Toward an integration of evolutionary biology and ecosystem science




Ecology Letters (2011) 14: 690–701


At present, the disciplines of evolutionary biology and ecosystem science are weakly integrated. As a result, we have a poor understanding of how the ecological and evolutionary processes that create, maintain, and change biological diversity affect the flux of energy and materials in global biogeochemical cycles. The goal of this article was to review several research fields at the interfaces between ecosystem science, community ecology and evolutionary biology, and suggest new ways to integrate evolutionary biology and ecosystem science. In particular, we focus on how phenotypic evolution by natural selection can influence ecosystem functions by affecting processes at the environmental, population and community scale of ecosystem organization. We develop an eco-evolutionary model to illustrate linkages between evolutionary change (e.g. phenotypic evolution of producer), ecological interactions (e.g. consumer grazing) and ecosystem processes (e.g. nutrient cycling). We conclude by proposing experiments to test the ecosystem consequences of evolutionary changes.


The integration of ecosystem science with community ecology and evolutionary biology is one of the most important frontiers for the conceptual unification of the biological sciences (Holt 1995; Levin 1998; Elser 2006; Loeuille 2009). It is vital for a better understanding of the feedbacks between organismal diversity and global material and energy flows (Falkowski et al. 2008; Menge et al. 2008) and for predicting both community and ecosystem responses to environmental change (Chapin et al. 2000; de Mazancourt et al. 2008).

In the past few decades, the disciplines of community ecology and evolutionary biology have become increasingly integrated (Johnson & Stinchcombe 2007; Haloin & Strauss 2008; Vellend 2010). Evolutionary biologists recognize that ecological interactions among species influence natural and sexual selection pressures and underlie many evolutionary processes, such as phenotypic evolution (Ackerly 2003), diffuse co-evolution (Strauss et al. 2005) and speciation (Schluter 2000). Similarly, community ecologists recognize that evolutionary processes are important for understanding population dynamics (Yoshida et al. 2003; Pelletier et al. 2007), community composition and assembly (Emerson & Gillespie 2008) and metacommunity dynamics (Urban et al. 2008). There is also mounting evidence that ecological and evolutionary dynamics can occur on similar timescales (Hairston et al. 2005; Schoener 2011) and that feedbacks between natural selection and community dynamics can alter both the adaptive evolution and coexistence of species (Odling-Smee et al. 2003; Loeuille 2009; Post & Palkovacs 2009).

Although integrative research is progressively blurring the disciplinary line between community ecology and evolutionary biology, evolutionary biology and ecosystem science remain disjunct (Elser 2006; Loeuille 2009; Loreau 2010a). Evolutionary biologists regard divergent environments as a key driver of adaptive evolution and species diversification (Schluter 2000), but rarely study how organisms drive the ecosystem changes that shape selective environments (Odling-Smee et al. 2003; Strauss et al. 2005; Dieckmann et al. 2007; Erwin 2008). Ecosystem scientists view organisms and their environment as parts of an interactive system (O'Neill et al. 1986; DeAngelis 1992; Sterner & Elser 2002), but rarely study how evolutionary changes affect the flux of materials and energy through ecosystems (Norberg et al. 2001; Menge et al. 2008).

The disciplinary links between ecosystem science and evolutionary biology are among the weakest in the biological sciences (Levin 1998; Elser 2006). For major biogeochemical processes, such as nitrogen fixation, phosphorus mineralization and carbon cycling, the reciprocal effects of biodiversity and evolutionary change are poorly understood (Falkowski et al. 2008; Menge et al. 2008). Similarly, very little is known about how evolution by natural selection can affect fundamental ecosystem functions, such as primary productivity (Cadotte et al. 2008), food chain efficiency (Dickman et al. 2008) and decomposition (Gessner et al. 2010; Boudsocq et al. 2011). Although global environmental change elicits a broad range of evolutionary responses (de Mazancourt et al. 2008), some of which may directly impact material cycling in natural ecosystems (Collins & Bell 2004), evolution by natural selection is not yet incorporated into climate change models as a source of potential uncertainty (Murphy et al. 2004). Environmental scientists (e.g. biochemists, physicists) rarely consider how commonly studied ecological processes, such as trophic cascades (Terborgh & Estes 2010), or evolutionary processes, such as phenotypic evolution (Ackerly 2003), influence ecosystem functions.

We can achieve a greater integration between evolutionary biology and ecosystem science by broadening several of the sub-disciplines of ecology and evolution, including the study of biodiversity and ecosystem function (Loreau 2010a), community genetics (Neuhauser et al. 2003; Whitham et al. 2006), ecological stoichiometry (Sterner & Elser 2002) and eco-evolutionary dynamics (Fussmann et al. 2007; Schoener 2011). The latter, eco-evolutionary dynamics, addresses the reciprocal effects of evolutionary and ecological dynamics (Fussmann et al. 2007; Schoener 2011), and has a broad conceptual framework needed to understand interactions between evolutionary, ecological and ecosystem changes (Post & Palkovacs 2009). Recent research on eco-evolutionary dynamics has addressed the ecological consequences of changes in the phenotype distribution of populations (Yoshida et al. 2003; Palkovacs & Post 2009; Becks et al. 2010), and the ecosystem consequences of adaptation (Lennon & Martiny 2008; Gravel et al. 2011) and adaptive divergence between populations or species (Harmon et al. 2009; Bassar et al. 2010). Eco-evolutionary dynamics posits that both directions of effect, from ecology to evolution and from evolution to ecology, are substantial in their influence (Schoener 2011), and, in some cases, can result in feedback loops between ecological, ecosystem, and evolutionary processes (Post & Palkovacs 2009).

The goal of this article was to suggest ways to integrate evolutionary biology and ecosystem science. First, we review research on the ecosystem consequences of biodiversity (i.e. species richness and trait diversity), and then elaborate on how phenotypic evolution by natural selection can directly and indirectly affect ecosystem functions. Second, we review several models at the interfaces of evolutionary biology and ecosystem science and propose a new eco-evolutionary model, which builds on theory of ecological stoichiometry, to examine how phenotypic evolution of primary producers can affect the cycling of nutrients in a simple model ecosystem. Finally, we discuss several experimental designs and model systems that are useful for experimentally testing the effect of evolutionary changes on ecosystem functions.

