Prey dispersal rate affects prey species composition and trait diversity in response to multiple predators in metacommunities


Correspondence author. E-mail:


1. Recent studies indicate that large-scale spatial processes can alter local community structuring mechanisms to determine local and regional assemblages of predators and their prey. In metacommunities, this may occur when the functional diversity represented in the regional predator species pool interacts with the rate of prey dispersal among local communities to affect prey species diversity and trait composition at multiple scales.

2. Here, we test for effects of prey dispersal rate and spatially and temporally heterogeneous predation from functionally dissimilar predators on prey structure in pond mesocosm metacommunities. An experimental metacommunity consisted of three pond mesocosm communities supporting two differentially size-selective invertebrate predators and their zooplankton prey. In each metacommunity, two communities maintained constant predation and supported either Gyrinus sp. (Coleoptera) or Notonecta ungulata (Hemiptera) predators generating a spatial prey refuge while the third community supported alternating predation from Gyrinus sp. and N. ungulata generating a temporal prey refuge. Mesocosm metacommunities were connected at either low (0·7% day−1) or high (10% day−1) planktonic prey dispersal. The diversity, composition and body size of zooplankton prey were measured at local and regional (metacommunity) scales.

3. Metacommunities experiencing the low prey dispersal rate supported the greatest regional prey species diversity (H’) and evenness (J’). Neither dispersal rate nor predation regime affected local prey diversity or evenness. The spatial prey refuge at low dispersal maintained the largest difference in species composition and body size diversity between communities under Gyrinus and Notonecta predation, suggesting that species sorting was operating at the low dispersal rate. There was no effect of dispersal rate on species diversity or body size distribution in the temporal prey refuge.

4. The frequency distribution, but not the range, of prey body sizes within communities depended upon prey dispersal rate and predator identity. Taken together, these results demonstrate that prey dispersal rate can moderate the strength of predation to influence prey species diversity and the local frequency distribution of prey traits in metacommunities supporting ecologically different predators.


Large-scale spatial processes may interact with niche-based community structuring mechanisms to determine local and regional assemblages of predators and their prey (Holt 1993). Across regional spatial scales, biogeographical history (Schluter & Ricklefs 1993), predator and prey dispersal distances (Cronin et al. 2000; Hammond, Luttbeg & Sih 2007) and the strength of environmental gradients (Werner & McPeek 1994; Chase 2003) interactively determine predator-prey spatial co-occurrence patterns. The taxonomic and functional diversity of predators represented in the landscape can further affect both prey assemblages and the spatial distribution of prey traits (Chalcraft & Resetarits 2003a; DeWitt & Langerhans 2003). At the local community scale, selective predators may exclude prey species (Sih et al. 1985; Spiller & Schoener 1998), facilitate prey coexistence through suppression of dominant competitors (Paine 1966; Leibold 1996; Shurin & Allen 2001), or alter prey densities via apparent competition (Holt 1977; Bonsall & Hassell 1997). Recent advances in metacommunity theory suggest that predator and prey dispersal rates among local communities can modify these effects on local and regional predator-prey assemblages (Holt 1993; Shurin & Allen 2001; Amarasekare 2006). However, empirical work in multi-trophic metacommunities has largely focused on the impacts of a single predator in constant prey environments (e.g. Holyoak & Lawler 1996; Kneitel & Miller 2003; Cadotte, Fortner & Fukami 2006 but see Warren 1996; Holyoak 2000; Cadotte & Fukami 2005) and has yet to adequately address how predator and prey dispersal may affect prey composition in response to predator functional diversity represented in the region.

In predator-prey metacommunities, local communities may support functionally dissimilar predators in both space and time (Cadotte & Fukami 2005; McCann, Rasmussen & Umbanhowar 2005; Amarasekare 2007; Hammond et al. 2007). Spatial heterogeneity in selective predator incidence can generate spatial prey refuges (Sih 1987; Sih & Wooster 1994) and thus promote regional predator-prey coexistence (Cadotte & Fukami 2005). Likewise, temporally variable predator occurrence from movement behaviour or colonization-extinction dynamics can yield temporal prey refuges (Sih 1987; Howeth & Leibold 2008). Prey dispersal rates in metacommunities may interact with this spatial and temporal variation in predator incidence to facilitate prey tracking of local predation (‘species sorting’; Leibold & Norberg 2004) or generate source-sink dynamics in prey (‘mass effects’; Holyoak & Lawler 1996).

