Habitat-former effects on prey behaviour increase predation and non-predation mortality



  1. Habitat-forming species can influence mortality on associated species via altering structural and non-structural abiotic conditions. Importantly, these effects can occur simultaneously and in opposite directions, although how they contribute to the net outcomes for predator–prey interactions remain unexplored.

  2. Seagrasses often have positive effects on associated fauna because their structure directly reduces predator encounter rates. However, we identified a ‘risky’ behaviour (shallower burial) in an infaunal bivalve at a high seagrass cover – likely induced by non-structural abiotic change – suggesting positive effects may be outweighed by risky behaviours. We determined whether the physical structure of the seagrass interacted with burial behaviour of clams to determine the predation and non-predation mortality and whether these interactions were mediated by the cover of the seagrass.

  3. Surveys on an intertidal sand flat in Tasmania, Australia showed that the highest densities of a dominant bivalve, Katelysia scalarina, occurred at low (33%) seagrass cover, but the lowest densities and the highest proportion of unburied clams occurred at high (100%) cover. A field experiment manipulating burial depth, seagrass cover and predator access demonstrated that unburied clams suffered very high predation and non-predation mortality compared to buried clams (~4x higher), which outweighed any positive effects of the seagrass structure in reducing predator access. Being unburied also had non-lethal consequences with surviving unburied clams having a reduced tissue biomass compared to buried clams.

  4. In this system, predation was driven by the availability of prey when they undertake a risky behaviour (shallow burial). However, significant changes in behaviour may only occur once a threshold of habitat-former density is reached. In this instance, changes in behaviour were likely due to seagrass effects on sediment redox potential, which decreased significantly above 33% seagrass cover.

  5. Our findings demonstrate that the negative effects of a habitat-former on the behaviour of associated species, via alteration of non-structural abiotic conditions, can outweigh any positive effects provided by increasing habitat structure as is commonly reported for habitat-formers.


Habitat-forming species have strong effects on community structure and typically support diverse food webs. The extensive facilitation of communities by habitat-forming species results from the creation of complex structure, and the substantial alteration of physical, biological and chemical resources available to associated species (Jones, Lawton & Shachak 1994, 1997; Bruno & Bertness 2001; Bruno, Stachowicz & Bertness 2003; Bulleri 2009). These changes can affect species behaviour and survivorship directly but can also have profound effects on context–dependent interactions such as competition and predation (Bruno & Bertness 2001; Pearson 2009; Byers, Wright & Gribben 2010). As noted by Crooks (2002), habitat-forming species will simultaneously alter structural (physical) and non-structural abiotic conditions. Typically, these two components of habitat modification are investigated separately, and we know little about how they interact to determine mortality risk.

Habitat-forming species change mortality risk for associated species in different ways. Often the physical structure created by the habitat-forming species provides a refuge for prey by decreasing predator–prey encounter rates (Norkko 1998; Anderson 2001; Byers 2002; Grabowski 2004). Importantly, the refuge value for prey can increase with increasing habitat-former density (Grabowski 2004; Griffen & Byers 2006). By contrast, when an intermediate predator also uses the habitat and benefits from the refuge or the structure that it provides, indirect negative effects of the habitat-forming species on basal prey abundance may occur because of increased foraging of the predator (Leonard 2000; Trussell, Ewanchuk & Matassa 2006; Miyashita & Takada 2007; Pearson 2009). In addition, the physical structure created by habitat-forming species can ameliorate non-structural abiotic conditions that are stressful for associated species (Bertness et al. 1999; Altieri, Silliman & Bertness 2007). However, in many cases, the abiotic changes caused by the habitat-former have negative effects on associated species (Posey, Wigand & Stevenson 1993; Neira et al. 2006; Gribben et al. 2009a,b), and the structural and non-structural abiotic changes that influence mortality can change with increasing habitat-former density (Gribben et al. 2013).