The ecosystem consequences of biodiversity and phenotypic evolution

Research on biodiversity and ecosystem functioning has built strong disciplinary linkages between community ecology and ecosystem science (Hillebrand & Matthiessen 2009; Reiss et al. 2009; Loreau 2010a) and has significant potential to improve linkages between ecosystem science and evolutionary biology (Loreau 2010a). Evolution by natural selection has produced a phenotypically and functionally diverse global species pool, and extensive research has explored how this diversity of predators, grazers, primary producers and decomposers can affect key ecosystem processes (Reiss et al. 2009; Gessner et al. 2010; Loreau 2010a). More work is needed to disentangle how variation in the structure of biodiversity at multiple hierarchical levels (e.g. community, species, population and individual) can affect a broad range of ecosystem functions (Balvanera et al. 2006; Loreau 2010a). The productivity of ecosystems, for example, can be influenced by the number of species (Loreau 2010a), the phylogenetic diversity of species (Cadotte et al. 2008) and the evolutionary history of species (Gravel et al., 2011).

A practical way of studying the ecosystem consequences of biodiversity is to use continuous measures of individual trait diversity to quantify and compare the functional diversity of species assemblages (Cianciaruso et al. 2009; Hillebrand & Matthiessen 2009; Reiss et al. 2009). Trait variation among individuals is a fundamental component of biodiversity and a target of natural selection (Bolnick et al. 2003), and, as a result, is a natural intersection point between evolutionary biology and ecosystem science. However, we need a better understanding of the underlying causes of phenotypic and fitness variation among individuals, and how changes in selection pressures can cause direct or indirect effects on ecosystem functions (Norberg et al. 2001; Menge et al. 2008).

It is well established that the product of evolution by natural selection (e.g. species richness and trait variability) can affect ecosystem functions (Hillebrand & Matthiessen 2009; Reiss et al. 2009; Loreau 2010a), but much less is known about whether the process of phenotypic evolution is an important causative agent of ecosystem change (Norberg et al. 2001; Schoener 2011). This can be studied empirically by identifying heritable ecosystem-effect traits and testing how their evolution can affect ecosystem functions. An ecosystem-effect trait is a trait that underlies an organism's direct or indirect effect on an ecosystem function (see Table 1). There are several challenging research questions regarding ecosystem-effect traits. Are they heritable? Are they a target of natural selection? Will their evolution have predictable impacts on ecosystem functions? If ecosystem-effect traits are neutral with respect to fitness, then they will vary randomly over the landscape and will not be predictable based on the spatial distribution of selection gradients. In the following sections, we discuss how the evolution of ecosystem-effect traits can directly or indirectly affect ecosystem functions by influencing processes at the environmental (abiotic and biotic conditions), population and community scale of ecosystem organization.

Table 1.  Some examples of ecosystem-effect traits (that are potentially heritable) that might directly or indirectly affect ecosystem functions
TraitTypeSpeciesEcosystem functionScaleDescription of mechanismCitation
  1. These ecosystem-effect traits can also be classified following Violle et al. (2007), as: Response Traits (RT), Functional Traits (FT), Performance Traits (PT) and Effect Traits (ET). In some cases, it is unknown which trait, from a mixture of traits (Mix), is responsible for the contrasting ecosystem effects of different organisms. As discussed in the text, phenotypic change can affect ecosystem functions by affecting processes at different scales of ecosystem organization, including processes at the environment level (E), population level (P) and community level (C).

Root and leaf litter chemistryRTMetrosideros polymorphaNutrient cyclingESoil fertility drives heritable changes in leaf litter chemistry, thereby affecting nutrient cyclingTreseder & Vitousek 2001
Nitrogen excretion rateETPoecilia reticulataRate of nitogen recyclingEGuppies from sites with a species-rich high-predation community excreted more ammoniumPalkovacs et al. 2009
Rate of photosynthesisPTChlamydomonasPrimary productivityPExperimental evolution under high carbon dioxide can affect the growth rate and phenotype of algeaCollins & Bell 2004
Number of exploitable carbon substratesPTStrains of marine bacteriaProductivityCResource homogeneity drives resource specialization, which lowers productivity of species poor communities on heterogenous substratesGravel et al. 2011
Susceptability to herbivoryFTPinus edulisNutrient cycling, decomposition rateCHerbivores graze preferentially on susceptible trees, thereby altering litter qualityClassen et al. 2007
Proportion of algae in dietFTPoecilia reticulataAccrual rate of algae on tilesCGuppies from sites with a species-rich high-predation community ate more algeaPalkovacs et al. 2009
Susceptability to virus infectionFTSynechococcusPhosphorus availabilityCThe evolution of resistance to viruses by Synechococcus influences the rate of nutrient recyclingLennon & Martiny 2008
Leaf tannin concentrationFTHybrids of Populus spp.Nitrogen mineralization, decomposition rateCBeavers (Castor canadensis) preferentially fell low tannin treesWhitham et al. 2006
Body sizePTOncorhynchus spp.Nutrient flux across ecosystem boundariesCSize-selective predation by bears can affect the amount of salmon-derived nutrients moving from streams to terrestrial environmentsCarlson et al. 2011
Life history, feeding behaviourMixPoecilia reticulataBiomass specific gross primary productivity, leaf decomposition, nitrogen fluxC, EGuppies from sites with contrasting predator communities had indirect effects on ecosystem functionsBassar et al. 2010
Foraging traits, body shape, behaviourMixGasterosteus aculeatusNet primary production, rate of PAR and UV light attenuationC, ECompetition for resources in pelagic and littoral habitats has led to adaptive divergence in traits associated with foraging efficiency and growthHarmon et al. 2009

Phenotypic evolution affects ecosystem functions via effects on environmental conditions

