Oecologia DOI 10.1007/s00442-016-3643-0
COMMUNITY ECOLOGY – ORIGINAL RESEARCH
Microtopographic refuges shape consumer‑producer dynamics by mediating consumer functional diversity Simon J. Brandl1,2,3 · David R. Bellwood1,2
Received: 6 October 2015 / Accepted: 20 April 2016 © Springer-Verlag Berlin Heidelberg 2016
Abstract Consumer-producer dynamics are critical for ecosystem functioning. In marine environments, primary production is often subject to strong consumer control, and on coral reefs, the grazing pressure exerted by herbivorous fishes has been identified as a major determinant of benthic community structure. Using experimental surfaces, we demonstrate that on coral reefs, microtopographic refuges decrease the overall grazing pressure by more than one order of magnitude. Furthermore, by functionally characterizing consumer communities, we show that refuges also restrict grazer communities to only one functional group, algal croppers, which selectively remove the apical parts of algae. In contrast, detritivorous fishes, which intensively graze flat and exposed microhabitats and can remove both particulate matter and entire stands of algal filaments, are almost entirely excluded. This preclusion of an entire ecosystem process (the removal of particulates) results in two distinct coexisting benthic regimes: communities within refuges are diverse and characterized by numerous algal
Communicated by Deron E. Burkepile. Electronic supplementary material The online version of this article (doi:10.1007/s00442-016-3643-0) contains supplementary material, which is available to authorized users. * Simon J. Brandl
[email protected] 1
ARC Centre of Excellence for Coral Reef Studies, James Cook University, Townsville, QLD 4811, Australia
2
College of Marine and Environmental Sciences, James Cook University, Townsville, QLD 4811, Australia
3
Tennenbaum Marine Observatories Network, Smithsonian Environmental Research Center, Edgewater, MD 21037, USA
types and juvenile scleractinian corals, while communities outside refuges support only low-diversity assemblages dominated by simple, unbranched filamentous turf algal mats. Although limited to the scale of a few centimeters, microtopographic refuges can, therefore, mediate the biotic control of community development by affecting both overall grazing rates and the functional diversity of consumer communities. We suggest that the coexistence of two distinct benthic regimes at a small spatial scale may be an important factor for ecosystem functioning and highlight the need to consider the ecological complexity of consumer-producer dynamics when assessing the status of coral reef ecosystems. Keywords Turf algae · Reef resilience · Herbivory · Phase shift · Rugosity
Introduction Herbivory is universally accepted as a critical ecosystem process in vegetated environments (Gruner et al. 2008). Through the disturbance induced by feeding, herbivorous organisms affect producer communities in terms of their biomass, diversity, and productivity (Olff and Ritchie 1998; Worm and Duffy 2003). However, herbivory is multifaceted, and numerous different aspects can influence the dynamics between consumer and producer communities. One crucial factor relates to the functional identity of herbivores. In almost every ecosystem, herbivore communities comprise multiple species that vary in their functional niches, which modulates their effects on producer communities (Sommer 1999). Similarly, producer avoidance of, or tolerance to, herbivory affects dynamics between consumer and producer communities (Duffy and Hay 1990; Rasher
13
Oecologia
et al. 2013). While intrinsic mechanisms to avoid being grazed, such as morphological or chemical defense mechanisms, have received considerable attention (Berenbaum 1995), extrinsic factors that permit producers to escape herbivore pressure are less well understood (Milchunas and Noy-Meir 2002). Extrinsic factors underlying grazer avoidance include both biological and physical protection from grazing, i.e., refuges (Duffy and Hay 1990). Biological refuges comprise positive interactions among plant species, in which species susceptible to grazing gain associational refuge by growing close to an unpalatable or non-preferred species (Pfister and Hay 1988; Stachowicz 2001). Physical refuges usually relate to physical or geological features, which reduce the accessibility of producers to herbivorous grazers (Milchunas and Noy-Meir 2002). Compared to