Oecologia

, Volume 149, Issue 2, pp 265–275

Spatial refuge from intraguild predation: implications for prey suppression and trophic cascades

Authors

    • Department of EntomologyUniversity of Maryland
    • Department of EntomologyWashington State University
  • Robert F. Denno
    • Department of EntomologyUniversity of Maryland
Community Ecology

DOI: 10.1007/s00442-006-0443-y

Cite this article as:
Finke, D.L. & Denno, R.F. Oecologia (2006) 149: 265. doi:10.1007/s00442-006-0443-y

Abstract

The ability of predators to elicit a trophic cascade with positive impacts on primary productivity may depend on the complexity of the habitat where the players interact. In structurally-simple habitats, trophic interactions among predators, such as intraguild predation, can diminish the cascading effects of a predator community on herbivore suppression and plant biomass. However, complex habitats may provide a spatial refuge for predators from intraguild predation, enhance the collective ability of multiple predator species to limit herbivore populations, and thus increase the overall strength of a trophic cascade on plant productivity. Using the community of terrestrial arthropods inhabiting Atlantic coastal salt marshes, this study examined the impact of predation by an assemblage of predators containing Pardosa wolf spiders, Grammonota web-building spiders, and Tytthus mirid bugs on herbivore populations (Prokelisia planthoppers) and on the biomass of Spartina cordgrass in simple (thatch-free) and complex (thatch-rich) vegetation. We found that complex-structured habitats enhanced planthopper suppression by the predator assemblage because habitats with thatch provided a refuge for predators from intraguild predation including cannibalism. The ultimate result of reduced antagonistic interactions among predator species and increased prey suppression was enhanced conductance of predator effects through the food web to positively impact primary producers. Behavioral observations in the laboratory confirmed that intraguild predation occurred in the simple, thatch-free habitat, and that the encounter and capture rates of intraguild prey by intraguild predators was diminished in the presence of thatch. On the other hand, there was no effect of thatch on the encounter and capture rates of herbivores by predators. The differential impact of thatch on the susceptibility of intraguild and herbivorous prey resulted in enhanced top-down effects in the thatch-rich habitat. Therefore, changes in habitat complexity can enhance trophic cascades by predator communities and positively impact productivity by moderating negative interactions among predators.

Keywords

Habitat complexityIntraguild predationMulti-trophic interactionsPredator diversitySalt marsh

Introduction

Complex food webs with high species diversity are thought to buffer communities against trophic cascades because enemy impacts often attenuate through a reticulate network of species interactions (Strong 1992; Polis and Strong 1996; McCann et al. 1998). Intraguild predators, by feeding from multiple trophic levels, contribute to the reticulate nature of the food web and thus can reduce the potential for predator effects to cascade down to lower trophic levels (Polis et al. 1989; McCann et al. 1998). Indeed, intraguild predation has been shown to weaken suppression of herbivore populations (Rosenheim et al. 1995; Snyder and Ives 2001; Finke and Denno 2002) with negative impacts on plant biomass (Snyder and Wise 2001; Finke and Denno 2004, 2005).

Understanding how trophic interactions are influenced by habitat structure and physical refuges could improve our understanding of the ability of predators to elicit trophic cascades. Structurally-complex habitats may provide a refuge for predators, weakening the intensity of antagonistic interactions among predators and enhancing prey suppression (Finke and Denno 2002; Corkum and Cronin 2004; Griffen and Byers 2006). Therefore, by reducing reticulate interactions among predators, habitat structure has the potential to release plants from herbivory and promote trophic cascades, even in species-rich ecosystems. However, this result may depend on the individual responses of predators and prey to habitat complexity, since structural complexity has also been shown to amplify negative interactions among predators and diminish combined effects on herbivores (Warfe and Barmuta 2004).

The ability of habitat complexity to buffer against trophic cascades by reducing intraguild predation and enhancing predator impacts on herbivores implies that a refuge exists for intraguild prey but not for herbivorous prey. Habitat complexity can differentially affect the behavior of intraguild predators, intraguild prey, and herbivorous prey, which could influence encounter probabilities, capture success, and ultimately impact predation rates (Savino and Stein 1982, 1989). Structurally-complex habitats have been shown to influence prey susceptibility to predation by modifying predator foraging behavior (Clark and Messina 1998; Grabowski 2004) and providing species-specific prey refuges (Eklov and Persson 1995). For example, in a simple habitat, relatively sedentary herbivores may more effectively avoid detection by top-predators than intraguild prey, which are commonly more active and thus more easily discovered (Perry and Pianka 1997; Rosenheim and Corbett 2003). By contrast, the sensory field of a top-predator may be obstructed in a complex-structured matrix providing relatively greater refuge for more mobile intraguild prey than less active herbivorous prey (Uetz 1979). Identifying predator and prey behaviors that contribute to the effectiveness of habitat complexity as a refuge from predation will help identify the types of systems in which cascading predator effects are likely to occur.

