You are on page 1of 12

Global Change Biology (2005) 11, 490–501, doi: 10.1111/j.1365-2486.2005.00919.

Global change alters the stability of food webs


M A R K E M M E R S O N *, M A R T I J N B E Z E M E R w 1 , M A R K D . H U N T E R z and T . H E F I N J O N E S §
*Department of Zoology, University College Cork, Lee Maltings, Prospect Row, Cork, Ireland, wNetherlands Institute of Ecology
(NIOO-KNAW), PO Box 40, 6666 ZG Heteren, The Netherlands, zInstitute of Ecology, University of Georgia, Ecology Building,
Athens, GA 30602-2202, USA, §Cardiff School of Biosciences, Cardiff University, PO Box 915, Cardiff, CF10 3TL, UK

Abstract
Recent research has generally shown that a small change in the number of species in a
food web can have consequences both for community structure and ecosystem processes.
However ‘change’ is not limited to just the number of species in a community, but might
include an alteration to such properties as precipitation, nutrient cycling and
temperature. How such changes might affect species interactions is important, not just
through the presence or absence of interactions, but also because the patterning of
interaction strengths among species is intimately associated with community stability.
Interaction strengths encompass such properties as feeding rates and assimilation
efficiencies, and encapsulate functionally important information with regard to
ecosystem processes. Interaction strengths represent the pathways and transfer of
energy through an ecosystem. We review the best empirical data available detailing the
frequency distribution of interaction strengths in communities. We present the
underlying (but consistent) pattern of species interactions and discuss the implications
of this patterning. We then examine how such a basic pattern might be affected given
various scenarios of ‘change’ and discuss the consequences for community stability and
ecosystem functioning.
Keywords: community, ecosystems, food webs, herbivore, persistence, plant, predators, prey, resilience,
stability

Received 9 September 2004; and accepted 15 October 2004

levels and global warming might effect change in species


Introduction
interactions at local scales. This is important because a
Pioneering early ecologists (for example Odum, 1953; disruption to the strength or arrangement of interspecific
MacArthur, 1955; Elton, 1958) held that the dynamics of interactions in food webs can have consequences for the
complex ecological systems should be more stable than stability of those same systems. For instance, in the early
simpler systems. They based this view on the premise 1970s, Rosenzweig (1971) warned against the artificial
that those ecological systems, which were more species enrichment of ecosystems (to increase productivity) in
rich, have more interspecies pathways along which terms of increased nutrient and energy supply. He
energy can flow. These pathways can be depicted showed that for two trophic-level systems, either
graphically as food web diagrams and energy flow scenario would destroy steady states in simple models.
may be characterized by the trophic interactions that Throughout this paper, we describe a scenario of global
take place among species within the food web. The change (increased productivity) brought about by
biological strength of these trophic interactions details coincident global increases in temperature and CO2
the conversion of abundance or biomass from one concentration. We explore this scenario of change using
trophic level to another. In this paper, we will explore a simple three-species food web consisting of a basal
how increased productivity brought about by rising CO2 resource (plant), primary consumer (herbivore) and
secondary consumer (omnivorous predator).
Correspondence: Mark Emmerson, tel. 1 353 (0)21 490 4190.
Although the study of species interactions has tradi-
fax: 1 353 (0)21 427 0562, e-mail: m.emmerson@ucc.ie tionally been concerned with their effects on community
1
Present address: Nature Conservation and Plant Ecology Group, stability (sensu May, 1973), recent research has focused
Wageningen University and Research Centre, Bornsesteeg 69, 6708 more on how ecosystem processes such as productivity,
PD Wageningen, The Netherlands. nutrient flux and nutrient retention are affected by the

490 r 2005 Blackwell Publishing Ltd


G L O B A L C H A N G E A LT E R S T H E S TA B I L I T Y O F F O O D W E B S 491

loss of species (Chapin et al., 2000; McCann, 2000; Loreau dynamical behaviour of model communities close to
et al., 2001; Madritch & Hunter, 2002). The transfer of equilibrium. The interaction between a predator and its
energy (or carbon) through an ecosystem represents one prey is defined by the coefficient aij and represents the
such ecosystem process and interaction strengths provide negative per-capita effect of a predator species j on a prey
a direct link between the diversity and ecosystem species i. The negative effects of species i on itself are
functioning of a community and its ecological stability denoted aii, while the positive effects of the prey species
(Duffy, 2002). Interaction strengths have been quantified i on the predator j are given by aji. A system of such n
in a number of ways and this general term has been used linear equations describing the dynamics of a set of
variously to describe both (i) the biological flux of species can be represented in matrix algebra terms so that
material or energy from one trophic level to another and the interaction coefficients (aij terms) are the elements of
(ii) per-capita effects of a species on the growth rate of an n  n matrix (here called A). Using matrix algebra, (1)
another. In this paper, we will focus on the latter use of above can be rewritten as:
the term interaction strength.
dX
Concerns over increased global rates of species ¼ X  ðr þ AXÞ;
dt
extinction are driven by the urgent need to understand
and predict the consequences of species loss for the where X and r are vectors containing the densities (Xi,n)
stable and reliable provision of ecosystem services and intrinsic rates (ri,n) of those species present in the
(Duarte, 2000; Engelhardt & Ritchie, 2001; Loreau community or food web. In fact, May (1973) explored
et al., 2001; Lerdau & Slobodkin, 2002). To understand the stability of these model systems at equilibrium; he
how communities might respond locally to global explored the dynamics of a matrix known as the
change (changes in productivity driven by increased Jacobian (C) whose elements (cij terms) were the
temperature and CO2 concentrations), requires that we product of the interaction coefficients contained in A
identify the basic arrangement of species interaction and an n  n matrix containing the vector of nontrivial
strengths in communities and investigate how one equilibrium population densities (Xi where Xi does not
measure of global change might affect that pattern at equal zero) on the diagonal and zeros elsewhere (so
the local ecological scale. To achieve this, we now that cij ¼ aij Xi ). Essentially, the Jacobian matrix de-
review and detail measures of interaction strength, and scribes the dynamics of a community close to an
present what we consider to be some of the best data equilibrium point and details population-level interac-
currently available detailing the patterns of interaction tions. May (1973) explored randomly constructed
strengths in ecological systems; this is necessary for our model communities by assigning interaction strengths
exploration of change. We then define some simple and (cij) from a uniform distribution to the elements of the
biologically reasonable scenarios of ecological change Jacobian (C). He found that in systems with randomly
and explore how the patterning of interaction strengths assigned trophic interactions, an increased species
might be affected by such changes using a Lotka– richness tended to decrease the stability of model
Volterra modelling framework. Our aim in using Lotka– communities. Pimm & Lawton (1977, 1978) investigat-
Volterra dynamics is to make qualitative forecasts ing simple model food webs subsequently showed that
regarding the consequences of change, not to make the patterning of species interactions (that is the
quantitative predictions. Our goal is to make a heuristic presence or absence of omnivorous links and therefore
exploration of these scenarios using this simple model- food web structure alone) could have dramatic effects
ling approach. Finally, we will discuss the consequences on the stability of such systems. Yodzis (1981) working
of ‘change’ for community and ecosystem stability. with a compiled set of 40 real, as opposed to model,
food webs found that the arrangement of species
interaction strengths was essential for food web
What is the pattern of interaction strengths in stability. When the detailed pattern of interaction
communities? strengths in stable food webs was disrupted the
The effects of species’ interactions on community stability resulting webs were, on average, less stable.
were first explored by May (1973) using linear stability These classic studies using linear stability analyses, all
analyses. May used Lotka–Volterra models of the form: made use of the Jacobian matrix for the determination of
0 1 food web stability (May, 1973; Pimm & Lawton, 1977,
dXi Xn
1978; Yodzis, 1981). From these theoretical studies, there
¼ Xi @ri þ aij Xj A; ð1Þ
dt j¼1
are two quantities that emerge as characterizing the
interactions among species, aij the interaction coefficient
where Xi and ri define the population density and and aij Xi the element of the Jacobian matrix at equili-
intrinsic rate of increase of species i, to explore the brium (it should be noted that the Jacobian exists