biological refuges, such physical refuges have received relatively little consideration, despite their demonstrated importance in both terrestrial (Shitzer et al. 2008) and aquatic (Hay 1981; Bergey 2005) environments. Importantly, physical refuges can vary dramatically in scale, ranging from entire islands (Milchunas and Noy-Meir 2002) to microtopographic structures on the scale of a few millimeters (Menge and Lubchenco 1981; Dudley and D’Antonio 1991; Bergey 2005). Although often inconspicuous, the latter can significantly influence consumer-producer dynamics, especially when grazing pressure is intense (Menge and Lubchenco 1981; Milchunas and Noy-Meir 2002; Bergey 2005). Primary producers in marine environments are subject to particularly rigorous consumer control (Gruner et al. 2008; Bennett et al. 2015), and among marine habitats, benthic communities experience the highest levels of grazing by herbivores (Poore et al. 2012). On tropical coral-dominated reefs, grazing pressure is particularly intense, and reductions of grazing pressure often lead to the rapid establishment of fleshy algae (Littler and Littler 1984; Smith et al. 2001; Burkepile and Hay 2008; Cheal et al. 2010; Rasher et al. 2013). As a consequence, tight links between strong consumer control by herbivorous fishes and the resilience of coral reefs have been established (Nyström et al. 2008; Graham et al. 2013); however, despite the apparent acceptance of this relationship as an ecological truism, recent work also suggests that the relationship between nominally herbivorous fishes and benthic algae is not a simple, linear relationship, but influenced by many different facets of consumer and producer ecology, as well as a plethora of environmental factors (Burkepile 2012; Burkepile and Hay 2010; Burkepile et al. 2013; Brandl et al. 2014; Brandl et al. 2015; Russ et al. 2015). Thus, more detailed work on the relationship between herbivorous grazers and the benthic community is required in order to develop a deeper understanding of consumer-producer dynamics on coral reefs (Adam et al. 2015).
13
One aspect of demonstrable importance for consumerproducer relationships is the presence and effect of grazing refuges in coral reef ecosystems. Several cases of biological refuges, in which readily consumed algae associate with species unpalatable to most herbivores, have been reported from coral reefs (Littler et al. 1986; Pfister and Hay 1988). Likewise, physical refuges exist at several different scales, ranging from regional refuges [inner shelf vs. outer shelf on the Great Barrier Reef (GBR) (Wismer et al. 2009)], to reef zones [inner flat or sandy plains vs. crest (Hay 1981; Fox and Bellwood 2007)]. However, microtopographic refuges from grazing pressure (i.e., the three-dimensional structure of the reef on the scale of a few centimeters) also represent a widespread type of physical refuge within coral reef systems, occupying up to 25 % of the available microhabitats on the reef crest and flat at two sites on Lizard Island (Brandl et al. 2015). While crevices appear to be critical for the recruitment and survival of scleractinian corals (Brock 1979; Nozawa 2008; Arnold et al. 2010; Brandl et al. 2014; Edmunds et al. 2014), and represent an important ecological axis for niche partitioning in herbivorous fishes (Robertson and Gaines 1986; Fox and Bellwood 2013; Brandl and Bellwood 2014; Brandl et al. 2015), no examination of the overarching effect of microtopographic refuges on consumer-producer dynamics exists to date. The purpose of the present study, therefore, was to provide an experimental evaluation of the effects of microtopographic refuges on grazing dynamics on coral reefs. In quantifying grazing pressure, average turf lengths, the functional diversity of grazer communities, and the community composition of the benthos in different microhabitats, we sought to provide answers to the following questions: 1. Do microtopographic refuges reduce the grazing pressure exerted by herbivorous fishes on the benthic community? 2. Is this exclusion evenly spread among fish species and functional groups? 3. How do changes in grazing pressure and grazer identity affect the benthic community? By answering these questions, we demonstrate that microtopographic complexity leads to two distinct grazing regimes on coral reefs, which coexist at the scale of a few centimeters and may represent a key feature of consumerproducer dynamics in reef ecosystems.
