Intermediate-consumer identity and resources alter a food web with omnivory

Authors


Department of Biological Sciences, California State University, Sacramento, 6000 J Street, Sacramento, CA 95819-6077, USA. Tel.: (916)278 3633. Fax: (916)278 6993. E-mail: kneitel@csus.edu

Summary

  • 1Omnivory is an important interaction that has been the centre of numerous theoretical and empirical studies in recent years. Most of these studies examine the conditions necessary for coexistence between an omnivore and an intermediate consumer. Trait variation in ecological interactions (competition and predator tolerance) among intermediate consumers has not been considered in previous empirical studies despite the evidence that variation in species-specific traits can have important community-level effects.
  • 2I conducted a multifactorial microcosm experiment using species from the Sarracenia purpurea phytotelmata community, organisms that inhabit the water collected within its modified leaves. The basal trophic level consisted of bacterial decomposers, the second trophic level (intermediate consumers) consisted of protozoa and rotifers, and the third trophic level (omnivore) were larvae of the pitcher plant mosquito Wyeomyia smithii. Trophic level number (1, 2 and 3), resources (low and high), omnivore density (low and high) and intermediate consumer (monoculture of five protozoa and rotifers) identity were manipulated. Abundance of the basal trophic level, intermediate consumers, and growth of the omnivore were measured, as well as time to extinction (intermediate consumers) and time to pupation (mosquito larvae).
  • 3The presence of different intermediate consumers affected both bacteria abundance and omnivore growth. At high resource levels, Poteriochromonas, Colpidium and Habrotrocha rosa reduced bacteria densities greater than omnivore reduction of bacteria. Mosquito larvae did not pupate at low resource levels except when Poteriochromonas and Colopoda were present as intermediate consumers. Communities with H. rosa were the only ones consistent with the prediction that omnivores should exclude intermediate consumers at high resources.
  • 4These results had mixed support for predictions from omnivory food web theory. Intermediate consumers responded and affected this community differently under different community structures and resource levels. Consequently, variation in species-specific traits can have important population- and community-level effects and needs to be considered in food webs with omnivory.

Introduction

Trait differences among species in food webs can have numerous implications for communities, including species coexistence and abundance patterns (Hunter & Price 1992; Leibold 1996; Chase, Leibold, & Simms 2000). Empirical and theoretical studies indicate that these differences are ecologically important in a variety of systems (Leibold 1989; McPeek 1998; Ritchie & Olff 1998; Steiner 2003; Kneitel & Chase 2004). For example, differences in competitive ability and predator tolerance among species within a trophic level can affect coexistence within the trophic level, abundances at multiple trophic levels, and the potential for trophic cascades (e.g. Leibold 1989; Kneitel & Miller 2002; Kneitel & Chase 2004). However, it remains unclear how trait differences will alter food webs with omnivory.

Omnivory, feeding on more than one trophic level, has been increasingly incorporated into our understanding of food webs and community structure (Fig. 1; Menge & Sutherland 1976; Polis & Strong 1996; Holt & Polis 1997; Morin 1999; Diehl & Feissel 2000; Mylius et al. 2001). Theoretical studies focus on the conditions that allow species coexistence between the intermediate consumers and the omnivore (Fig. 1). Three major predictions have been made concerning patterns of abundance and coexistence in food webs (Holt & Polis 1997; Diehl & Feissel 2000; Mylius et al. 2001). First, coexistence between the omnivore and an intermediate consumer is possible only if the intermediate consumer is a better competitor for the basal prey. Second, the presence of an intermediate consumer can facilitate or inhibit the growth of the omnivore depending on the resource level (Holt & Polis 1997; Diehl & Feissel 2000; Mylius et al. 2001). At low resource levels, the omnivore is expected to go extinct because of resource competition, and at high resource levels, the intermediate consumer is predicted to be driven to extinction by the omnivore. Finally, as the omnivore consumes the intermediate consumer, the intermediate consumers should be less abundant in the presence of an omnivore (Diehl & Feissel 2000). Given these narrow sets of traits for coexistence, it is surprising that many food webs with omnivory exhibit coexistence (e.g. Kneitel & Miller 2002; HilleRisLambers, van de Koppel, & Herman 2006). One explanation is that these models address a single intermediate consumer, while species with different traits typically occur in the intermediate trophic level (e.g. Hunter & Price 1992; Holt, Grover & Tilman 1994; Leibold 1996).

