Bottom-up effects of species diversity on the functioning and stability of food webs

Authors


Correspondence author. E-mail: anarwani@uvic.ca

Summary

1. The importance of species diversity for the stability of populations, communities and ecosystem functions is a central question in ecology.

2. Biodiversity experiments have shown that diversity can impact both the average and variability of stocks and rates at these levels of ecological organization in single trophic-level ecosystems. Whether these impacts hold in food webs and across trophic levels is still unclear.

3. We asked whether resource species diversity, community composition and consumer feeding selectivity in planktonic food webs impact the stability of resource or consumer populations, community biomass and ecosystem functions. We also tested the relative importance of resource diversity and community composition.

4. We found that resource diversity negatively affected resource population stability, but had no effect on consumer population stability, regardless of the consumer’s feeding selectivity. Resource diversity had positive effects on most ecosystem functions and their stability, including primary production, resource biomass and particulate carbon, nitrogen and phosphorus concentrations.

5. Community composition, however, generally explained more variance in population, community and ecosystem properties than species diversity per se. This result points to the importance of the outcomes of particular species interactions and individual species’ effect traits in determining food web properties and stability.

6. Among the stabilizing mechanisms tested, an increase in the average resource community biomass with increasing resource diversity had the greatest positive impact on stability.

7. Our results indicate that resource diversity and composition are generally important for the functioning and stability of whole food webs, but do not have straightforward impacts on consumer populations.

Introduction

The impact of species diversity on the stability of populations, communities and ecosystems has long been a central question in ecology (MacArthur 1955; May 1973; Tilman 1996). There are many definitions of ecological stability, and the influence of species diversity can depend on both the type of stability and the level of ecological organization being investigated (Tilman 1996; Steiner et al. 2005; Thébault & Loreau 2005; Ives & Carpenter 2007). Historically, concepts of stability have centred on mathematical notions of equilibrium (e.g. May 1973), in which a system is stable if it returns to equilibrium after a disturbance or if the rate of return is rapid (McCann 2000). Stability may also refer to the complexity of the dynamical behaviour, with high stability being conflated with a stable point equilibrium and low stability with chaos (e.g. McCann & Hastings 1997; Fussmann & Heber 2002). However, the prevalence of variable, non-equilibrium population dynamics in nature has led empiricists to resort to a less restrictive notion of stability (McCann 2000). In the last 15 years, empirical research on the topic has largely focused on the influence of diversity on temporal stability, defined as the inverse of temporal variability (or the coefficient of variation, CV) (Tilman 1996; Steiner et al. 2005; Tilman, Reich & Knops 2006; Vogt, Romanuk & Kolasa 2006; van Ruijven & Berendse 2007; Hector et al. 2010).

Defined in this way, diversity may have different impacts on the stability of populations and aggregate community properties (Lehman & Tilman 2000; Steiner 2001). Diversity may increase the stability (1/CV, or μ/σ) of aggregate community properties on a single trophic level either by increasing the mean total function (μ) or by reducing the variability of function (σ) over time (Lehman & Tilman 2000). Diversity can increase the total function (μ) (overyielding) via resource use complementarity, facilitation or a selection effect (Loreau & Hector 2001; Fox 2005) and/or decrease the variance (σ2) of the function by reducing the summed variances or summed covariances of the functioning of individual species. Increasing diversity will reduce the variability of an aggregate function relative to the case of a single species because of averaging among the species when species do not fluctuate synchronously. This is known as statistical averaging (Cottinham, Brown & Lennon 2001). Mean–variance scaling can also reduce the variability of an aggregate function when the variance of each species’ level of functioning scales with its mean as a power function with an exponent, z, >1 (i.e. the ‘Portfolio effect’, Doak et al. 1998; Tilman, Lehman & Bristow 1998). In this case, if community biomass remains constant as diversity increases, individual populations will have lower average biomasses and therefore variances (and summed variances) over time. Diversity can also reduce the variability of aggregate properties by reducing the summed covariances among populations (i.e. the ‘Insurance Hypothesis’, Yachi & Loreau 1999). This may occur when species are competing or responding uniquely or asynchronously to fluctuations in limiting factors in their environment. However, insurance effect is predicted to reduce the stability of individual populations as diversity increases (Lehman & Tilman 2000).

The majority of empirical studies on the influence of diversity on stability have focused on effects at a single trophic level (e.g. Tilman, Reich & Knops 2006; Hector et al. 2010), although the number of multitrophic studies is growing (reviewed in Schmid et al. 2009; Jiang & Pu 2009). In support of theory, grassland biodiversity experiments have shown that plant species diversity tends to increase and stabilize aggregate community biomass over time but destabilize population-level biomass in single trophic-level ecosystems (Tilman, Reich & Knops 2006; van Ruijven & Berendse 2007; Hector et al. 2010). More generally, species diversity has been found to have positive effects on average levels of a variety of ecosystem functions, including nutrient uptake, resource use efficiency, carbon fixation, respiration, decomposer activity, secondary production and resistance to disturbance (Cardinale et al. 2006; Balvanera et al. 2006; Worm et al. 2006; Ptacnik et al. 2008; reviewed in Schmid et al. 2009), and may stabilize them as well (Ptacnik et al. 2008).

The impacts of species diversity on populations, ecosystem functions and stability in multitrophic ecosystems, however, may differ substantially from those in single trophic-level systems (Thébault & Loreau 2005; Duffy et al. 2007). For example, predators are often mobile, consume their prey in a density-dependent fashion, may overexploit their prey or forage adaptively and may display complex feeding relations including omnivory, interference and intraguild predation (Holt & Polis 1997; Duffy et al. 2007). These aspects of trophic interactions are hypothesized to have impacts on food web structure and stability (Rosenzweig 1971; Holt & Polis 1997; McCann 2000; Kondoh 2003). Trophic interaction strengths are also known to impact food web stability, with weak interactions stabilizing food webs (McCann 2000; O’Gorman & Emmerson 2009; Rip et al. 2010). Trophic interaction strengths can be modified by species diversity and community composition in the resource trophic level (Kratina, Vos & Anholt 2007; Narwani & Mazumder 2010). As a result, food web structure, functioning and stability may depend on resource diversity, as well as the feeding selectivity of the consumer, and competition-resistance trade-offs among resources (Holt & Polis 1997; Thébault & Loreau 2003, 2005).

