in clear lagoonal waters. Within locations, sites were located within a total environmental gradient of less than 2.8 km, and each location was separated by less than 10 km distance. Palau provides an ideal area to study sedimentation impacts on benthic communities because the sedimentation gradient occurs over a small scale of less than 3 km which means that dispersal limitation is unlikely to occur and confound patterns of coral recruitment.

Figure 1.

(a) Location of sites situated within three Western locations of Palau: Ngardmau (North), Ngeremlengui (Central) and Ngermeduu (South). Study sites within the locations are indicated by coloured circles according to total sedimentation rate from turf pods (mg cm⁻² d⁻¹), (b) tiles and herbivore exclusion cages at initial deployment (0 months). (c) Representative images of succession of communities at 3, 6, 9 and 12 months on caged and uncaged tiles at inshore and offshore sites. (Online version in colour.)

(a). Environmental gradients

Terrigenous sediment within the study sites is derived from the erosion of volcanic soils from the adjacent catchments on the main island of Babeldaob [30]. At each of the sites, PVC tube sediment traps (5 cm diameter, 60 cm tall) were deployed to quantify rates of sedimentation (45 sediment traps in total) at each of the five sites within each location (n = 3 per site) to replicate the study design of Golbuu [31]. Cylinder traps were deployed for three-month periods in May 2016, August 2016, December 2016, March 2017 and June 2017. Following retrieval of cylinder traps after each period, sediments were collected for laboratory analysis and identical traps redeployed. To allow for comparisons among studies using different methods to quantify sedimentation rates, we conducted an additional experiment in May 2019 where we simultaneously deployed ‘TurfPods’ [32] and cylinder traps (n = 3 each per site) across all study sites (n = 15) for a 5 day period (electronic supplementary material, figure S2). TurfPods were constructed identical to [32] with 3 mm artificial turf used to trap sediments.

Flow rates were measured in March 2017 at each site to assess potential differences in flow-driven productivity among sites during calm conditions (wind speeds less than 5 km h⁻¹). Flow was assessed using the percentage dissolution rate of plaster balls made from tennis ball moulds [33]. Three plaster balls were weighed and then deployed at each site and the dissolution of each plaster ball was measured after one full tidal cycle in the field using the dry weight lost. Differences in flow rates were assessed using a nested analysis of variance (ANOVA) with site nested in location.

(b). Laboratory analysis

Sediments were rinsed with freshwater, dried at 55°C, ground using a mortar and pestle and dried at 105°C for 24 h to obtain total sedimentation rate (mg cm⁻² day⁻¹) following [34]. Samples were combusted at 550°C for 4 h and reweighed to determine organic material content. Carbonate content was determined following 32% hydrochloric acid digestion of carbonate material. Pearson's correlation analysis was performed to assess the relationship between fractions (organic, carbonate and terrigenous matter sedimentation rate) and total sedimentation rate using the ‘corrplot’ function in the ‘corrplot’ package [35] in R software v. 4.0.2 [36]. To allow for comparisons with future studies, we used linear regression to analyse the relationship between total sedimentation rates (g cm⁻² d⁻¹) from TurfPod and cylinder traps at each site. Given the close agreement between the two methods (R² = 0.87), we used the coefficient of the regression (electronic supplementary material, figure S2) to convert the total sedimentation rates from 2016–2017 cylinder traps into TurfPod sedimentation rates for ecological analysis.

(c). Ecological trajectories

To track benthic succession over time, we used ‘crevice tiles’ (figure 1b) to simulate patterns of algal colonization and succession in relation to microhabitats and herbivory [11,15]. Each tile measured 10 × 10 cm and had 24 equally spaced crowns and crevices, where each crown measured 1.2 cm length × 1.2 cm width × 1.0 cm depth. Within each site, tiles were assigned one of three treatments: caged, partially caged and open (figure 1b). Caged tiles were enclosed in PVC coated wire cages (20 cm width × 20 cm length × 30 cm height, galvanized 2.5 cm wire mesh openings) to mimic the impact of overfishing by excluding large herbivorous fish from accessing the tile [11,15]. Partial cages were constructed identical to cages but had two opposing vertical sides and the lid removed to allow access to grazing fish. Open tiles had no cage structure to allow for unimpeded herbivore access to the tile. Five replicate tiles from each treatment (caged, partial, open) were deployed onto sites 1, 3 and 5 of each location, resulting in 15 tiles per site, 45 tiles per location, 135 tiles across all sites/locations. All tiles were placed at a depth of approximately 5–6 m within each site and separated by greater than 2 m distance. Tiles were deployed in May 2016 and photographed at three, six, nine and 12 month intervals (August 2016, December 2016, March 2017 and June 2017).

