1Herbivory is particularly intense in marine environments, with a higher proportion of primary productivity removed than in terrestrial habitats. Experimental manipulation of large herbivores (fish, urchins) has clearly documented their grazing impacts on algal and seagrass beds. Grazing impacts of mesograzers (small invertebrates such as amphipods and isopods) are, however, less understood due to the practical difficulties in manipulating their abundance in field conditions.
2We developed a novel technique that successfully manipulated the abundance of herbivorous amphipods on macroalgae without the potential artefacts associated with exclusion cages or mesocosms. We then used this technique to test the effects of reduced amphipod grazing over extended periods on the structure of a temperate algal assemblage. We tested grazer effects on growth rates and epiphyte cover of the brown alga Sargassum linearifolium, and on developing assemblages on bare substrates.
3Large reductions in the abundance of herbivorous amphipods affected neither the growth rates of S. linearifolium, the cover of its epiphytes, nor the structure of algal assemblages. This result contrasts strongly to previous studies in mesocosms documenting strong impacts of mesograzers on community structure, and we discuss differences in the experimental approaches and biology of the systems that could give rise to the observed differences in grazer impacts.
4Synthesis. Marine macroalgae and seagrasses support very high densities of small herbivores whose ecological role in these habitats is poorly understood. We have provided the first, replicated experiment that directly manipulates their density in situ to quantify grazer impacts without caging artefacts. Our results indicate that strong impacts are not likely with the naturally occurring amphipod densities in the temperate algal bed studied. Further such experimental tests in field conditions are required to understand the properties of grazer and plant communities that can predict grazer impacts.
Herbivory is a fundamental ecological process that regulates the biomass and composition of primary producers, and thus the structure and functioning of natural systems. The intensity of herbivory, and its importance relative to other processes, varies widely among habitats, with herbivores removing primary production three times faster, on average, in aquatic than in terrestrial systems (Hay & Steinberg 1992; Cyr & Pace 1993; Shurin et al. 2006). The proportion of primary production removed may vary by an order of magnitude among habitats within aquatic or terrestrial systems, attributable to differences in plant traits (net primary production, nutritional quality; Cebrian 1999) and differences among the types and sizes of herbivores predominant in different habitats (Crawley 1989).
In terrestrial systems, there are well-documented effects of herbivores on plant communities by both large vertebrates, mostly grazing mammals, and small invertebrates, mostly insects (Crawley 1989). In contrast, the understanding of the impact of marine herbivory is almost entirely restricted to large herbivores. A long tradition of field experimentation has clearly established the important role of large grazers, particularly fish, urchins and gastropods, in structuring benthic communities (Hawkins & Hartnell 1983; Carpenter 1986; Fletcher 1987; Hughes 1994; Heck & Valentine 2006). The ecological role of small, mobile herbivores (amphipods, isopods, small gastropods and polychaete worms, collectively termed mesograzers; Hay et al. 1987) is less understood. Mesograzers are highly abundant in macroalgal and seagrass beds worldwide, with densities frequently in the tens of thousands per meter of habitat (Brawley 1992). They form a link to higher trophic levels (especially fish) (Taylor 1998a), and have been likened to insects on terrestrial plants due to their abundance, high diversity and small size relative to their host plants (Hay et al. 1987). Unlike for insects, however, detailed information is lacking on what most mesograzers actually consume in the field, and few studies have examined the grazing impacts of mesograzers on the abundance or structure of algal or seagrass communities.
The poor understanding of the ecological role of mesograzers in marine communities stems largely from the practical difficulties with experimentally manipulating the abundance of these small invertebrates in the field. To date, the strongest evidence for mesograzers altering community structure comes from caging experiments (Korpinen et al. 2007b), mesocosms (Duffy & Hay 2000), or from field experiments in which the densities of predatory fish are manipulated (Heck et al. 2006; Davenport & Anderson 2007). Such studies in seagrass systems routinely demonstrate the ability of small grazers to limit populations of algal epiphytes with associated benefits for seagrasses (Heck & Valentine 2006). In mesocosms simulating algal-dominated habitats, the effects of mesograzers on algal biomass during experimental periods have varied from strong (Duffy & Hay 2000; Bruno & O’Connor 2005) to minimal or undetected (O’Connor & Bruno 2007; Douglass et al. 2008). Given this variation in recorded grazing intensity, and potential artefacts (altered water flow, sedimentation, animal mobility; Miller & Gaylord 2007) associated with even well-designed caging or mesocosm studies, there remains a need for experimental manipulation of mesograzers in the field. The use of cages or mesh bags is problematic due to the small mesh size (approximately 1 mm) required to limit movement by mesograzers, and controls for these potential artefacts have been shown to differ from treatments lacking cages (Lotze et al. 2001).