Figure 1.

Simplified structure and feeding relationships in the pitcher-plant food web.

The inquiline community that occurs inside the leaves of Sarracenia purpurea has been well described in a number of studies (Addicott 1974; Heard 1994; Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002; Miller & Kneitel 2005). The modified pitcher leaves collect rainfall that then acts as a pitfall trap for insects and other invertebrates, the primary energy input to the system. Over 90% of the captured prey in Florida is the fire ant Solenopsis invicta (Kneitel & Miller 2002). The drowned prey is then broken down by bacteria, which make up the basal trophic level and are consumed by the rest of the community (Fig. 1). Numerous species of protozoans and rotifers make up the middle trophic level, and the most common top-predator (omnivore) is the larvae of the pitcher-plant mosquito Wyeomyia smithii that filter feeds on bacteria, protozoa and rotifers (Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002). The species of the middle trophic level share a common prey (bacteria) and predator (mosquito larvae), but it is unclear how the combination of resources (dead ants) and protozoan species identity will affect mosquito larvae growth and bacterial abundance.

Characteristics of the S. purpurea community are somewhat different from the structure of standard omnivory models. First, this community departs from the assumptions of many omnivory models that use per capita numerical responses (Holt & Polis 1997; Diehl & Feissel 2000). Mosquitoes, like many inhabitants of aquatic systems, have a complex life cycle where only a portion of their life cycle is spent in that habitat. Therefore, individual growth rate, survival and time to pupation can be used as surrogates of per capita growth rate (Morin 1984, 1987; McPeek 1998; Yee & Juliano 2006), which should respond similarly to standard model predictions. Second, variation in traits among the intermediate consumers, such as the trade-offs among competitive ability and predator tolerance in this system (Cochran-Stafira & von Ende 1998; Kneitel & Chase 2004; Miller & Kneitel 2005), may result in a departure from standard predictions. Food webs with intermediate consumers that are strong resource competitors and somewhat tolerant of predation should reflect most of the predictions from omnivory models. As strong competitors in the S. purpurea system are predator intolerant (Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002; Kneitel 2002), one departure from these models is that they are more likely to be driven to extinction in the presence of omnivores at low and high resource levels. Predator tolerant intermediate consumers in this system have higher per capita growth rates allowing them to overcome predator consumption rates (Kneitel 2002). When predator tolerant species (poor competitors) are present, two predictions arise: (1) at low resource levels, they are less likely to be driven to extinction and their presence should facilitate the omnivore's growth (rather than inhibit it), and (2) at high resource levels, they are expected to coexist with the omnivore (rather then meeting a fateful extinction).

This article describes a multifactorial microcosm experiment that examined protozoan and rotifer species responses to resources and predators, and how the presence of each of these intermediate consumers directly affected abundance levels of the basal trophic level and the growth rate of the omnivore. Examining this increased complexity in food webs and the role of heterogeneity of ecological traits will further our understanding of community structure, species interactions, and regulation in communities with omnivory (Hunter & Price 1992; Leibold 1996; Polis & Strong 1996). Specifically, four hypotheses were tested with organisms that inhabit the Sarracenia purpurea inquiline community. First, intermediate consumers are better competitors than the omnivore for shared resources. Second, intermediate consumers can affect omnivore growth differentially at different resource levels. Third, intermediate consumers are less abundant in the presence of the omnivore. Fourth, species-specific intermediate consumers (that exhibit trade-offs between competitive ability and predator tolerance) can affect the coexistence between omnivores and intermediate consumers.