Tile images were analysed using Coral Point Count with the Excel extensions (CPCe) program [37]. Fifty-points were assigned per tile, and organisms under each point was categorized into functional groups: bare tile, terrestrial sediment, carbonate sediment, epilithic algal matrix and crustose coralline algae (merged into ‘EAM’), turf, upright foliose macroalgae (UFM), encrusting macroalgae (EFM), non-coral invertebrates (such as ascidians and sponges, merged into ‘invertebrates’), cyanobacteria and corals. We differentiated terrigenous and carbonate sediments into ‘finer-grained and darker-coloured sediment’ and ‘coarser, lighter-coloured sediment’ using visual assessments of particle size, colour and apparent texture based on [19], and differentiated sparse filamentous turf algae (less than 5 mm) as EAM, while dense and/or tall (greater than 5 mm) mats of filamentous algae as ‘turf’ due to their different ecological functions [38]. From a grazing perspective, EAM is functionally analogous to short productive algal turfs (SPATs, [39]), while ‘turf’ are analogous to long sediment-laden algal turfs (LSATs, [39])

(d). Coral recruitment

At the last time point, after 12 months of succession, all tiles were transported in seawater to Palau International Coral Reef Center (PICRC) to quantify coral recruitment. Tiles were observed under a dissecting microscope and any coral recruits were recorded along with maximum width, and identified to the lowest taxonomic resolution possible and subsequently grouped into Acropora spp., Pocillopora spp., or ‘other growth form’ (comprising of Poritidae, Merulinidae and Lobophylliidae and other unidentifiable taxa) due to difficulties in taxonomic identification of coral recruits.

(e). Herbivore biomass

Herbivore assemblages were surveyed using 10 replicate 30 × 4 m transects per site by the same observer (P. J. Mumby). The identity, life phase (terminal, intermediate and juvenile phases), and body length (total length to the nearest centimetre) were recorded for each individual (Scaridae, Siganidae, Acanthuridae, Pomacentridae). The lengths of individual herbivores were converted to biomass based on allometric scaling relationships.

(f). Statistical analysis

We used multivariate, mixed effect permutation-based analysis of variance (PERMANOVA, [40]) to assess changes in community structure over time. PERMANOVAs were calculated on Bray–Curtis similarity matrices of the transformed per cent cover of functional groups (999 permutations) using PRIMER 7 with the add-on package PERMANOVA+ (PRIMER-E Ltd, Plymouth, UK). The multivariate model included ‘time’ as a fixed categorical factor (four levels: three months, six months, nine months and 12 months), ‘treatment’ as a fixed categorical factor (three levels: caged, open and partially caged), ‘sediment’ as a fixed continuous variable, ‘location’ as a random categorical factor (three levels: North, Central and South, ‘site’ as a random categorical factor nested in location (nine levels: inner, middle and outer reef for each location) and ‘tile’ replicate as a random categorical factor nested in ‘site’ and ‘treatment’. As no significant difference was observed between open and partial cages (PERMANOVA, p = 0.16), and both open and partial cages were different from caged treatments (PERMANOVA, p = 0.001 for both) in the final model, the treatment factor was pooled into two groups to be used in the final model: herbivory being present (open and partially caged tiles) and herbivory being absent (caged tiles). Estimated components of variance (ECV) were used to partition the relative contribution of herbivory and sediment to the final community structure following [40]. Principal coordinates analysis (PCoA) was used to visualize patterns of community succession based on the Bray–Curtis similarity matrix of benthic community structure using the ‘pcoa’ function in the ‘ape’ package [41] in R v. 4.0.2, and used to visualize contour plots of sedimentation rates using the ‘levelplot’ function in the ‘lattice’ package [42] in R v. 4.0.2 [36]. All results are presented as mean and standard error.