Here, we present a novel technique to exclude mesograzers from marine vegetation without the artefacts associated with caging or experimental enclosures. We use an insecticide incorporated into a slow-release matrix (plaster) with the aim of continually deterring invertebrates from patches of benthic habitat in situ. This allows for natural variation in the biological and physical factors that may affect the grazing intensity of small invertebrate herbivores. Insecticides are routinely used in assessments of insect herbivory on terrestrial plants (Carson & Root 2000), but previous attempts to use this technique underwater (Carpenter 1986; Brawley & Fei 1987; Brostoff 1988) have been hampered by the rapid recolonization of epifauna after single doses of insecticide, and, in some cases, confounded experimental designs where treatments were physically segregated from each other to avoid grazer recolonization from untreated plots. Natural densities of mesograzers are often re-established within 24–48 h after defaunation (Poore 2005) thus requiring very frequent applications of insecticide to maintain areas free from grazing.
We provide, for the first time, direct experimental manipulations of mesograzers in the field with experimental designs able to rigourously test hypotheses about grazer impacts on algal community structure. Our specific aims were to: (i) demonstrate the effectiveness of plaster blocks impregnated with carbaryl in reducing mesograzer densities in the field; (ii) test the effects of reduced mesograzer densities on the growth, biomass and composition of macroalgae and their epiphytes in a temperate algal bed dominated by the brown alga Sargassum linearifolium; and (iii) test the effects of reduced mesograzer densities on the establishment of an algal assemblage on a bare substrate. This last aim tests the hypothesis that mesograzers may alter algal assemblages primarily by consumption of the early life-history stages of algae, in a manner similar to grazing gastropods on intertidal rocky shores (Hawkins & Hartnell 1983).
study sites and organisms
Manipulative experiments were conducted at Shark Bay, Port Jackson, New South Wales, Australia (33°51′9″ S, 151°16′0″ E). The study site consists of a sandstone platform at 0.5–3 m depth on which dense beds of foliose macroalgae occur. The experiments focus on algal beds dominated by the brown macroalga Sargassum linearifolium (R. Brown ex Turner) which is abundant at the site (Poore & Steinberg 1999). The co-occurring brown alga Dictyopteris acrostichoides (J. Agardh) Boergese was also used in exclusion trials. The most abundant mesograzers inhabiting Sargassum beds are the gammaridean amphipods in the families Ampithoidae and Hyalidae. The ampithoids in particular are well-known to consume macroalgae (Poore et al. 2008), and have been shown to affect algal community structure in other systems (Duffy & Hay 2000). Their ecology at Shark Bay is well-described with studies on their abundance, distribution, feeding preferences, survival and dispersal (Poore & Steinberg 1999; Poore et al. 2000; Poore 2005).
exclusion of mesograzers in field conditions
We tested the suitability of the insecticide carbaryl (1-naphthyl-N-methylcarbamate) to exclude mesograzers in field conditions over extended periods of time. Carbaryl is a carbamate pesticide used previously in marine environments to remove pests in aquaculture (Roth et al. 1993), and mesograzers before experimental treatments in mesocosms (Duffy & Hay 2000; Douglass et al. 2008). Carbaryl is available commercially as a liquid pesticide or in solid form known as ‘carbaryl technical powder’, with a low solubility in water (110 mg L−1). Carbaryl is degraded rapidly by hydrolysis, photolysis and bacterial action with a half life of 5 h in seawater and sunlight (Armbrust & Crosby 1991).