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


492 M . E M M E R S O N et al.

whether a system is at equilibrium or not). The value aij range of soil food webs that differed in complexity. De
is a per-capita effect and a constant coefficient of the Ruiter et al.’s (1995) formulation of interaction strength
model, while aijXi will be dependent on the density of was calculated for both the negative effects of predators
species i (Xi). The elements of the Jacobian matrix at on prey ðaij Xi Þ and the positive effects of prey on
equilibrium are the partial derivatives of the growth predators ðaij Xi Þ, and was equivalent to the elements of
equations with respect to each species (dfi/dXj, where the Jacobian matrix at equilibrium. The frequency
fi 5 dXi/dt and dfi =dXj ¼ cij ¼ aij Xi ), and describe the distributions of these Jacobian elements also show strong
dynamic forces acting around the equilibrium point. skews towards weak interactions for both positive and
The quantity dfi =dXj ¼ aij Xi represents the effect of negative predator–prey interaction strengths. The stu-
species j on the growth rate (fi) of species i at dies on both interaction coefficients and interaction
equilibrium. Any process (whether intrinsic or extrin- strengths present a consistent pattern of predominantly
sic) that changes the size of species’ populations will weak effects in communities, although it should be
therefore have quantifiable effects on the elements of noted that some studies report differing patterns
the Jacobian matrix (aijXi terms). This provides a useful determined by species body size (Sala & Graham, 2002).
way of exploring how ‘change’ might affect the strength Using a nonlinear modelling approach, McCann et al.
of trophic interactions. Interaction strength is a generic (1998) showed that this empirical pattern of weak
term that refers to the interactions that exist among sets interactions could stabilize community dynamics when
of species and can be measured or quantified in any typically destabilizing links (for example omnivorous
number of ways. To avoid confusion we hereafter links) were weak. Neutel et al. (2002) obtained a similar
explicitly refer to the elements of the matrix A as result using linear stability analyses and showed that
interaction coefficients (per-capita effects) and the the stability of a food web was intimately associated
elements of the Jacobian matrix (C) as interaction with the strength of Jacobian elements in omnivorous
strengths (population-level effects). food web loops.
The empirical study of species’ interaction strengths In summary, a considerable body of evidence
is not trivial, requiring that a predator or consumer be indicates that the patterning of interaction strengths is
removed from a community and the numerical re- vital for food web and community stability. Empirical
sponse of its prey and predators be quantified. data show a strong skew towards weak interaction
Empiricists studying the strength of trophic interactions coefficients and weak Jacobian elements. Theoretical
are logistically constrained in the number of species’ studies suggest that, within food webs, it is the specific
interactions that they can investigate. This constraint is patterning of weak interactions that promotes food web
evident in a number of recent studies that have stability. How might the patterning of species interac-
quantified the strength of species interactions (Paine, tions be affected by predicted changes to the global
1992; Fagan & Hurd, 1994; Raffaelli & Hall, 1996; biosphere? Would a change to the patterning of species
Wootton, 1997). These studies represent some of the interaction strengths caused by such predicted changes
best empirical data available quantifying interaction be sufficient to cause instabilities in community
strengths, but because of the small sample sizes, the dynamics? We next consider a scenario of global change
authors have resorted to bootstrapping techniques to and examine how this scenario might affect the strength
obtain a reliable estimate of the variability associated and patterning of species interactions.
with interaction strength (Paine, 1992; Wootton, 1997).
A recent review (Berlow et al., 1999) deals comprehen-
sively with the behaviour of different empirical mea-
How could global change affect interaction strength?
sures of interaction strength. In this study, we concern
ourselves with a measure of interaction strength known Productivity is influenced by the interacting effects of
as the dynamic index or log ratio, which is equivalent to numerous global changes including elevated atmo-
the interaction coefficient in discrete time versions of the spheric CO2, rising temperatures, altered precipitation
Lotka–Volterra equation (aij) and describes the per-capita patterns, nitrogen deposition and species loss. Our aim
effects of a predator on its prey (see Berlow et al., 1999 is to examine (by whatever mechanism) changes in
for details). Data from a number of well-known studies primary productivity and consequent effects on trophic
(Paine, 1992; Fagan & Hurd, 1994; Raffaelli & Hall, 1996; interaction strengths. Clearly, taking into account all of
Wootton, 1997) indicate that these indices (which these interacting factors and their effects on productiv-
represent some of the best data currently available) are ity is not possible in the space of this paper. To make
skewed strongly towards weak interactions. our task somewhat more manageable, we only consider
De Ruiter et al. (1995) estimated Jacobian elements scenarios of increased global productivity associated
using biomass densities rather than density per se, for a with increasing concentrations of atmospheric CO2.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