Numerous aquatic food web studies have manipulated whole food web diversity by manipulating diversity at multiple trophic levels and generally have found that diversity stabilizes community properties but can have contrasting effects on population stability (Steiner et al. 2005; Vogt, Romanuk & Kolasa 2006; Jiang & Pu 2009). Such experiments, however, could not address the relative importance of the diversity of resources and consumers (but see Gamfeldt, Hillebrand & Jonsson 2005). Despite the abundant theory demonstrating the particular importance of bottom-up effects of diversity on food web properties and stability (MacArthur 1955; Pimm 1991; Fussmann & Heber 2002; Thébault & Loreau 2003, 2005), very few studies have empirically tested the influence of resource diversity on the stability of consumer populations (Luckinbill 1979; Petchey 2000; Gonzalez & Descamps-Julien 2004), community dynamics (Hairston et al. 1968, Haddad et al. 2011) or ecosystem functions in food webs. To our knowledge, no experimental work has yet investigated bottom-up effects of diversity on all three levels of ecological organization.

There are a number of mechanisms by which resource species diversity may affect the stability and availability of resources and therefore the stability of consumer populations (Petchey 2000). (1) Resource availability: Increased resource species diversity may lead to an increase in total resource biomass because of the selection or complementarity (Loreau & Hector 2001), which could have positive effects on generalist consumer abundance (Haddad et al. 2011). However, greater resource biomass may also destabilize generalist consumers via the paradox of enrichment (Rosenzweig 1971; Luckinbill 1979). Impacts of selection, facilitation or complementarity on specialist consumers would be tied to impacts on their particular resources, assuming no indirect effects or interaction modifications by non-prey species (Luckinbill 1979). (2) Resource reliability: Greater diversity may lead to greater stability of total resource biomass via statistical averaging (Doak et al. 1998; Tilman, Lehman & Bristow 1998) or insurance effects (Yachi & Loreau 1999). This would result in more stable populations of generalist consumers because of the increased reliability of resources (MacArthur 1955), but less stable populations of specialists because of the lower abundance (statistical averaging) or greater variability (insurance effect) of individual resources (Haddad et al. 2011). (3) Resource composition effects: Diverse resource communities are more likely to contain a species, or group of species, with any particular combination of traits via sampling effects. For example, diverse communities may be more likely to contain inedible species (‘Variance in edibility’ hypothesis, Leibold 1989). This may stabilize consumer populations (Vos et al. 2001), by reducing consumer foraging efficiency or by competing with edible species for limiting nutrients, thereby lowering their carrying capacity (Kretzschmar, Nisbet & McCauley 1993, Grover 1995; McCauley et al. 1999). However, it may also decrease the total biomass of consumers (μ), increasing their risk of stochastic extinctions. Conversely, when resource species are all edible (and/or consumers are generalists), diversity may increase the rate of resource capture via indirect effects or complementarity among resource species (Narwani & Mazumder 2010; Toscano et al. 2010), resulting in increased growth and total biomass. Increased consumption could be either stabilizing or destabilizing depending on its relative impacts on consumer biomass and variability.

Here, we tested the effects of resource species diversity, resource community composition and consumer feeding selectivity on the density and stability of resource and consumer populations in planktonic food webs. We also tested the impacts of these factors on the average and variability of a number of community properties and ecosystem functions including primary producer biomass (total biovolume and chlorophyll-a), total community biomass (particulate carbon), nutrient accumulation (particulate nitrogen and phosphorus) and the rate of primary production (we used δ13C as a proxy for the rate of carbon fixation) (Fry 1996; Brutemark et al. 2009). Our work differs from the majority of previous biodiversity experiments in four ways: (1) we specifically tested bottom-up impacts of species diversity, (2) we examined these impacts on three levels of ecological organization (populations, communities and ecosystems), (3) we tested the relative impacts of resource diversity and community composition, and (4) we investigated the importance of consumers feeding selectivity.

Materials and methods

Experimental Design and Sampling

The experimental design consisted of four monoculture and four polyculture phytoplankton community compositions (Table 1), and these community compositions were crossed by two consumer species. The two consumers were the cladoceran freshwater filter-feeders Daphnia pulex and Ceriodaphnia dubia. The monoculture phytoplankton species were edible for both consumers. The polycultures each contained five phytoplankton species, and the compositions of the four communities were chosen so as to comprise a gradient in palatability for the two consumers. Daphnia pulex is larger than C. dubia and is therefore able to consume larger particles (∼45 μm in diameter as opposed to 35 μm or less for C. dubia), making it a relative generalist. Community composition 1 contained species that were all edible for both consumers. The number of inedible species in each community (according to size, one axis >40 μm) increased from composition 1 to 4 (Table 1). Each treatment was replicated three times, yielding a total of 48 microcosms.