We developed an in situ slow-release method for carbaryl by incorporating the insecticide into a plaster matrix. Plaster blocks have been used previously to manipulate water quality in field conditions (e.g. pulses of heavy metals to hard substrate assemblages; Johnston & Keough 2002). We ran three trials to establish whether mesograzer densities would be reduced on macroalgae adjacent to plaster blocks containing carbaryl. Densities were contrasted to those on algae adjacent to plaster blocks containing no carbaryl. Plaster blocks containing carbaryl were made by dissolving 189 g of wettable carbaryl powder (80% carbaryl) in 1050 mL of fridge-chilled water which was then added to 1800 g of chilled plaster (methods as per Johnston & Keough 2002), resulting in 7.6% carbaryl by weight. The mixture was then combined with an electric kitchen beater until the consistency resembled a smooth paste. The paste was then poured into plastic moulds (100 mL) and wire hooks were placed in the centre of the plaster block to allow attachment of blocks to the substrate in the field. After initial drying periods of approximately 5 min the blocks were removed from the moulds and air dried for 4 days. Control blocks were made as above with no carbaryl added.
The first trial contrasted the densities of epifauna on S. linearifolium and D. acrostichoides 2 days after plaster blocks were placed adjacent (within 3 cm) to these algae in the field (individual thalli as replicates). This was repeated in a second trial that recorded densities 4 days after the placement of blocks. The plaster blocks had largely dissolved after 4 days (mean of 87 ± 1.3% of mass dissolved) so a final trial used a more slowly dissolving dental plaster to test the effectiveness of the exclusion technique over a 7-day period (S. linearifolium only). In each trial, plaster blocks were attached to the substrate using masonry nails, treatment and control blocks were interspersed, and all algae used were separated by at least 1 m. At the end of each trial, algal individuals were rapidly enclosed underwater in 1 L plastic containers. Samples were fixed in 5% formalin in seawater then shaken in freshwater, and rinsed through a 300 µm sieve to remove the associated epifauna (a technique that successfully collects 94–98% of amphipods; Poore & Steinberg 1999).Under a dissecting microscope, the numbers of amphipods, gastropods and isopods were recorded. In trials 1 and 2, the abundance of each taxonomic group (number of animals per gram of algae) was analysed with two-way analysis of variance (anova) with treatment (carbaryl vs. control) and algal species (S. linearifolium vs. D. acrostichoides) as factors. Densities in trial 3 were contrasted between treatments using a t-test.
While carbaryl is not known to affect algae (Shacklock & Croft 1981; Carpenter 1986), we grew algae in the presence of carbaryl and the absence of herbivores to ensure that our exclusion technique does not affect algal growth rates of a large foliose macroalga (S. linearifolium) or the early life-history stages of a mixed assemblage of macroalgae. Thirty small fragments of S. linearifolium with a single apical meristem (0.520 ± 0.039 g) were excised from individuals in the field, weighed, measured and then placed within 10 cm high cylindrical cages of 8 mm plastic mesh and returned to the field. Cages were attached to the sandstone substrate at Shark Bay using masonry nails and allocated randomly to one of three treatments as above (blocks with carbaryl, blocks without carbaryl and controls lacking a plaster block). After 7 days (by which the length of apices had increased by 24.5 ± 2.5% and the mass increased by 37.5 ± 4.2%) the fragments were collected, weighed and measured. Mesograzers were not observed to colonize these small fragments within the experimental enclosures.
Possible effects of carbaryl on early life-history stages of algae were assessed by allowing algal recruitment onto perspex settlement plates placed in the field for 3 weeks, and then culturing these early recruits in the laboratory under fluorescent light at 18 °C for 1 week. Five replicate plates were cultured in seawater, another five in seawater with low carbaryl concentrations (10 ppm or 0.01 g L−1) and another five plates were cultured in seawater with high carbaryl concentrations (100 ppm or 0.1 g L−1). Our low concentration was equal to Shacklock and Croft's (1981) high level, and the high concentration here was ten times higher than their trialled value. At the end of the experiment, the algae from each plate were removed, dried overnight in a drying oven at 60 °C and weighed.
effects of reduced amphipod grazing on sargassum linearifolium beds
The incorporation of carbaryl into plaster blocks was successful in greatly reducing the abundance of herbivorous amphipods inhabiting macroalgae (see Results). Consequently, we used this technique to test the effects of these grazers on: (i) the growth of S. linearifolium; (ii) the structure of algal assemblages in S. linearifolium beds; and (iii) the cover of epiphytes on S. linearifolium.