G L O B A L C H A N G E A LT E R S T H E S TA B I L I T Y O F F O O D W E B S 493

Predictions suggest that atmospheric CO2 levels will predict (Aber & Melillo, 2001). Temperature effects are
rise from around 350–700 ppm by the end of this particularly difficult to predict because while photo-
century (IPCC, 2001). Associated with this increase in synthesis has a bell-shaped response to temperature,
CO2 levels are: (1) direct effects of atmospheric carbon respiration increases exponentially (Larcher, 1995).
on plant photosynthesis and primary production; (2) Carbon gain relative to loss may therefore decline at
increases in average global temperature (global average high temperatures, resulting in a decline in production.
surface temperature has already increased by 0.6 1C in The responses of various ecosystems to the
the 20th century, IPCC, 2001), and (3) changing patterns interactive effects of changes in CO2, temperature and
of precipitation (Aber & Melillo, 2001). Coupled atmo- precipitation will depend, in part, on their relative
sphere–ocean general circulation models provide us sensitivities to those environmental factors. Simply put,
with the best predictions on how patterns of tempera- an ecosystem whose productivity is limited by
ture and precipitation will change as CO2 levels rise precipitation is more likely to respond to a change in
(IPCC, 2001). Most terrestrial regions will experience an precipitation than one that is limited by temperature
increase in mean annual temperature between 3 and (Schloss et al., 1999). Broad generalizations that operate
10 1C whereas rates of precipitation may either decrease across many different ecosystems therefore, become less
or increase by up to 2 mm day1, depending on region. robust as the number of climatic variables considered is
To begin integrating these climatic predictions into an increased. Our ability to make predictions is also limited
understanding of species interactions, we begin our because models that include multiple interactive effects
analysis with anticipated effects upon primary produc- of global change are still relatively rare (Aber & Melillo,
tion. Then we will suggest how changes in primary 2001). One noteworthy study is that by Yu et al. (2002).
production might influence the strength of trophic This study incorporates predicted changes in CO2
interactions. Critical to this last step is distinguishing concentration, precipitation and temperature from
between changes in the rate of primary production and seven general circulation models into predictions of
changes in the quality of the plant tissue that is ecosystem change in a forest transect in east China.
produced. Whether plants react in a qualitative or While all seven general circulation models predicted
quantitative way, or even both, is likely to have very increases in net primary productivity under elevated
different effects on species interactions. CO2, effects of precipitation were relatively weak
whereas productivity was negatively correlated with
Direct effects of CO2 on primary production. It is generally temperature. Broadleaf forests were predicted to
accepted that photosynthetic rates are limited primarily increase while conifer forests, shrubs and grasses were
by carbon (Drake et al., 1997; Norby et al., 1999; Aber & predicted to decrease. Overall, however, primary
Melillo, 2001). Hence, most models predict that net production was still seen to increase under all
primary production will increase as atmospheric scenarios of elevated CO2 (Yu et al., 2002).
concentrations of CO2 increase. Various models Despite Yu et al. (2002) and predictions made from the
predict changes that range from 0.7% to 1 32.4% bell-shaped relationship between photosynthesis and
(VEMAP, 1995). Overall, the average predicted increase temperature (Aber & Melillo, 2001), increases in
in net primary production is around 20%. In many temperature under rising CO2 levels may still have an
plant communities, primary production will increase overall positive impact on net primary production,
below ground as well as above ground (Tate & Ross, particularly if growing seasons are extended. A meta-
1997) and elevated CO2 may increase fine root analysis of available studies suggests that ecosystem
production of some forest trees by as much as 96% warming should increase plant productivity by an
(King et al., 2001). For most plant species, increased average of 19%, with a 95% confidence interval of 15–
rates of primary production are associated with 23% (Rustad et al., 2001). If true, then the effects of global
increases in relative growth rate (Saxe et al., 1998). For warming on productivity may operate in exactly the
example, in their study of 10 Acacia species, Atkin et al. same direction as the direct effects of CO2 on
(1999) reported that the relative growth rates of Acacia productivity.
trees increased by an average of 10% over a 12-week
period under elevated CO2 conditions. Quality vs. quantity. In a food chain context, the
predicted increases in primary production may
CO2-mediated changes in temperature and precipitation. appear to favour herbivore populations; however,
While the prediction of increased productivity in many empirical studies suggest that plant quality for
response to rising concentrations of CO2 appears to be a consumers will decline as CO2 levels rise (Bezemer &
relatively robust generalization, the effects of concomitant Jones, 1998). This will have implications for the strength
changes in temperature and precipitation are harder to of per-capita trophic interactions between a consumer

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


494 M . E M M E R S O N et al.

and its plant resource. Specifically, the accumulation of changes in temperature and atmospheric CO2
carbon under elevated CO2 dilutes concentrations of concentrations, there is even less information on the
nitrogen in plant tissues by 15–25% (Lincoln et al., 1993; effects of climate change on secondary consumers
Lindroth et al., 1995) thereby increasing C : N ratios (parasitoids and predators). In the few studies
(Ceulemans & Mousseau, 1994; Wilsey, 1996; Hughes & currently available on the direct effects of elevated
Bazzaz, 1997). Given that nitrogen is considered a CO2 on parasitoids, CO2 did not influence Cotesia
limiting factor for many herbivore species (Mattson, melanoscela, a parasitoid of the gipsy moth (Lymantria
1980), dilution of nitrogen under elevated CO2 may dispar), although pre-enclosure mortality of the
place further constraints on the growth rates of parasitoid was slightly increased when CO2 was
herbivores. In addition, the concentrations of carbon- elevated (Roth & Lindroth, 1995). Bezemer et al. (1998)
rich secondary metabolites sometimes increase under also showed that parasitism rates of M. persicae
elevated CO2 (Lindroth et al., 1995; Agrell et al., 2000). remained unchanged in elevated CO2. Using
While there appears to be considerable variation in the mathematical models Hassell et al. (1991) have shown
responses of plant chemical defences to rising CO2 that host–parasitoid relationships are altered by
levels (Hunter, 2001), some classes of tannin, such as environmental change; temperature elevation, for
condensed tannins, seem to respond most frequently. example, may differentially affect developmental rates
To summarize, in nearly every case examined to date, of hosts and parasitoids. Such differences can
foliar nitrogen concentrations decline under elevated potentially result in the breakdown of synchronization
CO2 and, when present, foliar concentrations of between the two populations, which, in turn, may have
condensed tannins increase (Fajer et al., 1989, 1991; major effects on population dynamics.
Johnson & Lincoln, 1991; Lincoln et al., 1993; Lindroth
et al., 1995; Jones & Hartley, 1999).
Models and methods
Herbivores confronted with low-quality plant tissue
would be expected to compensate by eating more. This
Simulating the effects of change
prediction usually holds true, particularly for chewing
insects (Lincoln et al., 1984, 1986, 1993; Fajer et al., 1989; We use a Lotka–Volterra framework, within which to
Lindroth et al., 1993, 1995; Salt et al., 1995; Docherty explore the effects of ‘change’ (defined here as an
et al., 1996; Kinney et al., 1997; Williams et al., 1997; increase in productivity) on interaction strength. To
Whittaker, 2000). Likewise, the area damaged by leaf- investigate how increased productivity might affect
mining insects may also increase, for example, the area species population sizes (Xi), per-capita interaction
of leaf mines on Quercus myrtifolia increase by over 25% coefficients (aij) and, in turn, the elements of the
under elevated CO2, apparently because nitrogen Jacobian matrix (aijXi), we determined the period
concentrations fall by over 11% (Salt et al., 1995; doubling bifurcations for a simple three species food
Stiling et al., 2003). Nonchewing insects have been less chain featuring omnivory in discrete time. The bifurca-
well studied and in general patterns are yet to emerge. tions occur when the stability of a system changes as a
Bezemer et al. (1998), for example, showed that peach model parameter passes through a critical value.
potato aphid (Myzus persicae) abundance was enhanced Essentially, the bifurcations describe how the dynamics
by elevated CO2 and temperature, but in a companion of a species population change as a function of the
study, Bezemer et al. (1999) found that plant and aphid species intrinsic rate of increase. The bifurcation
species significantly influenced the response. diagram describes how a species population will
As well as separating quality and quantity effects, it change from a stable equilibrium, to limit cycles of
is also crucial that we distinguish between overall varying amplitude and periodicity, to chaotic dynamics
levels of defoliation and per-capita consumption by as the intrinsic rate of the basal species increases. In the
insects. Increases in plant productivity and nitrogen- simple food chain investigated here, Species 1 is basal,
mediated declines in insect density can result in lower Species 2 is an herbivore and Species 3 is an omnivore
levels of defoliation on plants despite increases in per- feeding on Species 1 and 2 (see simple food web
capita consumption rates by herbivores (Hughes & detailed in Fig. 1). Species population dynamics were
Bazzaz, 1997, Stiling et al., 2003). A 2-year study of determined using a discrete time version of the Lotka–
herbivore communities under open-topped CO2 Volterra equation;
chambers in scrub oak forest has shown that all X1;tþ1 ¼ X1;t þ X1;t ðb1  a11 X1;t  a12 X2;t  a13 X3;t Þ;
herbivore species decline in density under elevated
X2;tþ1 ¼ X2;t þ X2;t ðd2 þ a21 X1;t  a22 X2;t  a23 X3;t Þ;
CO2 (Stiling et al., 2003).
X3;tþ1 ¼ X3;t þ X3;t ðd3 þ a31 X1;t þ a32 X2;t  a33 X3;t Þ;
While there is little information on how primary
consumers (herbivores) respond, in the longer term, to ð2Þ