An investigation of the links between habitat complexity and the strength of trophic cascades would provide additional insight into important issues concerning the biological control of herbivorous insect pests. The simplification of habitats resulting from agricultural intensification could diminish the potential for diverse predator assemblages to control insect pest populations and increase yield via trophic cascades due to frequent intraguild predation (Rosenheim et al. 1995). However, complex habitats may mitigate such antagonistic predator–predator interactions and thus enhance overall pest suppression and positively impact plant yield. Therefore, control of economically important pests might be achieved through the targeted use of habitat modification in association with the manipulation of predator assemblages (Landis et al. 2000).

We investigated the potential for habitat complexity to mediate predator–predator interactions and the strength of cascading predator effects on basal resources using an assemblage of terrestrial arthropods inhabiting Atlantic coastal salt marshes. In this system, Prokelisia planthoppers are the most common herbivores and they are consumed by a variety of invertebrate predators including hunting spiders (Pardosa littoralis), web-building spiders (Grammonota trivittata), and mirid bugs (Tytthus vagus). This predator complex includes strong intraguild predators (the hunting spider) that readily consume all other predators, weak intraguild predators (the web-building spider) that occasionally consume other predators, and strict predators (the mirid bug) that do not engage in intraguild predation (Finke and Denno 2002, 2005; Denno et al. 2004). Therefore, the opportunity exists for a diversity of trophic interactions among predators. In addition, the structural complexity of the vegetation varies tremendously across the marsh due to the differential accumulation of leaf litter (thatch) (Denno et al. 1996).

Previous studies of this system have shown that in structurally-simple habitats an increase in predator diversity can negatively affect prey suppression and dampen the strength of a trophic cascade, due to the increased probability for intraguild predation when predator diversity is high (Finke and Denno 2004, 2005). However, the indirect effects of habitat structure on primary productivity were not assessed, even though the ability of habitat complexity to mediate trophic interactions among a sub-set of predators in this system has been documented. Specifically, the presence of thatch provides refuge for Tytthus mirid bugs from intraguild predation by Pardosa wolf spiders, increasing the combined impact of these predators on the planthopper population (Finke and Denno 2002). In addition, complex habitats with thatch intensify the impact of Pardosa predation on planthopper populations by enhancing the numerical response of spiders to planthopper prey (Döbel and Denno 1994), encouraging predator aggregation (Denno et al. 2002), diminishing cannibalism (Langellotto 2002), and increasing capture efficiency (Denno et al. 2002; Langellotto 2002).

The objective of this study was to determine if habitat complexity, by providing a refuge for multiple predators from intraguild predation, might enhance the collective ability of the predator complex to limit planthopper populations and thus increase the overall strength of the trophic cascade. Both factorial experiments and behavioral studies were conducted to evaluate the interactive effects of habitat complexity and the predator complex on lower trophic levels. This research aims to advance our knowledge of the complex interactions that occur between natural enemy assemblages and the habitats where they reside, and ultimately determine the impact of such interactions on the occurrence of trophic cascades.

Materials and methods

Study system

Research was conducted using the terrestrial food web associated with intertidal salt marshes along the Atlantic coast of North America. The perennial cordgrass Spartina alterniflora is the most abundant plant species found within the intertidal zone of mid-Atlantic marshes where it often grows as extensive monocultures (Redfield 1972; Denno et al. 1996). Within this zone, the structural complexity of Spartina varies tremendously with elevation due to differences in the frequency of tidal flooding, nutrient subsidy, and litter decay (Redfield 1972; Denno et al. 1996). One of the major contributors to variation in structural complexity along this elevational gradient is the accumulation of dead Spartina leaf litter (thatch) at higher elevations (Denno et al. 1996). As a result, the structural complexity of the Spartina vegetation varies across habitats from architecturally-complex meadows where thatch is abundant (200–500 g dry mass/m2) to the structurally-simple vegetation of mud flats that is devoid of thatch (Redfield 1972; Finke and Denno 2002).

Spartina cordgrass serves as the only host plant for the most abundant herbivores on the marsh, the phloem-feeding planthoppers Prokelisia dolus and P. marginata (Hemiptera: Delphacidae) (Cook and Denno 1994; Denno et al. 1996). Densities of these small insects (3 mm in body length) frequently exceed 1,000 adults/m2 (Denno et al. 2002). Prokelisia planthoppers are trivoltine on mid-Atlantic marshes, with peaks of adult abundance occurring in May, July, and September, and overwintering occurs during the nymphal stage (Denno et al. 1996). Prokelisia eggs are deposited within the Spartina leaf blades and hatch after 2 weeks. Nymphs pass through five instars before molting to adults (Denno et al. 1996).