Recent empirical evidence suggests that the spatial frequency of prey species and prey traits maintained in metacommunities will depend in part upon prey dispersal rates among local communities (Warren 1996; Cadotte & Fukami 2005; Cadotte et al. 2006). For example, prey immigration may mediate the negative impact of local predator selectivity by delivering migrant prey representative of the range of species maintained in the regional pool, and may thereby reduce local prey extinction risk (Sih & Wooster 1994; Holyoak & Lawler 1996; Shurin 2001). Additionally, ecologically important prey traits directly selected by predation, including body size and avoidance behaviour, may only persist in a local community through prey immigration (Leibold & Norberg 2004; Urban et al. 2008). Although prey dispersal rate may be important in shaping prey species and trait composition at multiple scales, the relative effects of dispersal may depend on the breadth of functional diversity represented in the regional predator species pool.

In this study, we tested for interactive effects of prey dispersal rate and spatially and temporally heterogeneous predation from different predators on prey structure in experimental pond metacommunities. Pond metacommunity assemblages were modelled after two naturally co-occurring and differentially size-selective invertebrate predators, backswimmers Notonecta and whirligig beetles Gyrinus, and their zooplankton prey. Experimental metacommunities consisted of three pond mesocosm communities, where two of the three communities maintained a spatial prey refuge through constant predation from Gyrinus sp. or N. ungulata Say while the third community maintained a temporal prey refuge by alternating Gyrinus sp. and N. ungulata incidence in simulated colonization and extinction. Mesocosm metacommunities were connected at either low or high planktonic prey dispersal which reflected dispersal rates likely to alter zooplankton diversity based upon metacommunity models (Mouquet & Loreau 2002, 2003) and field studies (Michels et al. 2001; Cohen & Shurin 2003). Low dispersal approximated 7% of the demographic turnover rate of zooplankton in this experiment (0·10 day−1; estimated from Gillooly 2000), while high dispersal approximated 100% of the turnover rate. In accord with metacommunity theory, we hypothesized that local communities at low prey dispersal would be more diverse than communities at high dispersal (Mouquet & Loreau 2002, 2003), while high dispersal metacommunities would support a smaller, and less equitably distributed, regional prey species pool from regional homogenization (Mouquet & Loreau 2003). Due to differences in foraging of Gyrinus and Notonecta, we expected these predators to produce local prey assemblages with contrasting size distributions (following Chalcraft & Resetarits 2003b). Yet, we predicted that differences in prey size structure among the three local communities would be reduced at high dispersal as a consequence of regional homogenization.

Materials and methods

Study system

Ponds serve as a model system in which to address the influence of predator functional diversity on prey structure at local and regional spatial scales in metacommunities. Pond communities are nested within a landscape of terrestrial matrix where species diversity is jointly determined by regional species dispersal rates and the local environment (Shurin 2001; Cottenie et al. 2003; Howeth & Leibold 2010). Local pond communities are spatio-temporally connected in part by actively dispersed predators and passively dispersed prey (Caceres & Soluk 2002; Chase 2003; Chase & Ryberg 2004). Two functionally dissimilar invertebrate predators, the whirligig beetle Gyrinus (Coleoptera: Gyrinidae) and the backswimmer Notonecta (Hemiptera: Notonectidae), coexist locally in natural ponds (Clark 1928) and are patchily distributed among ponds in a metapopulation structure (Svensson 1985; Briers & Warren 2000). Adult Notonecta are relatively large (body length 11–12 mm; Rice 1954) and less gape-limited than Gyrinus (4–5 mm; Oygur & Wolfe 1991). Notonecta swims vertically through the water column while preying upon large-bodied zooplankton (Scott & Murdoch 1983; Arner et al. 1998); whereas Gyrinus remains on the water surface and opportunistically preys upon small-bodied zooplankton (Svensson 1985; Oygur & Wolfe 1991). Here, we constructed pond mesocosm metacommunities to resemble these natural predator-prey assemblages. We consider the triad of mesocosm communities connected by dispersal to be the ‘regional’ spatial scale or the ‘metacommunity,’ and each mesocosm community to be the ‘local’ spatial scale.

Experimental design

The pond mesocosm experiment was conducted at Michigan State University’s Kellogg Biological Station (Hickory Corners, Michigan, USA) from 23 May–21 August 2003. On 23 May, twenty-four 322 L polyethylene stock tank mesocosms were acid washed and filled with 20 L of silica sand for substrate and 300 L of well water. Well water was measured for ambient levels of total nitrogen (N) and phosphorus (P) using spectrophotometric methods. N (NaHNO3) and P (NaH2PO4) were subsequently added to each tank to achieve target nutrient levels of 2100 μg L−1 N and 150 μg L−1 P which reflect the upper range of N and P found in natural ponds of the region (Hall et al. 2007). Weekly additions of N (0·173 g tank−1) and P (0·009 g tank−1) maintained target levels and offset 5% day−1 loss of nutrients to the bottom substrate. On 29 May, mesocosms received 200 mL of a pond water inoculum containing a diverse regional assemblage of zooplankton, algal and bacterial communities representative of biota from 16 south-central (Barry and Kalamazoo Counties) Michigan ponds. After inoculation, communities were allowed to reach near-equilibrium for c. 1 month. All mesocosms were covered with 1 mm screen mesh lids to prevent exchange of organisms with the outside environment.