Natural selection can affect ecosystem functions by acting on traits that underlie the capacity of organisms to modify their biological, chemical and physical environment (Jones et al. 1997; Sterner & Elser 2002). For example, ecosystem engineers are organisms that control the availability of resources through the creation, maintenance or modification of habitats (Jones et al. 1997). Ecosystem engineering is an important mechanism of niche construction, the process by which organisms alter their environment and the selective regimes of future generations (Laland et al. 1999; Odling-Smee et al. 2003; Erwin 2008). The evolution of traits underlying the ecosystem engineering effects of organisms could have a broad range of impacts on ecosystem functions (Harmon et al. 2009). For example, the stoichiometric traits associated with resource demand in consumers (Sterner & Elser 2002; Klausmeier et al. 2004) can affect the ratio and recycling rate of the nutrients used by primary producers (Sterner & Elser 2002) and the spatial distribution of primary production (McIntyre et al. 2008). Similarly, primary producer traits, such as growth forms, rooting structures and mutualistic associations, influence the provizioning of habitat structure and the moderation of abiotic stress in the environment (Callaway et al. 2002; Bouma et al. 2010). Evolution in the stiffness of marsh grass stems, for example, might affect the dissipation of hydrodynamic energy from waves and the trapping of sediment, and, as a result, influence the abiotic environment of coastal salt marsh ecosystems (Bouma et al. 2010). Although the effects of ecosystem engineering on ecosystem functions are potentially large, both the causes of trait variation underlying these effects and the responses of these traits to natural selection are poorly understood.

Phenotypic evolution affects ecosystem functions via effects on populations

Evolution by natural selection can directly influence the demographic parameters of populations (e.g. finite rate of increase) by changing the performance traits of individuals (e.g. growth rate, survival and fecundity). Phenotypic evolution of performance traits could directly affect many ecosystem functions because demographic variation of populations is often an influential component of biomass flux in ecosystems (Loeuille 2009). For example, evolution in the photosynthetic traits of algae in response to changes in carbon dioxide concentrations could alter the rates of primary production and carbon sequestration in aquatic ecosystems (Collins & Bell 2004). Natural selection can also indirectly affect ecosystem functions by acting on performance traits that influence population or metapopulation dynamics (Hanski & Saccheri 2006; Pelletier et al. 2007; Ezard et al. 2009). Selection acting on heritable traits such as the body size of ungulates (Pelletier et al. 2007), the defense traits of algae (Becks et al. 2010) and the flight performance of butterflies (Hanski & Saccheri 2006) are examples of how rapid phenotypic evolution can affect population dynamics (Schoener 2011).

The influence of the rates of phenotypic evolution on contemporary ecological dynamics (e.g. population dynamics, species sorting) has been an important unifying concept for community ecology and evolutionary biology (Hairston et al. 2005; Schoener 2011). Ecological theories of metapopulation (Hanski & Gilpin 1997) and metacommunity dynamics (Holyoak et al. 2005) are now incorporating the idea that local adaptation and dispersal jointly shape the assembly of contemporary communities (Barton & Whitlock 1997; Urban et al. 2008). This integration is also illuminating how ecological and evolutionary processes jointly affect the distribution of phenotypic variation at the landscape scale (Leibold 2010; Hanski et al. 2011). A challenge for future studies is to determine whether the effects of phenotypic evolution on population dynamics have broader impacts on ecosystem functions at local and regional scales.

Phenotypic evolution affects ecosystem functions via effects on communities

The evolution of traits that underlie how species interact with one another (e.g. foraging traits, defense traits) could drive changes in community dynamics that, in turn, affect ecosystem functions. Predation is a good example of a species interaction that can strongly influence the structure of food webs and the flux of organic matter in ecosystems (Schmitz 2010; Terborgh & Estes 2010). The fitness of predator individuals often varies according to the match between foraging traits (e.g. body size or gape width) and prey traits. In Fig. 1, we illustrate how a difference in the mean phenotype (e.g. foraging trait) of a predator population (Fig. 1a) might alter the population dynamics of multiple prey species (Fig. 1b), reconfigure the strength of trophic interactions between predator individuals and their prey (Fig. 1c) and indirectly affect an ecosystem function (Fig. 1d). For example, the optimum foraging rate of the invertebrate predator Chaoborus depends on the match between its gape width and the size of its zooplankton prey (e.g. Daphnia) (Swift & Fedorenko 1975). An upward shift in the optimum gape width of Chaoborus (Fig. 1a), resulting from a change in the selection regime, could reduce the relative abundance of larger zooplankton species that are more vulnerable to predation. Note the change in the rank order of abundance of prey of different sizes in Fig. 1(b). The resulting network of trophic interactions between Chaoborus individuals (black circles) and zooplankton species (green, red and blue circles in Fig. 1c) could reduce the average size of grazers, and potentially increase the level of primary production by algae (Cyr 1998) (Fig. 1d).

Figure 1.

A hypothetical example of how phenotypic evolution of a predator can affect an ecosystem function. Panel (a) shows the phenotype distribution of a predator's gape width (e.g. Chaoborus) under two contrasting selection regimes that favour a different optimal gape width. The size of the black circles illustrates an individual predator’s phenotype. Panel (b) shows two hypothetical scenarios for the population dynamics of three prey species that differ in their size (A: green, B: red and C: blue), and, hence, vulnerability to gape limited predators (Rank of prey size: C > B > A). Note that the rank order of prey population abundance changes with a shift in the gape width of predators. Panel (c) shows how a shift in the gape-size distribution of predators can affect the strength of trophic interactions between individual predators and prey of different sizes. The density of the predators stays the same, but larger prey become more vulnerable due to the increased frequency of larger predators, as indicated by the increase in the number of trophic links between individual predators and the larger prey species. Panel (d) illustrates how the contrasting food web structure, illustrated by differences in the weight of the link between the predator population and the prey, might affect an ecosystem function (e.g. primary productivity).