Generalist wolf spiders (Araneae: Lycosidae), particularly Pardosa littoralis, are the major predators of planthopper nymphs and adults on mid-Atlantic coastal marshes (Döbel et al. 1990; Döbel and Denno 1994). Pardosa reaches a marsh-wide average of ∼300 individuals/m2 in late summer, and single spiders consume up to 70 planthoppers per 24 h (Döbel and Denno 1994). Pardosa spiders are known to aggregate in complex-structured habitats with large amounts of thatch, where densities reach >600 individuals/m2 (Döbel and Denno 1994; Langellotto 2002). Pardosa is also a documented intraguild predator of other common predators of planthoppers in the field, including the mirid bug Tytthus vagus (Hemiptera: Miridae) and the web-building spider Grammonota trivittata (Araneae: Linyphiidae) (Finke and Denno 2002; Denno et al. 2004). The mirid bug, which uses its piercing-sucking mouthparts to probe Spartina leaves for embedded planthopper eggs, is the most devastating natural enemy of planthopper eggs on mid-Atlantic salt marshes (Döbel and Denno 1994). Tytthus densities average ∼300 individuals/m2, but peak at >1,000 individuals/m2 (Finke and Denno 2002). Grammonota, the most abundant web-building spider on the marsh (200–1,500 individuals/m2; Denno et al. 2002), is also a predator of planthoppers adults and nymphs (Denno et al. 2004).

Intraguild predation by Pardosa is predominantly asymmetric in this system since both Tytthus and Grammonota consume mostly planthoppers (Finke 2005). Differences in the relative sizes of these predators likely contribute to the asymmetry in intraguild interactions, since the Pardosa wolf spider (7.0±0.5 mg wet mass) is on average many times larger than either Grammonota (1.11±0.4 mg wet mass) or Tytthus (0.48±0.02 mg wet mass) (Matsumura et al. 2004). However, size-structured predation does occur in spiders (Rypstra and Samu 2005), and the intraguild predation of smaller Pardosa individuals by larger Grammonota adults is possible.

Experimental design

Trophic cascades in simple and complex habitats

To assess the ability of a diverse predator assemblage to elicit a trophic cascade on Prokelisia planthopper population size and Spartina cordgrass biomass in simple and complex-structured habitats, a factorial manipulation of predator assemblage presence (a combination of Pardosa wolf spiders, Grammonota web-building spiders, and Tytthus mirid bugs present versus predators absent) and habitat complexity (thatch-free versus thatch-rich vegetation) was conducted in mesocosms. Each mesocosm contained ten greenhouse-reared Spartina culms (Environmental Concern, Saint Michael’s, Md., USA) transplanted into sand-filled pots (30 cm diameter, 0.07 m2) and caged within a clear plastic cylinder (cellulose butyrate, 22 cm diameter × 30 cm height) sunk into the sand and covered by a screened lid (0.6×0.6 mm holes, 85% light transmission). Mesocosms were divided among seven watering pools maintained in the laboratory under 1,000-W sodium-vapor lamps suspended 2 m above. Treatments were randomly assigned to mesocosms within watering pools in a randomized complete block design. All herbivores and predators for this experiment were obtained at our major study site in Tuckerton, Ocean Co., New Jersey, USA (for detailed site description, see Denno et al. 2002).

Habitat-complexity treatments were applied by placing either 25 or 0 g (dry weight) of field-collected Spartina thatch onto the bottom of each mesocosm, interdigitated among the live culms of Spartina to mimic the natural situation. On 23 August 2004, 20 field-collected planthoppers (285 adults of Prokelisia dolus/m2) were released into each mesocosm and predator treatments were established the following day. Predators were released at densities and stage/size-class mixes that approximated the natural community of predators in the field at the start of the study [5 Pardosa (71 individuals/m2), 10 Grammonota (142 individuals/m2), and 10 Tytthus (142 individuals/m2)]. In addition to the predator assemblage treatments (assemblage present versus absent), three treatments consisting of each predator species alone with planthoppers (Pardosa only, Grammonota only, or Tytthus only) and one treatment consisting of Spartina plants alone were also included. Each of these treatments was crossed with the presence of thatch. The single predator species treatments were included to account for any species-specific effects of habitat complexity on prey capture success that may occur independently of the presence of other predator species, and the plant-only treatment was included to account for any direct effects of the presence of thatch on plant biomass. Each of the 12 treatment combinations was replicated 7 times for a total of 84 mesocosms.

Predators were released into mesocosms in an additive treatment design (i.e. treatments with multiple predator species contained the summed number of individuals used in each of the single predator treatments). An additive design was used because the overall abundance of predators on the marsh increases with predator species richness (correlation between total predator abundance and predator species richness from 41 samples taken throughout the growing season in four major habitat types across the Tuckerton marsh, r2=0.598, P<0.001; unpublished data). In addition, although the additive design confounds the presence of interspecific interactions with overall predator density, an additive design controls for changes in intraspecific interactions and is thus an appropriate design for detecting the occurrence of intraguild predation (Sih et al. 1998, Joliffe 2000).