Modification of non-structural abiotic conditions by the habitat-forming species can change species behaviour in ways that also increase predation risk. Such behaviour may occur when the abiotic conditions are modified to such an extent that the fitness cost of not responding behaviourally outweighs any additional predation risk associated with behavioural change. For example, phytotoxins produced by the invasive wetland shrub Lonicera maackii resulted in tadpoles making more trips to the water surface increasing their exposure to predators (Watling et al. 2011), while foraging by rodents increased in an invasive shrub, possibly due to reduced moon illumination and temperature within the habitat (Mattos & Orrock 2010). Similarly, shallow burial by clams in an invasive seaweed was caused by reduced concentrations of dissolved oxygen in the water (Wright et al. 2010). Although shallow burial is typically a risky behaviour in clams (Whitlow, Rice & Sweeney 2003; Meyer & Byers 2005), the net effects of this behaviour may be only partly due to predation (Byers, Wright & Gribben 2010), and non-predation mortality may also occur due to the extra costs of maintaining shells closed without the support of surrounding sediment. Parsing the predation and non-predation mortality consequences of behavioural changes will advance our understanding of the mechanisms by which habitat-formers control species interactions.

In this study, we investigated how the behaviour of a common infaunal bivalve, Katelysia scalarina (hereafter Katelysia), was affected by different densities of the habitat-forming seagrass, Zostera muelleri (hereafter Zostera), and how changes to Katelysia behaviour interacted with Zostera cover to influence both predation and non-predation mortality. Like other habitat-formers, seagrasses commonly facilitate associated fauna by directly reducing predation pressure, which often decreases with increasing seagrass cover (Orth, Heck Jr & van Montfrans 1984; Irlandi & Peterson 1991; Bologna & Heck 1999; Goshima & Peterson 2012). However, increased rates of predation on basal prey in seagrass can occur (Skilleter 1994). Seagrasses can also change the burial behaviour of prey species by altering abiotic conditions in ways that may make them more susceptible to predation (Skilleter 1994; Wright et al. 2010). In Zostera, Katelysia often occurs unburied on the sediment surface potentially exposing it to a range of invertebrate (e.g. crabs and whelks) and vertebrate (e.g. fish and birds) predators. Specifically, we tested the hypothesis that increasing seagrass cover increases risky behaviour resulting in higher rates of mortality. First, we described patterns of clam abundance and behaviour (burial depth) among different levels of seagrass cover. Secondly, we experimentally determined the effects of clam behaviour, seagrass cover and predators on Katelysia survivorship. Last, we investigated the potential abiotic mechanisms determining changes in clam behaviour.

Materials and methods

Study species and system

Zostera is a common seagrass species in sheltered and moderately exposed sand and silt in estuaries around Tasmania, Australia (Edgar 2000) (Note that Z. muelleri is thought to be synonymous with Z. capricorni; Les et al. 2002). Zostera has fibrous creeping roots from which the shoots extend vertically. On intertidal sand flats, the shoots are typically 20–30 cm in length. Our study was conducted on a sand flat (mid-intertidal) located in the George Town Conservation Area (41°5′23.1″S, 146°48′46.7E″), Tasmania, approximately 50 km north of Launceston, Australia. At the time of the study, approximately 70% of the sand flat (which was ~ 2 hectares in area) was covered in seagrass of variable cover. The sand flat at George Town also contained an extensive population of Katelysia, an infaunal bivalve obtaining a maximum length of 40 mm (Edgar 2000). Several predators which likely consume Katelysia occur on the sand flat, including shorebirds (especially pied oystercatchers, Haematopus longirostris; sooty oystercatchers, Haematopus fuliginosus), the European green crab, Carcinus maenas and predatory whelks (Bedeva hanleyi and Cominella lineolata).