Advances in the field of community genetics have revealed that ecosystem functions are strongly influenced by heritable genetic variation within individual species, particularly those species that are dominant or play an important functional role in communities (Neuhauser et al. 2003; Whitham et al. 2006; Johnson 2011). For example, Whitham et al. (2006) describe how genetic variation underlying the production of tannins in poplar leaves can alter the rates of leaf-litter decomposition and nitrogen mineralization (Whitham et al. 2006). The rate of decomposition is a fundamental ecosystem function and is determined by the interaction between the quality of organic matter (e.g. leaf-litter) and the activity of multiple decomposer species (Gessner et al. 2010). Beavers (Castor canadensis) are an important agent of natural selection in this ecosystem because they preferentially fell trees with a low tannin concentration in their leaves. In this case, selection acts on a functional trait (i.e. tannin concentration) and indirectly affects decomposition by altering the community composition of trees and the chemistry of leaf-litter reaching decomposer communities (Whitham et al. 2006). Community genetics has been criticized for its inability to produce a genuine of theory of ecosystem evolution, partly because it does not address feedbacks between changes in community structure and phenotypic evolution (Loreau 2010b). However, research in this field has great potential for integrating evolutionary biology and ecosystem science, particularly if it focuses on how the processes that structure phenotypic and genetic variation (e.g. natural selection, drift, etc.) can influence the dynamics of ecosystem functions.

Is phenotypic evolution an important driver of ecosystem change and does it result in eco-evolutionary feedbacks?

There are many examples of how phenotypic evolution might affect ecosystem functions (Table 1), but are such effects large or small compared with other biotic and abiotic drivers of ecosystem functions? In a seminal paper, Hairston et al. (2005) proposed a method for partitioning variation in population growth rate into, roughly, an evolutionary component (i.e. phenotypic variation) and an ecological component (i.e. density-dependent population dynamics and changes in response to variation in the abiotic environment). This method could also be used to partition variation in an ecosystem function into a component associated with phenotypic change and a component associated with other ecological and environmental drivers of the ecosystem function. Such analyses could reveal whether evolutionary changes are important for predicting the dynamics of ecosystem functions. Phenotypic evolution is an important driver of population dynamics (Schoener 2011) and could turn out to be an under-appreciated driver of ecosystem functions (Norberg et al. 2001).

At present, it is unclear whether the effects of evolutionary processes on ecological dynamics and ecosystem functions have an important or trivial feedback effect on evolutionary dynamics (Post & Palkovacs 2009; Schoener 2011). For example, do the rapid evolutionary changes that occur in communities (Thompson 1998) cause changes to ecosystems that are sufficiently persistent and intense to alter the evolutionary trajectory of organisms? We agree with Schoener (2011) that a large and interdisciplinary research effort in eco-evolutionary dynamics is needed to address such questions.

Integrative modeling approaches at the interfaces between evolutionary biology and ecosystem science

Theory at the intersection of evolution and ecosystem functioning is rare (Loeuille 2010; Loreau 2010a). However, there is longstanding debate about how natural selection might maximize the flux of energy through ecosystems (Lotka 1922; Odum 1971; Loreau 2010a). For example, in models with a simplified ecosystem structure comprized of nutrients, producers, consumers and decomposers, grazing by herbivores can maximize the rate of primary production under a wide range of conditions, provided that plant fitness is tied to productivity rather than biomass (Cohen et al. 2000; Loreau 2010a). If plant populations vary in their tolerance to grazing, and there is spatial structure in the plant-herbivore community, then evolutionary change in the plants will be mediated by a balance between their spatial aggregation and patch size (de Mazancourt & Loreau 2000). In such a scenario, the optimal level of plant tolerance allowing grazing optimization to occur is difficult to attain by individual level selection alone. Thus, the emergence of enhanced plant productivity via grazing is governed by a balance between selection at the group and individual level (Loreau 2010a). Such theory implies that selection at the individual level might constrain the maximization of certain ecosystem functions, but this idea has not been experimentally tested.

Community evolution models are beginning to integrate theory from community ecology, evolutionary biology, and ecosystem science (Fussmann et al. 2007; Loeuille 2009; Loreau 2010a). These models generate realistic food web topologies, include important ecological traits (e.g. body size) in their design and, in some cases, allow for the evolution of ecologically relevant parameters (Loreau 2010a). The models arising from ecological theory are based on a few simple rules governing species interactions, and are often developed to compare with empirical data about food web structure or to test plausible mechanisms of food web stability (Rossberg et al. 2005; Gross et al. 2009). Dieckmann et al. (2007) proposed a class of community evolution models based on adaptive dynamics, whereby community structure evolves through recurrent evolutionary branching (i.e. speciation) and the invasibility of new phenotypes is determined based on equilibrium dynamics. These models allow for frequency and density-dependent selection by explicitly including population dynamics, but have a relatively simplified framework for modeling phenotypic evolution. In addition, they assume that evolutionary dynamics occur at a much slower rate than ecological dynamics (Fussmann et al. 2007), making them difficult to apply to the analysis of eco-evolutionary dynamics (Cortez & Ellner 2010).

Recently, Loreau (2010b) lauded community evolution models for their integration of ecological and evolutionary thinking and for their encapsulation of the essential elements of a genuine theory of ecosystem evolution. A significant strength of these models is that they simplify biodiversity dynamics by representing species with key functional traits (e.g. body size) that are subject to evolutionary change and underlie species interactions. However, they face several challenges for understanding ecosystem dynamics. First, their insights about community and ecosystem dynamics come largely from theory, because community evolution models are extremely difficult to test with experiments and observational data from natural ecosystems. Second, these models assume a predictable relationship between the phenotypic variation of organisms and ecosystem processes, even though there is still considerable uncertainty about how the variability and evolution of traits can affect population dynamics (Pelletier et al. 2007; Becks et al. 2010) and other ecosystem processes (Bailey et al. 2009; Harmon et al. 2009; Bassar et al. 2010). Third, most of these models do not appreciate the capacity of organisms to alter selective regimes by transforming both the biotic and abiotic environment of ecosystems. Feedbacks between organisms and the abiotic environment are included in some analytical models (Goudard & Loreau 2008; Kylafis & Loreau 2008), but there is no consensus about how the non-trophic effects of species on their environment will impact the evolution of species interactions (Loreau 2010a). Despite the limitations of community evolution models, they are a welcome addition to the dearth of integrative theory linking evolutionary dynamics with the physical and biochemical diversity of ecosystems.