On 23 November 2004, after more than two planthopper generations, the size of the planthopper population, Spartina biomass, and predator density were assessed. Herbivore and predator densities were determined by visually counting all living planthoppers (nymphs and adults), Pardosa, Grammonota, and Tytthus in each mesocosm. Spartina biomass was determined by harvesting all live aboveground vegetation from each mesocosm and by drying the vegetation in an oven for 3 days at 55°C, and weighing it. Density of planthoppers, Spartina biomass, and number of predators per mesocosm were scaled up to units per m2.

The interactive effects of the predator assemblage and habitat complexity on planthopper suppression and cascading effects on Spartina biomass were determined by two-way analyses of variance (PROC MIXED) with predator presence (the assemblage present or absent), the structure of the habitat (thatch present or absent), and their interaction as class effects in the model. Block was modeled as a random source of variation where significant (SAS 2002). To verify that thatch did not have direct effects on Spartina productivity, the effect of habitat complexity on Spartina biomass in the absence of consumers (Spartina plants only) was determined by one-way analysis of variance with the presence or absence of thatch as the main class effect in the model and the block modeled as a random source of variation (SAS 2002). To assess any indirect effects of habitat complexity on planthopper population size or plant biomass mediated through changes in the capture rates of planthoppers by individual predator species, separate one-way analyses of variance were done on each response variable (planthopper population size or plant biomass) for each predator species independently (Pardosa only, Grammonota only, and Tytthus only) with the presence or absence of thatch as the main class effect in the model and block modeled as a random source of variation (SAS 2002).

The potential for habitat complexity to provide a refuge for predators from intraguild predation was assessed by comparing predator survival when interspecific predator–predator interactions such as intraguild predation were precluded (single predator species treatments which lacked heterospecific predators) to the survival of predators when interspecific interactions were possible (predator assemblage treatments where heterospecific predators were present) in thatch-rich and thatch-free habitats. A two-way analysis of variance on the total number of predators surviving at the end of the study was performed with heterospecific predator presence (pooled single predator species treatments versus predator assemblage), habitat complexity treatment (thatch absent or present), and their interaction as class variables in the model (PROC MIXED, SAS 2002). Due to the additive treatment design, initial predator abundance in the predator assemblage was equal to the combined number of predators released in each of the single predator species treatments. Therefore, to compare total predator survival in the single predator species treatments (interspecific interactions precluded) to that in the predator assemblage (interspecific interactions possible), total predator abundance in the single predator species treatment was calculated by summing the number of predators surviving in the Pardosa-only, Grammonota-only, and Tytthus-only replicates within each block. Separate two-way analyses of variance were also performed on the density of each individual predator species (Pardosa only, Grammonota only, or Tytthus only) remaining at the end of the experiment with heterospecific predator presence (the focal predator species alone in the single predator treatment versus the focal predator in combination with heterospecific predators in the predator assemblage), habitat complexity treatment (thatch present or absent), and their interaction as class variables (PROC MIXED, SAS 2002).

Means were compared by performing t-tests with Bonferroni adjustment of P-values to account for multiple comparisons. Data were log-transformed when necessary to meet assumptions of analysis of variance including normality and homogeneity of variances.

Effect of habitat complexity on prey capture, encounter and mobility: a laboratory assessment

To verify that habitat structure provides refuge for predators from intraguild predation and to assess the role of predator and prey behavior in this effect, behavioral observations of Pardosa wolf spiders (intraguild predator), Tytthus mirid bugs (intraguild prey), and Prokelisia planthoppers (herbivorous prey) in the presence and absence of thatch were conducted in laboratory mesocosms. We focused our behavioral observations on Pardosa, Tytthus, and Prokelisia because Tytthus are at high risk from intraguild predation by Pardosa, resulting in diminished Prokelisia planthopper suppression in the presence of both predators (Finke and Denno 2002), whereas the intraguild predation of Grammonota by Pardosa is less strong (Denno et al. 2004).

Mesocosms were sand-filled pots enclosed in a 30 cm high × 7.5 cm diameter plastic cellulose butyrate tube cage (0.0044 m2) containing one Spartina transplant. Habitat complexity treatments were applied to our mesocosms prior to the addition of invertebrates. Complex-habitat treatments were established by adding 3 g (dry mass) of field-collected thatch to each mesocosm, whereas simple habitats received none. Experimental communities were comprised of one Pardosa wolf spider, three Tytthus mirid bugs, and five Prokelisia planthoppers. Arthropods used for the behavioral observations were field-collected from Tuckerton, NJ and maintained on caged Spartina plants for ∼24 h until used in the trial. Treatment animals were released into the cage sequentially (first Prokelisia, then Tytthus, and finally Pardosa) and allowed at least 1 h of settling time.

