Towards a food web perspective on biodiversity and ecosystem

CHAPTER 8

Towards a food web perspective onbiodiversity and ecosystemfunctioningBradley Cardinale, Emmett Duffy, Diane Srivastava, Michel Loreau,Matt Thomas, and Mark Emmerson

8.1 Introduction

One of the most common questions asked byresearchers across a variety of scientific disciplines is‘How does the number of nodes connected togetherinto a network influence the efficiency and reliabilityof that network?’. While social scientists and epide-miologists might think of ‘nodes’ and ‘connections’as people interacting within a social network, com-puter scientists, neurologists, and civil engineerswould instead think of servers connected together ina world-wide web, synapses connecting neurons inthe brain, or hubs connecting to other hubs in atransportation network or telecommunications grid(Albert and Barabasi 2002, Newman 2003). Regard-less of the particular study system, all of theseindividuals ask similar questions about how thenumber of nodes and connections among nodesinfluence the efficiency and reliability by whichinformation, disease, energy, or matter is transmittedthroughout that network.Within the field of ecology, one of the oldest and

most fundamental questions asked by researchers is‘How does the number of species interacting withina food web influence the efficiency and reliabilityby which energy and matter are transmittedthrough that web?’. Research on this topic can bebroadly divided into two foci. Historically, muchattention in ecology has focused on identifyingthose taxa that are the most influential nodes in afood web. For many years, it has been thought thatsome subset of species might represent ‘hubs’ ofinteractions and/or exhibit such strong interactions

that they exert a disproportionate influence overfood web dynamics. This idea has fueled muchdebate over the prevalence of omnivory in foodwebs (Polis and Strong 1996, Thompson et al. 2007,Yodzis 1984) and whether the increased number offeeding links that result from omnivory increases ordecreases the stability of energy flow through afood web (McCann et al. 1998, MacArthur 1955).Identifying species that represent influential nodeshas also been one of the primary goals in the searchfor ‘ecosystem engineers’ (Jones et al. 1994), ‘key-stone species’ (Paine 1966, Power et al. 1996) orother types of ‘strong interactors’ (Wootton andEmmerson 2005) that might have cascading effectson the diversity and biomass of species at a varietyof different trophic levels (Paine 1966, Carpenteret al. 1987, Elser et al. 1988).In the 1990s, ecologists began to pursue a slightly

different perspective on food webs. This perspec-tive focused not on the cascading impacts of indi-vidual species, but rather on how the number ofspecies that comprise any single trophic level mightcontrol fluxes of energy and matter. Research in thisarea was generally referred to as Biodiversity effectson Ecosystem Functioning (BEF for short), and wasoften justified on grounds that (1) loss of biologicaldiversity ranks among the most pronounced chan-ges to the global environment (Sala et al. 2000,Pimm et al. 1995), and (2) reductions in diversity,and corresponding changes in species composition,may alter fluxes of energy and matter that underlieimportant services that ecosystems provide to

105

humanity (e.g. production of food, pest/diseasecontrol, water purification, etc. Daily 1997, Chapin etal. 1998). While the value of BEF research for conser-vation biology andmanagement has been questionedby some (Schwartz et al. 2000, Srivastava and Vellend2005), there is a more fundamental reason for therecent prominence of this topic. BEF is one of the fewresearch topics in ecology that examines how bio-logical variation per se acts as an independent variableto regulate key community and ecosystem-level pro-cesses (Naeem 2002b). Understanding the ecologicalconsequences of variation among species has shownmuch potential to complement our historical focus onthe ecological impacts of highly influential species.

Although the BEF paradigm has evolved consid-erably over the past 15–20 years andbeen increasinglyapplied to a variety of organisms and ecosystems,studies have continued to focus mostly on simplified‘model’ communities. In fact, the typical experimenthasmanipulated an average of just seven species in anaverage of just one trophic group (Fig. 8.1(a)). Suchminimal levels of complexity are far from the realitiesof natural food webs, where, even for some of thesimplest communities, species interact within webscomposed of hundreds of species spanning manytrophic levels (Lafferty et al. 2006, Polis 1991,Martinez1992). At present, it is unclear whether such over-

simplifications are justified, or alternatively, whetherthey have led ecologists to potentially erroneousconclusions. However, what is clear is that a largebody of research in ecology has shown that interac-tions of species across trophic levels can have cas-cading impacts that influence the diversity andbiomass of organisms at numerous levels in a foodweb.At the very least, this suggests that the past focusof BEF on diversity within single trophic levels maybe insufficient to quantitatively predict, and perhapseven qualitatively reflect, the ecological consequencesof diversity loss.In this chapter, we continue with the development

of an idea that originated with other authors whohave argued that, in order to understand howextinction alters the functioning of whole ecosys-tems, ecologists will likely need to merge modernparadigms of BEF with much more classic ideas infood web ecology that consider not only the func-tional role of diversity within trophic levels, but theinteractions of species across trophic levels (Duffyet al. 2007, Bruno and Cardinale 2008, Petchey et al.2004a). Our chapter is organized as follows. In Sec-tion 8.2 we briefly review five hypotheses about howfluxes of energy and matter through a food webmight depend on the diversity of species comprisinga web. Those hypotheses are divided into those that

01 2 3 4

Trophic level

5 6

5

10

15

Num

ber

spec

ies

20

25 100

80

60

40

20

% S

tud

ies

0

(b)(a)

Figure 8.1 (a) Summary of the biological complexity of biodiversity-ecosystem functioning (BEF) studies performed to date. On the x-axis is thenumber of trophic levels included in different experiments. On the left-hand y-axis (plotted as grey bars) is the mean number of species per trophiclevel. On the right hand y-axis (plotted as triangles) is the percentage of studies that have included 1, 2, or more trophic levels. Note that 93 per centof BEF experiments have focused on a single trophic level composed of a mean seven species. (b) An example of the complexity of a real, yet still relativelysimple natural food web in a salt marsh (from Lafferety et al. 2007). Note that within this system there are dozens of species (nodes) and hundreds of feedinglinks (lines connecting nodes) among plants, herbivores, predators and parasites that span six or more trophic levels. Figure reproduced with permissionfrom K. Lafferty.