Integrating models of eco-evolutionary dynamics with ecological stoichiometry is a useful way to illustrate how phenotypic evolution of organisms can affect nutrient cycling (Sterner & Elser 2002). The study of nutrient cycling is an important intersection between evolutionary biology and ecosystem science (Elser 2006), because the evolution of traits governing resource acquisition and utilization can have profound impacts on global biogeochemical cycles (Klausmeier et al. 2004; Falkowski et al. 2008). Considerable research has addressed the food web consequences of stoichiometric variation among species (Hall 2009), but much less is known about how evolution of stoichiometric traits could affect ecosystem functions (Boudsocq et al. 2011).

Here in, we develop an eco-evolutionary model to investigate how evolutionary change in the resource uptake rates of primary producers can affect nutrient dynamics in a simple model ecosystem. In terms of developing a predictive theory for ecosystem functioning, our model describes ecosystem dynamics as the rates at which elements are transferred among abiotic (e.g. nutrients) and biotic (e.g. species) states in the environment. We characterize evolutionary dynamics as the rate of change in an organism's stoichiometric traits (e.g. nutrient uptake), and ecological dynamics as changing resource densities.

Our model combines the basic elements of the ecological stoichiometry models of Grover (2002) with the adaptation to competition for non-substitutable resources model of Fox & Vasseur (2008). Following Grover (2002), we consider the dynamics of a pentagonal food-web consisting of two non-substitutable resources (R1 and R2), two producers (YP1 and YP2), and a generalist consumer (C). For simplicity, the producers have contrasting and fixed resource demands (expressed as: yield coefficients, ePij) and nutrient specific uptake rates U that can evolve over time (Table 2).

Table 2.  Initial and final values for the variables, and values for the parameters used in the model simulation
SymbolDescriptionInitial valueFinal value
RiResource i density (e.g. R1, R2)400, 533337, 337
YPjProducer j density0.5, 0.00.375, 0.376
CConsumer density0.50.532
UP1Uptake rate of YP1 on Ri0.5, 0.50.75, 0.25
UP2Uptake rate of YP2 on Ri0.5, 0.50.25, 0.75
ajConsumer attack rate on YPj1, 1
mjProducer j mortality0.1, 0.1
dConsumer mortality0.05
eP1iYP1 yield coefficient for resource i0.0025, 0.0075
eP2iYP2 yield coefficient for resource i0.0075, 0.0025
eCiConsumer yield coefficient for resource i0.0005, 0.0005
lConsumer leakage rate0.5
vjAdditive genetic variance of producer j traits0.00026, 0.00026

The population growth rate of producers eqn 1 is determined by their loss rates from consumers (aj), density-independent mortality (mj),


and a growth rate on their limiting nutrient (GYj, eqn 2).


In eqn 2, we model a simple linear tradeoff in the uptake rates of both resources by setting, for example, the uptake rate of R2 to (1 -; UP1). The producers have contrasting affinities for non-substitutable nutrients defined by their nutrient-specific yield coefficients (ePij : i producers per unit j resources).

The population growth rate of consumers eqn 3


is determined by density-independent mortality (d) and growth on the limiting resource (GC, eqn 4). The latter is affected by both the efficiency of the consumer's resource consumption (l), the stoichiometry of the consumer's prey (ePij), and the conversion efficiency on different resources in their diet (eCi).


For simplicity, resource dynamics are solved by assuming mass balance in the system eqn 5.


Evolutionary change is included by allowing, for example, the uptake rate (U) of a producer species to change over time to maximize its per capita growth rate in the current environmental conditions eqn 6.


Here in, vj is a constant and represents additive genetic variance when the uptake is a quantitative genetic trait (Fox & Vasseur 2008), and so both the fitness gradient and vj simultaneously determine the rate of evolutionary change. Due to the discontinuous partial derivative, eqn 6 is approximated following Fox & Vasseur (2008).

An application of the model

The model illustrates how evolutionary change in the uptake rates of primary producers can affect competitive interactions among species and influence the resource dynamics of a system. Using the initial conditions provided in Table 2, we simulated the dynamics of two resources, two producers and a consumer.

Dynamics of the simulation

We started the model simulation with a single producer (YP1) with a lower yield coefficient on R1 than R2 (Table 2), and after some time (t = 2000), we introduced a second producer (YP2) with a lower yield coefficient on R2 than R1. During the first part of the simulation (t < 2000), YP1 adapts to the environment by increasing its uptake rate on R1. The system reaches a stable ecological equilibrium relatively quickly, as shown by the population trajectories of YP1 and C (Fig. 2a). At this point, there is a high ratio of producer : consumer biomass, and a low R1 : R2 ratio of the available nutrients (Fig. 2a). The introduction of YP2 (t = 2000) has three main effects on the system. First, YP2 rapidly takes advantage of the underutilized R1 in the system and changes the ratio of supplied nutrients (note the change of the supply point in Fig. 2b). Second, YP2 increases the stoichiometric quality of the consumer's food, and, as a result, increases the consumer density and predation pressure on both species (apparent competition). This leads to a reduction in the producer : consumer biomass ratio. Third, as YP2 adapts to the nutrient environment (Fig. 2c) it alters the stoichometry of the available nutrients (Fig. 2b).

Figure 2.