Variation in clam abundance and behaviour

Initially, we investigated patterns of clam abundance and behaviour as a function of seagrass per cent cover. We counted the number of live and dead clams to determine differences in the abundance and proportion of live clams across different seagrass densities. The number of clams alive and dead was determined for four categories of seagrass cover (0, 33, 66 and 100% ± 5%) in 0·25 m2 quadrats (N = 8 quadrats/seagrass treatment). Quadrats were placed in patches of different seagrass cover treatments, which were interspersed to reduce any spatially confounding effects. Clams on the surface were counted throughout the entire quadrat. Because of the high abundance of buried clams, they were subsampled using a core (20 cm diameter, total area 314 cm2; 3-mm mesh size) placed in the centre of each quadrat to a depth of 5 cm. The number of clams recovered in each core was then scaled up to 0·25 m2 and added to the number of clams on the surface to give a total number of clams/0·25 m2. Only dead clams that appeared recently deceased (clams that still had both valves intact and still had the creamy shell colour of live clams; longer dead clams often had green colorations on shells from colonization by microbiota; PE Gribben and JT Wright, personal communication) were counted. As a measure of behaviour, we assigned individual clams to one of two categories based on burial depth in the sediment: fully buried (not visible on the surface) and unburied (any part of the shell unburied and visible on the surface). Most unburied clams in our surveys were 50–100% unburied so our unburied treatment was a realistic representation of unburied behaviour.

We used generalized linear models (GLMs) to relate variation in live clam abundance, the proportion of live to total number of clams/quadrat (live + dead) and the proportion of unburied clams/quadrat as a function of seagrass cover (fixed factor). Because the abundance of clams can be affected by a multitude of factors and not necessarily related to differential rates of mortality among habitats, the proportion of live clams to total number of clams provided an initial estimate of differences in potential mortality rates of clams among seagrass treatments, while standardizing for differences in the abundance of live clams across the seagrass treatments (see 'Results'). We used GLMs as examinations of the distributions of residuals, and plots of residuals vs. means (Quinn & Keough 2002) on raw data, indicated the distribution of residuals was strongly non-normal and heteroscedastic. Transformation of the data did not reduce heteroscedasticity of variances or improve normality of the residuals. A Poisson error structure with a log link was appropriate for analysing the abundance data, and a binomial error structure for the proportion data – based on the degree of data dispersion (determined following Zuur et al. 2009). Wald chi-square tests were used for tests of significance. Pairwise post hoc tests using least significant difference adjustments for multiple comparisons were used to determine differences among seagrass per cent covers when significant effects of seagrass density were detected.

Effects of seagrass cover, clam behaviour and predator access on clam mortality

The surveys indicated that clam behaviour (burial) differed as a function of seagrass cover. To examine how seagrass cover and clam behaviour interact to determine overall mortality on clams, we conducted an experiment in which we manipulated seagrass cover (four levels; 0%, 33%, 66% and 100% cover), clam burial (two levels: buried and unburied) and predator access (three levels: full cages, half cages and open cages). We established the experiment within a bed of ~100% seagrass cover using a split-plot partly nested design. Within this bed, we marked out 24 plots each 1 × 1 m. Plots were randomly allocated a single seagrass cover treatment applied to the whole plot (i.e. seagrass cover was the between plot factor, and plots were replicates of seagrass cover; N = 6 plots/seagrass cover treatment). A single replicate of each clam burial × predator access treatment combination (within plot factors) was added to every plot resulting in N = 6 replicates of each clam burial × predator access treatment combination for each seagrass cover. Seagrass cover treatments were created by removing seagrass (rhizomes and shoots) in a regular way until there was an even cover of the required seagrass cover treatment in the plot. 100% seagrass cover plots were prodded by hand to mimic the disturbance that seagrass removal created. Any clams found in the plots while we established the treatments were removed, although we did not extensively search the plots to minimize disturbance. Plots were maintained for 2 months prior to the start of the experiment to allow them to settle following the physical removal of seagrass. During this period, the plots were checked every ~ 2 weeks and weeded as required to maintain the correct seagrass treatment. Cages were 14 cm wide × 14 cm long × 12 cm high with sides, tops but no bases, and were constructed from plastic mesh (mesh size 3 mm). Full cages excluded all predators; half cages had small openings (8 × 3 cm) in two opposing sides and allowed access by small mobile predators such as whelks and crabs but excluded larger predators such as birds, fish and rays, and open cages had sides to prevent clams from escaping but no top and allowed access by all predators. All cages were pushed 5 cm into the sediment. The unburied clam treatment was created by inserting a piece of plastic mesh (5 × 5 mm mesh size) onto the sediment surface inside the cages. The buried clam treatment had no mesh base. To ensure the seagrass treatment was the same between cages with and without the mesh base, we carefully hooked seagrass up through the mesh. During the experiment, sediment accumulated in some of the unburied treatment cages resulting in clams of variable burial depth (ranging from ~ 50 to 100% unburied) both within and between cages, which reflected the natural situation observed for clams in seagrass beds.