On 15 August 2001, behaviors were recorded during a 15-min observational period for one focal individual of each of the three species present. For species where more than one individual was present, focal animals were chosen pseudorandomly as the third Prokelisia planthopper or the second Tytthus mirid bug to be seen by the observer following the settling period. Recorded behaviors included the number of captures of Tytthus and Prokelisia by the Pardosa spider, the number of encounters of Tytthus and Prokelisia with Pardosa, the mobility (total time spent walking) of Pardosa, Tytthus and Prokelisia, and the location of Pardosa. The encounter and capture rates of Prokelisia by Tytthus were not recorded since Tytthus are predators of planthopper eggs which are embedded within the leaf blade. The frequency and duration (to 0.1 s) of individual behaviors were recorded using a handheld computer (Psion Workabout, PsiWin) with event recording software (The Observer Mobile, Noldus). Treatments were conducted randomly throughout the day for a total of ten replications.

To determine whether the presence of thatch differentially influenced the susceptibility of intraguild prey and herbivorous prey to Pardosa predation and if susceptibility could be attributed to differences in prey behaviors, the interactive effects of prey identity (Prokelisia planthopper or Tytthus mirid bug) and habitat complexity (thatch present or absent) on the capture rate of prey by Pardosa (no. of captures/min), the encounter rate of prey with Pardosa (no. of encounters/min), and prey mobility (total time spent walking) were assessed by performing separate 2×2 ANOVAs (PROC MIXED, SAS 2002) for each response variable. To determine whether habitat complexity influenced prey susceptibility by modifying Pardosa foraging behavior, spider mobility (time spent walking) and location (time spent on the Spartina plant) were compared in the presence and absence of thatch by separate one-way ANOVAs (PROC MIXED, SAS 2002).

Results

Trophic cascades in simple and complex habitats

Habitat complexity enhanced the cascading effects of a diverse predator assemblage including intraguild predators on herbivore population size (significant predator assemblage by habitat complexity interaction; F1,20=5.51, P=0.0293) and plant productivity (significant predator assemblage by habitat complexity interaction; F1,22=6.55, P=0.0307). Specifically, when the predator assemblage was present, planthopper abundance was lower in the presence of thatch than when thatch was absent (Fig. 1a) and this effect cascaded down to positively impact Spartina biomass (Fig. 1b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0443-y/MediaObjects/442_2006_443_Fig1_HTML.gif
Fig. 1

The interactive effects of the absence (filled circle) or presence (open circle) of a diverse predator assemblage (5 Pardosa, 10 Grammonota, and 10 Tytthus bugs) and the structural complexity of the salt marsh habitat (thatch absent or present) on: a the density of Prokelisia dolus planthoppers and b the aboveground living biomass of Spartina (note log10 scale in both panels). The presence of thatch enhanced planthopper suppression by the predator assemblage (F1,20=5.51, P=0.0293) resulting in greater plant biomass (F1,22=6.55, P=0.0307). Means±1 SEM are shown (n=28)

This positive effect of habitat complexity on the cascading effects of the predator assemblage was the result of thatch diminishing antagonistic interactions among predators and not due to a direct effect of thatch on herbivores or plants or to the enhanced foraging success of individual predator species in the presence of thatch. The presence of thatch did not impact Spartina biomass in the absence of both herbivores and predators (F1, 6=0.14, P=0.7233), nor did thatch influence the suppression of planthopper populations by Pardosa, Grammonota, or Tytthus individually (F1,6=0.78, 0.31, 0.05, respectively; P>0.40) or their indirect effects on plant biomass (F1,6=0.10, 2.25, 0.40, respectively; P>0.18).

Habitat complexity diminished the overall occurrence of intraguild predation including cannibalism in the predator assemblage, resulting in greater total abundance of predators in the assemblage when thatch was present as compared to when thatch was absent (significant heterospecific predator presence by habitat complexity interaction; F1,24=4.30, P=0.0491) (Table 1a, Fig. 2a). Pardosa was not susceptible to intraguild predation by Tytthus, and predation of Pardosa by Grammonota was infrequent, therefore Pardosa’s survival was not significantly affected by the presence of other predator species or its interaction with habitat complexity (nonsignificant heterospecific presence by habitat complexity interaction; Table 1b, Fig. 2b). However, habitat complexity provided a refuge for Pardosa from a combination of cannibalism and intraguild predation from Grammonota, resulting in significantly greater Pardosa survival in the presence of thatch than in the absence of thatch (significant main effect of habitat complexity; Table 1b, Fig. 2b). Grammonota was susceptible to intraguild predation by Pardosa, resulting in lower Grammonota survival in the presence of other predator species than when alone (significant main effect of heterospecific presence; Table 1c, Fig. 2c). In addition, thatch provided a refuge for Grammonota from both Pardosa predation and cannibalism, as evidenced by its higher density in mesocosms with thatch than those without it (significant main effect of habitat complexity; Table 1b, Fig. 2c). Therefore, thatch enhanced Pardosa and Grammonota survival whether these predators were alone or in the presence of heterospecific predators. In single-predator species treatments this effect was due to a reduction in cannibalism, while both reduced cannibalism and intraguild predation likely contributed to the effect in mixed predator treatments. Surprisingly, the survival of Tytthus mirid bugs was not affected by the presence of heterospecific predators, habitat complexity, or their interaction in this study (Table 1d, Fig. 2d). This result was likely a phenological issue since the final Tytthus density was extremely low (<2 nymphs per mesocosm for all treatments) and only newly hatched first instar nymphs were recovered, suggesting that the majority of mirids were hidden within leaf tissue as eggs at the time that final Tytthus densities were assessed.
Table 1