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


G L O B A L C H A N G E A LT E R S T H E S TA B I L I T Y O F F O O D W E B S 495

rates of Species 2 and 3, set equal to 0.01 and 0.001,


respectively (in the absence of food these species
populations could not increase as for the basal auto-
trophs. Consequently, they have a negative intrinsic
population growth rate, which is offset by the food that
they eat). The biological justification for this decline
with increasing trophic height being that death rate is
strongly correlated with a species body size. Body size
tends to increase with trophic height and so death rate
would decline. Finally, t is time. Intraspecific terms also
have different meanings for the different species
present in the system. For the basal species a11 5 r1/K1,
while for nonbasal species intraspecific competition is
set equal to 0.1 (a22 and a33). Therefore, as r1 increases,
the strength of basal species intraspecific competition
increases relative to nonbasal species intraspecific
competition.
In this simple food chain, the secondary consumer
(omnivorous predator) is capable of feeding on both the
primary consumer (herbivore) and the basal resource
(plant). A gradient of productivity was established by
incrementing the intrinsic rate of increase (0.01 incre-
ments), of the basal species r1 in the food chain. r1 was
incremented over the interval [0, 4]. To produce the
bifurcations, at each value of r1, the three discrete time
equations’ of system (2) above, were iterated for 40 000
time steps. Initial conditions for each run were set at
X1 5 1, X2 5 0.5, X3 5 0.01, reflecting a biologically
plausible pyramidal population structure (population
density tends to decrease with increasing trophic height
Fig 1 Period doubling bifurcations for a three species omnivor- and is negatively correlated with increasing body size).
ous food web featuring (a) primary producer, (b) primary The population density over the last 500 time steps of
consumer and (c) omnivorous secondary consumer (for the food this series is represented in one dimension as a function
web diagram also shown filled circles indicate the position of the of the basal species intrinsic rate. Over this time interval,
species in the food web). The omnivorous secondary consumer the system of equations converged either to an equili-
feeds on both the primary producer and the primary consumer.
brium value or onto stable attractors in the system.
Three scenarios are represented: (i) Interspecific and consumer
We explored three different scenarios to investigate
species intraspecific interactions are considered constant. (ii)
Herbivorous interactions (aij) are considered a function of basal
how the persistence of this simple three species food
species intrinsic rates (r1), because as plant growth increases, chain might be affected given an increase in basal
plant quality declines and plant consumers must ingest more to species productivity. These differing scenarios repre-
compensate. (iii) Both herbivorous interactions and consumer sent situations where:
assimilation efficiencies are considered to be functions of the (i) Interspecific and nonbasal species intraspecific
primary producers intrinsic rate. This simulates increased plant interactions (a22 and a33) are considered constant over
quality at low levels of productivity and decreased quality at the range of basal species intrinsic rates. The per-capita
high levels of plant productivity (see accompanying text for effects of prey on predators (aji) are related to the per-
further details). capita effects of predators on prey (aij) by a predator’s
conversion efficiency so that aji 5 e  aij, where e is 0.1.
where birth and death processes have slightly different For the simulations presented here, the interaction
meanings for different trophic levels. b1 defines the coefficient matrix A was parameterized in the following
positive per-capita birth rate of Species 1; basal species way:
population growth is logistic so that b1 5 r1/K1, r1 being 2 3
r
the intrinsic rate of increase of Species 1 and K1 the K 0:5 0:05
carrying capacity which is set equal to 1. For nonbasal A ¼ 4 0:05 0:1 0:3 5:
species, d2 and d3 define the negative per-capita death 0:0005 0:03 0:1