Sixty circular plots of 30 cm diameter were established in the S. linearifolium beds at Shark Bay in June 2006. Twenty plots were randomly allocated to each of three treatments: (i) the addition of a plaster block with carbaryl; (ii) the addition of a plaster block lacking carbaryl (to test for any effect of plaster alone); and (iii) an un-manipulated control. Plot labels and carbaryl blocks were attached to the substrate with masonry nails, with blocks replaced every 7 days (before complete dissolution). Carbaryl release from the blocks is expected to be effective in deterring amphipods from plots this size (all algae within 15 cm of the block) as it was effective in reducing amphipods from individuals of S. linearifolium that were, on average, 26 cm tall. Plots were spatially interspersed and at least 1 m apart. Each week for 10 weeks, the percentage cover of all foliose macroalgae and the height of five individuals of S. linearifolium per plot were recorded.
At the end of 10 weeks, algae within the plots were destructively sampled after the final estimate of percentage cover. The individual S. linearifolium closest to the plaster block in the centre of the plot was enclosed within a plastic bag for analyses of epifauna and epiphytes. Half of the plots were then scraped of all foliose algae for analyses of algal biomass per plot (n = 10 per treatment). The S. linearifolium individuals were fixed in 5% formalin and rinsed in freshwater to census epifauna as above. The epiphyte load on these individuals was measured by recording the species and biomass of large epiphytes readily removed from the host thallus and the percent cover of encrusting and small filamentous epiphytes on the stems and fronds. Epiphyte cover was visually estimated on all stems, and on 50 randomly selected fronds within each of three regions of the thallus (base, 0–10 cm from holdfast; mid, 11–20 cm from holdfast; and top, > 21 cm from holdfast). Three height regions were chosen due to the likely influence of tissue age on epiphyte cover (Jennings & Steinberg 1997). Sargassum linearifolium has apical meristems and thus basal tissue is oldest.
The height of S. linearifolium (mean of five individuals per plot) and the percent cover of five dominant algal species (those with > 1% cover, pooling treatments) were analysed with a repeated-measures anova with treatment and sampling week as factors. The biomass of these algal species per plot, biomass of foliose epiphytes, cover of epiphytes on the stem, and densities of amphipods on S. linearifolium at the end of the experiment were analysed with anova with treatment as the single factor. The cover of epiphytes on the fronds was analysed with nested anova with treatment as a fixed factor, and algal individual nested within treatment. All of the epiphyte analyses were run separately for each of the height regions on the S. linearifolium thallus (where present).
effects of reduced amphipod grazing on bare substrates
In addition to possible effects on existing foliose macroalgae, mesograzers could affect algal assemblages by the consumption of early life-history stages present during the establishment of an assemblage. To test this hypothesis, we used the exclusion technique to reduce grazer densities on bare substrates. Ninety sandstone blocks (the local rock type), 10 × 10 cm and 3 cm thick, were attached to plastic mesh, and then nailed to the substrate at Shark Bay in June 2005. Thirty blocks were randomly allocated to each of three treatments: (i) the addition of plaster blocks with carbaryl; (ii) the addition of plaster blocks lacking carbaryl; and (iii) no plaster block. The plaster blocks were replaced every 7 days.
After 3 months, the sandstone blocks were placed individually in bags and returned to the laboratory. Ten blocks from each treatment were rinsed extensively in freshwater to remove epifauna which was fixed in 5% formalin. All blocks were then preserved for analysis of algal cover. Percent cover was estimated by placing a clear plastic sheet with 81 points over the block. The cover of dominant algal taxa and the densities of epifauna were contrasted among treatments using one-way anova.
Anova and t-tests were performed using systat 10 (SPSS Inc. 2000). Data were assessed for normality and homogeneity of variance using frequency histograms of residuals, and plots of residuals vs. means, respectively. Logarithmic transformations were made where appropriate. The significance level was taken as P < 0.05.
exclusion of mesograzers in field conditions
The inclusion of carbaryl into a slow-release matrix was successful in greatly reducing the abundance of grazing amphipods inhabiting macroalgae in all trials (Fig. 1a). The reductions in mean density were 77% over 2 days in trial 1, 61% over 4 days in trial 2, and 94% over 7 days in trial 3, with all contrasts of abundance between algae adjacent to carbaryl blocks significantly different from the control blocks (Table S1 in Supporting Information). The treatment was equally effective on both S. linearifolium and D. acrostichoides with no significant interactions between treatment and algal species (Table S1).