Mixed model analysis of variance results for the effects of heterospecific predator presence (the focal predator species alone versus the focal predator in combination with the assemblage of other predator species) and habitat complexity (thatch absent or present) on the density of all predator species pooled, the hunting spider Pardosa, the web-building spider Grammonota, and the mirid egg predator Tytthus

Source of variation

df

F

P

Total predator density

 Heterospecific predator presence

1,24

17.50

0.0003

 Habitat complexity

1,24

2.77

0.1090

 Heterospecific presence × Habitat complexity

1,24

4.30

0.0491

Pardosa density

 Heterospecific predator presence

1,18

0.02

0.8799

 Habitat complexity

1,18

5.29

0.0337

 Heterospecific presence × Habitat complexity

1,18

2.84

0.1090

Grammonota density

 Heterospecific predator presence

1,18

67.46

0.0001

 Habitat complexity

1,18

18.51

0.0004

 Heterospecific presence × Habitat complexity

1,18

2.87

0.1073

Tytthus density

 Heterospecific predator presence

1,18

0.56

0.4657

 Habitat complexity

1,18

0.05

0.8226

 Heterospecific presence × Habitat complexity

1,18

1.96

0.1783

https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0443-y/MediaObjects/442_2006_443_Fig2_HTML.gif
Fig. 2

The interactive effect of the absence (filled circle) or presence (open circle) of heterospecific predators and habitat complexity (thatch absent or present) on the density (no. per m2) of: a total predators, bPardosa wolf spiders, cGrammonota web-building spiders, and dTytthus egg predators. Total predator density in the absence of heterospecific predators was calculated as the sum of the number of predators surviving in each of the three single predator-species treatments. There was a significant interactive effect of habitat complexity and the presence of the heterospecific predator species on total predator survival (F1, 24=4.30, P=0.049), suggesting a refuge from intraguild predation in complex habitats. Means±1 SEM are shown. Note the log10 scale in a and d

Effect of habitat complexity on prey capture, encounter and mobility: a laboratory assessment

Behavioral observations confirmed that intraguild predation occurred in this system and that the intensity of intraguild predation was mediated by the complexity of the habitat. Pardosa wolf spiders captured more Tytthus mirid bugs than Prokelisia planthoppers (significant main effect of prey identity; F1,16=36.66, P<0.001), and the presence of thatch diminished the capture rate of Tytthus, the intraguild prey, but not Prokelisia, the herbivorous prey (significant prey identity × habitat complexity interaction; F1,16=5.38, P=0.02) (Fig. 3a). The discrepancy in capture rates corresponded to differences in the encounter rates of prey with Pardosa. The wolf spider encountered Tytthus more often than Prokelisia (significant main effect of prey identity; F1,16=33.45, P<0.001), and the number of encounters with Tytthus declined in the presence of thatch, whereas the number of encounters with Prokelisia were unchanged (significant prey identity by habitat complexity interaction; F1,16=3.79, P=0.055) (Fig. 3b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00442-006-0443-y/MediaObjects/442_2006_443_Fig3_HTML.gif
Fig. 3

The interactive effects of prey identity (herbivorous Prokelisia planthopper, filled triangle or intraguild Tytthus, open triangle) and habitat complexity (thatch absent or present) on: a the capture rate of prey by the intraguild predator Pardosa (no. per min), b the encounter rate of prey by Pardosa (no. per min), and c prey mobility (total seconds spent walking). Behaviors of individuals in 20 experimental mesocosms were recorded over a 15-min observational period. Means±1 SEM are shown