To analyse the interactions between sedimentation and herbivory, generalized additive mixed effects models (GAMMs) were calculated from the percentage cover assessments at the last time point after 12 months of succession using the ‘glmmTMB’ package [43] in R v. 4.0.2 [36]. GAMMs allowed for the final percentage cover of benthic groups to be modelled against sedimentation with separate, smoothed trends for when herbivores are present and absent. GAMMs were fit using the beta_family and a logit-link function in the ‘glmmTMB’ package [43], and splines fit to the predictor using basis splines using the ‘splines’ package [36]. The model structure for GAMM analysis was identical to that of PERMANOVA, except excluded ‘time’. We modelled the response of the dominant functional groups: turf, EFM, UFM, EAM, terrestrial sediment and invertebrates.

The relationship between sedimentation and herbivore biomass was analysed using GAMMs with ‘Location’ and ‘Site nested in Location’ as random effects using the ‘gamm4’ package [44] in R v. 4.0.2 [36].

Coral recruit density at the final time point was analysed using binomial general linear mixed effects models (GLMM) with a zero-inflated negative binomial distribution using the ‘glmer.nb’ function in the ‘lme4’ package [45] in R v. 4.0.2 [36] with ‘location’ and ‘site nested in location’ as random effects. A separate GLMM was performed for the lowest taxonomic resolution (‘Acropora spp.’, ‘Pocillopora spp.’, and ‘other’) as well as the total number of coral recruits to assess if coral recruit abundance changed with exposure to different levels of sedimentation in the absence and presence of herbivory.

3. Results

(a). Environmental gradients

Total sedimentation rates recorded by cylinder traps ranged nearly 15-fold across the environmental gradient, from 0.24 ± 0.2 s.e. to 3.56 ± 1 mg cm⁻² d⁻¹ (electronic supplementary material, figure S1). Comparisons between cylinder traps and TurfPods (electronic supplementary material, figure S2) indicate a calibrated TurfPod total sedimentation rate ranging between 0.03 ± 0.1 s.e. and 1.32 ± 0.2 mg cm⁻² d⁻¹. At North and Central locations, sites closest to the river within were dominated by terrigenous/organic content, while sites further away from rivers were dominated by carbonate fractions (electronic supplementary material, figure S1). The South location had high terrigenous/organic content across all sites with the exception of the site furthest away from the river (electronic supplementary material, figure S1). Flow rate (as measured by plaster balls) differed significantly among sites (F = 8.29^14,30 p < 0.001), yet post-hoc multiple comparisons revealed no consistent significant differences among sites with the exception of a single site (Central Site 4) which had consistently higher flow rates.

(b). Ecological trajectories

PERMANOVA analysis revealed a significant interaction between herbivory and time (PERMANOVA, p = 0.001) in driving community structure over the 12-month study (electronic supplementary material, table S1). Visual exploration of benthic communities from the PCoA (figure 2a) reveals clear differences between high (greater than 0.6 mg cm⁻² d⁻¹) and low sediment (less than 0.6 mg cm⁻² d⁻¹) regimes (axis 1) and between –herbivory (caged) and +herbivory (uncaged) tiles (axis 2, figure 2b). After 12 months, high sediment sites were dominated by either terrestrial sediments and sponges/ascidians (−herbivory treatment) or by long turfs (+herbivory treatment, figure 2b). By contrast, low sediment environments were dominated by carbonate and EAM in the +herbivory treatment (figure 2b) and by UFM and cyanobacteria in the −herbivory treatment (figure 2b). Ecological trajectories for all treatments appeared to stabilize by the 12-month time period (figure 2a).

Figure 2.