Methods

experimental system

Protozoan and rotifer species that inhabit the leaves of the pitcher plant Sarracenia purpurea have been studied in both laboratory and field settings (Fig. 1; Addicott 1974; Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002, 2003; Miller, Kneitel & Burns 2002). These species are affected by resource density and predation pressure (larvae of the pitcher-plant mosquito Wyeomyia smithii) (Addicott 1974; Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002). Several protozoan species have been found to be tolerant of predation, but were weaker resource competitors (e.g. Bodo and Poterioochromonas; Cochran-Stafira & von Ende 1998; Kneitel 2002). Other species have been found to be better resource competitors, but more vulnerable to predation (Colpoda, Colpidium and Habrotrocha rosa; Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002; Kneitel 2002). Resource densities have also been found to have strong effects on protozoan and bacteria populations (Kneitel & Miller 2002), as well as mosquito larvae growth (Istock, Wasserman & Zimmer 1975).

The five most common species were the focal species of this study: the protozoans Bodo, Colpidium, Colpoda and Poterioochroma, and the pitcher obligate rotifer, Habrotrocha rosa. All species are heterotrophic with the exception of the mixotroph, Poterioochroma. Prior to the experiments described below, these species and a population of W. smithii, were isolated from field samples taken from a S. purpurea population growing in the Crystal Bog, approximately 4 km north of Wilma, Florida, in the Apalachicola National Forest. Protozoa and rotifers were kept in 50-mL vials that contained 30 mL of sterile water and approximately 20 dead sterilized ants; species’ monocultures were transferred weekly to fresh media.

experimental design

The basic design for this laboratory experiment consisted of four communities (Fig. 2) having 1, 2 or 3 trophic levels: (1) bacteria were grown alone; (2) bacteria grown with monocultures of each of five protozoan and rotifer species; (3) bacteria grown with two densities of Wyeomyia smithii larvae [low (one larvae) and high (three larvae)]; and (4) bacteria grown with a monoculture of each protozoan and rotifer species and the two densities of Wyeomyia smithii larvae. Each of the communities were grown at low (five ants) or high (25 ants) resource levels. Four replicates of each treatment were established for each of the 36 community types in 20 mL of medium (deionized water with a bacterial inoculum, see below) in 50 mL vials. The amount of medium approximated the volume found in the field (mean = 21·1 mL, SE = 7·9). Resource levels reflected ambient densities in the field, but predator levels were somewhat lower than that found naturally. Higher larvae densities result in all protozoan and rotifer species being driven to extinction quickly in vials (Kneitel, unpublished data). A survey of published and unpublished data finds mosquito larvae presence and abundance patterns highly variable within sites [n = 13, % pitchers occupied = 75·2 (SE = 20·6), mean abundance/pitcher = 12·6 (SE = 6·8), and range = 0–63·7 (SE = 30·6); Addicott 1974, Heard 1994, Miller et al. 1994, Harvey & Miller 1996, Kneitel & Miller 2002; Kneitel 2002]. Similar patterns of resource variation have been found in this system (mean = 8·9 ants/pitcher, SD = 3·5; Kneitel & Miller 2002; J. M. Kneitel unpublished data, T. E. Miller unpublished data).

Figure 2.

Experimental design of experiment. There were four communities and each were inoculated with a pitcher-plant bacteria culture. Each community with protozoa contained a monoculture of one of the protozoa and rotifer species. Communities with larvae contained low and high densities and all communities were grown at low and high resource levels. Arrows refer to the direction of consumption.

Bacteria from natural pitcher-plant samples were used to inoculate treatments using a single drop (approximately 0·05 mL) and 12 h were allowed for populations to establish. Approximately 500 individuals of each protozoan and rotifer species were inoculated into treatments. Bacteria also entered into the communities with protozoa, but bacterial species composition did not differ among treatments (Kneitel, unpublished data). Wyeomyia smithii larvae were washed repeatedly and placed in sterile water for 24 h prior to initiating treatment vials. The water was checked to ensure prevention of protozoan contamination. Larvae were randomly chosen for treatments with no initial size differences among treatments (mean = 2·41 mm, SD = 0·56 mm; anova: larval density, F1,188 = 0·158, P = 0·69; resource, F1,188 = 0·58, P = 0·45; protozoan species, F5,186 = 1·44, P = 0·21).