Amphipods were the most abundant mesograzers inhabiting S. linearifolium during the trials (6.10 ± 0.7 individuals per gram, pooling trials). The other potential grazers – isopods (0.28 ± 0.07 individuals per gram) and gastropods (1.42 ± 0.20 individuals per gram) – were far less abundant and not consistently reduced in abundance by the carbaryl treatment (Fig. S1). The abundance of isopods was reduced by the carbaryl treatment on S. linearifolium but not D. acrostichoides in trial 1 (post hoc test after treatment × species interaction) but did not differ between treatments in trials 2 and 3 (Table S2). Gastropod abundance did not differ among treatments in trials 1 and 3 but were marginally more abundant on carbaryl treated algae in trial 2 (Table S2).
Carbaryl had no measurable effects on the growth of S. linearifolium with neither increases in length (F2,24 = 0.59, P = 0.56) nor biomass (F2,24 = 2.7, P = 0.09) differing among apices grown in the presence of carbaryl, the presence of control blocks, or the absence of any plaster block in the field (growth details in Table S2). Similarly, the growth of algae that had settled on perspex plates was unaffected by our carbaryl treatments in laboratory culture (algal mass, F2,12 = 0.76, P = 0.49).
effects of reduced amphipod grazing on sargassum linearifolium beds
At the end of the experiment in S. linearifolium beds, mean amphipod abundance on S. linearifolium adjacent to carbaryl blocks was significantly reduced (by 80%) (Fig. 1b, F2,35 = 13.96, P < 0.001, log-transformed). Considering only those amphipod families known to be herbivorous, the reduction in mean densities was 86% (treatment, F2,35 = 13.0.1, P < 0.001). The abundance of neither isopods nor gastropods was affected by the carbaryl treatment (Table S1).
The reduction in amphipod grazing did not affect either the growth, biomass or cover of S. linearifolium, or the cover or biomass of any co-occurring macroalgae (Fig. 2). Sargassum linearifolium increased in mean height (± SE, pooling treatments) from 14.9 ± 0.3 cm to 25.8 ± 0.9 cm (Fig. 2a), with a significant effect of sampling week (F8,424 = 70.86, P < 0.001) but neither treatment (F2,53 = 1.81, P = 0.18) nor the interaction between week and treatment (F16,424 = 0.56, P = 0.89) showed any significant effect in the repeated-measures of anova. Only four co-occurring species (Lobophora variegata, Corallina officinalis, Colpomenia peregrina and Dilophus marginatus) had a mean percentage cover > 1% (pooling treatments and weeks). The percent cover for all species except S. linearifolium varied among sampling weeks, with no effect of reducing amphipod abundance (Fig. 2, Table 1). At the end of the experiment, the same five species comprised 98% of algal biomass (pooling treatments and date). The reduction of amphipod abundance did not affect the final biomass (wet weight) of any of these species (F2,27 < 3.02, P > 0.07, all data log-transformed).
Table 1. Repeated measures anova contrasting the percentage cover of abundant macroalgae among treatments that manipulated amphipod density in the beds of Sargassum linearifolium for 10 weeks. Treatments were additions of a plaster block with carbaryl, blocks lacking carbaryl and controls with no plaster block. All data except those for S. linearifolium are log-transformed and P-values are those with the Greenhouse-Geisser adjustment (as recommended by Quinn & Keough 2002). The data for Colpomenia peregrina lack the first week in which none of this species was present (degrees of freedom for the within subject terms are thus 8, 16 and 456)
Treatment × Week
Encrusting coralline and green filamentous algae were the only epiphyte taxa with greater than 1% cover on the fronds of S. linearifolium (pooling treatments) with the reduced abundance of amphipods having no effect on percent cover of either (Fig. 3, Table 2). The stems were more highly epiphytised, with cover of encrusting corallines on the basal and mid regions of stems highest on the algae adjacent to carbaryl blocks (base, F2,55 = 3.61, P = 0.03; mid, F2,53 = 4.41, P = 0.02) but not differing significantly from the treatment including the plaster control block (post hoc tests, Fig. 3a). The cover of encrusting algae in the top regions of the thallus did not differ among treatments (F2,32 = 0.22, P = 0.80), nor did the cover of green filamentous algae in any region (Fig. 3b, base; F2,55 = 1.29, P = 0.28; mid, F2,53 = 0.19, P = 0.83, top, F2,32 = 0.30, P = 0.74, log-transformed).