Table 1.   Resource community compositions. Inocula of phytoplankton species were obtained from the University of Texas Culture Collection (UTEX) and the Canadian Phycological Culture Centre at the University of Waterloo (CPCC). The diameter estimates are in μm of ‘estimated spherical diameter’ (ESD). These estimates were determined from ≥195 natural units of each species using a Bench Top Model FlowCAM®
CompositionSpecies (source and identifier)Diameter (ESD)
Monoculture 1Scenedesmus acutus (UTCC 10)7
Monoculture 2Pseudokirchneriella subcapitata (CPCC 37)8
Monoculture 3Cyclotella sp. (UTCC 537)8
Monoculture 4Rhodomonas minuta (UTCC 344)11
Polyculture 1Scenedesmus acutus (UTCC 10)7
Pseudokirchneriella subcapitata (CPCC 37)8
Cyclotella sp. (UTCC 537)8
Rhodomonas minuta (UTCC 344)11
Cryptomonas erosa (CPCC 446)16
Polyculture 2Cyclotella sp. (UTCC 537)8
Cryptomonas erosa (CPCC 446)16
Ankistrodesmus falcatus (UTEX 101)37
Staurastrum Pingue (UTEX 1606)44
Fragilaria crotonensis (UTCC 269)123
Polyculture 3Pseudokirchneriella subcapitata (CPCC 37)8
Rhodomonas minuta (UTCC 344)11
Stichococcus bacillaris (UTCC 177)12
Asterionella formosa (CPCC 605)42
Pediastrum simplex (CPCC 431)62
Polyculture 4Ankistrodesmus falcatus (UTEX 101)37
Asterionella formosa (CPCC 605)42
Staurastrum pingue (UTEX 1606)44
Pediastrum simplex (CPCC 431)62
Fragilaria crotonensis (CPCC 269)123

Microcosms consisted of 2-L media bottles containing sterile COMBO medium (Kilham et al. 1998). COMBO is a standard freshwater culture medium with a nitrate concentration of 50 μmol L−1 and a phosphate concentration of 1000 μmol L−1. Prior to inoculation, we measured the density and average biovolume (estimated as the equivalent spherical diameter, ‘ESD’) of ≥85 natural units (cells, colonies or filaments) of each monoculture in the species pool at stationary phase using a Bench Top Model FlowCAM® (Fluid Imaging, Yarmouth, ME, USA). Experimental monocultures then received a biovolume of 500 000 μm ESD of the appropriate species, and polycultures received 100 000 μm ESD of each species in the mixture. We inoculated the phytoplankton 1 week prior to the addition of zooplankton (week 0). Daphnia pulex treatments received nine adults, and C. dubia treatments received 14 adults. The microcosms were placed randomly into an environmental growth chamber at 20 °C on a 12-h light: 12-h dark cycle.

We swirled the microcosms daily. Larger phytoplankton species settled out of suspension faster than smaller species, which may also have acted as a defence against grazing. We collected 250-mL samples weekly for twelve weeks, replacing samples with sterile standard COMBO medium (Kilham et al. 1998). Two observers checked each microcosm for the presence of zooplankton prior to sampling. When zooplankton were extinct, we replaced them at the original low density. We replaced zooplankton upon extinction to overcome the effects of stochastic extinctions at low density. Re-additions of consumers at low density did not appear to substantially alter community dynamics. Communities that could not support consumers continually experienced extinctions regardless of re-additions, while for communities in which zooplankton re-established after extinction, inoculation densities represented <2% of the maximum abundance (Appendix S3, Supporting information). If zooplankton were present at low density (<10 animals), we visually counted all of the animals in the microcosm or else the number of animals in the 250-mL sample. When there were greater than eight animals in the sample, we preserved a 40-mL subsample in sugared-buffered formalin (10%) for enumeration under a microscope. On weeks 5 and 9, we re-inoculated all of the microcosms with 1/10th of the original phytoplankton densities. Upon sampling, we took 10-mL subsamples for phytoplankton enumeration using the FlowCAM® within 48 h. We filtered 60-mL subsamples onto Whatman® GF/F grade glass microfiber filters (Whatman PLC, Maidstone, Kent, UK) and stored them at −20 °C for later chlorophyll-a analysis. We extracted the chlorophyll-a in 95% ethanol by shaking for 20 min and refrigerating for 24 h at 4 °C in the dark. We measured the chlorophyll-a using a Turner Designs Triology Laboratory Fluorometer (Turner Designs Inc., Sunnyvale, CA, USA) calibrated to manufacturer-recommended standards.

In addition to consumer and resource dynamics, we also monitored a number of ecosystem-level variables over time. We measured particulate organic carbon, nitrogen and δ13C (as a proxy for the rate of primary production) by filtering an 80-mL subsample from each microcosm onto ashed Whatman® GF/C grade glass microfiber filters. We dried the filters in an oven at 60 °C for 12–24 h and measured nutrient levels and carbon isotopes using an ECS 4010 CHNSO Elemental Analyzer (Costech Analytical Technologies Inc., Valencia, CA, USA) coupled to a Delta V continuous flow stable isotope ratio mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA). We took 20-mL subsamples for particulate phosphorus analysis, filtering 10 mL through 0·45-μm nitrocellulose filters for the measurement of dissolved phosphate, and 10 mL of the unfiltered sample for the measurement of total phosphate. Particulate phosphate was calculated as the total minus the dissolved phosphate. We digested the subsamples using potassium persulphate in an autoclave and measured the orthophosphate on a QuickChem® 8000 nutrient autoanalyzer (Lachat Instruments, Loveland, CO, USA).

Statistical Analyses

We used nested anovas to explore treatment effects, with zooplankton species identity, resource diversity and phytoplankton community composition (nested within diversity) treated as fixed effects. We asked whether these factors had significant effects on the mean and variability (CV) of zooplankton density, chlorophyll-a, total phytoplankton community biovolume, total particulate carbon, nitrogen and phosphorus. We also asked whether these factors affected the CV of population biovolume within microcosms and the average δ13C signature. We used data transformations when necessary to improve the normality and homoscedasticity. When consumer identity had a significant interaction with resource community composition or diversity, we analysed the data for each consumer separately (Table 2).