(a) PCoA ordination of sites (+ herbivory = circles, –herbivory = squares) with vector overlays display the dominant benthic functional groups, and contour plots indicates sedimentation rates (mg cm⁻² d⁻¹) at each site (dashed line indicates threshold between high sedimentation [greater than 0.6 mg cm⁻² d⁻¹] and low sedimentation, [less than 0.6 mg cm⁻² d⁻¹]). (b) Principal coordinates analysis (PCoA) of benthic community structure through time, where sites are separated by herbivory (+herbivory = green circles, –herbivory = red squares) and by sedimentation rates (high sedimentation [greater than 0.6 mg cm⁻² d⁻¹] = closed symbols, low sedimentation, [less than 0.6 mg cm⁻² d⁻¹] = open symbols). Trajectories of succession are indicated by lines where time points 1, 2, 3 and 4 equate to 3, 6, 9 and 12 months. (Online version in colour.)

(c). Interactive drivers of benthic communities

To quantify the effects of the environmental gradient of sediment loading and herbivory on the cover of individual functional groups on tiles, we used GAMM models at the final timepoint (12 months). The six dominant functional groups (EAM, EFM, UFM, invertebrates, terrestrial sediment, turf algae) all varied significantly and nonlinearly across the sediment gradient (figure 3) and showed significant interactions with herbivory.

Figure 3.

(a) Relationships between total sedimentation rate (mg cm⁻² d⁻¹) and percentage cover of benthic functional groups derived from general additive mixed effects models (GAMMs) under models with +herbivory (green lines) and –herbivory (red lines) where error envelope represents ±95% CI, (b) relationship between the probability of coral recruits (mean ± s.e.) and increasing total sedimentation rate (mg cm⁻² d⁻¹) for a) coral taxa (Acroporidae, Pocilloporidae, and ‘other’) and under +herbivory (green lines) and –herbivory (red lines) predicted by zero-inflated, negative binomial generalized linear mixed effect models (GLMMs) where error envelope represents ±95% CI. (Online version in colour.)

Cover of epilithic algal matrix was two fold higher in the presence of herbivory (42.3 ± 10%) than when it was excluded (21.9 ± 10.1%) at low levels of sedimentation (p < 0.001). Yet epilithic algal matrix declined significantly with increasing sediment rate regardless of herbivory (figure 3a). Cover of encrusting fleshy macroalgae was 38% lower when herbivores were present (5.7 ± 2.9%) than when herbivores were excluded (14.9 ± 7.2%) at low levels of sedimentation (p < 0.001), and declined significantly with increasing sediment rate regardless of herbivory (figure 3a). Cover of upright fleshy macroalgae was 2.5-fold higher when herbivores were present (14.9 ± 7.2%) than when herbivores were excluded (5.7 ± 2.9%) at low levels of sedimentation (p < 0.001), and declined significantly in the +herbivory treatment with increasing sediment rate (figure 3a). Cover of invertebrates was consistently low (2–3% cover) across the sedimentation gradient in the +herbivory treatment, but increased when herbivores were excluded under high rates of sedimentation (GAMM, p < 0.001) to a maximum of 14.0 ± 5.4 (figure 3a); a sevenfold increase than when herbivores were present. Cover of terrestrial sediments was consistently low across the sedimentation gradient in the + herbivory treatment (less than 12% cover), but increased when herbivores were excluded under high rates of sedimentation (p < 0.001) to a maximum of 54.2 ± 10.9% cover, a approximately fivefold increase than when herbivores were present (11.4 ± 4.9%, figure 3a).

Cover of turf algae increased with increasing sedimentation rate in the +herbivory treatment, but exhibited a nonlinear response in the −herbivory treatment, declining under high rates of sedimentation (p < 0.001, figure 3a). Notably, the impact of sedimentation on functional groups was nonlinear and occurred at either low or high sedimentation rates. Epilithic algal matrix, upright fleshy macroalgae and encrusting fleshy macroalgae were only impacted when herbivores were excluded at low sedimentation rates, while at high sediment thresholds (approx. 0.8 mg cm⁻² d⁻¹), they declined to less than < 2% cover regardless of herbivory. Above approximately 0.8 mg cm⁻² d⁻¹, tiles were dominated by terrestrial sediment (max 70 ± 10% cover), turf (max 38 ± 13% cover) and non-coral invertebrates (max 14 ± 6%) when herbivores were excluded (figure 3a), and dominated by turf (max 92 ± 4% cover) when herbivores were present (figure 3a). Below approximately 0.8 mg cm⁻² d⁻¹, communities were dominated by turf (max 51 ± 11% cover), upright fleshy macroalgae (max 18 ± 5% cover) and encrusting fleshy macroalgae (16 ± 6% cover) when herbivores were excluded, and by epilithic algal matrix (max 41 ± 13% cover) and encrusting fleshy macroalgae (max 8 ± 3% cover, figure 3a) when herbivores were present.