Table 2. Analyses of variance contrasting the percent cover of encrusting coralline and green filamentous algae on the fronds of Sargassum linearifolium at the end of the experiment that manipulated amphipod density in S. linearifolium beds. Analyses were run separately for the base, mid-region and top of the thallus. All data were log-transformed
Encrusting coralline algae
Green filamentous algae
Large foliose epiphytes that could be removed and weighed were relatively scarce (on average 3.2 ± 0.7% of the host biomass) with only three species comprising 95% of all the biomass (C. peregrina, D. marginatus and Rhodymenia sp.). The reduction of amphipod abundance did not affect the biomass of any of these species at the end of the experiment (F2,55 < 0.48, P > 0.55, log-transformed).
effects of reduced amphipod grazing on bare substrates
Amphipod abundance on bare substrates adjacent to carbaryl blocks was reduced by 99% in contrast to control treatments (Fig. 1c, F2,26 = 19.8, P < 0.001, log-transformed). Isopods were also deterred by our treatments (Fig. S1, F2,26 = 6.5, P = 0.005, log-transformed), but were rare (4.6 ± 1.5 individuals per block) in comparison to amphipods (72.4 ± 16.3 per block) and gastropods (137.5 ± 23.2 per block). There were no treatment effects on gastropod abundance (Fig. S1, F2,26 = 1.5, P = 0.25, log-transformed).
After 3 months, the bare sandstone blocks had been extensively colonized by macroalgae (mean of 95 ± 0.9% cover). The reduction in amphipod abundance did not affect either the total cover of algae (F2,75 = 0.31, P = 0.73), or the percent cover of any of the most abundant taxa (Fig. 4, Ulva sp., F2,75 = 1.19, P = 0.31, reflected and log-transformed; red filamentous algae, F2,75 = 0.60, P = 0.55, log-transformed; Colpomenia, 95% confidence intervals overlap, skew in data not fixed by transformation).
impact of reduced amphipod grazing in sargassum beds
Our novel technique to use plaster as a slow release matrix for an insecticide proved a highly successful method of locally reducing abundant herbivorous amphipods (by 80–99%) from marine habitats, thus allowing tests of the impacts of their grazing in spatially replicated experiments in situ without the artefacts associated with caging. The difficulties in doing this are largely responsible for the poor understanding of mesograzer impacts on seagrass and algal habitats, limiting comparisons to larger herbivores and to ecologically similar herbivores on land (particularly insects).
The first such experiment, conducted in a temperate algal bed in New South Wales, Australia, failed to detect any grazing impacts. This contrasts strongly to the well-documented effects of macrograzers (urchins, fish and gastropods) on the same subtidal reefs (Fletcher 1987; Wright et al. 2005). Intense grazing pressure is required to affect the cover of large brown algae in this habitat, with more than twice the natural densities of a local macrograzer (the urchin Heliocidaris erythrogramma) needed before the cover of Sargassum spp. is affected (Hill et al. 2003). Impacts from mesograzers could arise, however, from relatively little consumption if valuable tissues (e.g. meristems) are favoured, or by the removal of early life-history stages (as epiphytes or on bare substrates). Our experiments found no such effects, neither of reduced growth in S. linearifolium nor of reduced cover of epiphytes or other macroalgae. This is in contrast with the well-documented effects of mesograzers on seagrass epiphytes (Heck & Valentine 2006) and of small invertebrate herbivores on periphyton in freshwater systems (Hillebrand 2008).
Our results are also in strong contrast with those from the few systems in which there have been attempts to quantify mesograzer effects on the performance or composition of macroalgae in the field. The recent experiments by Davenport & Anderson (2007), in which mesograzer densities were manipulated by altering fish predation, documented clear impacts of mesograzers on the kelp Macrocystis pyrifera. Mesograzers are also important determinants of assemblage structure in the beds of the brown alga Fucus vesiculosus in the Baltic Sea (experimental studies using cages reviewed by Korpinen et al. 2007a). In the hard-substrate habitats of coastal North Carolina, an extensive series of manipulations in experimental mesocosms has shown that mesograzers can profoundly alter benthic assemblage structure (Duffy 1990; Duffy & Hay 2000; Bruno & O’Connor 2005). Observed grazing intensity, however, is variable and these studies include those in which there was little top-down control of algae by grazers despite a 10-fold increase in grazer abundance during the experiment (Douglass et al. 2008), and those in which there were effects on assemblage structure but not algal biomass (O’Connor & Bruno 2007).