Differences between prey in their susceptibility to predation and the ability of habitat complexity to provide refuge from predation were likely due to differences in prey mobility rather than Pardosa foraging behavior. Tytthus were more active than Prokelisia overall (significant main effect of prey identity; F1,16=55.97, P<0.001) (Fig. 3c), and were more likely than Prokelisia to diminish their activity in the presence of thatch (significant prey identity by habitat complexity interaction; F1,16=4.33, P=0.039) (Fig. 3c). There was no evidence that habitat complexity influenced the foraging behavior of Pardosa, since there was no difference in either spider mobility (F1,18=1.75, P=0.1899) (Table 2a) or the time the spider spent foraging on the plant (F1,18=0.34, P=0.5614) (Table 2b) in the presence versus absence of thatch.
Table 2

The impact of habitat complexity (thatch absent or present) on: the mobility of the wolf spider Pardosa (total time spent walking) and its location (time spent on the plant) during a 15-min observational period in mesocosms containing Tytthus bugs and Prokelisia planthoppers

Pardosa behavior

Mean±SE

F1,18

P

Time spent walking (s)

 Presence of thatch

 

1.75

0.1899

 Thatch absent

119.56±17.65

  

 Thatch present

159.76±19.35

  

Time spent on plant (s)

 Presence of thatch

 

0.34

0.5614

 Thatch absent

82.10±33.29

  

 Thatch present

52.33±23.70

  

Discussion

We found that complex-structured habitats enhanced planthopper suppression by an arthropod predator assemblage and positively affected Spartina cordgrass biomass (Fig. 1). This trophic cascade resulted from the moderating effects of habitat complexity on antagonistic predator–predator interactions (intraguild predation and cannibalism) and not from any direct effect of habitat complexity on planthopper population size or an indirect effect of habitat complexity on planthopper suppression by mediating the foraging success of individual predator species.

Total predator density was diminished when a mix of predator species was present compared to that when only a single species was present (pooled sum of single species treatments) (Table 1a, Fig. 2a), suggesting the occurrence of intraguild predation. Moreover, the adverse effect of heterospecific predator presence on total predator abundance was moderated in complex-structured habitats. The same response was not found when each predator species was considered individually. However, Grammonota and Pardosa both showed trends toward diminished susceptibility to intraguild predation in the complex habitat, indicating that the refuge effect on total predator survival was largely due to the combined effect of thatch on Pardosa and Grammonota survival. Had Tytthus been in an active and visible stage (nymphs or adults) rather than in the embedded egg stage at the time of the predator census, the interactive effects of thatch and the presence of heterospecific predators on total predator density (Fig. 2a) would likely have been even stronger. Notably, a significant heterospecific predator presence by habitat complexity interaction on the number of Tytthus surviving has been found repeatedly in previous experiments (e.g. F1,33=4.38, P=0.04; F1,36=21.36, P=0.0001; Finke and Denno 2002). Nonetheless, we still found a significant refuge for predators from intraguild predation in the thatch-rich habitat despite the non-representative result for Tytthus.

The abundance of Pardosa, averaged across predator diversity treatments, was significantly higher in the presence of thatch (Table 1b, Fig 2b). The increase in Pardosa’s abundance in the complex-structured habitat can be attributed to both a reduction in the intraguild predation of smaller Pardosa by Grammonota as well as cannibalism (Langellotto 2002), since Pardosa did not reproduce during the timeframe of this study. However, the effects of cannibalism were likely minimal in this case, since Pardosa’s abundance in the single predator species treatment was similar across habitat types. Despite the higher survival of Pardosa in the thatch-containing treatment overall (averaged across the single species and predator assemblage treatments), planthopper suppression was not affected by the presence of thatch when only Pardosa was present, perhaps due to the lack of a significant difference in Pardosa abundance in the single-species treatments (see Fig. 2b, comparison of black-filled circles). Previously, thatch was found to intensify the impact of Pardosa predation on planthopper populations in open field plots (Denno et al. 2002). In light of our results, this effect is likely attributable to the aggregation of Pardosa spiders in thatch-rich habitats rather than increased capture efficiency (Denno et al. 2002; Langellotto and Denno 2004), because immigration was precluded in the current mesocosm study.

Grammonota density was reduced in the presence of other predators, indicating that this web-building spider was susceptible to intraguild predation by Pardosa, as has been shown in other studies (Denno et al. 2004). However, intraguild predation and cannibalism were diminished and the survival of Grammonota was enhanced in the complex habitat with thatch (Table 1c, Fig. 2c). Because Grammonota did not reproduce during this study, the impact of thatch on Grammonota’s abundance was attributable to the spatial refuge provided by complex habitats. Other studies have shown that elevated web-spider densities in complex-structured habitats result from an increase web attachment sites (McNett and Rypstra 2000). Here, we found that a decrease in antagonistic interactions with other predators, specifically Pardosa, occurred in the presence of thatch, thus providing an alternative explanation for the accumulation of web-building spiders in complex-structured habitats.