Table 2.   Results of nested anova for effects of species diversity (‘Diversity’), phytoplankton community composition (nested within diversity, ‘Phyto’) and consumer species identity (‘Zoop’) on the average and coefficient of variation of (a) zooplankton density, (b) chlorophyll-a, (c) particulate carbon, (d) particulate nitrogen, (e) particulate phosphorus and (f) δ13C. Transformations are indicated in parentheses when they were necessary to improve the normality or homoscedasticity. Significant interactions among phytoplankton community composition and zooplankton consumer species are shown, followed by separate results for each consumer species
 Sum Sqd.f.F valuePR2
(a) Zooplankton density
Mean (Square-root-transformed)
 Diversity/Phyto*Zoop383·7006, 324·9530·001 
 Daphnia pulex    0·951
  Diversity65·1901, 1610·3380·0050·503
  Diversity/Phyto348·7106, 169·217<0·0010·448
 Ceriodaphnia dubia    0·826
  Diversity0·0121, 160·0010·981<0·001
  Diversity/Phyto1486·5006, 1612·695<0·0010·826
CV
 Diversity/Phyto*Zoop2·6746, 323·2420·013 
 D. pulex    0·610
  Diversity0·3781, 162·0990·1670·051
  Diversity/Phyto4·1256, 163·8230·0150·559
 C. dubia    0·805
  Diversity0·0481, 160·5020·4890·006
  Diversity/Phyto6·2206, 1610·909<0·0010·799
(b) Chlorophyll-a
Mean (log10-transformed)
 Diversity/Phyto*Zoop1·9606, 326·880<0·001 
 D. pulex    0·886
  Diversity2·3831, 1635·366<0·0010·251
  Diversity/Phyto6·0166, 1614·881<0·0010·635
 C. dubia    0·921
  Diversity2·2001, 1679·793<0·0010·396
  Diversity/Phyto2·9156, 1617·619<0·0010·525
CV (log10-transformed)
 Diversity/Phyto*Zoop0·4486, 324·5390·002 
 D. pulex    0·834
  Diversity0·1651, 1611·2300·0040·117
  Diversity/Phyto1·0166, 1611·513<0·0010·717
 C. dubia    0·807
  Diversity0·6811, 1637·455<0·0010·451
  Diversity/Phyto0·5376, 164·9170·0050·356
(c) Particulate carbon
Mean (log10-transformed)
 Diversity/Phyto*Zoop3·7106, 324·5780·002 
 D. pulex    0·825
  Diversity0·5571, 1616·0450·0010·175
  Diversity/Phyto2·0656, 169·918<0·0010·649
 C. dubia    0·806
  Diversity0·4591, 1628·159<0·0010·342
  Diversity/Phyto0·6226, 166·3650·0010·463
CV
 Diversity/Phyto*Zoop0·4666, 322·5750·038 
 D. pulex    0·703
  Diversity0·4281, 1611·1230·0040·116
  Diversity/Phyto0·6576, 162·8440·0440·587
 C. dubia    0·638
  Diversity0·1361, 166·2460·0230·252
  Diversity/Phyto0·6706, 165·2800·0040·386
(d) Particulate nitrogen
Mean (log10-transformed)
 Diversity/Phyto*Zoop0·8636, 324·3480·003 
 D. pulex    0·772
  Diversity0·3441, 1616·4010·0010·139
  Diversity/Phyto0·5016, 163·9890·0120·633
 C. dubia    0·716
  Diversity0·4411, 169·7640·0070·291
  Diversity/Phyto2·0116, 167·4120·0010·425
CV
 Diversity/Phyto*Zoop0·9016, 323·8920·005 
 D. pulex    0·833
  Diversity0·3141, 1611·5310·0040·121
  Diversity/Phyto1·8516, 1611·343<0·0010·712
 C. dubia    0·671
  Diversity0·7321, 1614·6570·0010·301
  Diversity/Phyto0·8976, 162·9940·0370·369
(e) Particulate phosphorus
Mean (log10-transformed)
 Diversity0·1071, 323·2770·0800·042
 Zoop0·3471, 3210·6160·0030·135
 Diversity/Phyto0·5996, 323·0560·0180·233
 Diversity*Zoop0·0811, 322·4900·1240·032
 Diversity/Phyto*Zoop0·3946, 322·0110·0930·153
CV
 Diversity0·1701, 323·82500·0590·034
 Zoop0·3801, 328·5430·0060·077
 Diversity/Phyto2·6456, 329·904<0·0010·533
 Diversity*Zoop0·0291, 320·6470·4270·006
 Diversity/Phyto*Zoop0·3186, 321·1890·3370·064
(f) δ13C
Mean
 Diversity/Phyto*Zoop23·16, 322·6540·033 
 D. pulex    0·930
  Diversity52·21, 1644·082<0·0010·193
  Diversity/Phyto198·86, 1627·986<0·0010·736
 C. dubia (violated normality)    0·927
  Diversity146·91, 1685·368<0·0010·390
  Diversity/Phyto202·56, 1619·610<0·0010·537

We aimed to determine the mechanism by which resource biomass was stabilized in polycultures. We estimated the exponent (z) of the population biovolume mean–variance scaling relationship to determine whether statistical averaging contributed to community stability (Tilman, Lehman & Bristow 1998; Steiner et al. 2005). We also tested whether the total population biovolume, summed variances or summed covariances varied among the treatments (Lehman & Tilman 2000). Problems with using population covariance as a measure of synchrony have been noted (Isbell, Polley & Wilsey 2009), and so we also calculated synchrony sensuLoreau & de Mazancourt (2008):

image(eqn 1)

where inline image is the variance of the whole community biovolume and σi is the standard deviation of the biovolume of species i over time. Synchrony can have values between zero and one, with a value of zero indicating perfect asynchrony and a value of one indicating perfect synchrony. It should be noted that this metric will display high ‘synchrony’ not only when species’ biomass dynamics are positively related, but also when community biomass dynamics are dominated by an individual species. Dominance in a community can reduce the stabilizing impacts of statistical averaging or an insurance effect (Doak et al. 1998; Yachi & Loreau 1999; Gonzalez & Descamps-Julien 2004), and it may reflect a positive effect of selection on average total functioning. We calculated the Simpson index, D, as a measure of dominance, where D = Σpi2, and pi is the proportion of species i in a given microcosm. We entered average dominance, summed variance, summed covariance, synchrony and mean biovolume into a multiple regression to determine the magnitude and direction of their effects on the CV of total community biovolume and to ascertain how much of the variance in the CV was explained by each predictor.