variation in mesograzer densities and composition
Variation in the observed impacts of mesograzers could relate to differences in the methods used to measure such impacts, as described above, with evidence from the few direct field manipulations quite mixed. However, variation in the biological components of the different systems may also be very important. In particular, mesograzer abundance, which is often highly variable in space and time (Taylor 1998b; Poore 2005), should be an important determinant of grazing rates and thus of impact on algal assemblages. A comparison of abundances among study systems (Table 3) provides no clear relationship between the likelihood of detected impacts and amphipod densities that vary by two orders of magnitude. We found no grazing effects in an algal bed where mean amphipod densities range from 2–12 amphipods per gram, while other studies detected impacts at lower densities, or at densities much higher than recorded here (20–219 individuals per gram) (Table 3).
Table 3. A comparison of the amphipod densities used in experimental studies testing for impacts of herbivorous amphipods. Studies chosen were those that measured effects on the performance of individual macroalgal species, or on the biomass and/or composition of algal assemblages that included abundance data expressed as numbers of individuals per gram wet mass of algae. Studies are sorted by increasing densities
When examining algal succession, we detected no differences between the substrates with almost complete amphipod exclusion and the un-manipulated substrates with mean amphipod densities of 102 ± 20.3 per 100 cm2 of substrate. These results are similar to those of Carpenter (1986), who found little effect of mesograzers on coral reef algae at much lower densities than ours (4.7–14.1 individuals of the amphipods Ampithoe spp. per 100 cm2). In contrast, Brawley & Adey (1981) described effects on algal assemblages in a mesocosm with densities of up to 5.2 Ampithoe ramondi per 100 cm2. In each of these experiments on early successional stages, the final recorded densities may be misleading if mesograzers only occupy the experimental units once there is sufficient cover of algae.
Much of the variation in mesograzer abundance is assumed to relate to variable predation pressure, although experimental evidence for this hypothesis is somewhat mixed (Duffy & Hay 2000) and there is at least some evidence that mesograzers can be food limited (Edgar & Aoki 1993). Differences in mesograzer impacts could thus arise from differences in the patterns of predation between our system, those in other regions, and those in experimental mesocosms, where it has been suggested that the limited mobility of predators may result in overestimates of the effects of predation (Douglass et al. 2008). The well-studied habitats in North Carolina feature strong seasonal variation in fish abundance, with amphipods becoming very abundant (approximately 130 individuals per gram of algae) during winter when fish are largely absent (Duffy & Hay 2000). In the temperate algal beds of New South Wales, carnivorous fish have been shown to limit amphipod abundance (Kennelly 1983), but predation pressure on amphipods in Sydney Harbour is probably low due to intense fishing activity in this highly urbanized region (and thus bias our results toward detecting an effect of mesograzers).
Predicting grazing rates depends not only on mesograzer abundance, but their composition (Duffy et al. 2001), body size and the likelihood of feeding on macroalgal tissue vs. the epiphytic flora associated with their hosts (Duffy 1990). Douglass et al. (2008) attributed the limited top-down control of algae by mesoherbivores in a mesocosm to the fact that the grazer community did not include amphipods from the family Ampithoidae – a group well-known to consume macroalgal tissue. A similar absence cannot explain our failure to detect any effects of reduced amphipod densities. The fauna at Shark Bay includes seven species of the Ampithoidae (Poore & Lowry 1997), and other families known to consume macroalgal tissue: Hyalidae, Aoridae and Biancolinidae (Poore et al. 2000; Taylor 2006). In combination, these species form a high proportion of the amphipod fauna that we manipulated (51% of individuals in the current experiment, and 64% and 89% of individuals in the surveys from Shark Bay presented in Poore & Lowry (1997) and Poore et al. (2000). The fauna also includes the amphipod Peramphithoe parmerong (at densities up to 2.2 individuals per gram, Poore & Steinberg 1999), a member of the genus for which most records of significant damage to large brown algae in the orders Fucales and Laminariales have come (Poore et al. 2008). We are also confident that we manipulated grazing rates on the algae as we observed no compensatory increases in any other group of mesograzers (isopods and gastropods) associated with our reductions in amphipod densities in our long-term experiments (O’Connor & Bruno 2007).