Panel (a): The system starts with the parameters and values provided in Table 2. The abiotic concentration of R1 (N, solid blue line) is initially lower than R2 (P, dashed blue line) because of YP1’s high R1 : R2 ratio. Initially, YP1 (solid green line) increases rapidly in density whereas C (red line) decreases. The system exhibits short lived predator–prey cycles that are rapidly dampened, resulting in a stable ecological equilibrium by t = 50. YP1 then adapts to the environment by increasing its uptake rate of R2 at the expense of R1 (Panel c). This adaptation allows it to increase its growth rate but interactions with the consumer keep its density relatively constant. Increases in its growth rate fuels an increase in C’s density and a subsequent decrease in the densities of both N and P. The invasion of YP2 (dashed green line) disturbs the system in several ways. The high P content in the environment allows YP2 to rapidly increase in density, causing a decrease in the density of producers YP1 due to apparent competition. The addition of YP2 actually increases the growth rate of YP1 due to increased consumption of YP1 and increasing the R1 concentration (Panel d). In addition, the invasion of YP2 shifts the adaptive peak of YP1 closer to its U1 value (Panel c). The system cycles briefly and then reaches a new stable ecological equilibrium with all three species. Further adaptation by YP2 leads to an increase in its growth rate (Panel c), and over time, this increases the transfer of R2 into the producer trophic level (relative to R1) until the concentrations of R1 and R2 are equal. At this point neither YP1 nor YP2 has an advantage. This change in the abiotic condition causes greater decreases in YP1 than is compensated by an increase in YP2, and so the total producer density decreases. Meanwhile, the density of C increases as the ratio of YP1 : YP2 approaches 1 : 1 at which point there is no stoichiometric mismatch between the consumer and its diet. Panel (b): ZNGI for YP1 (solid green line) and YP2 (dashed green line) showing the concentrations of R1 and R2 at which growth equals death due to density independent mortality and consumption by C. Here we show only a single pair of ZNGIs (at t = 2100), but the ZNGIs evolve in our model such that the resource supply point always allows for coexistence (data not shown). The ZNGIs shift based on consumer density and with changes in the uptake rate of nutrients. The blue points represent the concentrations of R1 (N)and R2 (P) at different time steps. At t = 0, R2 > R1. The rapid increase in YP1 exacerbates the nutrient imbalance [R2] >> [R1] by the time the first ecological equilibrium is reached at t = 50. The adaptation of YP1 decreases both P and R1 equally until t = 2000. The invasion of YP2 (at t = 2000) decreases the R2 in the environment, due to the stoichiometry of YP2, and increases the R1 in the environment, due to increased YP1 mortality. The second ecological equilibrium is reached at approximately t = 2100. As mentioned above, the system converges to an ecological and evolutionary equilibrium at (t = 3500), when the producer species become identical in the uptake rates. Panel (c): Shows the time course of uptake rates of R1 for each species (green lines). The optimal uptake rate (black lines), vary based on the ratio of R1 : R2 in the environment, and is always the uptake rate at which growth is maximized. The optimal growth rate can be considered the ‘peak’ of an adaptive landscape. In our model, fitness falls away from this adaptive peak linearly from 100 to 0% as uptake rates vary from on 0 to 1. From the start of the simulation until t = 2000, YP1 adapts to the local environment, which itself is being modified by the adaption of YP1. The invasion of YP2 shifts the optimal uptake rates of both species. From t = 2100 to t = 3500, both species adapt to the changing nutrient environment. Panel (d): Shows the R1 : R2 ratio of the excreted nutrients (red dashed line) of a consumer with a stoichiometry given by the black dotted line (e.g. R1 : R2 is 1 : 1 for sake of simplicity and with no loss of generality). Feeding solely on YP1 (solid green line) results in a preferential R1 excretion in the environment, whereas feeding only on YP2 (dashed green line) results in preferential R2 excretion into the environment.

The phenotypic evolution of producers

We can visualize the stoichiometric phenotypes of organisms using zero net growth isoclines (ZNGIs) (Fox & Vasseur 2008). In our simulation, the producers adapt their uptake rates to the changing nutrient dynamics (Fig. 2c), such that their ZNGIs evolve over time (Fig. 2b). When YP2 is introduced, YP1 has almost reached its optimal uptake rate (black lines in Fig. 2c). The two producers can coexist because the nutrient supply point equilibrates at the intersection of their ZNGIs (Fig. 2b). As YP2 evolves, by reducing its uptake of R1 in favour of R2 (Fig. 2c), it becomes a better competitor for the available nutrients in the system and slowly increases in abundance (Fig. 2a). As there are few constraints to adaptation, the species converge to have completely overlapping ZNGIs by the end of the simulation, but evolve divergent nutrient uptake strategies that reflect their initial differences in resource demand (Table 2).

The effects of consumers

Consumers impose density-dependent mortality on the producer population and alter the stoichiometry of nutrients via nutrient recycling. As shown in Fig. 2(d), the consumer excretes proportionally more R1 than R2 (R1 : R2 > 1) when grazing on YP1, and vice versa when grazing on YP2 (R1 : R2 < 1), meaning that consumers affect nutrients in the environment via the interaction between their resource demand and the eco-evolutionary dynamics determining primary producer dynamics. Our example only considers a single non-evolving consumer, but one could also model the outcome of competition between two consumers with different stoichiometric traits.

In sum, our model illustrates how the composition and phenotypic evolution of a producer species can alter the dynamics of an ecosystem process (e.g. nutrient recycling) and the ecological structure of the community (e.g. biomass ratio of roducers : consumers). The model can be run with and without evolutionary change and eco-evolutionary feedbacks, and could be tested against empirical data, from either a natural ecosystem or an experiment, to assess the consequences of phenotypic evolution for nutrient cycling.

Experimental approaches for testing the ecosystem consequences of evolutionary diversity and dynamics

In this section, we propose ways to experimentally test the ecosystem consequences of the evolutionary history of organisms (i.e. the product of selection) and of the contemporary evolution of ecosystem-effect traits (i.e. the process of evolution by natural selection). To start with, we describe an experimental design that we refer to as a common gardening experiment, which is useful for quantifying ecosystem effects.

Common gardening experiments

In a classical common garden experiment, the environmental conditions are held constant for each genotype, to quantify how much phenotypic variation is attributable to genetic differences among organisms. Our proposed common gardening experiment is a type of common garden experiment that is specifically designed to quantify the effects of organisms on a common type of ecosystem. In a common gardening experiment, the phenotypes (or genotypes) of organisms are held constant for each ecosystem type to quantify how much ecosystem variation is attributable to phenotypic (or genetic) differences among organisms. ‘Common gardening’ refers to the idea that the ecosystems in the experiment begin from a common starting point and are modified, in potentially different ways, by the activities of the organisms that are used as treatments in the experimental design. Recently, there have been several examples of common gardening experiments that have been used to determine the ecosystem consequences of phenotypic divergence and phenotypic evolution (Lennon & Martiny 2008; Harmon et al. 2009; Palkovacs & Post 2009; Bassar et al. 2010; Gravel et al. 2011). However, many more common gardening experiments are needed to understand the broad range of ecosystem functions that could be influenced by the evolutionary history of organisms and by contemporary evolutionary dynamics.