We performed all statistical analyses using v. 2.9.2 of R software (R Development Core Team 2009).

Results

Resource diversity had a positive effect on the density of the generalist consumer, Dpulex, (Fig. 1a, Table 2a), but not the specialist (Fig. 2a, Table 2a). Contrary to our predictions, diversity did not significantly stabilize the population density of either consumer species (Figs 1b and 2b, Table 2a), although there was a trend towards greater stability with diversity for the generalist (Figs 1b). We re-ran the analyses for D. pulex, omitting treatments in which the consumer continually went extinct (M2 and P3), and the results were not substantially altered (see Appendix S6, Supporting information). Resource diversity stabilized and increased the phytoplankton community biovolume (Fig. 3a,b), while it destabilized and increased the average of resource population biovolume (Fig. 3c,d). Resource diversity had positive effects on the mean level of the majority of community properties and ecosystem functions, regardless of the consumer treatment (Figs 1c,e,g,k and 2c,e,g,i,k, Table 2b–f). Diversity also stabilized all community and ecosystem-level variables by reducing their coefficients of variation (Figs 1d,f,h,j and 2d,f,h,j, Table 2b–f, effects on phosphorus are marginally non-significant).

Figure 1.

 Effects of resource species diversity and community composition for microcosms containing the consumer Daphnia pulex. Effects on the mean and coefficient of variation are plotted for zooplankton density (a, b), chlorophyll-a (c, d), particulate carbon (C) (e, f), particulate nitrogen (N) (g, h), particulate phosphorus (P) (i, j) and δ13C (k). Monocultures are represented by ‘M’ and polycultures by ‘P’ in the category axis labels (see Table 1). The mean value for each community composition is plotted as a dot ± standard error (SE) bars (n = 3). Mean values for monocultures and polycultures are plotted as black horizontal lines ± SE bars (grey horizontal lines) (n = 12).

Figure 2.

 Effects of resource species diversity and community composition for microcosms containing the consumer Ceriodaphnia dubia. Effects on the mean and coefficient of variation are plotted for zooplankton density (a, b), chlorophyll-a (c, d), particulate carbon (C) (e, f), particulate nitrogen (N) (g, h), particulate phosphorus (P) (i, j) and δ13C (k). Monocultures are represented by ‘M’ and polycultures by ‘P’ in the category axis labels (see Table 1). The mean value for each community composition is plotted as a dot ± standard error (SE) bars (n = 3). Mean values for monocultures and polycultures are plotted as black horizontal lines ± SE bars (grey horizontal lines) (n = 12).

Figure 3.

 (a) The coefficient of variation of total phytoplankton community. (b) The mean phytoplankton community biovolume over time. (c) The phytoplankton population coefficient of variation over time, averaged across populations within a microcosm. (d) The mean phytoplankton population biovolume over time, averaged across populations within a microcosm. (e) The log of the sum of population variances within a microcosm. (f) The log of the absolute value of the mean covariance among the pairs of phytoplankton species within a microcosm over time, multiplied by the sign of the covariance. (g) The synchrony (sensu Loreau and deMazancourt 2008) among the pairs of populations of phytoplankton within a microcosm over time. (h) The mean edible biovolume (<40 mm in longest axis length) within a microcosm. All functions were taken over a 12-week time period. Bars represent the mean across replicates (n = 3), and the error bars display ± 1 SE. Dark bars represent values for Daphnia pulex, and light bars represent values for Ceriodaphnia dubia.

More specifically, resource diversity had significant positive effects on chlorophyll-a for both consumers (Figs 1c and 2c, Table 2b) and negative effects on the CV of chlorophyll-a (Figs 1d and 2d, Table 2b). The trends in the particulate carbon, the CV of carbon (Figs 1e,f and 2e,f), particulate nitrogen and the CV of nitrogen (Figs 1g,h and 2g,h) mirrored the trends in chlorophyll-a and the CV of chlorophyll-a, respectively (Table 3). Diversity significantly increased the mean but reduced the CV of both carbon and nitrogen for both consumers (Fig. 1e–h and 2e–h, Table 2c,d). Treatment effects on particulate phosphorus were qualitatively similar to those of carbon and nitrogen, although the data were more variable (Figs 1i,j and 2i,j, Table 2e). Diversity had positive effects on δ13C for both consumers (Figs 1k and 2k, Table 2f).

Table 3.   Summary of effects of resource species diversity on the dependent variables measured
Dependent variableDaphnia pulexCeriodaphnia dubia
  1. Positive effects are indicated by a ‘+’, and negative effects are indicated by a ‘−’. Significance at α = 0·05 is indicated with an asterisk. Marginally non-significant effects, 0·05 < P < 0·1, are indicated with ‘m-ns’. Non-significant effects (P > 0·10) are indicated with ‘ns’. No discernible effect (no trend) is indicated with a ‘0’.