experimental sampling

Protozoa and bacteria were sampled after 24 h and then every 3 days for 21 days. Sampling also occurred on days 25 and 35 where I sampled a larger volume (1 mL) for protozoa to determine if protozoan and rotifer species went extinct. This experimental length is sufficient to capture equilibrium responses to treatments (McGrady-Steed & Morin 1996; Kneitel & Miller 2002). Protozoan abundance was sampled by taking 0·1 mL samples and counting densities in a Palmer counting cell; H. rosa was sampled using 1 mL in a Sedgewick rafter cell. The same amount of water taken for samples was replaced with sterile water. Abundance (individuals per mL) and time to extinction were followed in all replicates and used as dependent variables. Care was taken during sampling to avoid removal of ant resources, but the removal of bacteria and intermediate consumers without replacement may have affected productivity over the course of the experiment. As the same methods were used in sampling all replicates, reduction in productivity would have occurred across all treatments and over time. None the less, care should be taken in interpreting resource effects.

Bacterial abundance was sampled by gently shaking each vial, then removing a 0·05 mL sample. This sample was then serially diluted with sterile water, and 0·1 mL of a 10−5 dilution was then spread on Luria-broth agar plates (Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002). These plates were incubated in a growth chamber at 26 °C for 48 h, and abundance was measured as the number of colonies growing on the plate. To test the hypothesis that the intermediate consumers were better competitors for the basal prey than the omnivore, I compared each intermediate species effect on bacteria (mean over time) in the absence of mosquito larvae with the high-density larvae effect (mean over time) in the absence of intermediate consumers. It was logistically difficult to tease apart the relative effects of predation and competition in the intermediate consumer–omnivore treatments. Therefore, relative competitive ability was assumed to be a function of greater reduction in bacterial abundance levels.

Mosquito larvae were measured by photographing individuals and measuring their length (mm) using the public domain NIH Image program (developed at the US National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image). Individuals were measured prior to experimental initiation and then every 7 days for 21 days. Growth rates were measured as the slope of larval length (mm) increase over time. Five individuals died over the course of the experiment: larvae were replaced in those replicates, but were not used in the analyses. Time to pupation was also measured in all treatments. While sampling for the rest of the community ended after 35 days, mosquito larvae were maintained to determine if pupation would occur. Those larvae not pupated by 60 days were likely not to pupate (Istock et al. 1975) and were recorded as 60 days.

statistical analyses

Three-way anovas were conducted on total bacterial abundance, mosquito larvae growth rates (mm), and larval time to pupation (days). Predator density, resource levels and intermediate consumer identity were used as the independent variables. Post-hoc tests were conducted comparing bacterial abundance in monoculture treatments vs. mosquito larvae treatments at low and high resource levels. Two-way repeated measures anovas were conducted on protozoan abundances and time to extinction using predator densities and resource levels as the independent variables. Bacterial abundance and each of the intermediate consumer abundances were the dependent variables that were log transformed to correct for non-normality. Where appropriate, Tukey–Kramer post-hoc tests were conducted on individual treatment effects.

Results

bacterial abundance

Bacterial abundance was generally between 105 and 107 cells per mL, similar to values found in natural pitcher plants (Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002). Total bacterial abundance was affected by mosquito larvae density, resource addition, and intermediate-species identity (Table 1). Resources increased bacteria abundance and larval density decreased bacteria abundance (Fig. 3). Individual protozoan and rotifer species decreased bacterial abundance to varying degrees. Bodo, Poterioochromonas, Colpidium and H. rosa significantly decreased bacterial abundance compared with the control, and Colpoda was the only species that had no significant effect (Table 1, Fig. 3). Intermediate consumer effects on bacteria compared with mosquito larvae varied among resource levels. None of these comparisons were significantly different at low resources. At high resources, Poterioochroma (F = 7·43, P = 0·01), Colpidium (F = 15·77, P < 0·001) and H. rosa (F = 10·73, P = 0·002) decreased bacterial abundance significantly more than mosquito larvae. There were no significant interactions among any of the treatments. Lack of a significant species identity–larval density interaction indicated a lack of any trophic cascades.

Table 1. F-values from the results of three-way anova on bacterial abundance (mean over time), Wyeomyia smithii growth rate, and days to pupation
SourceBacterial abundanceLarval growth rate (mm day−1)Larval time to pupation (days)
  1. *P < 0·05, **P < 0·01, ***P < 0·001. Letters in parentheses indicate direction of significant effect: D = decrease and I = increase.