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


496 M . E M M E R S O N et al.

(ii) As the intrinsic rate of a plant increases, plant interactions of the primary and secondary consumers
quality may decline. To compensate for this a herbivore (Jonsson & Ebenman, 1998).
must compensate and ingest more plant material to
fulfil its nutritional requirements. Therefore, as the
Results
basal species’ intrinsic rate increases so too will the
herbivores per-capita effects (aPH), where P and H refer
Population size
to plant and herbivore, respectively. To simulate this
scenario, we assumed herbivorous interactions (a12 and We constructed the period doubling bifurcations for
a13) in the simple food chain to be functions of the basal each of the scenarios ((i)—(iii)) (Fig. 1). When interac-
species’ intrinsic rate so that: tion coefficients are independent of basal species
aPH ¼ Fðr1 Þ; ð3Þ intrinsic rate (Fig. 1, a–c, i) the abundance of both
primary and secondary consumer is low, being largely
where unresponsive to the behaviour of basal species popula-
r1 tion dynamics. However, the simple food web is not
Fðr1 Þ ¼ l þ u : ð4Þ
b þ r1 feasible when the intrinsic rate of the basal species is
43 (there are no positive population densities). When
Here, l and u are constants, approximately defining the
herbivorous interaction coefficients are considered to be
lower and upper bounds to the interaction coefficient
a function of basal species intrinsic rate (Fig. 1, a–c, ii),
and were set at 0.2 and 0.8 for the primary consumer
basal species population dynamics remain largely
and 0.002 and 0.008 for the secondary consumer,
unaffected. Nonbasal species populations, on the other
respectively, b is a constant, set equal to unity. For
hand, increase monotonically until the first basal
interaction coefficients, this choice of l and u give a
species period doubling occurs. Subsequent increases
range from 0.2 to 1 for the primary consumer, and
in basal species intrinsic rate result in a slight decline in
0.002 to 0.01 for the secondary consumer. Using this
nonbasal species population size and an increased
function, the interaction coefficient asymptotes as the
variability of each species population. When both,
intrinsic rate of the basal species increases. The
predator assimilation efficiency and interaction coeffi-
biological justification for choosing this function is that
cient are considered functions of basal species intrinsic
herbivores are only capable of handling food at some
rate (Fig. 1, a–c, iii), the population size of nonbasal
maximal rate. This is essentially a Holling Type II
species both declines substantially and becomes more
functional response, but in this study, rather than
variable as basal species intrinsic rate increases. The
ingestion being a function of prey density, it is a
decline in average population size and coincident
function of prey growth rate.
increase in variability mean, that each population of
(iii) At low levels of productivity, plant quality
these species may well be more prone to stochastic
should be high, when prey intrinsic growth rates are
environmental perturbations as productivity increases.
low, the per-capita effects of herbivores on prey are
This is especially so when the scenarios that lead to a
consequently small. However, the benefits to predators
suggested increase in productivity also predict in-
should be large, because plant quality is high. It is
creased occurrence of extreme climatic events. Despite
difficult to represent this if it is assumed that the
this, both scenarios ((ii) and (iii)) result in consumer
ecological efficiency of the predator remains constant.
population sizes that are still on average much larger
To simulate an increase in plant quality at low levels of
than for Scenario (i) where interaction coefficients and
plant productivity we assume that the ecological
assimilation efficiencies are independent of basal
efficiency of a predator is also a function of plant
species productivity.
intrinsic growth rate so that:
e ¼ Fðr1 Þ; ð5Þ
Patterns of stability
where
The mechanisms detailed above such as increased
r1
Fðr1 Þ ¼ u  l : ð6Þ consumption with increased productivity, underlie
b þ r1
herbivorous interactions and have implications for the
As plant productivity increases, so the ability of a stability of the simple three species food chain
predator to convert plant biomass into predator examined. Stability in the present context refers to the
biomass decreases asymptotically reflecting the fact persistence of species in the food chain. For all three
that plant quality declines. Here, we set l and u both at scenarios, persistence is determined largely by basal
0.3 this provides a biologically plausible range of 0.3– species population dynamics at high levels of produc-
0.06 for the ecological efficiency used in the herbivorous tivity. When interaction coefficients are insensitive to

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


G L O B A L C H A N G E A LT E R S T H E S TA B I L I T Y O F F O O D W E B S 497

basal species productivity the chain does not persist


when the basal species intrinsic rate exceeds 3 (r143).
When only interaction coefficients are considered to be
a function of basal species intrinsic rate, no species are
lost from the food chain up to an intrinsic rate of 3.16
(and the food chain persists). When both interaction
coefficient and assimilation efficiency are considered as
functions of r1, the chain does not persist above
r1 5 3.11. At no point in the simulations do either
species 2 or 3 become extinct, affecting the persistence
of the entire food chain. This potentially could occur
with either the primary consumer becoming extinct,
leaving the omnivorous secondary consumer or the
omnivorous secondary consumer itself could become
extinct, leaving only the basal species and the primary
consumer. Despite the three differing scenarios, we find
that there is little qualitative difference in terms of
species persistence. However, the differing scenarios do
affect consumer population sizes differently. What,
then, are the implications for the patterning of Jacobian
elements (aijXi)?

Patterning of interaction strengths


To examine the distribution of predatory interaction Fig. 2 The consequences of Scenarios (i)–(iii) (see text) for (a)
coefficients or Jacobian elements for a simple three- Patterning of Jacobian elements (c12) as a function of primary
species food chain is inappropriate, as the three producers intrinsic rate (r1) using Lotka–Volterra dynamics in
predatory food chain interactions detailed here do not discrete time (Eqn (2)), and (b) Stability of three species food
constitute a comprehensive distribution. Instead, we web measured as return time (1/Re(l1)) to equilibrium. For the
determination of return time the model parameterizations
consider what happens to the Jacobian elements in the
suggested by Scenarios (i)–(iii) are used to parameterize
food chain as basal species productivity increases. Put
Lotka–Volterra dynamics in continuous time (see Eqn (1)).
simply, the patterning of Jacobian elements (here a12 see
Fig. 2a) mirrors the patterning of each species popula-
tion size. As basal species productivities increase, and but rather exist in proximity to some form of cyclic or
each species population undergoes bifurcations, so the chaotic attractor.
Jacobian elements become more variable. For Scenario
(ii), Jacobian elements become increasingly negative
Continuous time
(stronger) with increases in productivity. As interaction
coefficients increase as a function of basal species Until now, we have only considered the discrete time
intrinsic rate and assimilation efficiency remains con- version of the Lotka–Volterra equation ((2) above) and
stant, each consumer population size increases. For we have seen that the system converges to an
Scenario (iii), the Jacobian elements become weaker but equilibrium point or stable attractor in phase space. If
still increase in variability. This decline in the strength we use the same parameter values for the interaction
of Jacobian elements is a consequence of the predatory coefficients and birth and death processes using the
interaction coefficients increasing as basal species continuous form of the Lotka–Volterra equation (Eqn
productivities increase while the predator’s assimila- (1)), then a local stability analysis of the omnivorous
tion efficiency declines, resulting in a smaller predator food chain can be carried out. A stability analysis
population size overall. Given the scenarios we explore determines the ability of a community to return to
here, the patterning of the Jacobian elements seems to equilibrium following a small perturbation that moves
have little consequence for the persistence of the food a community away from the equilibrium point. Such a
chain, which remains feasible over the gradient of basal stability analysis reveals that in close proximity to an
species productivities. At high productivities, the attractor the food chain is always stable (see Fig. 2b),
simple food chain and its constituent species popula- although if a perturbation moves a community a
tions do not exist at a simple equilibrium (Fig. 1a–c), large distance from the attractor (which defines the