Mean
 Zooplankton density+*0
 Chlorophyll-a+*+*
 Carbon+*+*
 Nitrogen+*+*
 Phosphorus+ m-ns+ m-ns
 δ13C+*+*
CV
 Zooplankton density−ns0
 Chlorophyll-a−*−*
 Carbon−*−*
 Nitrogen−*−*
 Phosphorus−m-ns−m-ns

Resource community composition had significant effects on consumer density and CV for both the specialist and generalist consumers (Table 2a, Figs 1a,b and 2a,b). The generalist, D. pulex, had its greatest abundance and stability in polyculture 1 (Fig. 1a,b, Table 1), in which all species were in the edible size fraction (Appendices S2e and S3a, Supporting information). The generalist had its lowest abundance and stability in monoculture 2 (Appendix S2b, Supporting information), followed closely by polyculture 3 (Appendix S2g, Supporting information), which were both dominated by the phytoplankton species P. subcapitata (Fig. 1a,b). Resource community composition also had significant effects on average chlorophyll-a (Figs 1c and 2c), carbon (Figs 1e and 2e), nitrogen (Figs 1g and 2g), phosphorus (Figs 1l and 2l) and δ13C (Figs 1k and 2k) for both consumers (Table 2b–f). In all cases, except for the mean zooplankton density for D. pulex and the CV of chlorophyll-a for C. dubia, composition explained more variability than species diversity per se (Table 2).

The specialist, C. dubia, had its greatest density and stability in monoculture 1, followed closely by polyculture 1, which were both dominated by the species S. acutus (Fig. 2a,b, Appendix S3a, Supporting information). To determine whether the proportion of edible phytoplankton determined zooplankton density for each consumer in polyculture, we regressed the average zooplankton density on the average proportion of edible phytoplankton species (<40 μm average ESD). There was a positive effect of the relative edible biovolume for the specialist (slope = 880·900, t = 2·852, d.f.=10, P = 0·017, R= 0·449), but not for the generalist (slope = 190·400, t = 1·416, d.f. = 10 P = 0·187, R= 0·167). Consumer density (square-root-transformed) was negatively correlated with total chl-a for the generalist (R = −0·549, t = −3·082, d.f. = 22, P = 0·005), but not the specialist (R = 0·066, t = 0·315, d.f. = 22, P = 0·756). Edible resource biovolume increased with species diversity in both consumer treatments (FD. pulex 1,16 = 37·449, P < 0·001; FC. dubia 1,16 = 65·142, P < 0·001) and was significantly affected by community composition (FD. pulex 6,16 = 9·050, P < 0·001; FC. dubia 6,16 = 49·624, P < 0·001, Fig. 3h).

The density of the generalist grazer D. pulex was negatively related to the proportion of edible algal biovolume over time within polycultures 3 and 4 (Appendix S4a,c, Supporting information). The decline in relative edible biovolume indicates that species larger than 40 μm increased with D. pulex density, suggesting compensatory dynamics among edible and inedible fractions of algal biomass with fluctuations in D. pulex density. Such compensation, however, was not generally evident for C. dubia treatments. In polyculture 3, however, C. dubia density increased rapidly before week 3 (Appendix S4b, Supporting information) and then declined with the proportion of edible biovolume by week 4. In this transition, the decline of an edible species, P. subcapitata, was compensated for by the increase in an inedible species, P. simplex (Appendix S3c, Supporting information).

Effects of Resource Diversity on Resource Community Stability

The two measures of phytoplankton community biomass, total biovolume and chlorophyll-a, were positively related (log–log correlation Pearson’s R = 0·875). The stability of most community properties and ecosystem functions were positively related to chlorophyll-a and so we tested for stabilizing effects of phytoplankton population biovolume on aggregate properties and ecosystem functions via impacts on total resource community biovolume. Diversity tended to cause a decline in the CV of total phytoplankton community biovolume for both consumers (Fig. 3a), although the effect was not significant for the specialist (F1,16 = 1·499, P = 0·239, P = 0·166, Fig. 3a). Resource community composition had significant effects on the CV of community biovolume for both consumers (log-transformed, FD. pulex 6,16 = 9·760, P < 0·001; FC. dubia 6,16 = 13·520, P < 0·001). Mean community biovolume increased with diversity (log-transformed, F1,32 = 165·222, P < 0·001, Fig. 3b), and community composition also had significant effects (F6,32 = 29·302, P < 0·001). Conversely, the mean CV of population biovolume increased with diversity (F1,32 = 100·277, P < 0·001, Fig. 3c), but was also affected by community composition (F6,32 = 12·728, P < 0·001). The average population biovolume increased with diversity (log-transformed, F1,32 = 4·569, P < 0·001, Fig. 3d) and was also significantly affected by composition (F6,32 = 29·4729, P < 0·001).

The mean–variance scaling exponent, z, equalled 1·726 (F1,134 = 4078·7, P < 0·001, R= 0·968), suggesting that mean–variance scaling should stabilize community biomass if community biomass does not increase with species diversity and is evenly distributed among species (Lehman & Tilman 2000). However, we found that both the community and average population biovolume increased with diversity (Fig. 3b,d) and that biomass was unevenly distributed among resource species (Appendix S1, Supporting information), reducing the influence of this stabilizing mechanism. The summed community variance increased with diversity (log-transformed, F1,32 = 119·710, P < 0·001) and was affected by community composition (F6,32 = 18·943, P < 0·001) (Fig. 3e).

The covariance of phytoplankton in polyculture depended on the interaction between the consumer species and resource composition (sign*log(absolute value)-transformed, F3,16 = 7·795, P = 0·002). For the generalist, there was no significant effect of composition (F3,8 = 2·368, P = 0·147), but for the specialist there was (F3,8 = 2279·762, P < 0·001) (Fig. 3f). There was also a significant interaction effect of consumer and composition on synchrony (F3,16 = 3·512, P = 0·040, Fig. 3g). Composition had a significant effect on synchrony for the generalist (F3,8 = 10·387, P = 0·004) and a marginally non-significant effect for the specialist (F3,8 = 3·502, P = 0·069) (Fig. 3g). Community dominance (Simpson’s Index, D) was significantly affected by community composition (F3,16 = 15·619, P < 0·001, Appendix S1, Supporting information). The synchrony and covariance data are generally in agreement, showing that polyculture 2 had positive temporal coherence among species, polyculture 3 showed compensation among species and polyculture 4 showed compensation for D. pulex, but synchrony for C. dubia treatments (Fig. 3f,g). The main discrepancy was for polyculture 1, in which a single species (S. acutus) was dominant for most of the time series causing high synchrony, but in which a single peak in the density of another species (P. subcapitata) caused negative covariances (Appendix S3a, Supporting information).