Omnivore density (d.f. = 1)12·01*** (D) 1·24 27·28*** (I)
Resources (d.f. = 1)98·56*** (I)18·87*** (I)174·54*** (D)
Species identity (d.f. = 5)18·06*** (D) 2·10  2·46* (D)
Omnivore × resource 1·41 3·35 40·29***
Omnivore × species 0·76 0·89  1·63
Resource × species 1·2 0·43  1·96
Omnivore × resource × species 0·82 1·31  0·58
Figure 3.

Bacterial abundance (mean over time ± SD) in the presence of different consumers. Abbreviations refer to: Bo, Bodo; Po, Poterioochromonas; Ci, Colpidium; Co, Colpoda; Ro, Habrotrocha rosa (rotifer); and HLarv, high larval density (omnivore) without intermediate consumers.

protozoan abundance and extinction times

Protozoan species differed in their responses to the treatments (Table 2). Bodo time to extinction was not affected by the treatments, but abundance decreased with predator addition and increased with resource addition (Table 2; Fig. 4a). Poterioochromonas time to extinction and abundance was not affected by either resources or larval predation (Table 2; Fig. 4b). Colpidium and Colpoda responded similarly: abundance was positively affected by resource addition and negatively affected by larval density increases (Table 2). For both species, an interaction between larval density and resources resulted from a strong negative predator effect occurring only at low resources (Table 2; Fig. 4c,d). Time to extinction reflected these results, whereby, the high resource treatment allowed these species to persist with the omnivore. Similarly, H. rosa abundance significantly decreased with omnivore density; however, abundance was also negatively affected by increasing resources; in the presence of the omnivore or high resources, time to extinction occurred quickly (Table 2; Fig. 4e).

Table 2. F-values from the results from (a) two-way anova on days to extinction, and (b) repeated measures two-way anova on abundance, for each of the intermediate consumer species
SourceIntermediate consumers
BodoPoterioochromonasColpidiumColpodaH. rosa
  1. * P < 0·05, ** P < 0·01, *** P < 0·001. Letters in parentheses indicate direction of significant effect: D = decrease and I = increase.

(a) Time to extinction
Omnivore (d.f. = 2) 2·96 1·0040·92*** (D)41·27*** (D)38·42*** (D)
Resource (d.f. = 1) 0·97 1·0039·80*** (I)48·60*** (I)23·91*** (D)
Omnivore × resource (d.f. = 2) 0·16 1·0011·59***13·18***18·88***
(b) Abundance
Between subjects
Omnivore (d.f. = 1)16·86*** (D) 0·7034·05*** (D)36·58*** (D)10·08** (D)
Resource (d.f. = 1)23·49*** (I) 2·0774·47*** (I)77·39*** (I) 8·34** (D)
Omnivore × resource (d.f. = 1) 1·87 0·30 0·9410·17** 9·49**
Within subjects
Time34·19***21·91*** 0·35 2·65 4·61**
Time × omnivore 3·68* 2·82*10·85*** 3·02* 2·76
Time × resource19·27***24·07*** 2·94* 2·58 1·85
Time × omnivore × resource 1·50 2·66 4·98** 3·83* 2·15
Figure 4.

Mean abundance over time (± SE) of the five intermediate consumers in the six treatments.

wyeomyia smithii larvae growth

Growth rates (mm day−1) of W. smithii, the omnivore, increased in response to resource addition, but no other treatment affected growth in size significantly (Table 1). Time to pupation (another measure of larval growth) was strongly affected by all treatments (Table 1). In general, resource addition decreased the time to pupation. Increasing larval density (or increasing intraspecific competition) slowed the time to pupation, but an increase in resources reduced time to pupation at both larval densities, resulting in a significant larval density–resource interaction (Table 1). When the presence of each protozoa and rotifer species was individually compared with the control, the presence of the Poterioochromonas and Colpoda significantly reduced W. smithii time to pupation at low resource levels (Fig. 5).

Figure 5.

Mean time to pupation at high larval densities (± SD) in the presence of different protozoa species at different resource levels. Asterisks above protozoa species indicate they are different than the None treatment.