The experiment started on 24 November 2011. The day before the experiment began, we collected 1700 buried clams (size range 20–25 mm) from a nearby unvegetated area to ensure they were all of the same starting condition. All clams were marked with a permanent marker to distinguish them from other clams that may have been in our plots. Ten marked clams were added to each cage. We did not need to forcibly bury clams in the buried treatments, as they buried within 15 min once placed on the surface in all seagrass treatments. After 13 weeks, we recovered all live and dead clams from the cages. We also thoroughly hand searched the plots outside cages to see whether any clams had escaped.

We used four-factor split-plot anova to determine the effects of seagrass cover, clam burial, predator access, (all fixed factors) and plot (nested within seagrass; random factor) on total clam mortality. For unreplicated split-plot designs, there is no test for the plot within seagrass term and interactions involving the plot within seagrass term (Quinn & Keough 2002). However, these terms were included in the analysis to provide the appropriate denominator for calculating F-ratios for tests of fixed effects. Because there was a significant clam burial × predator access interaction, Tukey's tests (α < 0·05) were conducted separately within each treatment combination of those factors.

For clams that were collected alive at the end of the experiment, we also determined whether seagrass cover and burial behaviour affected final dry shell and tissue biomass. Once collected, clams were transported to the laboratory where the shell length (maximum anterior–posterior axis) was measured using vernier callipers prior to opening the shells and removing the tissue. Tissue from individual clams was placed on small pre-weighed foil trays. Shells and tissue were dried in an oven at 60 °C for 48 h and reweighed. We used four-factor split-plot anova to determine the effects of seagrass cover, clam burial, predator access, (all fixed factors) and plot (nested within seagrass; random factor) on shell and tissue biomass. Shell and tissue biomass of individual clams were standardized by dividing by its shell length prior to analyses. Analyses were conducted on means of standardized values per cage as some cages only had one clam remaining alive at the end of the experiment.

Predator surveys

To determine the abundance of clam predators foraging at our site, we surveyed densities of bird, crab and whelk predators during the experiment. Birds were surveyed prior to the experiment (18 October 2011) and three times during the experiment (15 December 2011; 2 January 2012; and 14 February 2012). On each date, bird surveys were done for 15 min at low tide at three sites that were ~ 250 m apart using a stationary visual technique. We counted all shorebirds, gulls and black swans, but only report species that are known consumers of clams (pied oystercatcher, Haematopus longirostris and sooty oystercatcher, Haematopus fuliginosus) and species known to consume clam predators (crabs and whelks). The latter were white-faced heron Egretta novaehollandiae, Pacific Gull Larus pacificus and Eastern Curlew Numenius madagascariensis. Only birds that were on the ground were counted, and these birds were typically foraging during our surveys. Crab surveys were done using baited traps (standard mesh traps, 80 × 20 × 60 cm with a mesh size of 10 mm). Six traps were set within the experimental area on four nights (15 December 2011; 21 December 2011; 12 January 2012 and 22 February 2012) over a single high tide. Approximately 250 g of pilchards were used as bait in each trap. The only crabs trapped were invasive Carcinus maenas. These were measured (maximum carapace width) and their sex determined. Density estimates of whelks (only Bedeva hanleyi occurred in our surveys – the scavenging gastropod Nassarius pauperatus is common but separate feeding experiments revealed it was not a predator on Katelysia) were obtained in three ways. First, they were counted in the cores used to determine initial clam abundance in different seagrass covers. Secondly, at the completion of the experiment, they were counted in one core (20 cm diameter) taken in each plot. Numbers of whelks were low in both these surveys as the total area sampled was relatively small (~ 1 m2 each time). Consequently, to more rigorously determine patterns of abundance of Bedeva, we counted the number of Bedeva in 0·25 m2 quadrats placed in both seagrass (= 40) and unvegetated areas (= 30), shortly after the completion of the experiment (8 March 2012).