We re-ran the statistical analyses using detrended data for the aggregate community variables and zooplankton densities in order to remove the effects caused by potential linear temporal trends in the data. Overall, there were no qualitative changes in the results of the effects of diversity and phytoplankton community composition. However, the interactions between phytoplankton community composition and zooplankton treatment for zooplankton density and particulate carbon became insignificant. This indicated that phytoplankton communities had more similar impacts on both consumers when the effects of time were removed from the analyses. The effects of community composition and diversity on summed population and community variances and population synchrony were unchanged by detrending the data. The only exception was for the population covariances. Some of the significantly negative covariances became undistinguishable from zero, or even positive (polyculture 3 for C. dubia). This elimination of negative covariances after detrending resulted from the removal of the effects of shifts in dominance in which one species begins with high biomass but declines over time, while another species gradually takes over dominance of the community.

Discussion

We found that the bottom-up effects of resource species diversity on community biomass, ecosystem functioning and stability were generally positive. Resource diversity enhanced mean resource community biovolume, chlorophyll-a, carbon, nitrogen and δ13C, regardless of consumer feeding selectivity. Diversity also stabilized these functions over time, in support of previous findings from single trophic-level experiments in grasslands (Tilman, Reich & Knops 2006; Hector et al. 2010) and in aquatic food webs (Steiner et al. 2005; Vogt, Romanuk & Kolasa 2006). By contrast, resource diversity did not significantly stabilize consumer density. Overall, this suggests that while species diversity at the bottom of a food web may have dominant impacts on community properties and ecosystem functioning, these impacts may not penetrate upwards through the food web.

Resource species diversity is predicted to stabilize consumer populations by either increasing the availability of resources (overyielding, Lehman & Tilman 2000; Petchey 2000), ensuring long-term availability of edible species (portfolio or insurance effects, Tilman, Lehman & Bristow 1998; Yachi & Loreau 1999; Petchey 2000), reducing the strength of trophic interactions (McCann 2000; Kratina, Vos & Anholt 2007, Narwani & Mazumder 2010) or reducing the carrying capacity of edible resources (Kretzschmar, Nisbet & McCauley 1993; McCauley et al. 1999). There are also a number of ways in which resource diversity may destabilize consumers. Increased total resource abundance may lead to a paradox of enrichment (Rosenzweig 1971). The stability of particular resource species may decline because of reduced abundance or greater variability owing to increased competition (Lehman & Tilman 2000, Haddad et al. 2011). Finally, trophic interaction strengths may actually increase with resource species diversity depending on community composition (Narwani & Mazumder 2010), and this can destabilize consumer–resource dynamics (McCann 2000, Rip et al. 2010).

Contrary to our expectations and despite a stabilizing effect of diversity on algal community abundance, neither of the consumer population densities was stabilized by diversity. This may reflect a compensatory dynamics between edible and inedible phytoplankton species. Compensation can stabilize total algal biomass, while consumer density continues to cycle out of phase with edible algal biovolume, similar to ‘cryptic’ population cycles in the context of rapid evolution (Yoshida et al. 2003, 2007; Jones et al. 2009). For D. pulex, such compensation appeared to occur in compositions 2 and 4, in which consumer density oscillated out of phase with the proportion of edible algal biovolume (Appendices 3a,c). C. dubia, however, did not display similar coupled oscillations in mixture. Decoupling of the impacts of diversity on phytoplankton and zooplankton can also occur when the dominant phytoplankton species cannot support zooplankton density. In polyculture 3, both D. pulex and C. dubia densities were dramatically reduced over time, concurrent with an increase in the dominance of a single phytoplankton species, suggesting poor edibility or quality of this dominant species (Nelson, McCauley & Wrona 2001). For C. dubia, the dominant (P. simplex) was inedible, but for D. pulex, the dominant (P. subcaptiata) was neither inedible nor particular low in food quality (Appendix S5, Supporting information, high ratios indicate low quality). Nevertheless, in both cases, a phytoplankton species escaped grazer control and either reduced the abundance of other edible species (Kretzschmar, Nisbet & McCauley 1993; Murdoch et al. 1998) or competitively eliminated them.

Daphnia pulex density increased with resource diversity but the CV of its density did not decline, indicating that temporal variance also increased with diversity. Daphnia pulex is a generalist and therefore may have benefitted from complementarity among resources species (DeMott 1998; Petchey 2000; Duffy et al. 2007; Narwani & Mazumder 2010). Increased resource availability may have also been destabilizing via the paradox of enrichment (Rosenzweig 1971). Also, some resource communities were dominated by individual species, and on average, resource population variability increased with diversity, suggesting that increased variability in D. pulex may have resulted from interactions with individual resource species (Luckinbill 1979). Supporting this idea, the context-dependence effect of Fox’s Price equation partition (2010) on population variances (which measures the changes in individual species’ variances between high- and low-diversity communities) showed that the dominant species in polycultures 1 and 3 (S. acutus and P. subcapitata, respectively) were more variable in polyculture than they were in monoculture (see Appendix S2b, Supporting information). Finally, earlier work has shown that in communities with multiple palatable resource species, D. pulex may have an elevated rate of consumption (Narwani & Mazumder 2010). This may have strengthened consumer–resource interactions, destabilizing temporal dynamics (McCann 2000; Rip et al. 2010).