On 26 June, six adult individuals each of Gyrinus or Notonecta were added to one of the two mesocosms supporting constant food webs in each metacommunity. Half (= 2) of the cyclical predation communities received Gyrinus while the remaining half received Notonecta. The predator species in the cyclical predation communities were switched every 14 days, where the two predators never coexisted in time within the same local community. Six adult individuals of the appropriate species were added at the start of each new cycle. Of the two predators, only Notonecta reproduced during the experiment. Juvenile notonectids were removed from both cyclical and constant predation communities during the predator switching events in order to maintain standardized predation pressure.

Low and high prey dispersal treatments were initiated in mesocosm metacommunities on 4 July. To simulate prey dispersal within a metacommunity of the low dispersal treatment, water was manually moved among the three communities once per week with an integrated depth polyvinyl chloride (PVC) sampler (5% tank water volume per week, roughly equivalent to 0·71% day−1; 8 L week−1 transferred from each of the two constant communities to the community supporting cyclical predation and 8 L subsequently transferred back to each constant community; Fig. S1a, Supporting information). This low dispersal rate approximates natural levels of zooplankton dispersal found among hydrologically isolated ponds in south-central Michigan (Cohen & Shurin 2003). To achieve high dispersal within a metacommunity (31·1 L day−1, 10% tank water volume per day, roughly equivalent to 70% per week) water was transferred among the three communities with a centrifugal water pump through 19 mm PVC pipe (Fig. S1b, Supporting information). Water was pumped from the constant environment tanks to the variable tank first and then allowed to flow back to equilibrate water levels, in a scheme similar to the manual-transfer used in the low dispersal treatment. Based upon data presented in Michels et al. (2001), this rate approximates high zooplankton dispersal among natural ponds that are hydrologically connected. Each community in the low dispersal treatment supported a pump and PVC loop to control for possible pump-induced disruption of predator-prey interactions and any negative effects on plankton mortality sustained in the high dispersal treatment (Fig. S1a, Supporting information). Predator exclusion screens, consisting of overlapping 6 mm mesh hardware cloth sheets, covered the ends of the PVC pipes to prevent inter-community exchange of invertebrate predators. All water pumps were activated for 1 min daily during the evening (19.30–19.31 h), when zooplankton were distributed throughout the water column during diel vertical migration.


Mesocosm metacommunities were sampled for zooplankton (i) on 25 June prior to imposition of predation and dispersal treatments; (ii) on 3 July with the predation treatment imposed but prior to the dispersal treatment; (iii) and subsequently every 7 days through 22 August with both predation and dispersal treatments imposed. Tanks were sampled with a 1 L depth-integrated PVC sampler to remove 10·5 L of water from each tank. Ten liters of the sample were filtered through 80 μm mesh to isolate zooplankton. Zooplankton samples were cleaned manually to rid of debris and were preserved in 5% acid-sucrose iodine solution for later microscopic enumeration in the laboratory. Following Smith (2001), cladoceran zooplankton were identified to species for Alona affinis, Bosmina longirostris, Ceridodaphnia reticulata, Chydorus sphaericus, Daphnia pulex, Scapholeberis mucronata and Simocephalus serrulatus (species hereafter referred to by genus) and copepods were identified to the orders Calanoida and Cyclopoida. Further, adult and juvenile stages of Ceriodaphnia, Daphnia, Scapholeberis and copepods (copepodite stage) were assessed independently. Body lengths of c. 25 individuals of each species and stage were measured to obtain a mean species and stage-specific body length.

Regional and local prey diversity

Three different prey diversity measures: (i) species diversity (Shannon index, H’, in Magurran 2004 using log10); (ii) species richness (S); and (iii) species evenness (Shannon index, J’, in Magurran 2004 using log10) were calculated at local and regional spatial scales for each predator-prey metacommunity (Fig. S1c, Supporting information). Regional diversity measures represented the sum of unique species or taxonomic groups (for S) or individuals (for H’ and J’) sampled in the three communities of the metacommunity on a sample date. Local species richness did not differ among mesocosm metacommunities at the start of the experiment prior to imposition of prey dispersal and predation treatments (one-way anova; F7,16 = 0·94, = 0·51). Therefore, to evaluate predation and dispersal treatment effects on prey diversity, local and regional diversity measures were each averaged over weeks 3–6 (to align with density analyses below) and the final two sample dates of the experiment, weeks 5–6. Mean values were assessed for normality with a Shapiro–Wilks test prior to analysis with one-tailed Welch’s t-tests (regional) and split-plot anovas (local) in the R statistical environment (R Development Core Team 1996) and SAS v9·1 (SAS Institute, Cary, North Carolina, USA), respectively. One-tailed t-tests were employed because of an a priori hypothesis based upon metacommunity theory.