Investigation of mechanisms determining clam behaviour

We investigated the roles of habitat modification by seagrass in determining clam behaviour. In terms of habitat modification, we focussed on the effect of seagrass cover on sediment condition. Redox potential is often used as a broad measure of sediment conditions (Gribben et al. 2009a), and we determined redox potential in all experimental plots. In each plot, we took three measurements (all measurements were taken outside experimental cages and at least 20 cm away from them) immediately prior to the end of the experiment above. Redox was measured using a SenTix ORP redox combination platinum electrode with (3 mol L−1) KCl, Ag+-free reference electrolyte and SenTix D-82362 pH meter. The electrode was calibrated using RH28 buffer solution, and all redox readings corrected for EHRef = −263 mV, that is, reported redox potentials are vs. the hydrogen electrode, EH0 = 0 mV. We used nested anova to determine how redox potential varied as a function of seagrass cover (fixed factor) and plot nested within seagrass (random factor). Tukey's post hoc tests were used to determine differences among seagrass treatments.


Clam abundance and behaviour

The mean density of live clams varied significantly as a function of seagrass cover (Wald χ2 23·674, d.f. = 3, < 0·001). Clam density was highest at 33% seagrass cover followed by 66% and 0% seagrass cover, and was lowest at 100% seagrass cover (Fig. 1a). There were significant differences in clam density between all per cent covers (all P < 0·044) except 33% and 66% covers (P = 0·245), and 66% and 0% seagrass covers (P = 0·353). The mean proportion of live clams was also dependent on seagrass cover (Wald χ2 = 69·741, d.f. = 3, < 0·001) and followed a similar pattern to mean density of live clams (Fig. 1b). There were significant differences between all seagrass covers (all P < 0·010) except 33% and 66% covers (P = 0·687). The mean proportion of unburied clams was also dependent of seagrass cover (Wald χ2 = 33·290, d.f. = 3, < 0·001) and increased as seagrass cover increased: there was ~ sixfold increase in unburied clams between 0% and 100% cover (Fig. 1c). The proportion of unburied clams was significantly different between all seagrass covers (all P < 0·013) except 0% and 33% seagrass cover (P = 0·169).

Figure 1.

Mean (±SE) (a) total number of live Katelysia scalarina, (b) proportion of live Katelysia to total number of clams (live + dead) and (c) proportion of live clams unburied in 0%, 33%, 66% and 100% cover of Zostera muelleri (= 8). Bars sharing a letter do not differ at = 0·05.

Effects of seagrass cover, clam behaviour and predator identity on clam mortality

Total mortality was approximately four times greater for unburied compared to buried clams (Fig. 2; Table 1). There was a significant interaction between predator access and burial depth (Table 1). Within the interaction term, clam mortality was significantly higher in unburied compared to buried clams for all three predator access treatments (Fig. 2a,b; Table 1). For buried clams, mortality was significantly higher in half cages compared to full and open cages, which did not differ from each other. For unburied clams, mortality was highest in open cages followed by half cages and lowest in full cages (Fig. 2a,b). The effect of seagrass cover on mortality was marginally non-significant (Table 1), but there was a trend for mortality of unburied clams to increase above 0% cover (Fig. 2b).

Figure 2.

Mean (±SE) total proportion mortality of (a) buried and (b) unburied Katelysia scalarina in full, half and open cages placed in 0%, 33%, 66% and 100% cover of the seagrass Zostera muelleri (= 6 cages/treatment, = 10 clams/cage).

Table 1. Summary of the results of the four-factor anova examining the effects of seagrass cover (Seagrass), Predator access treatments (full cage = FC; half cage = HC; open cage = OC), Burial depth (buried and unburied clams) and Plot (Seagrass) (n = 6 plots/seagrass treatment) on the mortality of Katelysia scalarina. Significant P-values in bold
Factord.f.MS F P Denominator
  1. Tukey's done within the predator × burial interaction: unburied > buried for all caging treatments; HC > FC = OC for buried clams, OC > HC > FC for unburied clams.