Discussion

The results of this study show that the presence of different intermediate consumers affect coexistence and the functioning of the omnivore food web. Three (Poterioochromonas, Colpidium and H. rosa) out of the five intermediate consumers reduced bacterial abundance levels more than the omnivore. The presence of two (Poterioochromonas and Colpoda) species at low resource levels facilitated growth in the omnivore grew slowly and seldom pupated. Most of the intermediate consumers’ abundance were somewhat suppressed by the omnivore at low resource levels, and three predator intolerant species (Colpidium, Colpoda and H. rosa) were driven to extinction. Overall, trait variation among the intermediate consumers resulted in varying responses to adjacent trophic levels (bacteria and mosquito larvae) and species coexistence.

Species’ trait variation within a trophic level can affect various aspects of food web dynamics, including patterns of abundance and coexistence (e.g. Leibold 1989, 1996; Kneitel & Chase 2004). Each of the intermediate consumers’ densities varied in response to resource addition and predator densities, which is consistent with previous studies (Cochran-Stafira & von Ende 1998; Kneitel & Miller 2002, 2003; Miller et al. 2002). In turn, the presence of these species differentially affected the basal resource (bacteria) and the omnivore (Wyeomyia smithii larvae) in this experimental food web, with some support for theoretical predictions (Holt & Polis 1997; Diehl & Feissel 2000).

Intermediate consumers are predicted to be superior resource competitors than the omnivore for coexistence in food webs (Holt & Polis 1997; Diehl & Feissel 2000). Relative competitive ability was assumed to be a function of bacterial abundance reduction, and the superior competitors (Colpoda, Colpidium and Habrotrocha rosa; Cochran-Stafira & von Ende 1998; Kneitel 2002) were expected to reduce these levels most. At low resource levels, larvae and intermediate consumers significantly reduced bacterial abundance compared with the control, but larvae and intermediate consumers did not differ in the magnitude of their effects. While there were no differences here, it should be noted that the intermediate consumer populations were sustainable, whereas mosquito larvae typically did not survive to pupation. At high resource levels, Poterioochromonas, Colpidium and Habrotrocha rosa significantly reduced bacteria to a greater degree than the omnivore, mosquito larvae. The absence of Colpoda and presence of Poterioochromonas may have differed from predictions because they had been evaluated in competition experiments with other protozoa and not relative to mosquito larvae (Cochran-Stafira & von Ende 1998; Kneitel 2002). None the less, three of the intermediate consumers were better resource competitors, reflecting the prerequisites for omnivore–intermediate consumer coexistence (Holt & Polis 1997; Diehl & Feissel 2000).

Omnivores are expected to have negative effects on intermediate consumers, and at high resource levels, the omnivore is expected to exclude the intermediate consumer (Holt & Polis 1997; Diehl & Feissel 2000). Most of the results were inconsistent with these predictions; most intermediate consumers coexisted with the omnivore at high resource levels. Morin (1999) similarly found that resource addition resulted in coexistence between an intermediate consumer and omnivore. In both studies, the high resources treatment may have been in the middle region of the full range, but at higher nutrient concentrations in both systems, anoxic conditions easily develop (Kneitel, unpublished data). Only one species, H. rosa, was driven to extinction at high resources. This is not surprising as this species is strongly predator intolerant (Kneitel & Miller 2002, 2003; Miller & Kneitel 2005), and again, the presence of strong resource competitors that are predator intolerant reflected the predictions from omnivory food-web models. However, the results were quite different in the low resource environment.

Three of the intermediate consumer species (Colpidium, Colpoda and H. rosa) were driven to extinction quickly in the presence of the omnivore at low resources. Further, Poterioochromonas abundance and time to extinction and Bodo time to extinction were not affected by the presence of mosquito larvae in this study. These results are inconsistent with previous theoretical predictions (Holt & Polis 1997; Diehl & Feissel 2000), but consistent with the hypotheses in the present study. Previous work in this system have found that these species tend to be more tolerant of mosquito-larvae predation (Kneitel & Miller 2002, 2003), but weaker competitors (Kneitel 2002) while Colpidium, Colpoda and H. rosa tend to be less tolerant of predation by mosquito larvae (Kneitel & Miller 2002, 2003; Kneitel 2002). Consequently, under low resource levels, species’ abilities to tolerate predators appear to be important for coexistence, and not competitive ability, contrary to models predictions.