In order to evaluate the effect of prey dispersal rate and predation regime on prey composition in the three communities of each metacommunity, a permutational multivariate analysis of variance was performed in permanova v6 (Anderson 2001, 2005). A split-plot design was not possible in permanova, and thus a two-way analysis was employed. Juvenile and adult counts were combined for each prey species, except for the copepodites which were excluded because they were not assigned to order. The permanova was conducted on Bray–Curtis dissimilarity measures calculated from log10 (x + 1) transformed species density data generated from a mean of sampling weeks 3–6. In the Bray–Curtis distance measure, abundant and rare species contribute equally to the dissimilarity between sites. Thus, Bray–Curtis is especially appropriate for normalized density data (Legendre & Legendre 1998). To test for differences in variance in species composition between dispersal and predation treatments, we calculated the Bray–Curtis distance of the replicates from their treatment centroids, and compared the mean distances between treatments using permutation anova in permdisp2 (Anderson 2004). permanova and permdisp2 tests were performed using 10 000 unrestricted permutations of Bray–Curtis distance measures. For effects shown to be statistically significant at α = 0·05, post hoc pairwise comparisons with the permanovat-statistic assessed treatment differences.

To determine the response of individual prey taxa to dispersal rate and predation regime, an indicator species analysis, IndVal (Dufrene & Legendre 1997), was employed in PC-ORD v4 (McCune & Mefford 1999). In this method, an indicator value (I.V.) represents a percentage score corresponding to a species strength of treatment level specificity, as determined by species densities, where a score of 100 is a perfect predictor or ‘indicator’ of the treatment level. The cyclical and two constant predation communities were evaluated independently for dispersal and predation treatment level indicators. For prey communities under cyclical predation, one Gyrinus-Notonecta predation cycle was evaluated by averaging species density data for each of two 2-week predation periods, sampling weeks 3–4 and weeks 5–6. For prey communities under constant predation, species density data for weeks 3–6 were averaged for the analysis. Juvenile and adult counts were combined (copepodites excluded), and species density data were log10 (x + 1) transformed. Statistical significance of a species’ I.V. was evaluated by comparing the observed I.V. to a null distribution generated through 10 000 Monte Carlo permutations. The proportion of simulated I.V. greater than or equal to the observed I.V. determined significance for an α of 0·05.

Spatial and temporal prey refuges

A measure of beta diversity, Sorenson’s dissimilarity (Magurran 2004), was used to evaluate effects of dispersal rate and predator identity on species composition between the two communities in the spatial refuge (constant predation) and within the community in the temporal refuge (cyclical predation) (Fig. S1c, Supporting information). For the temporal refuge, beta diversity was calculated from mean density data taken from weeks 3–4 and weeks 5–6 to encompass alternating predation from Gyrinus and Notonecta. For the spatial refuge, beta diversity was calculated from mean density weeks 3–6 to align with the temporal analysis. To detect responses of individual species to predator identity in the spatial and temporal refuges within each dispersal treatment, this same density data was contrasted between Gyrinus and Notonecta regimes. Diversity and log10 density values were analysed with one-tailed Welch’s t-tests in the R statistical environment.

Regional and local prey size structure

The role of prey dispersal rate on the distribution of prey body sizes within the predator-prey metacommunity was assessed by calculating the proportion of individuals within body size classes at the local and regional scale. Each species and stage was assigned to a body size class based upon mean body length (mm), where size classes were defined by log10 0·1 mm increments. Species and life-history stage densities (juvenile and adult, where applicable) were averaged over weeks 3–6 and weeks 5–6, and subsequently were converted to proportional representation for each body size class. Size class proportions for each community and metacommunity were arcsine square root transformed and analysed for the effects of dispersal rate and body size class using anova in statistica v6.1 (StatSoft Inc., Tulsa, OK, USA).


Regional and local prey diversity

Prey dispersal rate altered prey species diversity at regional spatial scales in predator-prey metacommunities (Fig. 1a; Table 1). The high dispersal metacommunities were less diverse (H’) than low dispersal metacommunities. High dispersal significantly decreased regional species evenness (J’); yet there was only a marginal effect of dispersal rate on regional richness (S, Fig. 1c and e). These regional effects were strongest at the end of the experiment, averaged over weeks 5–6. At the local spatial scale, the influence of dispersal rate on diversity measures was less apparent (Table 1). Diversity was reduced in high dispersal communities relative to low dispersal communities, but this effect was only marginally significant for weeks 5–6 and NS for weeks 3–6 (Fig. 1b). There was no impact of dispersal rate or predation on local species richness or evenness for weeks 5–6 (Fig. 1d and f). However, there were marginal effects of dispersal, and a dispersal by predation interaction, on local prey species richness for weeks 3–6.