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


498 M . E M M E R S O N et al.

equilibrium point) the system may not return to that and primary productivity, while consumer controlled
same attractor. Stability is determined by evaluating the models predict that the interaction strength of the
real parts of all eigenvalues l of the Jacobian matrix (C), highest trophic level should increase with productivity
if these are all negative then the food web is stable. The (Chase et al., 2000). We have envisaged a scenario
reciprocal of the dominant eigenvalue (1/Re(l1)) is a whereby, primary consumer consumption increases as
measure of resilience or the return time to equilibrium plant productivity increases because quality declines.
following a small perturbation from the equilibrium It could also be argued that secondary consumer
point). A long return time indicates a system with a low consumption of primary consumers will increase. In
resilience and vice versa. The behaviour of the Lotka– insect systems, primary consumers would be required to
Volterra system in continuous time is quite different; eat more, with the consequence that growth is slowed,
we do not see the system settling onto limit cycles or this could lead to a higher exposure of primary
chaotic orbits. When interaction coefficients are invar- consumers to their natural enemies (slow growth–high
iant with respect to basal species growth rate (r1), and mortality hypothesis). We have not considered such
primary consumer population size is small, the Jaco- effects in the present study.
bian elements are weak and return times are long, Similarly, we have not explicitly considered the
irrespective of productivity (Scenario (i)). When the dynamical consequences of ecological stoichiometry.
interaction coefficients increase with intrinsic rate, but The premise of the present paper is that resource
assimilation efficiency remains constant, return times quality will decline because of increased plant produc-
decline asymptotically as Jacobian elements become tivity (simplistically assuming a scenario of elevated
stronger (Scenario (ii)). For Scenario (iii), when both CO2 and increased temperature) and that consequently
interaction coefficients and assimilation efficiency are a consumers will have to eat more to sustain their
function of r1, the Jacobian elements detailing the nutritional requirements. In the present study, we do
herbivorous interactions become weaker as productiv- not explicitly explore the dynamics of nutrient flows
ity increases and return times increase. The implication across trophic levels and an exploration of such models
is that when Jacobian elements are weak, the system might reveal quite different dynamics (see Sterner&
takes longer to recover from small perturbations close Elser (2002) for a comprehensive review). Surprisingly,
to equilibrium. Weak interactions might, therefore, this is an area where there is an equal paucity of
delay the recovery of food webs. fundamentally important empirical data.
To predict the consequences of global change will
require that we are able to extrapolate the findings of
Discussion
simplistic modelling studies such as those detailed here
Using Lotka–Volterra models, we have incorporated to large-scale ‘real’ systems. Yet, usable and relevant
some additional basic assumptions regarding the likely descriptions of real systems describing the size and
response of consumer–resource interactions to different content of different nutrient pools, biomass and trophic
scenarios of change. We find that these assumptions structure simply do not exist. Clearly, there is an urgent
have implications for the stability of real communities. need to collect such relevant data. Unfortunately, such
Increased herbivore consumption associated with a studies would be very descriptive and consequently
decline in plant quality results in an overall decline in difficult to fund. Predicting the consequences of global
herbivore population size when a predator’s conversion change will require a combination of theoretical and
efficiency is also considered a function of basal species empirical approaches. Well-resolved descriptions of
intrinsic rate (ri). A decline in herbivore densities is also ‘real’ systems are desperately needed if we are to make
seen in empirical studies investigating the effects of CO2 predictions regarding the consequences of global
enrichment on plant growth (Hughes & Bazzaz, 1997, change as envisaged here.
Stiling et al., 2003). Evidence for increased per-capita Recently, Emmerson & Raffaelli (2004) have studied
consumption is mixed; McNaughton et al. (1989) for the effects of body-size patterns in food webs and have
example have shown that there is a positive relationship used body-size ratios of predators and prey to predict
between primary production and the amount of that the strength of trophic interactions in food webs using
productivity that is consumed by herbivores in grass- Lotka–Volterra models as used here. They found that
lands (while McNaughton’s study involved a compar- predator prey body-size ratios raised to the power of 3/
ison of sites that varied in productivity, there was no 4 could be used to parameterize complex food webs for
explicit test in McNaughton’s work of changes in which body-size data were known. They hypothesized
quality). In theoretical studies, there have been contrast- that the 3/4 scaling of the predator–prey body-size ratio
ing viewpoints. Resource-controlled models assume no was a reflection of basal metabolic rate scaling with
relationships between the strength of consumer control body size. Basal metabolic rate is also strongly