Unexpectedly, the density of Tytthus did not differ across treatments (Table 1d, Fig. 2d), but treatment effects were likely masked in this experiment because most individuals were in the egg stage during the census. Previous studies have documented that Tytthus is extremely susceptible to intraguild predation by Pardosa (Finke and Denno 2002, 2003), and that it finds refuge from intraguild predation in thatch-rich habitats (Finke and Denno 2002), a finding that was confirmed by our behavioral observations.

Habitat complexity differentially affected predator–predator and predator–prey interactions, highlighting the importance of individual predator and prey behavior in mediating the refuge effect of thatch. Behavioral observations confirmed that intraguild predation is common in this system and that Pardosa spiders encounter and capture more Tytthus bugs than Prokelisia planthoppers overall (Fig. 3a, b). This result is likely due to the finding that mirid bugs are more active than planthoppers (Fig. 3c). Planthoppers are relatively sedentary herbivores which spend extended periods of time motionless with their feeding stylets inserted into cordgrass phloem. In contrast, Tytthus bugs are active foragers that scurry along leaves in search of planthopper eggs (D. Finke, personal observation). Importantly, thatch provided a refuge from predation for the intraguild prey Tytthus but not herbivorous planthopper prey. We found no evidence for changes in the foraging behavior of Pardosa in the presence of thatch (Table 2). Therefore, this differential refuge effect was attributable to the decreased amount of time that Tytthus spent walking in the thatch-rich habitat, perhaps due to a reduced foraging area since planthopper eggs are concentrated at the top of the plants in the presence of thatch (Hines 2004), thereby diminishing both encounter and capture rates by the visually-orienting Pardosa (Fig. 3).

Notably, the mobility of Prokelisia planthoppers, and thus susceptibility to predation, was unaffected by the presence of thatch (Fig. 3). Although we did not measure the behavioral responses of Grammonota to Pardosa under different habitat structures, encounter rates are likely reduced in the presence of thatch. For example, this web-building spider readily detects and rapidly retreats from Pardosa when approached (Denno et al. 2004; Matsumura et al. 2004), and the greater diversity of hiding sites provided by a thatchy habitat likely promotes its escape. Therefore, the mediating effect of habitat complexity on the intensity of predator–predator interactions and prey suppression likely depends on the foraging behavior and microhabitat use of the individual predator and prey species in the system (Rosenheim et al. 2004; Schmitz et al. 2004).

Theory predicting the dynamics of communities dominated by intraguild predation suggests that intraguild prey should be superior to intraguild predators in exploiting a shared prey resource (Holt and Polis 1997; Borer et al. 2003). In this context, our results suggest that an increase in habitat complexity will ultimately result in enhanced predator–predator interactions via exploitative competition, due to reduced rates of intraguild predation. As a result, the abundance of intraguild prey (Tytthus and Grammonota) should be relatively greater than that of the intraguild predator (Pardosa) in thatch-rich habitats. Indeed, a survey of natural marsh habitats documented an increase in the relative abundance of Grammonota web-building spiders (Denno et al. 2002) and Tytthus mirid bugs (Finke and Denno 2002) compared to Pardosa wolf spiders in complex habitats with thatch. Thus, the relative success of intraguild prey, the superior exploiters of the shared prey, should further contribute to the strengthened top-down control of herbivore populations in complex-structured habitats (Denno et al. 2005).

Here, we discovered that habitat complexity reduced the occurrence of reticulate interactions among predators and enhanced conductance of predator effects through the food web to primary producers. Specifically, the presence of thatch diminished intraguild predation and cannibalism and increased prey suppression by the arthropod–predator complex, resulting in greater Spartina cordgrass biomass due to relaxed herbivory by planthoppers. Greater top-down suppression in the complex habitat was due to the finding that the intraguild prey were better able to avoid predation in the presence of thatch whereas the herbivorous prey remained equally susceptible to predation irrespective of habitat-type. As a result, preserving the structural complexity of habitats, both natural and managed, could contribute to greater herbivore suppression and elevated rates of primary production by damping antagonistic interactions among natural enemies (Agrawal and Karban 1997; Roda et al. 2000; Norton et al. 2001; Finke and Denno 2002; Langellotto 2002; Corkum and Cronin 2004; Griffen and Byers 2006). However, the extent to which habitat complexity increases top-down effects likely depends on the specific behavioral responses of predators and prey to the structure of the habitats in which they reside (Finke and Denno 2002; Corkum and Cronin 2004; Warfe and Barmuta 2004; Griffen and Byers 2006).

Acknowledgments

This manuscript greatly benefited from the comments of Pedro Barbosa, James Cronin, Galen Dively, William Fagan, Charles Mitter, Oswald Schmitz, and two anonymous reviewers. This research was supported by National Science Foundation grants to R.F.D. (DEB-0313903) and D.L.F. (DEB-0410572). All experiments comply with current USA laws.

Copyright information

© Springer-Verlag 2006