Figure 1.

 Regional and local prey diversity in predator-prey metacommunities. (a) Regional prey species diversity (H’) as a function of low and high prey dispersal rate. (b) Local prey species diversity as a function of prey dispersal rate and predation from Gyrinus sp. (black bars), Gyrinus sp.-Notonecta ungulata (striped bars) and Notonecta ungulata (gray bars). (c) and (d) Regional and local prey species richness (S). (e) and (f) Regional and local prey species evenness (J’). Closed circles and dashed line represent the mean response for low and high dispersal treatments. Statistically significant dispersal effects: + < 0·1; *< 0·05. Statistical analysis reported in the results. Values are mean sampling weeks 5 and 6 + 1 SE, = 4.

Table 1.   Regional and local prey species diversity (H’), richness (S) and evenness (J’) in predator-prey metacommunities. Diversity measures were contrasted across dispersal rate (low, high) and predation (Gyrinus sp., Notonecta ungulata, Gyrinus sp.-Notonecta ungulata) treatments.
Response variableTreatmentd.f.t, FP-valuet, FP-value
weeks 3–6weeks 5–6
  1. Results reported from one-tailed Welch’s t-tests (regional) and split-plot anovas (local) on mean diversity values from weeks 3 to 6 and weeks 5 to 6. Significance levels for probability values: *< 0·1; **< 0·05.

 Diversity (H’)Dispersal5·38−2·100·04**−2·720·02**
 Richness (S)Dispersal3·74−1·850·07*−1·670·09*
 Evenness (J’)Dispersal5·98−1·460·11−3·110·01**
 Diversity (H’)Dispersal1, 60·290·614·460·08*
Predation2, 121·740·221·000·40
Dispersal × Predation2, 122·450·131·540·25
 Richness (S)Dispersal1, 65·160·06*2·880·14
Predation2, 120·290·761·140·35
Dispersal × Predation2, 123·190·08*0·780·48
 Evenness (J’)Dispersal1, 60·360·572·550·16
Predation2, 121·790·210·880·44
Dispersal × Predation2, 121·080·372·490·12

Prey dispersal rate influenced local prey species composition in metacommunities (permanova; dispersal F1,18 = 2·12, = 0·05; predation, F2,18 = 1·14, = 0·35; dispersal × predation, F2,18 = 0·34, = 0·98). There was no effect of predation regime, nor an interactive effect of predation and dispersal, on community composition. Dispersal rate did not alter the variance in community composition (permdisp2; dispersal F1,18 = 0·84, = 0·37; predation, F2,18 = 3·77, = 0·04; dispersal × predation, F2,18 = 1·41, = 0·28). In contrast, predation affected community variance, with Gyrinus and Notonecta communities differing from one another and where Notonecta communities exhibited the greatest dispersion (mean Bray–Curtis within-treatment dissimilarities: G = 26·15, N = 35·28, G-N = 31·94; post hoc treatment contrasts; G × N, = 2·68, = 0·02, G-N × G; = 1·28, = 0·22; G-N × N, = 1·45, = 0·17). There was no interactive effect of predation regime and dispersal rate on community variability.

The indicator analysis identified key species responsible for dispersal-induced changes in community composition (Table 2). Two small-bodied zooplankters, Bosmina and Chydorus, were significant indicators of the low dispersal rate reflecting their greater densities in low dispersal metacommunities (Fig. 2a–f; Fig. S2, Supporting information). The remaining species in the experiment showed no trends in abundance by dispersal treatment (Fig. S3, Supporting information). Only one species, the large-bodied Daphnia, served as a significant indicator of predator identity. Daphnia was more abundant in Gyrinus communities than in Notonecta communities (Fig. 2g and h; Fig. S2, Supporting information). Daphnia additionally served as an indicator of the low dispersal rate for communities under cyclical predation (Fig. 2i).

Table 2.   Results of the indicator species analysis assessing the strength of prey taxa association with prey dispersal rate (D; L = low, H = high) and predator identity (P; G = Gyrinus sp., N = Notonecta ungulata) by predation regime (constant, cyclical) in predator-prey metacommunities.
  1. The observed indicator value (I.V.) reflects the indicator score for the listed treatment level and represents the largest of the two I.V. values calculated for each of the two treatment levels (L or H; G or N), with 100 being a perfect indicator score. The indicator value from randomized groups (I.V.rand) is relayed for comparison. Prey are ranked by mean size from smallest (Chydorus) to largest (Daphnia). Significance levels for probability values: *< 0·1; **< 0·05.

Figure 2.