By contrast, neither C. dubia mean density nor variability was significantly affected by resource species diversity. The specialist’s density was correlated with the edible fraction of phytoplankton community biovolume, and this may have decoupled zooplankton and total phytoplankton dynamics in polyculture 3 inedible biomass was high. The biovolume of edible species generally increased with diversity however, and edible phytoplankton species dominated the biomass of other polycultures (polycultures 2 and 4, Appendices S4a,c, Supporting information). Low food quality may also have limited C. dubia abundance (Nelson, McCauley & Wrona 2001), although the C:N and C:P molar ratios of particulate matter indicate that only polyculture 4 may have been exceptionally phosphorus limited (Appendix S5b,d, Supporting information). Finally, it is possible that the low density of inedible species in these compositions may have lowered the specialists’ density because of interference with grazing of edible species (Kretzschmar, Nisbet & McCauley 1993; Narwani & Mazumder 2010). Overall, a decoupling of total algal biomass dynamics and C. dubia density was likely due to the presence of inedible species in three of four polycultures.

Our results contrast with those of one other study that has investigated the bottom-up effects of diversity on consumer population stability for consumers with varying feeding selectivities (Haddad et al. 2011). Haddad et al. (2011) found that there was often no significant effect of resource diversity on generalist populations, but that specialists were generally destabilized by diversity. They concluded that this was likely due to increased variances in the populations of palatable resources required by specialist consumers. Previous aquatic microcosm studies have provided mixed results. Some have shown no significant effect of resource diversity on consumer abundance (Fox 2004), or declines in abundance (Gonzalez & Descamps-Julien 2004; Fox 2006), and stability either has not been affected by diversity (Gonzalez & Descamps-Julien 2004) or has been positively affected (Hairston et al. 1968; Petchey 2000). Microcosm experiments with Daphnia and Ceriodaphnia have shown that greater resource diversity results in higher mean resource biomass, mainly by increasing the biomass of inedible species (Steiner 2001). Resource diversity can also increase community biomass stability in the presence of protist predators (Jiang, Joshi & Patel 2009). While the effects of resource diversity on resource community biomass, ecosystem functions and stability appear to be general, greater attention to the mechanisms by which resource diversity may impact consumer dynamics is warranted.

The community properties and ecosystem functions that we measured in this experiment were closely tied to phytoplankton community biomass (chlorophyll-a). Resource community biomass was most strongly stabilized by an increase in the average resource community biomass (see also Tilman, Reich & Knops 2006; Hector et al. 2010). While the mean–variance scaling relationship suggested that diversity should have reduced summed variances (z > 1), summed variances increased with diversity. Uneven species distributions and increased average population biovolume may have prevented a stabilizing effect (Doak et al. 1998; Tilman, Lehman & Bristow 1998). Dominance also decreases the stabilizing impact of statistical averaging (Doak et al. 1998), and static dominance by one species reduces the importance of an insurance effect on stability (Yachi & Loreau 1999; Gonzalez & Descamps-Julien 2004). Conversely, if the dominant species tends to have high biomass and stability, they may increase the biomass and stability of the phytoplankton community and the community as a whole (Petchey 2000; Gamfeldt, Hillebrand & Jonsson 2005; Steiner et al. 2005). In this experiment, dominance was likely positively related to stability through the influence of dominant, stable, fast-growing species, such as S. acutus and P. subcapitata (Fig. S1a,b, Supporting information). Finally, asynchrony resulting from compensatory dynamics between edible and inedible species was stabilizing in some compositions (Appendix S4b,c, Supporting information, Gonzalez & Descamps-Julien 2004, Gonzalez & Loreau 2009, Isbell, Polley & Wilsey 2009, Hector et al. 2010), while synchrony among species was destabilizing in others (Zhang & Zhang 2006, Romanuk et al. 2010, Tirok & Gaedke 2010). We did not find any evidence to support the hypothesis that generalists are more likely to cause synchrony in resource dynamics than specialists (Tirok & Gaedke 2010).

Community composition generally had greater impacts on the average and variability of the functions that we measured than diversity. The effects of individual compositions on consumers and community and ecosystem properties reflect the balance between the impact of resource species interactions on coexistence and individual species’ effect traits (Hillebrand & Matthiessen 2009). For example, polyculture 1 was dominated by S. acutus, and the high density and low variability of both consumers in this composition may have been due to the dominant effect of this species via a selection effect (Appendix S3a, Supporting information, Steiner et al. 2005). Nevertheless, both the abundance of D. pulex and the biovolume of S. acutus were greater in polyculture 1 than in the S. acutus monoculture, signalling a positive effect of complementarity among phytoplankton species (Gamfeldt, Hillebrand & Jonsson 2005). Daphnia pulex attained its lowest density in monoculture on P. subcapitata and had a similarly low density in polyculture 3, which was generally dominated by this species (Appendix S3c, Supporting information), suggesting that this species’ traits dominated this composition’s effects (e.g. poor quality, unpalatability, toxicity or environmental effects – on pH for example). For both consumers, polyculture 4 had the greatest level of coexistence among resource species. In this community, ecosystem functions may have benefited from niche partitioning and insurance effects resulting from competition-resistance trade-offs.

In summary, resource species diversity had positive effects on the abundance and stability of resource community biomass and a number of ecosystem functions in these multitrophic ecosystems. While diversity is clearly important for the functioning and stability of food webs, we did not detect straightforward bottom-up impacts on consumer stability, suggesting that resource diversity may have both stabilizing and destabilizing impacts on some consumers. Finally, community composition generally had a greater impact on food web properties than resource diversity, indicating that biodiversity–ecosystem functioning and stability research would benefit from greater focus on species’ interactions and functional effect traits.

Acknowledgements

We thank N. Arseneau, M. Bryan, R. Holberger, A. Lee, S. Mazumder and H. McNally for technical help. We especially thank D. Varela for allowing generous use of her FlowCam. We thank J. Isakovic, L. Weider, C. Kerfoot, L. Zelazny, G. Abrusan and N. Yan for cultures of organisms. Thanks to P. Venail for helpful conversations on the paper. This project was supported by NSERC Discovery and NSERC IRC grants to AM and an NSERC PGS-D scholarship to AN.

Ancillary