Common gardening experiments can be preceded by a common garden experiment to determine whether ecosystem variation is attributable to either the genetic determinants of phenotypic variation or to phenotypic plasticity. Some traits can change in response to an environmental factor early in organism's development (i.e. exhibit phenotypic plasticity), but after some period of time they become fixed and insensitive to environmental variation [a type of age-specific environmental canalization: Debat & David (2001)]. Organisms could be grown in a common environment until a phenotype of interest is relatively stable, and then the organisms could be subsequently used as experimental treatments in a common gardening experiment. Such experiments could test, for example, whether the different ecosystem-effects of closely related species (or populations) result from adaptive genetic divergence or non-genetic differences in phenotypically plastic traits (Seehausen 2009).

Ecosystem consequences of contrasting evolutionary histories

Several recent common gardening experiments have found that phenotypic differences among organisms can have a wide range of community and ecosystem effects (Harmon et al. 2009; Bassar et al. 2010). In these experiments, it is hard to identify specific ecosystem-effect traits because it is unclear which traits caused the observed ecosystem divergence (see ‘Mix’ category in Table 1). In such cases, the evolutionary origins of the phenotypic variation can be informative for predicting which traits are likely to cause ecosystem effects. In adaptive radiations, for example, the traits under divergent selection are often those used to exploit resources in the natural environment. As a result, we might expect trait divergence between species in an adaptive radiation to cause strong and contrasting effects on ecosystems (Harmon et al. 2009). In comparison, the divergence of traits that are not closely associated with ecosystem properties, such as the evolution of mating traits driven by sexual selection (Arnqvist 1998), will probably have negligible impacts on ecosystem functions. Overall, we are a long way from predicting whether the traits under divergent selection are the same as those that cause contrasting ecosystem effects.

A few recent studies have used a period of experimental evolution to examine how organisms with different evolutionary histories can affect community dynamics (Becks et al. 2010) and ecosystem effects (Gravel et al. 2011). Gravel et al. (2011) experimentally evolved twenty strains of marine bacteria into specialists or generalists, with respect to the number of utilizable carbon substrates. For each group of strains (i.e. ancestors, generalists and specialists) Gravel et al. (2011) established communities in microcosms along a gradient of species richness and measured community productivity (i.e. bacterial metabolic activity) on a mixed medium of carbon substrates. They found that the slope of the relationship between species richness and productivity was steeper for communities made up of specialists, illustrating the influence of variation in the evolutionary history of community members on ecosystem functions (Gravel et al. 2011). This is a nice example of a simplified common gardening experiment, because it used a highly controlled and narrowly defined ecosystem (a mixed carbon substrate) and only measured one ecosystem function (productivity).

To date, common gardening experiments have not tested whether the ecosystem effects of species with different phenotypes (or genotypes) have sufficient strength and persistence to affect future evolutionary change, and, as a result, none have experimentally demonstrated an eco-evolutionary feedback. In a recent common gardening experiment, Harmon et al. (2009) found that sticklebacks with different phenotypes had contrasting effects on zooplankton community structure, gross primary productivity and the rate of light extinction in the water column, but do such contrasting effects at the community and ecosystem level influence the selection pressures on sticklebacks in future generations? This is not a trivial question to answer, but it could be addressed by doing a common gardening experiment followed by a selection experiment in the modified ecosystems. For example, one could do a common gardening experiment in replicated ponds using sticklebacks with two divergent phenotypes as experimental treatments. We would expect the pond ecosystems to diverge over time (Harmon et al. 2009), but would this generate divergent selection environments for the next generation of sticklebacks? This could be tested by removing the founding stickleback populations from all of the ponds, adding a common set of juvenile sticklebacks back into the same ponds and measuring the relationship between the fitness of individuals and a quantitative trait under natural selection (i.e. a fitness function) in the ecosystems that were modified during the common gardening experiment. Such an experiment could experimentally test whether the modification of an ecosystem by an organism could change the selection pressures of that same organism in subsequent generations.

Ecosystem consequences of contemporary evolutionary dynamics

Most of the previous common gardening experiments have tested how historical evolutionary changes can affect contemporary ecosystem functions (Harmon et al. 2009; Bassar et al. 2010; Gravel et al. 2011). As a result, they have temporally separated the process of natural selection from the evaluation of ecosystem effects. However, one can allow selection to occur over the course of a common gardening experiment to directly test the ecosystem effects of variation in (1) the strength of natural selection (i.e. the selection differential), or (2) the response of a population to selection (e.g. adaptation over multiple generations).

Ecosystem effects of variation in the strength of selection

A useful way to assess the impact of phenotypic change on ecosystem functions is to manipulate the strength of selection on a target organism in a common gardening experiment. The experimental work on predator–prey dynamics of rotifers (e.g. Brachionus) and algae (e.g. Chlamydomonas or Chlorella) provides a good model system for such experiments (Yoshida et al. 2003; Becks et al. 2010). Becks et al. (2010) found that the selective regime caused by rotifer grazing reduced the amount of variation in a heritable defense trait of algae, which altered the nature of predator–prey dynamics. In this experiment, the strength of predator-induced selection was manipulated by altering the initial amount of phenotypic variation in the defense traits of the prey population. For practical reasons, these experiments are conducted in chemostats that maintain a highly controlled environment and food supply. However, it would be interesting to investigate how different evolutionary dynamics might impact the functioning of these simple ecosystems (e.g. nutrient cycling, pH). The allocation of resources to defense might divert energy away from growth and reproduction (i.e. affect demographic variation), and the subsequent change in draw down of carbonic acid by phytoplankton photosynthesis might affect other strains or species. If the genetic composition of one strain affects the persistence, activity or coexistence of other strains, then strain diversity could have significant effects on ecosystem functioning (Isbell et al. 2009).