 Density of indicator species through time in metacommunities, standardized by initial densities in the absence of dispersal (Disp) and predation (Pred) treatments. A density value of zero reflects no change in density from initial pre-treatment conditions (no Disp, no Pred). Bosmina density in low prey dispersal (closed circles) and high prey dispersal (open circles) communities under (a) Gyrinus sp. (G) (b) Notonecta ungulata (N) and (c) Gyrinus sp.-Notonecta ungulata (G-N) predation. (d–f) Chydorus and (g–i) Daphnia. Treatment sequence through time (week; left to right): no predation and no dispersal; predation and no dispersal; predation and dispersal for weeks 1 through 6. Values are mean log10(x + 1) density + 1 SE, = 4. Nonstandardized density values reported in Fig. S2, Supporting information.

Spatial and temporal prey refuges

Beta diversity values in both the spatial and temporal refuges were lower on average in the high dispersal metacommunities, indicating an increase in compositional similarity, but they did not significantly differ from the low dispersal metacommunities (spatial refuge, beta diversity mean ± SE; low dispersal, 0·61 ± 0·14, high dispersal, 0·56 ± 0·11; one-tailed t-test, = −0·32 d.f. = 5·82, = 0·38; temporal refuge, low dispersal, 0·61 ± 0·10, high dispersal, 0·50 ± 0·16; = −0·59, d.f. = 4·92, = 0·29).

There were differences in prey species densities as a function of predator identity in spatial, but not temporal, prey refuges within low and high dispersal metacommunities (spatial, Table S1, Supporting information; temporal, one-tailed t-test; all species, > 0·1; Table S2, Supporting information). In the spatial prey refuge of low dispersal metacommunities, Chydorus, Daphnia and Scapholeberis supported higher densities in the presence of Gyrinus, suggesting an effective refuge from direct or indirect effects of Notonecta predation (Fig. 3a–f, Table S1, Supporting information). There was no difference in the densities of these species between the two community types in the high dispersal metacommunities. Simocephalus was the only species that supported different densities in the two constant predation communities of high dispersal metacommunities, where densities were significantly lower in the presence of Notonecta (Fig. 3g and h).

Figure 3.

 Species that exhibited differential density responses to predator identity in the spatial prey refuges of low and high dispersal metacommunities. (a) Chydorus density in low dispersal and (b) high dispersal spatial refuges in the constant presence of Gyrinus sp. (G) and Notonecta ungulata (N), (c) and (d) Daphnia, (e) and (f) Scapholeberis and (g) and (h) Simocephalus. Values are mean log10(x + 1) density for weeks 3 through 6 + 1 SE, = 4. Statistical significance: + < 0·1; *< 0·05. Statistical analysis reported in the results.

Regional and local prey size structure

The distribution of prey body sizes within predator-prey metacommunities was shaped by an interaction between prey dispersal rate and predator identity (Fig. 4). At the regional scale, there was an effect of body size class, but not prey dispersal rate, on the proportion of individuals represented in each prey size class over weeks 5–6 (two-way anova; weeks 3–6: dispersal, F1,36 = 0·01, = 0·94; size class, F5,36 = 1·90, = 0·12; dispersal × size class, F5,36 = 1·44, = 0·23; weeks 5–6: dispersal, F1,36 = 0·03, = 0·86; size class, F5,36 = 3·51, = 0·01; dispersal × size class, F5,36 = 0·76, = 0·58, Fig. 4a). In both low and high dispersal metacommunities there were fewer individuals representing the larger body size classes. There was no influence of dispersal rate on mean body size in the region (Table S3, Supporting information). In communities under Gyrinus predation, neither body size class nor dispersal rate affected the distribution of body sizes (two-way anova; weeks 3–6: dispersal, F1,36 = 0·04, = 0·84; size class, F5,36 = 0·89, = 0·50; dispersal × size class, F5,36 = 1·31, = 0·28; weeks 5–6: dispersal, F1,36 = 0·05, = 0·82; size class, F5,36 = 1·37, = 0·26; dispersal × size class, F5,36 = 0·70, = 0·62, Fig. 4b). Further, mean body size in Gyrinus communities was unaffected by dispersal rate (Table S3, Supporting information). In contrast to Gyrinus communities, there was a significant interaction between dispersal rate and size class on size distributions in Notonecta communities (two-way anova; weeks 3–6: dispersal, F1,36 = 0·32, = 0·57; size class, F5,36 = 5·92, < 0·001; dispersal × size class, F5,36 = 5·43, < 0.001; weeks 5–6: dispersal, F1,36 = 0·08, = 0·78; size class, F5,36 = 8·52, < 0·001; dispersal × size class, F5,36 = 5·54, < 0·001; Fig. 4c). Generally, there were fewer individuals in the larger size classes under Notonecta predation. At high dispersal, however, more individuals were maintained in the large size classes. Thus, the high prey dispersal rate significantly increased mean body size in Notonecta communities over weeks 3–6, and marginally increased body size over weeks 5–6 (Table S3, Supporting information). In communities under cyclical Gyrinus-Notonecta predation, there was a marginal effect of body size class on size distributions over weeks 5–6, but there was no effect of dispersal rate (two-way anova; weeks 3–6: dispersal, F1,36 = 0·05, = 0·83; size class, F5,36 = 1·64, = 0·17; dispersal × size class, F5,36 = 0·39, = 0·85; weeks 5–6: dispersal, F1,36 = 0·07, = 0·80; size class, F5,36 = 2·38, = 0·06; dispersal × size class, F5,36 = 0·39, = 0·85; Fig. 4d). Similar to Notonecta communities, there were fewer individuals represented in the larger size classes. Dispersal rate did not impact mean prey size in cyclical predation communities (Table S3, Supporting information).