variation in algal palatability and growth rates
In addition to variation among mesograzers, differences in the identity and productivity of their food sources will contribute to the likelihood of grazer control of algal assemblages (Bruno et al. 2005). Sargassum linearifolium is not known to contain non-polar secondary compounds (Poore & Steinberg 1999), and is palatable to a wide range of locally occurring mesograzers (Taylor & Steinberg 2005). Other abundant macroalgae in our plots, however, are less palatable. Lobophora variegata and D. marginatus produce non-polar metabolites and are low-preference foods for both meso- and macrograzers (Poore & Steinberg 1999; Taylor & Steinberg 2005) and C. officinalis is heavily calcified and unlikely to be consumed by amphipods. Only C. peregrina is likely to be readily consumed by all amphipods as it is highly palatable to P. parmerong (Poore & Steinberg 1999). On the bare substrates, green alga from the genus Ulva dominated, and while often assumed to be highly palatable, locally occurring species are of low preference to several mesograzers in this habitat (Taylor & Steinberg 2005).
The epiphytes of S. linearifolium also included unpalatable taxa – encrusting coralline algae and D. marginatus– and we cannot assume that epiphytes are more palatable than their macroalgal hosts (Duffy 1990; Poore 1994). Grazer effects on unpalatable epiphytes may still arise by the consumption of early life-history stages in biofilms rather than by grazing of larger epiphytes once established. The increased cover of encrusting coralline algae on stems of S. linearifolium with reduced amphipod densities supports this hypothesis; however, these results were not repeated for the fronds where amphipods are more likely to reside. Our results therefore are in contrast with those from studies showing strong mesograzer control of epiphyte loads. Supporting this finding is the fact that, with exception of one species, the abundances of herbivorous amphipods on S. linearifolium are uncorrelated with epiphyte cover (Poore et al. 2000).
With no evidence of grazer control of algal biomass, we are left with the conclusion that any tissue loss as a result of amphipod grazing is simply insufficient to greatly impact upon algae, given their growth rates during the experiment. Tissue loss as a fraction of primary production will relate to mesograzer traits (as discussed above) and rates of primary production, and declines in algal fitness will depend on the degree to which algae can compensate for tissue loss (Honkanen & Jormalainen 2002). Our experiments were conducted at a time of active growth of S. linearifolium with height increasing by approximately 40% during the experiment. Effects of grazing may be evident only during periods of low productivity (Nicotri 1977) and further experiments are required to assess whether grazing rates are affected by nutrient supply to the macroalgae (as is common in algal and seagrass beds worldwide; Hughes et al. 2004). Impacts of grazing may also be evident with experiments conducted over a longer duration, although previous mesocosm experiments have detected effects on algal assemblages within 1 (Duffy 1990), 2 (Ruesink 2000), 3 (Bruno & O’Connor 2005; O’Connor & Bruno 2007) or 4 weeks (Duffy & Hay 2000) – all periods much shorter than our 10-week experiment.
With an experimental technique to manipulate the abundance of small, mobile herbivores in marine habitats without the artefacts of caging, the opportunity now exists to conduct experiments in situ to establish the role of these herbivores in vegetated habitats worldwide. This will allow for quantitative comparisons with the well-studied larger grazers in marine habitats, and with small, invertebrate grazers in terrestrial and freshwater systems. The first use of this technique in a temperate algal bed in Australia detected no grazing impacts on macroalgae or their epiphytes, contrasting with previous mesocosm and caging studies in other regions. Further experiments are required to assess the degree to which the observed variation in mesograzer impacts depends on the differences in grazer abundance and composition, and on differences in the composition and productivity of algal assemblages.
Authors would like to thank René Reinfrank and Candida Barclay for technical assistance; Nicole Hill, David Roberts, James Smith, Nicolle Spyrou and Allie Syriatowicz for support in the field; Richard Taylor for discussions that were influential in the initial design of this project, and Emmett Duffy and an anonymous referee for comments that improved this manuscript. The research was supported by Australian Research Council Discovery Project DP0556372 to AGBP.