We can directly target the ecosystem consequences of selection by studying how variation in the strength of selection affects the distribution of ecosystem-effect traits in a population. This could be done by nesting a selection experiment, for example one that uses candidate genes that underlie adaptive phenotypic traits (Barrett et al. 2008), inside a common gardening experiment (Harmon et al. 2009). Specifically, one could measure how variation in the selection differential over the course of the selection experiment might have contrasting effects on ecosystem functions. The recent advances in the genomic resources of Daphnia open vast opportunities to do such experiments and to study how natural selection acting on specific genes (Routtu et al. 2010; Colbourne et al. 2011) might affect ecosystem functions. We already know a great deal about how changes in the life-history and stoichiometric traits of plankton might affect primary productivity and nutrient dynamics in aquatic ecosystems (Sterner & Elser 2002; Hall 2009). In addition, some of these traits can evolve rapidly in response to selection (Gorokhova et al. 2002; Weider et al. 2005). For example, divergent selection acting on Daphnia production rate can cause rapid genetic changes in the length of intergenic spacers (IGS) of ribosomal DNA (Gorokhova et al. 2002), which is related to Daphnia’s life-history and stoichiometric traits (Weider et al. 2005). The rapid developments in environmental genomics will undoubtedly provide valuable resources for integrating evolutionary biology with ecosystem science (Vandenkoornhuyse et al. 2010).

Ecosystem effects of population responses to selection

There is very little experimental evidence showing how the evolution of adaptive traits over multiple generations can directly affect an ecosystem function. In an elegant example, Lennon & Martiny (2008) found that the rapid evolution of resistance in a photosynthetic cyanobacteria (Synechococcus) to an infectious virus (Myoviridae) seemed to mediate nutrient availability in the environment via top-down control by viruses (Table 1). In this case, the evolution of resistance, which occurred over the course of an experiment, is an example of adaptation by natural selection (Elena & Lenski 2003) because beneficial mutations most likely arose in the host populations of Synechococcus and increased in frequency in response to virus-mediated selection (Lennon & Martiny 2008). This experiment is unique because it measured the ecosystem effects of a population response to selection over the course of a common gardening experiment (Lennon & Martiny 2008), as opposed to measuring the consequences of population dynamics resulting from variation in the strength of selection (Becks et al. 2010).

A significant challenge for future studies is to isolate how various mechanisms of evolutionary change can affect ecosystem functions. We can do this by identifying the genetic basis of ecosystem-effect traits, using selection experiments to determine how these traits respond to selection and by doing common gardening experiments to quantify the ecosystem effects.


The integration of community ecology and evolutionary biology has progressed considerably, in part, because of the realization that the outcomes of ecological interactions among species are contingent on their evolutionary histories (Johnson & Stinchcombe 2007; Haloin & Strauss 2008; Schoener, 2011). G. E. Hutchinson promoted this idea by describing ecological systems as analogous to theatres, in which the actors (i.e. species or individuals) have roles determined by their evolutionary history, and the acts are played out in an unscripted fashion that is contingent on the environmental setting (i.e. the local theatre) (Hutchinson 1965). Modern ideas about eco-evolutionary dynamics call for a more nuanced analogy – one that might stimulate greater integration between ecosystem science and evolutionary biology. For instance, we can consider that the roles of actors (i.e. their phenotypes) in local theatres (i.e. ecosystems) change over generations in response to direct pressures from actors’ peers and audiences (i.e. the agents of selection). Actors’ roles can evolve because of reciprocal interactions (i.e. eco-evolutionary feedbacks) between actors and their peers (i.e. the community) or between actors and the structural components of the theatre (e.g. the abiotic environment of ecosystems). In addition, actors can influence the development and renovation of the theatre (i.e. ecosystem modification and engineering) and this, in turn, can increase the number of actors in the play (i.e. via niche construction) and affect the outcome of future plays.

At present, ecosystem science and evolutionary biology lack a cohesive theoretical and experimental framework. We can achieve greater integration between evolutionary biology and ecosystem science by expanding the scope of several research fields in ecology and evolution. At a time of unprecedented human alteration of the environment, it is crucial to integrate evolutionary biology and ecosystem science to achieve a deeper understanding of both the causes and consequences of biodiversity change in natural ecosystems.


We thank the ProDoc SNF grants and the Eawag Directorate for funding for the summer school ‘Interaction between ecological and evolutionary processes in aquatic ecosystems’. We also thank the organisers Helmut Buergmann and Hitoshi Araki and the invited speakers: J. Elser, N. Hairston Jr., A. Hendry, E. Litchman, L. De Meester and E. W. Triplett. We also thank J. Fox, N. Loeuille, T. Ingram, R. Svanbäck, R. El-Sabaawi, B. Miner, K. Wagner and P. Ramsay for their comments, corrections and insights on earlier drafts. B.M. was supported by SNF grant number 31003A-125006: ‘Food-web and ecosystem responses to global change’. We also thank the three anonymous reviewers who significantly improved this manuscript.


  • 1Community genetics: The study of interspecific genetic interactions and their effects on community composition and ecosystem functioning.
  • 2Eco-evolutionary dynamics: The effects of ecological dynamics on evolutionary change and the effects of evolutionary dynamics on ecological change. The effects in both directions can be mediated by changes in ecosystem functions.
  • 3Eco-evolutionary feedback: The feedback between ecological and evolutionary processes whereby an organism affects population, community or ecosystem dynamics in a way that affects its own evolutionary trajectory.
  • 4Ecosystem: A unit that includes all of the organisms (i.e. the ‘community’) in a given area interacting with the physical environment, so that a flow of energy leads to clearly defined trophic structure, biotic diversity and material cycles (i.e. exchange of materials between living and non-living parts).
  • 5Ecosystem function: The flux of energy, organic matter or nutrients in an ecosystem, including the flux of biomass associated with trophic interactions. Functions are expressed as a rate of change of an ecosystem property.
  • 6Ecosystem property: A metric expressing the amount of energy, organic matter or nutrient concentration in an ecosystem.
  • 7Ecological stoichiometry: The study of the balance of energy and the ratios of key elements (primarily C, N and P) in ecological interactions.
  • 8Community property: Any descriptor of the biological community in an ecosystem such as species composition, species richness and phenotypic diversity.
  • 9Niche construction: The process by which organisms modify components of their environment, such as resource distribution or habitat space to affect selection pressures on themselves or other organisms in an ecosystem.
  • 10Zero net growth isocline (ZNGI): Resource supply level at which the growth rate of a population is zero because its rate of reproduction and mortality are equal.