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


G L O B A L C H A N G E A LT E R S T H E S TA B I L I T Y O F F O O D W E B S 499

temperature dependent and clearly scenarios of global communities. When ‘change’ is envisaged as an
warming also have implications for the strength of increase in productivity: (1) herbivorous interaction
trophic interactions via a quite different mechanism coefficients (per-capita effects) will become stronger as
than those we envisage here. The temperature depen- consumers fulfil their dietary requirements, faced with
dence of basal metabolic rate clearly has implications a fast-growing poor-quality food resource. While per-
for food web dynamics and ecosystem functioning. capita interactions could increase, the strength of
Linking the study of ecosystem dynamics through the population-level effects (described by the Jacobian
study of interacting populations, metabolic theory and matrix) might decline if population sizes coincidentally
ecological stoichiometry would be a fruitful area of decrease. Neutel et al. (2002) using complex soil food
research. webs have shown that the pattern of weak interactions
In the present study, we chose a simple food web and in omnivorous loops is vital for food web stability. A
a range of parameter values that produced feasible disruption to the patterning of such interactions in
communities (where all species populations have more complex food webs than the simplistic web we
positive densities) and have explored how changes to explore here could have dramatic consequences for
these parameter values affect community feasibility. At food web and ecosystem stability. (2) Average popula-
first, this might seem trivial but feasibility is of tion size of consumers will likely decline as a
paramount importance. The exploration of complex consequence of poor-quality food resources; (3) Jaco-
food web dynamics is problematic mainly because bian elements will become weaker as a consequence of
finding parameter values that result in feasible com- reduced population size; and (4) species populations
munities is very difficult. We explored three scenarios will become more variable and may take longer to
that might affect a community’s coordinates in para- recover from environmental or anthropogenic pertur-
meter space and investigated whether predicted sce- bations. This last point (4), is important as extreme
narios of change might lead a community along some stochastic environmental effects are predicted to in-
trajectory into an unfeasible region, where one or more crease with predicted scenarios of global warming. The
species populations would decline or become extinct. In combined effect of small population size, increased
general for the scenarios we have explored, increased population variability and increased occurrence of
productivity does not result in decreased community stochastic environmental effects will likely result in an
persistence. However, we have investigated just one increased chance of extinction for the resulting popula-
possible food web structure. There are many possible tions. In conclusion, empirical validation of our
(simple) food web scenarios that remain to be explored, simulations is necessary but on the basis of this study
for example those that include apparent competition or communities may become inherently less stable, that is
intraguild predation. less resilient, smaller, more variable and hence more
Our discussion of change has been limited to the prone to extinction as they respond to predicted global
effects of increased productivity on the strength of a environmental change.
class of consumer resource interactions (plant–herbi-
vore). We have not considered other forms of change
for example; changes in the size distribution of Acknowledgements
harvested populations, and how this might affect the Special thanks to Bob Paine, Tim Wootton, Bill Fagan, Dave
strength or existence of size-based feeding relation- Raffaelli and Peter deRuiter for the provision of raw data. John
ships. We have not considered the effects of increased Pitchford for valuable comments. We thank all those that
O3 on ultraviolet light radiation levels and how this participated in discussion during the ‘Food webs in a changing
world’ meeting, Texel, the Netherlands. We also thank three
might affect species in food webs detrimentally. There
anonymous reviewers for constructive and useful comments.
is a range of climatic factors that we have not
considered and importantly, these remain to be ex-
plored. We have not considered how species might References
adapt or respond evolutionarily to possible changes in
the global biosphere and such scenarios remain an area Aber JD, Melillo JM (2001) Terrestrial Ecosystems, 2nd edn.
Academic Press, San Diego.
ripe for research. Here, we show that simplistic models
Agrell J, McDonald EP, Lindroth RL (2000) Effects of CO2 and
can provide important mechanistic insights into the
light on tree phytochemistry and insect performance. Oikos,
causes of empirically observed patterns in real com- 88, 259–272.
munities and the types of patterns we might expect to Atkin OK, Schortemeyer M, McFarlane N et al. (1999) The
see as the global biosphere changes. response of fast- and slow-growing Acacia species to elevated
Even within the limited scope of this study, we can atmospheric CO2: an analysis of the underlying components
suggest a number of consequences of ‘change’ for of relative growth rate. Oecologia, 120, 544–554.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


500 M . E M M E R S O N et al.

Berlow EL, Navarrete SA, Briggs CJ et al. (1999) Quantifying occidentalis (Thysanoptera: thripidae) and the common milk-
variation in the strengths of species interactions. Ecology, 80, weed, Asclepias syriaca. Oecologia, 109, 286–290.
2206–2224. Hunter MD (2001) Effects of elevated atmospheric carbon
Bezemer TM, Jones TH (1998) Plant–insect herbivore interactions dioxide on insect-plant interactions. Agricultural and Forest
in elevated atmospheric CO2: quantitative analyses and guild Entomology, 3, 153–159.
effects. Oikos, 82, 212–222. Intergovernmental Panel on Climate Change (IPCC) (2001)
Bezemer TM, Jones TH, Knight KJ (1998) Long-term effects of Climate Change 2001: Synthesis Report. Cambridge University
elevated CO2 and temperature on populations of the peach Press, New York.
potato aphid Myzus persicae and its parasitoid Aphidius Johnson RH, Lincoln DE (1991) Sagebrush carbon allocation
matricariae. Oecologia, 116, 128–135. patterns and grasshopper nutrition: the influence of CO2
Bezemer TM, Knight KJ, Newington JE et al. (1999) How general enrichment and soil mineral limitation. Oecologia, 87, 127–134.
are aphid responses to elevated atmospheric CO2. Annals of the Jones TH, Hartley SE (1999) A protein competition model of
Entomological Society of America, 92, 724–730. phenolic allocation. Oikos, 86, 27–44.
Ceulemans R, Mousseau M (1994) Tansley review no 71: effects Jonsson T, Ebenman B (1998) Effects of predator–prey body size
of elevated atmospheric CO2 on woody plants. New Physiol- ratios on the stability of food chains. Journal of Theoretical
ogist, 127, 425–446. Biology, 193, 407–417.
Chapin F III, Zavelta E, Eviner V et al. (2000) Consequences of King J, Pregitzer K, Zak D et al. (2001) Fine-root biomass and
changing biodiversity. Nature, 405, 234–242. fluxes of soil carbon in young stands of paper birch and
Chase JM, Leibold MA, Downing AL et al. (2000) The effects of trembling aspen as affected by elevated atmospheric CO2
productivity, herbivory, and plant species turnover in grass- and trophospheric O3. Oecologia, 128, 237–250.
land food webs. Ecology, 81, 2485–2497. Kinney KK, Lindroth RL, Jung SM et al. (1997) Effects of CO2 and
de Ruiter PC, Neutel A-M, Moore JC (1995) Energetics, patterns NO3 availability on deciduous trees: phytochemistry and
of interaction strengths, and stability in real ecosystems. insect performance. Ecology, 78, 215–230.
Science, 269, 1257–1260. Larcher W (1995) Physiological Plant Ecology, 3rd edn. Springer
Docherty M, Hurst DK, Holopainen JK et al. (1996) Carbon Verlag, Berlin.
dioxide induced changes in beech foliage cause female beech Lerdau M, Slobodkin L (2002) Trace gas emissions and species-
weevil larvae to feed in a compensatory manner. Global Change dependent ecosystem services. Trends in Ecology and Evolution,
Biology, 2, 335–341. 17, 309–312.
Drake B, Gonzalez-Meler M, Long SP (1997) More efficient plants: Lincoln DE, Couvet D, Sionit N (1986) Response of an insect
a consequence of rising atmospheric CO2. Annual Review of herbivore to host plants grown in carbon dioxide enriched
Plant Physiology and Plant Molecular Biology, 48, 607–637. atmospheres. Oecologia, 69, 556–560.
Duarte CM (2000) Marine biodiversity and ecosystem services: Lincoln PE, Fajer ED, Johnson RH (1993) Plant-insect herbivore
an elusive link. Journal of Experimental Marine Biology and interactions in elevated CO2. Trends in Ecology and Evolution, 8,
Ecology, 250, 117–131. 64–68.
Duffy JE (2002) Biodiversity and ecosystem function: the Lindroth RL, Arteel GE, Kinney KK (1995) Responses of three
consumer connection. Oikos, 99, 201–219. saturniid species to paper birch grown under enriched CO2
Elton CS (1958) The Ecology of Invasions by Animals and Plants. atmospheres. Functional Ecology, 9, 306–311.
Methuen, London. Lindroth RL, Kinney KK, Platz CL (1993) Responses of
Emmerson MC, Raffaelli DG (2004) Predator–prey body size, deciduous trees to elevated atmospheric CO2: productivity,
interaction strength and the stability of a real food web. Journal phytochemistry, and insect performance. Ecology, 74, 763–
of Animal Ecology, 73, 399–409. 777.
Engelhardt KAM, Ritchie ME (2001) Effects of macrophyte Lincoln DE, Sionit N, Strain BR (1984) Growth and feeding
species richness on wetland ecosystem functioning and response of Pseudoplusia includens (Lepidoptera: Noctuidae) to
services. Nature, 411, 687–689. host plants grown in controlled carbon dioxide atmospheres.
Fagan WF, Hurd LE (1994) Hatch density variation of a Environmental Entomology, 13, 1527–1530.
generalist arthropod predator: population consequences and Loreau M, Naeem S, Inchausti P et al. (2001) Biodiversity and
community impact. Ecology, 75, 2022–2032. ecosystem functioning: current knowledge and future chal-
Fajer ED, Bowers MD, Bazzaz FA (1989) The effects of enriched lenges. Science, 294, 804–808.
carbon dioxide atmospheres on plant–insect herbivore inter- MacArthur RH (1955) Fluctuations of animal populations and a
actions. Science, 243, 1198–1200. measure of community stability. Ecology, 36, 533–536.
Fajer ED, Bowers MD, Bazzaz FA (1991) The effects of enriched Madritch MD, Hunter MD (2002) Phenotypic diversity influ-
CO2 atmospheres on the buckeye butterfly, Junonia coenia. ences ecosystem functioning in an oak sand hills community.
Ecology, 72, 751–754. Ecology, 83, 2084–2090.
Hassell MP, May RM, Pacala SW et al. (1991) The persistence of Mattson WJ (1980) Herbivory in relation to plant nitrogen
host–parasitoid associations in patchy environments. 1. A content. Annual Review of Ecology and Systematics, 11, 119–
general criterion. The American Naturalist, 138, 568–583. 161.
Hughes L, Bazzaz FA (1997) Effect of elevated CO2 on May RM (1973) Stability and Complexity in Model Ecosystems, 2nd
interactions between the western flower thrips, Frankliniella edn. Princeton University Press, Princeton.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501