Seagrass30·1432·9080·060Plot (seagrass)
Predator access20·35713·416 <0·001 Predator × plot (seagrass)
Burial depth11·67434·182 <0·001 Burial × plot (seagrass)
Plot (seagrass)200·049 No test 
Predator × burial20·2048·112 0·001 Predator × burial × plot (seagrass)
Seagrass × burial30·0881·7930·181Burial × plot (seagrass)
Seagrass × predator60·0180·6650·678Predator × plot (seagrass)
Burial × plot (seagrass)200·049 No test 
Predator × plot (seagrass)400·027 No test 
Seagrass × predator × burial60·0170·6680·661Predator × burial × plot (seagrass)
Predator × burial × plot (seagrass)360·025 No test 

Of the dead clams we recovered during the experiment, in the open and half cages, 38% had drill holes indicative of predation by Bedeva, and only 25% were cracked or chipped, indicative of Carcinus or possibly bird predation (data not shown).

For live clams collected at the end of the experiment, overall mean (±SE) dry tissue biomass was 20% lower in unburied (3·05 ± 0·10 g mm−1) compared with buried clams (3·76 ± 0·13 g mm−1) (Fig. 3). Dry tissue biomass varied significantly with burial depth (F1,111 = 52·609, P < 0·001) but not seagrass cover (F3,20 = 2·935, P = 0·058) or cage type (F2,111 = 1·275, P = 0·284). Dry shell biomass did not vary with burial depth (F1,111 = 0·053, P = 0·818), seagrass cover (F3,20 = 0·703, P = 0·553) or cage type (F2,111 = 0·104, P = 0·901) (data not shown).

Figure 3.

Mean (±SE) dry tissue biomass (mg tissue mm−1 shell length) of (a) buried and (b) unburied Katelysia scalarina recovered in full, half and open cages placed in 0%, 33%, 66% and 100% cover of the seagrass Zostera muelleri (= 4–6 cages/treatment, means are averaged within cages, = 1–10 clams/cage).

Predator surveys

Pied oystercatcher were resident at our site throughout the experiment with 10·8 ± 1·4 individuals present on each date (mean ± SE pooled across sites) while sooty oystercatcher were uncommon (1·0 ± 0·6, mean ± SE) individuals on each date. The mean ± SE abundance on each date of white-faced heron was 7·5 ± 3·5; Eastern Curlew was 6·0 ± 3·5 and pacific gull was 5·3 ± 2·4. The mean ± SE abundance of invasive Carcinus on the four nights (total no. caught per night pooled across traps) was 13·3 ± 2·7. The average ± SE (mm) maximum carapace width on the four nights was 32·4 ± 0·8 (range from 25 to 47 mm, n = 53). The male:female sex ratio was 25:28. Carcinus were observed at times in both open and half cages during the experiment. In the initial surveys, one individual of the whelk Bedeva was found in both the 66% and 100% seagrass plots, but none in the 0% and 33% seagrass plots. In the final surveys (cores), there were two Bedeva in the 33%, one each in the 66 and 100% and none in the 0% seagrass cover treatments. In randomly placed quadrats, Bedeva density was 1 ± 0·4 m−2 in seagrass and 0·4 ± 0·2 m−2 in bare sand. Bedeva were observed in both open and half cages during the experiment, although we did not record their abundance.

Investigation of mechanisms determining clam behaviour

Sediment redox decreased with increasing seagrass cover, although the effect of seagrass cover was only marginally non-significant (F3, 20 3·069, P = 0·051) (Fig. 4). Tukey's tests showed that redox was significantly lower at 0% and 33% cover compared to both 66% (Tukey's tests, 0% vs. 66%, P = 0·045; 33% vs. 66% P = 0·049) and 100% seagrass cover (Tukey's tests, 0% vs. 100%, P = 0·001; 33% vs. 100%, P = 0·001). There was no difference in redox between 0% and 33% (Tukey's tests, P = 0·95) and 66% and 100% (Tukey's test, = 0·450). Redox varied significantly with plots(seagrass) (F20,48 =2·731, P = 0·002).