Materials and methods Data collection To assess the effects of microtopographic refuges on fish grazing patterns and the benthic community, we
Oecologia
constructed two distinct grazing surfaces made from powdered coral rubble and cement. The first, flat surface consisted of a flat rectangle (length 115 mm, width 65 mm, height 15 mm). The second, complex surface, featured the same dimensions but included six evenly spaced vertical cylinders (radius 10 mm, height 35 mm; spaces between cylinders 25 mm) in two rows of three, to simulate microtopographic refuges as found on coral reefs. Specifically, this microhabitat was created to mirror the structural framework of dead, corymbose branching corals (e.g., genera Acropora, Pocillopora or Stylophora), which frequently form a mosaic of upright branches (mirrored by the cylinders) and interstitial spaces among these branches. Such structures are widely distributed on coral reefs (Kramer et al. 2014; Brandl et al. 2015). This design differs from previous work that examines concealed microhabitats on coral reefs (Nozawa 2008; Roff et al. 2015), in that the interstitial spaces can be flushed by wave action, therefore limiting the accumulation of particulates and the associated effects on benthic algal dynamics (Clausing et al. 2014). However, given the complexity of coral reef habitats, a range of other configurations and dimensions of microtopographic refuges seem also feasible and should be considered in the future. On the experimental surfaces, three distinct microhabitats were identified (Fig. 1): flat and exposed (flat tiles without cylinders, henceforth “flat”), complex and exposed (the outside and top of cylinders, henceforth “exposed”), and complex and concealed (the inside of cylinders and the flat area between, henceforth “concealed”). To ensure uniformity among surfaces, we used standardized moulds made from neutral cure sealant, dry cornstarch, and mineral spirits (2:2:1 by volume). Flat surfaces had only the recess (10 mm) for the flat rectangular base. For complex surface moulds, holes (depth 35 mm, radius 10 mm) were cut out of the mould using a drill press with a hole-saw fitting. To cast surfaces, a mixture of powdered coral rubble (dried for 72 h, and pulverized using a sledge hammer), river sand, and cement (1:1:1) was mixed with water and poured into the moulds. This mixture was selected following a pilot study on Lizard Island in 2012 that revealed no differences in the benthic community between natural tiles of dead coral and the rubble-cement mixture (cf. Hixon and Brostoff 1985). After setting the concrete mixture for 48 h, surfaces were extracted from the moulds and each surface was set centrally in a 750-ml disposable plastic container filled with a standard concrete mixture. After setting for 48 h, the containers were tied together for stability, producing grazing arrays of 20 surfaces (ten of each type). Six arrays were created yielding a total of 120 surfaces with 180 grazing microhabitats (flat n = 60, exposed n = 60, concealed n = 60). Arrays were deployed in groups of three at two different sites along the reef crest between Bird Islets and
South Island (near Lizard Island, northern GBR) in the beginning of June 2013. The arrays were left on the reef for 35 weeks to establish natural benthic communities. Subsequently, the lengths of epilithic turf filaments from each microhabitat were measured using vernier callipers. To do so, the turf filament closest to a haphazardly chosen point was straightened using tweezers (if visibly not erect) and its vertical extension measured to the nearest millimeter. Ten measurements were taken from each microhabitat (total n = 1800). No measurements were taken within 5 mm of the edge of the rectangular base to avoid edge effects. Photographs were taken of each microhabitat from a fixed distance of approximately 10 cm to quantify the benthic community composition. Finally, the foraging activity of fish assemblages grazing on the microhabitats was monitored using remote underwater videos (GoPro Hero III) in January/February 2014. Each array was filmed for 3–4 h on at least 5 nonconsecutive days, resulting in a total observation period of 111 h and an average of 18.52 ± 0.07 h per microhabitat. Subsequently, both benthic photographs and grazing videos were analyzed in the lab. On each photograph, we quantified the proportional cover of different benthic organisms from a planar view of the respective microhabitats (all of the flat microhabitat, upward-facing circular surfaces for exposed microhabitats, horizontal surface between bases of cylinders for concealed microhabitats). Benthic organisms were assigned to a functional category (Steneck & Dethier 1994), which could be accurately determined from a planar photograph and their proportional cover estimated using the software ImageJ. Categories included different types of algae (brown fleshy, brown foliose, crustose coralline, green filamentous, green fleshy, green calcified, simple unbranched filamentous turfs, red filamentous, red foliose, red corticated), cyanobacteria, sessile invertebrates (sponges, ascidians), scleractinian corals, and fecal pellets. Videos were analyzed by counting the number of bites taken on each microhabitat by all large mobile herbivorous fish species larger than 5 cm (sensu Choat et al. 2002). This threshold was implemented to exclude juvenile individuals, which may artificially inflate the grazing impact on a given surface as a result of numerous, rapid, small-sized bites. Care was taken not to deploy arrays in damselfish territories and only few bites were taken by small, sedentary species throughout the study. Although we cannot exclude the potential contribution of nocturnal grazers or micrograzers, previous work on the GBR suggests that the effect of these groups is limited on the GBR compared to herbivorous fishes (Hoey and Bellwood 2011). Various categorical schemes have been used to classify herbivorous fishes into functional groups (e.g., Duffy and Hay 1990; Bellwood et al. 2004; Green and Bellwood 2009; Cheal et al. 2010; Hoey et al. 2013). However, these
13
Oecologia Fig. 1 Description of the experimental setup. Two surfaces (panel i) were created, resulting in three distinct microhabitats (panel ii). Flat microhabitats (a) are flat and exposed, exposed microhabitats (b) are complex and exposed on the outside or top of cylinders, and concealed microhabitats (c) are on the inward-facing surfaces of cylinders and in the flat area between cylinders. For one array (panel iii), 20 surfaces (comprising 30 microhabitats) were strung together using iron chain. Three arrays were deployed at each of the two sites (panel iv), resulting in a total of 60 replicates per microhabitat type. As the area surrounding the created surfaces (d) was made from a different material (lacking the powdered coral rubble), it was not included in the study
schemes differ in their functional group assignments of many species, making the use of a single classification scheme inherently arbitrary. Therefore, in order to consider a broad suite of different traits influencing a species functional role, and to improve our understanding of functional groupings, a wide range of published data were used to characterize fish species in functional groups. These data include morphology, diet, and gut short-chain fatty acid (SCFA) profiles (Table S1). For the morphology, five traits previously linked to foraging microhabitat utilization (body depth, eye diameter, snout length, snout angle, and head angle), were considered (Brandl and Bellwood 2013).