(d). Herbivore biomass

Herbivore biomass declined linearly with increasing total sedimentation rate (GAMM, p < 0.001, χ² = 37.29, d.f. = 2) from a 45.5 g m² at low sediment sites to 1.4 g m² at high sediment sites (electronic supplementary material, figure S3).

(e). Coral recruitment

The most abundant recruits comprised Poritidae, Merulinidae and Lobophylliidae—categorized as ‘other’—followed by Pocillopora and Acropora (figure 3b). Coral recruitment declined with increasing sedimentation rate for all groups (figure 3b). Recruits were absent above rates of 0.8 mg cm⁻² d⁻¹ for both Pocillopora (GLMM, p = 0.04, z = −2.108) and Acropora (GLMM, p = 0.003, z = −2.914), while ‘other’ corals (GLMM, p = 0.04, z = −1.998) were encountered at low probability above rates of 0.8 mg cm⁻² d⁻¹ (figure 3b). The probability of recruitment (all taxa summed) was highly variable but was significantly higher when herbivores were present (76 ± 33%) than when herbivores were excluded (37 ± 18%) under low sediment rates (GLMM, p = 0.04, z = 2.055, figure 3b). Both treatments declined with increasing sediment rates above a threshold of approximately 0.8 mg cm⁻² d⁻¹ (GLMM, p = 0.002, z = −3.074, figure 3b).

4. Discussion

Sedimentation and overfishing are two of the greatest local stressors facing Pacific coral reefs [46,47]. While the individual impacts of sedimentation and overfishing are increasingly understood in recent years, the interaction between the threats is unclear due to the cascading impacts of sedimentation on both the physical environment and herbivory. We identify critical thresholds in sediment loading that drive changes in benthic community structure and overwhelm the effects of herbivory, inhibiting coral recruitment and impacting on coral reef recovery following disturbance.

Early succession on coral reefs can play a critical role in determining the overall composition of community structure after disturbance [13,15], with some early successional communities such as dense macroalgae becoming stable through reinforcing feedback loops. Under low levels of sedimentation, the presence of herbivores leads to the dominance of recruitment ‘inducers’ [sensu 13] such as crustose coralline algae, short algal turfs and epilithic algal matrix (combined here as EAM), while exclusion of herbivores leads to overgrowth of recruitment ‘inhibitors’ [sensu 13] such as long turfs, macroalgae (foliose and encrusting) and heterotrophic invertebrates [13–18]. Expanding these results across an environmental gradient of inshore–offshore, our results identify key coral inducer thresholds of sediment loading that interact with herbivory to alter ecological trajectories of succession. Beyond a threshold of 0.8 mg cm⁻² d⁻¹, substrates were dominated by recruitment inhibitors (long turf and terrestrial sediments) regardless of herbivory, leading to no net coral recruitment on tiles after 12 months. Below this threshold, corals were able to recruit, but the exclusion of herbivores resulted in the proliferation of encrusting and foliose macroalgae and a approximately 50% reduction in coral recruitment.

Under increasing sediment loading, the cover of macroalgae (crustose coralline algae, encrusting fleshy macroalgae and upright fleshy macroalgae) declined significantly regardless of herbivory, and was entirely absent under the highest sediment loading. The extent to which coral reef macroalgae are light limited under increasing turbidity is largely unknown, yet experimental evidence indicates that CCA are sensitive to prolonged (30 day) exposure to exceptionally low light levels (less than 0.1 mol photons m⁻² d⁻¹) under simulated dredging scenarios [48]. Research from temperate algal assemblages indicates that increases in turbidity can result in shifting species composition of algal canopies due to species-specific differences in light-use efficiency [49]. The different nonlinear responses for key macroalgal functional groups (EFM and UFM) under increased light limitation points to differences in the relative sensitivity of the different benthic algal communities to light. Separating the effects of reduced herbivory from potential shading effects from herbivore exclusion cages is challenging. Our previous experimental work using identical cages indicates a 17% reduction in mid day light under clear-water conditions [50], which suggests that shading from herbivore exclusion cages may exacerbate light limitation under higher sediment loading. The potential for reduced macroalgal growth through shading at low sediment levels would negatively offset increases in macroalgal cover from herbivore release, which implies that the reduced coral recruitment in herbivore exclusion treatments may be underestimated.