Figure 4.

Mean (±SE) redox potential in 0%, 33%, 66% and 100% seagrass cover plots measured at the end of our main experiment (N = 4 plots/treatment with 3 replicate readings/plot).


Habitat-forming species can simultaneously alter mortality risk via their physical structure (i.e. by reducing predator–prey encounter rates) and/or by modification of non-structural abiotic conditions, which affect behaviour. Here, changes in the behaviour (shallow burial) of the clam, Katelysia scalarina, associated with the seagrass Zostera muelleri, resulted in higher mortality compared to buried clams, and this was due to both predation and non-predation sources. Our surveys indicated that a low percentage cover of seagrass was beneficial to clams – abundances and proportion of live clams were highest at 33% seagrass cover. However, at higher per cent seagrass covers, unburied behaviour increased with a concomitant decrease in the abundance and proportion of live clams. Interestingly, our experiment showed that predation was similar across seagrass cover treatments when behaviour was held constant, indicating that predators were responding to the availability of prey that are at risk (i.e. unburied) rather than the cover of seagrass. Thus, although low seagrass cover had more live clams, the increasing unburied behaviour of clams at higher seagrass cover had strong negative effects on mortality and these negative effects outweighed any positive effects of high seagrass cover. Thus, indirect effects of seagrass cover on clam behaviour appeared more important in determining mortality than the physical effects of seagrass.

Changes in prey behaviour increase predation risk in a range of ecosystems. Young trout in freshwater ecosystems increase foraging in risky habitats when food availability is low (Biro, Post & Parkinson 2003). Colonization of coral reef heads by macrophytes can result in fish leaving the protective shelter of coral, even when it exposes them to predators (Feary 2007). Eutrophication in soft-sediment ecosystems can change abiotic conditions altering the burial behaviour of clams, which may increase predation risk (Taylor & Eggleston 2000; Tallqvist 2001). These studies provide examples of how external factors cause changes in prey behaviour. However, habitat-formers themselves can induce risky behaviour, particularly for infaunal bivalves. For example, the dense roots and modification of the abiotic environment by seagrasses and other macrophytes can reduce the burial depth of clams (Skilleter 1994; Wright et al. 2010). However, few studies have determined whether such changes in behaviour increase predation rates. Despite well documented positive effects of seagrass structure in reducing predation (Peterson 1982; Irlandi 1994; Goshima & Peterson 2012), we have demonstrated that increased predation resulting from changes in clam behaviour in seagrass outweigh the potential refuge benefits that seagrasses provide.

Our different caging treatments attempted to separate rates of predation by different predators in our system. The strong effects of predators on unburied clams appeared largely additive and were generally highest in the open cages when all predators had access compared to half cages when only smaller predators (whelks and crabs) had access. For buried clams, predation mortality was higher in half cages compared to full and open cages. The reasons for this are unclear but half cages may have provided intermediate predators a refuge from predation, allowing them more time to forage on buried clams in relative safety. Indeed, the invasive green crab, Carcinus maenas, was found in half cages on several occasions. The use of half cages by predators may also explain the marginally non-significant difference in predation between buried and unburied clams for this caging treatment. Alternatively, these findings may reflect an unknown caging artefact. However, because mortality of unburied clams in half cages was intermediate between open and full cages, any artefact is unlikely.

In the full-cage treatments, Katelysia on the surface suffered higher rates of non-predation mortality and had lower tissue biomass compared to buried clams. The increased exposure to higher temperatures at low tide and/or the extra cost of maintaining closed shells without the benefit of surrounding sediments could also have contributed to higher non-predation mortality and lower condition of unburied clams (Newell & Hidu 1982). Both lethal and non-lethal effects occurred irrespective of seagrass cover indicating that for unburied clams Zostera does not reduce abiotic stress. In addition, changes in redox potential may contribute to increased non-predation mortality and lower condition in unburied clams. Changes in sediment conditions (e.g. reduced dissolved oxygen and redox potential) do affect the survivorship of infaunal clams (Wright & Gribben 2008; Gribben et al. 2009b). However, both non-predation mortality and lower condition were relatively constant across seagrass cover whereas redox decreased with increasing cover, suggesting limited direct effects of redox on these.