Figure 4.

 Prey size structure in predator-prey metacommunities. (a) Proportion of individuals in each log10 zooplankton body size class (mm) in low prey dispersal (black bars) and high prey dispersal (gray bars) regions. (b) Proportion of individuals in each body size class under Gyrinus sp. (c) Notonecta ungulata and (d) Gyrinus sp.-Notonecta ungulata predation. The log −0·3 and 0·0 mm size classes were not represented by any species. Values are mean sampling weeks 5 and 6 + 1 SE, = 4. Effects of body size class, S, and prey dispersal rate, D, on size distribution. Statistical significance: + < 0·1; *< 0·05; NS, not significant. Statistical analysis reported in the results.


Prey dispersal rate altered prey species composition and tempered the relative impacts of predation on prey community structure and trait diversity in pond metacommunities. Low dispersal metacommunities maintained greater regional prey species diversity and evenness relative to high dispersal metacommunities. Only weak negative effects of high dispersal were detected on local species diversity, however. Although there was no influence of predation regime on prey diversity in the metacommunity, predator identity impacted prey composition in local communities under constant predation. Likewise, predator identity affected the structure of the regional prey trait pool (body sizes), while dispersal rate controlled local prey trait diversity in response to selective predation. Taken together, these findings highlight the important effects of prey dispersal rates in moderating the strength of predation to influence prey species diversity and trait composition in metacommunities supporting ecologically different predators.

Regional and local prey diversity

Regional prey diversity was negatively affected by dispersal rate via a less equitable distribution of prey species at high dispersal, and not through substantial dispersal-induced declines in regional richness (Fig. 1). This result runs counter to metacommunity theory for a single trophic level which predicts that regional richness will strongly decrease at high dispersal rates from regional homogenization and dominance by a regionally superior competitor, but is in alignment with this same theory that suggests evenness will be reduced at high dispersal (Mouquet & Loreau 2003). The lack of regional richness response to dispersal rate is in accord with regional richness patterns produced in multi-trophic metacommunities where vertical structuring (Kneitel & Miller 2003) and patch-type heterogeneity (Howeth & Leibold 2010) maintained richness at high dispersal; thereby suggesting that regional coexistence of multiple predators may be maintaining richness and preventing global competitive exclusion in this experiment.

At the local spatial scale, high dispersal prey communities were marginally less diverse than low dispersal prey communities, yet there was no effect of dispersal rate on local richness or evenness at the end of the experiment. Low densities and local extinction of Bosmina and Chydorus at high dispersal was primarily responsible for the marginally significant reduction in diversity (Table 2; Fig. S2, Supporting information). The response of these two small-bodied grazer species at high dispersal may have been a result of increased exploitative competition with large-bodied grazers such as Daphnia, which are often superior resource competitors (Leibold 1996). Based upon temporal density dynamics, it is not apparent whether source-sink relations were operating in Bosmina and Chydorus at the low dispersal rate to sustain higher densities. A result of lower diversity in high dispersal communities is predicted by competitive metacommunity theory (Mouquet & Loreau 2002, 2003) and is congruent with findings of other metacommunity studies (reviewed in Cadotte 2006). While dispersal rate marginally influenced local prey species diversity, there was no effect of predation regime in the three local communities. Lack of detectable predation effects suggests that the rates of dispersal were high enough to override impacts of multiple predators in the metacommunity. There was, however, some influence of predator identity on community structuring as revealed through differences in prey species densities maintained in communities under constant Gyrinus and Notonecta predation.

Spatial and temporal prey refuges

Dispersal rate modified the impacts of constant predation from functionally dissimilar invertebrate predators in the spatial prey refuge (Fig. 3). The low prey dispersal rate in the metacommunity resulted in a more effective spatial refuge by maintaining two distinct community assemblages under ecologically different predators. The observed differential density responses of prey species, notably Chydorus, Daphnia and Scapholeberis, to constant Gyrinus and Notonecta predation at low dispersal suggests that the low dispersal rate facilitated species sorting. High dispersal may have homogenized species densities through a degree of mass effects, which is supported by the relatively uniform densities of these three prey species in the spatial refuge (Fig. 3). This regional mixing at high dispersal contrasts with evidence for species sorting in response to constant predation that has been demonstrated in pond zooplankton communities experiencing rates of immigration that were comparable to, or exceeded, the high dispersal rate in this study (Cottenie & De Meester 2004; Howeth & Leibold 2008). However, these metacommunity experiments evaluated species sorting in response to fish predators (Lepomis) that are both more effective and perhaps more strongly size-selective. In this study, local invertebrate predation pressure was not as strong of a community structuring agent relative to the high regional prey dispersal rate.