13
Trait measurements were regressed against standard length in a dataset comprising 260 individuals in 99 species and residual values were averaged for each species. For the dietary data, due to differences among sources in the functional categorizations of dietary items, a broad classification scheme was created to ensure congruent dietary classifications across taxa. Categories included sediment, organic material, filamentous algae, fleshy/thallous algae, and other items such as benthic invertebrates or foraminifera. For the SCFA profiles, the overall amount of dominant SCFAs (acetate, butyrate, isovalerate, and proprionate) present in the gut segment with the highest SCFA concentrations was
Oecologia
used, as well as the proportional composition of SCFAs. SCFAs provide detailed insights into both the ingestion and utilization of dietary components (Clements and Choat 1995). Statistical analysis To assess differences in the overall grazing pressure (number of bites/day per square centimeter) and the length of turf algal filaments (millimeters) on the three microhabitat types, Bayesian mixed models (BMMs) were performed specifying the microhabitat type (flat, exposed, concealed) as a fixed effect. To account for the non-independence of microhabitats within an array, array was specified to have a random effect. In addition, site was included as a fixed effect as the estimation of random variance can be unreliable if only two levels are available. For the comparison of grazing pressure among microhabitat types (overall grazing pressure model), a Poisson error distribution with a log-link function was specified, appropriate for the count data obtained for this part of the analysis. Furthermore, bite counts for each microhabitat were modeled against an offset specifying the overall observation period (time in days) and the area available for grazing (area in square centimeters) to account for differences in observation time among arrays and differences in area available for grazing among microhabitats. For the comparison of algal filament length (turf length model), a Gaussian error distribution with a link function was used, as data were approximately normally distributed. To investigate the consumer community feeding on different microhabitat types, as well as the benthic communities, two non-metric multidimensional scaling ordinations (nMDS) were performed on Manhattan distance matrices. Subsequently, permutational multivariate ANOVAs (PERMANOVA) with 999 permutations were performed to investigate compositional differences among microhabitat types, again fitting array as a random effect. We also calculated fish species richness (i.e., the diversity of consumers interacting with the benthic community), and functional group richness for the benthic communities, respectively, and compared these patterns among microhabitat types using BMMs with a Poisson error distribution and uninformative priors. Furthermore, we evaluated the functional composition of consumers feeding on the different microhabitat types. We combined the three trait datasets (morphology, diet, and SCFAs) and normalized data to a mean of 0 and a SD of 1. We then calculated pairwise distances among species using Gower’s distance metric and divided the community into functional groups using a hierarchical clustering analysis with Ward’s method. This resulted in an optimal clustering size of two broad functional groups.
In order to compare the grazing pressure exerted on the different microhabitat types by each of the two functional groups, we used two zero-inflated Poisson BMMs (one for each group). We modeled the number of bites against two offset variables specifying the observation period and grazing area (as for the overall grazing pressure model above). Microhabitat type was formulated as a fixed effect and specified to interact with both the Poisson part of the response (counts of fish bites) and the binomial part denoting the probability of an observation being zero. Due to the presence of near complete separation in the data, we used weakly informative Cauchy distributed priors on the fixed effects (Hadfield 2010). More detail on model specifications (including specified priors and chain lengths) is provided in the Electronic Supplementary Material (ESM). All analyses were performed in the software R, using the packages vegan (Oksanen et al. 2013), FD (Laliberté and Legendre 2010), and MCMCglmm (Hadfield 2010).