Our results indicate that rapid early successional trajectories under high sedimentation rates in the first three to six months (figure 1) may have resulted in hostile conditions for initial settlement of macroalgal propagules [51,52], with tiles rapidly becoming dominated by terrestrial sediments in early stages of succession under high sediment loading (figure 2). Turf algal assemblages are less sensitive to increased sediment than some other phototrophs [16,26], and interact with sediment loading to inhibit coral recruitment [16]. Non-coral invertebrates (dominated by sponges, bryozoans and ascidians) exhibited low cover at low-sediment levels, but increased nonlinearly in the absence of herbivory at high sediment levels, indicating either increased heterotrophic capacity under higher sediment loading [53] in the absence of grazing herbivores, and/or decreased competition from macroalgae under increasing light limitation [49]. Importantly, our results point to nonlinear trajectories and differing thresholds among benthic groups in response to increased sediment loading. Benthic groups that are more tolerant to these certain aspects of sedimentation than others can succeed in high sedimentation environments either through competitive dominance or simply by surviving when others cannot [54].

Surveys of herbivorous fish assemblages at our study sites indicate that herbivore biomass at the lowest sediment lagoonal sites was comparable to adjacent offshore reef slopes [50], and declined linearly with increasing sedimentation rates (electronic supplementary material, figure S3). In offshore reef slopes throughout Micronesia, high levels of herbivore biomass are linked to lower macroalgal cover and increased coral recruitment [55]. The interaction between herbivory and sedimentation across the environmental gradient precludes simplistic interpretations of reduced herbivory at inshore sites, and increases in turbidity, deposited sediments and altered resource availability will likely result in altered functional roles of herbivory across the environmental gradient. Previous research from the Great Barrier Reef indicates that while parrotfish (Scaridae) biomass decreases from offshore to inshore in response to increased sediment loading, rabbitfish (Siganidae) biomass shows the opposite trend [56], resulting in different functional responses of herbivore assemblages. While herbivory plays a key role in mediating recruitment inhibiting taxa under low-sediment levels, functional changes in herbivore assemblages may in part explain the nonlinearity and interactive effects between sediment and herbivory across the environmental gradient (figure 3a).

At the highest levels of sedimentation, crevices within herbivore exclusion cages were infilled by terrestrial sediments (figure 1), limiting space to which coral larvae and other organisms preferentially recruit [11]. While the presence of cages could theoretically act as a sediment ‘trap’ by altering flow dynamics, our previous study using an identical cage design found no difference in water motion inside and outside of cages [50], indicating that herbivore exclusion cages do not result in increasing sediment deposition through either decreased flow or by limiting resuspension of deposited sediments on tiles. In the presence of herbivory, tiles were dominated not by sediment, but sediment-laden turf algae, which instead suggests that fish may displace sediment through swimming behaviours, or actively remove sediment through feeding behaviour and subsequent off-reef transportation [57]. These results are consistent with a recent study of sediment-algal dynamics that found a 10-fold increase in sediment loading in caged tiles than open tiles [58]. Sediment smothering of turfs can result in the development of anoxic sediment, which may create a negative feedback that entirely inhibits benthic settlement [59]. The high cover of turf algae at sites with high sediment loading—apparently facilitated by fish activities—may provide an important trophic link to higher trophic levels [60], and implies that inshore coral reefs may support diverse food chains and require fish to do so.