The temporal refuge under cyclical predation precluded prey tracking of predator identity and did not yield distinct prey assemblages through time at either the low or high prey dispersal rate. The local communities were homogenized by the tempo of alternating predation and the rates of prey dispersal, thereby preventing species sorting through time from occurring. Dispersal rates in the cyclical predation communities were higher than those in the constant predation communities given the order of dispersal events in the metacommunity, and these elevated rates may have reduced the opportunity for prey sorting. Any such effects, however, were likely to be small compared to the differences between dispersal treatments. The influence of temporal refuges on prey incidence and density will depend in part upon the time between predator pulses (Sih 1997) and the generation time of prey relative to prey immigration rate. The 2-week predator pulse cycle encompassed c. 1·5 generations of zooplankton prey and may have favoured the smaller-bodied species with shorter turnover times. Thus, predator pulses on longer time scales might yield contrasting local prey responses by reducing the potential for synergistic predator effects (Sih, Englund & Wooster 1998; Schmitz 2007) and altering the strength of resource competition between prey species. Colonization-extinction events in natural Gyrinus and Notonecta metapopulations may indeed occur over greater time periods than those reflected in this experiment. Observational studies find that extinction and recolonization of these predators in ponds ensues on an annual basis, where extinction is largely driven by changes in habitat conditions (Svensson 1985; Briers & Warren 2000). Therefore, in natural pond zooplankton metacommunities, enhanced opportunities likely exist for species sorting to operate in prey with generation times shorter than 1 year and consequently for temporal refuges to support distinct prey assemblages.

Regional and local prey size structure

The frequency distribution of prey body sizes in local communities depended upon predator identity and prey dispersal rate. Constant Notonecta predation altered the local frequency of prey body sizes, yielding a right-skewed distribution reflective of strong selective predation on larger-bodied prey (Fig. 4). The size-selective Notonecta predation significantly depressed Daphnia densities, and marginally depressed Simocephalus densities (Table 2). The negative impact of Notonecta predation on large-bodied prey is consistent with previous findings of Notonecta consumptive effects on zooplankton communities, and on Daphnia in particular (Scott & Murdoch 1983; Murdoch, Scott & Ebsworth 1984; Arner et al. 1998). The effect of selective Notonecta predation in the larger prey size classes was attenuated in the temporal refuge by alternating Notonecta and Gyrinus predation. The effects of size-selective Notonecta predation on local prey body size distributions scaled to the metacommunity, where there were fewer individuals represented in the larger size classes at both low and high prey dispersal. This impact of local predator-prey interactions on regional trait distributions has implications for the role of species sorting in shaping the frequency of prey traits in the regional species pool. Perhaps most importantly, however, the selective predation pressure in Notonecta communities was muted by high prey immigration with a consequent shift in the relative proportion, but not the range, of body sizes in the community. This dampening effect on the larger size classes resulted in a significant increase in mean prey size in high dispersal Notonecta communities. A dispersal rate-induced change in mean trait value emphasizes that regional dispersal can control local trait diversity in metacommunities experiencing heterogeneous ecological selection.


This is one of the first studies to empirically demonstrate that prey dispersal rate can moderate predation pressure and influence the local frequency distribution of ecologically important prey traits in metacommunities consisting of multiple predators. The low rate of prey dispersal maintained the greatest species diversity at regional spatial scales in metacommunities. Prey dispersal rate additionally determined the effectiveness of the prey refuges, where species and trait composition differences between the two constant predation communities were greater at the low dispersal rate from species sorting. Although the metacommunity was not fully homogenized by high dispersal, enough global mixing occurred to dampen the negative effects of selective local predation on prey species and trait densities in the communities supporting constant predation. The results emphasize the importance of the spatio-temporal distribution of multiple, functionally distinct predators on prey metacommunity structure. They further highlight that prey dispersal rate in the metacommunity has the potential to alter the diversity of the regional prey species and trait pool.


A. Downing, E. Garcia, P. Geddes and G. Mittelbach generously assisted with field work and logistics. We thank P. Amarasekare, T. Keitt and S. Sarkar for comments on earlier versions of the manuscript. Funding was provided by NSF DEB 0235579, 0640302 and 0717370 to M. Leibold, and a doctoral dissertation improvement grant NSF DEB 0508068 to J. Howeth. This is Kellogg Biological Station contribution number 1561.