Shallow burial in infaunal bivalves typically occurs in response to changes in abiotic conditions, particularly low dissolved oxygen (Norkko & Bonsdorff 1996; Tallqvist 2001; Wright et al. 2010). Such conditions often occur via algal respiration and microbial degradation of accumulated plant detritus (D'Avanzo & Kremer 1994), although thick root mats can also reduce clam burial depth (Skilleter 1994). Our field sampling indicated that unburied behaviour increased and sediment redox decreased with increasing seagrass cover, suggesting a link between seagrass sediment modification and behaviour. However, this behaviour did not manifest itself in the main experiment. When added to plots, all clams in the burial treatment fully buried within 15–30 min and only two clams were observed unburied during the experiment. One explanation for the lack of unburied behaviour in the experiment is that sediment effects on behaviour may accrue gradually through time and that the experiment was not long enough for behavioural responses to manifest themselves. This may be exacerbated because we started with clams in good condition (i.e. buried clams from unvegetated habitat). At the very least, the full burial of clams across all seagrass densities indicated a limited role for the physical structure of seagrass in directly affecting clam behaviour.

Despite previous evidence for a key role of seagrass cover in mediating predation risk (see reviews by Orth, Heck Jr & van Montfrans 1984; Williams & Heck 2001), the protective function of seagrass varies between species (Heck & Crowder 1991) suggesting an important role for seagrass traits in mediating predator–prey interactions. In our study, the absence of strong seagrass cover effects may be due to the traits of Zostera (e.g. blade length, biomass). Zostera had relatively short blades at our site (~20 cm) and although it occurred at 100% cover, its short stature may not be effective at inhibiting predators compared to other seagrasses with longer blades. Indeed, for the open cages, predation generally increased with seagrass cover when unburied clams were exposed to a full range of predators. In addition, seagrass cover or patch size can have negative effects on the traits of associated species by reducing water flow and food supply (Reusch & Williams 1999). However, we know little about how the traits of habitat-forming species affect prey behaviour and influence predation risk.

In our study, seagrass structure had no effect in mediating the predation risk of the large adult clams we used in our experiment. By contrast, seagrass physical structure often facilitates the recruitment of bivalves by providing a surface for colonization and/or a refuge from predation compared to nearby unvegetated areas (see Orth, Heck Jr & van Montfrans 1984; Williams & Heck 2001 for reviews). Indeed, juvenile Katelysia appeared more common in Zostera compared to unvegetated sediments, although we did not quantify recruitment in this study (Gribben PE and Wright JT Pers. Obs.). Recruitment into seagrass may explain the relatively high density of Katelysia in 100% seagrass cover, despite high levels of unburied behaviour and suggests that the effects of seagrass on Katelysia may change with ontogeny; positive effects on recruitment but negative effects on the mortality of larger Katelysia because of shallower burial. Reduced growth of bivalves in seagrass beds has been described (Bologna & Heck 1999), but we found no effect of seagrass on Katelysia tissue biomass.

Here, we have demonstrated that the effects of habitat-forming species on prey behaviour can have stronger negative consequences for predation and non-predation risk relative to any positive effects of habitat structure. Importantly, seagrass cover had consequences for both predation and non-predation mortality, but only via changes in clam behaviour. Increased predation in habitat-forming species as a function of increasing cover may be particularly prevalent where the habitat-formers are strong determinants of abiotic conditions and prey behaviour responds strongly to changes in abiotic condition, such as in marine and aquatic soft-sediment, and terrestrial soil-based ecosystems. Future studies should determine the direct and indirect effects of habitat-former traits and density on predator–prey interactions. Such studies will provide a rigorous understanding to the processes by which habitat-forming species control the outcome of species interactions.


We thank Ian Jermyn, Signe Latzel and Rebecca Mueller for helping build cages, setting up experiments, sorting samples and weighing clams. Ralph Cooper (Birds Tasmania) and Kris Carlyon (DPIPWE) provided assistance with information on George Town birds and permits. Jeff Ross provided use of the redox probe.