Another related prediction is that omnivore growth can be facilitated or inhibited by the intermediate consumer (Diehl & Feissel 2000). Previous empirical work has shown that omnivores experience extinction in low resource environments, but the populations are sustained when intermediate consumers are present (Liess & Diehl 2006). This response has been attributed to a trade-off among two resources in feeding and growth efficiency (Morin 1999; Diehl & Feissel 2000; Liess & Diehl 2006). In the present study, the larval omnivore rarely survived to pupation at low resource levels, except when Colpoda or Poterioochromonas were present. This implies that both the presence and identity of the intermediate consumer are critical for the omnivore's survival and growth at low resources. The variation in W. smithii larval growth might have been a function of differences among intermediate consumers including growth rates, abundance levels, digestibility (Jeschke, Kopp & Tollrian 2002) and nutritional value (Saunders & Lewis 1988). In addition, Poterioochromonas is mixotrophic (heterotrophic and autotrophic), which may have contributed an added input of energy through photosynthesis.

Resource levels were an important factor in determining the abundance of bacteria, protozoa and rotifers, as well as growth rates and time to pupation for Wyeomyia smithii. Similar to previous studies in this system (Istock et al. 1975; Kneitel & Miller 2002; Miller et al. 2002), there were strong bottom-up effects: most of the intermediate consumer species were buffered from extinction by resource addition, and mosquito larvae had higher growth rates and reduced times to pupation. Their responses were quite different: mosquito larvae did not pupate (in the absence of intermediate consumers), and intermediate consumer species responses were highly variable (Table 2). These effects at high resources levels, however, may have also resulted from changes in bacterial quality or composition.

This microcosm experiment used intermediate consumers to understand the role of trophic-level heterogeneity on an omnivore food web, providing a glimpse into the variable outcomes resulting from different intermediate consumers. Heterogeneity within a trophic level is important in a number of aquatic and terrestrial systems (e.g. Leibold 1989; Huntly 1991; Ritchie & Olff 1998; Persson et al. 2001; Steiner 2003). Most food-web studies with omnivores, however, do not take into account how variation in community dynamics is generated by trait variation among intermediate consumers. The results of this study clearly indicate that resources and intermediate consumers with different ecological traits (competitive ability and predator resistance) will have important effects on both the basal trophic level and omnivore abundance. In addition, these results may also explain patterns of coexistence and abundance observed in the pitcher plant inquiline community (Addicott 1974; Kneitel & Miller 2002; Kneitel 2002). For example, the weak negative relationship between larvae and protozoan abundance may be explained, in part, by resource and species-specific trait variation (Addicott 1974; Kneitel & Miller 2002; Kneitel 2002).

Studies of food webs with omnivory have primarily examined the conditions that facilitated coexistence between an omnivore and an intermediate consumer. Recent studies have shown that in many systems, the interactions can be more complicated (e.g. Myulis et al. 2001; HilleRisLambers et al. 2006; Liess & Diehl 2006). The results of this study show that species identity affects patterns of abundance and coexistence in a food web with omnivory. This implies that future work needs to examine multiple intermediate-consumer species while taking into consideration the diversity of traits among those intermediate consumers. Further, theoretical studies should explicitly include trade-offs between competition and predator tolerance among intermediate consumers, a common feature of many ecosystems (Kneitel & Chase 2004). Finally, integrating these species-specific differences in intermediate consumer traits will enable ecologists to better understand the functioning of omnivory in food webs (Hunter & Price 1992; Polis & Strong 1996).

Acknowledgements

I am grateful for Brian Lawton's extraordinary assistance in the laboratory. I am also indebted to Tom Miller, Alice Winn, Jon Chase, David Hoekman, Peter van Zandt, and two reviewers for comments on earlier drafts. This study was funded by National Science Foundation grants DEB-0083617 and DEB-0091776 to T.E. Miller.

Ancillary