Coral recruitment pulses are critical in driving recovery following disturbance [61]. Our results find a clear impact of sediment on coral recruitment after 12 months of deployment, which is consistent with a previous field survey indicating declines in density of coral juveniles with increasing sedimentation [62]. Sedimentation can reduce and inhibit coral recruitment by impacting the process at multiple stages; from fertilization [22] to post-settlement survival [24,25]. We identify a clear coral recruitment threshold where coral recruitment was only found in low sedimentation environments (less than 0.8 mg cm⁻² d⁻¹), and probability of recruitment was higher in open herbivory environments dominated by short turf algal assemblages and crustose coralline algae (EAM). As our study documented recruitment 12-months following tile deployment when recruits had passed post-settlement bottlenecks [11], we are unable to separate the effects of sediment loading on coral settlement rates and the role of benthic succession in driving post-settlement mortality. Although difficulties in taxonomic identification of small (less than 10 mm) coral recruits in highly diverse coral assemblages precludes high-resolution analysis of recruitment patterns, our results indicate differences in recruitment among broad taxonomic groupings. In particular, Acropora and Pocillopora appear more sensitive to sedimentation and were inhibited at low levels of sedimentation (less than 0.8 mg cm⁻² d⁻¹), whereas corals merged into the ‘other’ category exhibited a higher tolerance and threshold to sedimentation (greater than 0.8 mg cm⁻² d⁻¹). This may reflect environmental filtering of typically more sensitive taxa such as clear-water Acropora spp. in inshore reefs [63], or higher recruitment rates of other taxa (e.g. Merulinidae and Lobophylliidae). Further studies are needed to identify patterns of dispersal, settlement and recruitment of corals across inshore reefs to identify potential early life-history bottlenecks.

Coral reefs in high sedimentation environments are typically defined as experiencing sedimentation rates above 10 mg cm⁻² d⁻¹ [54,64]. While many studies over the past decades have deployed sediment traps to study the effects of sedimentation on coral reefs, sediment dynamics are complex and different approaches can provide substantially variable results [32,65], which hinders the interpretation and comparison among ecological studies. In particular, cylinder traps may over estimate sediment accumulation by trapping fine sediments above the benthos [65], resulting in a disconnect between measured sedimentation rates and processes occurring at the turf-sediment boundary such as coral recruitment [32]. By deploying TurfPods and cylinder traps across an inshore to offshore gradient at our study sites, our results indicate that while cylinder traps over estimate sedimentation rates, the relationship between cylinder trap sedimentation rates and TurfPod sedimentation rates was consistent (R² = 0.87). The highest rates of sedimentation at our inshore study sites measured by TurfPods during the short-term experiment (1.72 ± 0.44 mg cm⁻² d⁻¹, electronic supplementary material, figure S2) is slightly lower, yet within the range of sedimentation rates from TurfPods deployed on the inshore Great Barrier Reef (averaging 2.33 mg cm⁻² d⁻¹ [32]). Future studies that adopting the TurfPod methodology will provide further insight into ecological mechanisms and thresholds of sedimentation on coral reefs. Our results indicate that high sedimentation rates in Palau supports low diversity assemblages, low rates of coral recruitment, and marginal reef growth [62].

While widespread mortality of coral reefs through increasing sea surface temperatures represents a relatively recent phenomena over the past decades [3], the historical impacts of widespread catchment clearing and increased sediment loading in the Pacific stems at least as far back as the nineteenth century (e.g. [66]), and archaeological evidence points to exploitation of herbivores in Palau for over a millenia [67]. Our results suggest that local management of coral reefs may have measurable positive impacts on benthic assemblages, resulting in higher net rates of coral recruitment. Indeed, improved management of catchments in the island of Babeldaob between 2006 and this study in 2016 has resulted in significant reductions in sedimentation rates on the adjacent inshore reefs that have cascading effects on coral recovery following disturbance. Management of local stressors may not ameliorate the impacts of future global disturbances under a warming climate, but can positively influence recovery of trajectories of coral assemblages in the twenty-first century.

Data accessibility

Data are available from the Dryad Digital Repository: https://dx.doi.org/10.5061/dryad.f7m0cfxtg [68].

Authors' contributions

G.R., P.J.M. and A.W. collected the data; A.W. and G.R. analysed the data; A.W. wrote the first draft, all authors contributed to the final draft.

Competing interests

We declare we have no competing interests.

Funding

We received no funding for this study.