Results Overall grazing pressure and turf filament length We found marked differences in grazing pressure among the three investigated microhabitat types. The overall grazing pressure was highest on flat microhabitats [2.86 ± 0.15 (mean ± SE) bites day−1 cm−2], followed by exposed microhabitats (0.81 ± 0.05 bites day−1 cm−2), while the lowest grazing pressure occurred on concealed microhabitats (0.17 ± 0.02 bites day−1 cm−2) (Fig. 2a). There was no difference among sites, and random variance was small (Table S2). Although not assessed quantitatively, we detected no intra-specific difference in the size of individuals feeding on the respective microhabitats. Therefore, we considered the bites being taken on the different surfaces as being of approximately similar volume. There were also strong, but opposite, differences in the average turf filament length among the three microhabitat types. Flat microhabitats had the shortest turf filaments (2.48 ± 0.11 mm), closely followed by exposed microhabitats (2.78 ± 0.07 mm), while concealed microhabitats had by far the longest turf filaments (6.90 ± 0.17 mm) (Fig. 2b). As for grazing pressure, site had no effect and random variance from the different arrays was small (Table S3). We also found a distinct difference in the mean number of species feeding on the three microhabitat types (species richness). While flat and exposed microhabitats were statistically indistinguishable [flat, 3.85 ± 0.16 (mean ± SE) number of species feeding on a given microhabitat over the entire observation period; exposed, 3.18 ± 0.17 species], concealed microhabitats had substantially less grazer
13
Oecologia Fig. 2 Predicted values from the grazing pressure model (a) and the turf length model (b), comparing the three microhabitat types (n = 180). Flat microhabitats are subject to the highest grazing pressure, followed by exposed microhabitats. Concealed microhabitats experience the lowest grazing pressure. In contrast, turfs are longest in concealed microhabitats, while exposed and flat microhabitats both have shorter turfs. Plots show the predicted mean ± 95 % credible intervals (CIs). Statistical significance can be assumed where CIs do not overlap. Values in parentheses indicate the mean ± 95 % confidence intervals of the observed values
(a)
Mean grazing pressure ± 95% CIs (bites.d-1.cm -2) 1 2 3
0
4
(2.86 ± 0.30) Flat (0.81 ± 0.11) Exposed (0.17 ± 0.04) Concealed
(b)
0
Mean turf algal filament length ± 95% CIs (mm) 2 4 6
8
(2.48 ± 0.22) Flat (2.78 ± 0.14) Exposed (6.90 ± 0.34) Concealed
species richness (1.93 ± 0.14 species) (Table S4). Interestingly, patterns of mean functional group richness of the benthic communities showed the opposite trend, with flat microhabitats supporting a markedly lower number of functional groups [3.15 ± 0.15 (mean ± SE) groups per microhabitat], compared to both exposed and concealed microhabitats (exposed, 4.333 ± 0.14 functional groups; concealed, 4.75 ± 0.15 functional groups) (Table S5). Again, site had no effect on the observed patterns and random variance was small (Tables S4, S5). Therefore, site data were pooled for subsequent analyses. Community compositions There were distinct differences among the examined microhabitats, in terms of grazer communities and the benthos, separating concealed microhabitats from exposed and complex microhabitats, which were consistently similar (Fig. 3). The nMDS ordination for the grazer
13
communities showed a clear separation between concealed microhabitats and the two other microhabitat types, but not between flat and exposed microhabitats (Fig. 3a). The separation was driven largely by rabbitfishes (Siganus corallinus, Siganus punctatus, Siganus punctatissimus), and the two surgeonfish species Zebrasoma scopas and Acanthurus nigrofuscus, fishes that characteristically grazed in concealed microhabitats. In contrast, flat and exposed microhabitats were characterized by scraping parrotfishes (Scarus frenatus, Scarus globiceps, Scarus niger, Scarus psittacus, Scarus schlegeli, Chlorurus spilurus) and the surgeonfish species Acanthurus nigricauda and Ctenochaetus striatus. This separation of microhabitat types was statistically significant in the PERMANOVA (pseudo F2,174 = 37.12, P