G L O B A L C H A N G E A LT E R S T H E S TA B I L I T Y O F F O O D W E B S 501

McCann K, Hastings A, Huxel GR (1998) Weak trophic Saxe H, Ellsworth DS, Heath J (1998) Tansley Review No. 98.
interactions and the balance of nature. Nature, 395, 794–798. Tree and forest functioning in an enriched CO2 atmosphere.
McCann KS (2000) The diversity stability debate. Nature, 405, New Phytologist, 139, 395–436.
228–233. Schloss AL, Kicklighter DW, Kaduk J et al. (1999) Comparing
McNaughton SJ, Oesterheld M, Frank DA et al. (1989) Ecosys- global models of terrestrial net primary productivity (NPP):
tem-level patterns of primary productivity and herbivory in comparison of NPP to climate and the Normalized Difference
terrestrial habitats. Nature, 341, 142–144. Vegetation Index (NDVI). Global Change Biology, 5, 25–34.
Neutel A-M, Heesterbeek JAP, de Ruiter PC (2002) Stability in Sterner RW, Elser JJ (2002) Ecological stoichiometry: The biology
real food webs: weak links in long loops. Science, 296, 1120– of elements from molecules to biosphere. Princeton University
1123. Press, Princeton, NJ, USA.
Norby RJ, Willschleger SD, Gunderson CA et al. (1999) Tree Stiling P, Moon DC, Hunter MD et al. (2003) Elevated CO2 lowers
responses to rising CO2 in field experiments: implications for relative and absolute herbivore density across all species of a
the future forest. Plant Cell and Environment, 22, 683–714. scrub-oak forest. Oecologia, 134, 82–87.
Odum EP (1953) Fundamentals of Ecology. Saunders, Philadelphia. Tate KR, Ross DJ (1997) Elevated CO2 and moisture effects on
Paine RT (1992) Food-web analysis through field measurement soil carbon storage and cycling in temperate grasslands. Global
of per capita interaction strength. Nature, 355, 73–75. Change Biology, 3, 225–235.
Pimm SL, Lawton JH (1977) Number of trophic levels in VEMAP (1995) Vegetation/ecosystem modelling and analysis
ecological communities. Nature, 268, 329–331. project: comparing biogeography and biogeochemistry mod-
Pimm SL, Lawton JH (1978) On feeding on more than one els in a continental-scale study of terrestrial ecosystem
trophic level. Nature, 275, 542–544. responses to climate change and CO2 doubling. Global
Raffaelli DG, Hall SJ (1996) Assessing the relative importance of Biogeochemical Cycles, 9, 407–437.
trophic links in food webs. In: Food Webs: Integration of Patterns Whittaker RJ (2000) Scale, succession and complexity in island
and Dynamics (eds Polis GA, Winemiller KO), pp 185–191. biogeography: are we asking the right questions? Global
Chapman & Hall, New York. Ecology and Biogeography, 1, 75–85.
Rosenzweig ML (1971) Paradox of enrichment: destabilization Williams RS, Lincoln DE, Thomas RB (1997) Effects of elevated
of exploitation ecosystem in ecological time. Science, 171, CO2-grown loblolly pine needles on the growth, consumption,
385–387. development, and pupal weight of red-headed pine sawfly
Roth SK, Lindroth RL (1995) Elevated atmospheric CO2 effects larvae reared within open-topped chambers. Global Change
on phytochemistry, insect performance and insect parasitoid Biology, 3, 501–511.
interactions. Global Change Biology, 1, 173–182. Wilsey BJ (1996) Plant responses to elevated atmospheric CO2
Rustad L, Campbell J, Marion G et al. (2001) A meta-analysis of among terrestrial biomes. Oikos, 76, 201–205.
the response of soil respiration, net nitrogen mineralization, Wootton T (1997) Estimates and tests of per capita interaction
and aboveground plant growth to experimental ecosystem strength: diet, abundance, and impact of intertidally foraging
warming. Oecologia, 126, 543–562. birds. Ecological Monographs, 67, 45–64.
Sala E, Graham MH (2002) Community-wide distribution of Yodzis P (1981) The stability of real ecosystems. Nature, 289,
predator–prey interaction strength in kelp forests. Proceedings 674–676.
of National Academy of Sciences, USA, 99, 3678–3683. Yu M, Gao Q, Liu Y et al. (2002) Responses of vegetation
Salt DT, Brooks GL, Whittaker JB (1995) Elevated carbon dioxide structure and primary production of a forest transect in
affects leaf-miner performance and plant growth in docks eastern China to global change. Global Ecology and Biogeogra-
(Rumex spp.). Global Change Biology, 1, 153–156. phy, 11, 223–236.

r 2005 Blackwell Publishing Ltd, Global Change Biology, 11, 490–501

You might also like