106 B I OD I V E R S I T Y , E CO S Y S T EM FUNC T I ON I NG , AND HUMAN WE L L B E I NG

contrast diversity effects within different trophiclevels versus those that focus on diversity effectsacross trophic levels. In Section 8.3 we outline theempirical support for or against these hypotheses,emphasizing that most are still unresolved and inneed of testing. In the final Section 8.4, we outlinejust a few of the areas of research that we believewill be fruitful as ecologists move towards an inte-gration of BEF into food-web ecology.

8.2 Five early hypotheses aboutmulti-trophic biodiversity and ecosystemfunction

8.2.1 Diversity effects within trophic levels

8.2.1.1 Top-down effects of diversity grow increasinglystrong at higher trophic levels

C1 Cn

H1 Hn

P1 Pn

N

+++

++

+

Early hypotheses proposed that species extinctionfrom higher trophic levels was likely to have greaterimpacts on the functioning of ecosystems thanextinction from lower trophic levels (Table 8.1).Duffy (2002) argued that three characteristicspotentially make ecological processes more sensitiveto extinction by consumers than plants: (1) becausespecies at higher trophic levels have lower popula-tion sizes and are under stronger anthropogenicpressure than most wild plants, higher trophic levelsface greater risks of extinction and higher rates ofspecies loss; (2) consumer assemblages have loweroverall richness and higher degrees of resourcespecialization, leading to less ‘functional redun-dancy’ and limited potential for surviving species tocompensate for processes performed by lost coun-terparts; and (3) unlike plants, consumers often have

impacts on processes that are disproportionate totheir abundance or biomass. Duffy’s (2002) paperwas one of the first to call for a merger of BEF andfood-web theory, and the hypotheses put forth inthat paper were useful, in part, because theyrepresented an alternative to those posed by anumber of other authors. For example, some haveargued that extinction at higher trophic levelsmay, in fact, have less impact on ecological pro-cesses than extinction at lower trophic levels.These arguments have usually been based on theidea that animals are more generalized in their useof resources than historically appreciated, eitherbecause the extent of omnivory and intra-guildpredation has been underestimated (Rosenheim etal. 1995, Holt and Polis 1997, Polis and Holt 1992),or because animals can ‘switch’ among differentprey species by moving across habitats (Polis et al.1997, McCann et al. 2005). Resource generalizationhas been proposed to dampen the effects of con-sumer diversity on prey populations (Finke andDenno 2005, Snyder and Ives 2003).

8.2.1.2 Increasing diversity of a resource reducesthe strength of top-down control by consumers

C

–

R1 Rn

The majority of BEF studies performed to date havetaken a ‘top-down’ perspective, meaning that theyhave examined how diversity within a giventrophic level impacts the fraction of resources con-sumed, and production of biomass, by that focaltrophic level. In contrast, diversity may also have‘bottom-up’ effects on the dynamics of food webs,meaning that the diversity of resources may influ-ence how efficiently those resources are consumedand converted into biomass by higher trophic levels(Table 8.1). At least three hypotheses have beenproposed to explain how resource diversity mightinfluence trophic dynamics: (2.1) the variance inedibility hypothesis argues that a more diverse preyassemblage is more likely to contain at least onespecies that is resistant to consumers (Leibold 1989,

T OWARD S A FOOD WEB P E R S P E C T I V E ON B I OD I V E R S I T Y AND E CO SY S T EM FUNC T I ON I NG 107

Table8.1

Five

early

hypotheses

aboutmulti-trophicBEF.Whatdo

thedata

say?

Section

Hypothe

sis

Keyreference(s)

Section

Balanceof

eviden

ceCe

rtainty

Keyreferences

8.2.1

Diversity

effectswithin

trophiclevels

8.2.1.1

Top-downeffectsof

diversity

grow

increasin

gly

strong

athigher

trophiclevels.

Duffy

(2002)

8.3.1.1

Thebalanceof

evidence

isnotconsistent

with

thishypothesis.

Recent

meta-analyses

have

foundno

differencein

the

directionor

magnitude

ofdiversity

effectsforgroups

of

producers,herbivores,d

etritivores,o

rpredators.In

fact,

theretendsto

beconsiderablegeneralitysuch

thatdecreases

inspeciesrichnessdecrease

theefficiency

ofresource

captureandtheam

ount

ofbiom

assproduced

byanygiven

trophicgroup.

Medium

tohigh

Balvaneraet

al.(2006),

Cardinaleet

al.(2006a,

2007)

8.2.1.2

Increasin

gdiversity

ofresourcesreducesthe

strength

oftop-downcontrolb

yconsum

ers.

Leibold(1989),

Duffy

(2002),

Ostfeld

and

LoGiudice

(2003),R

oot

(1973)

8.3.1.2

Thebalanceof

evidence

isconsistentwith

thishypothesis.

Summariessuggestthat

consum

ptionof

lower

byhigher

trophiclevelsisreducedwhenaresource

base

ismore

diverse.

Note,

however,thatmostof

theavailabledata

comes

from

studiesthat

have

notdirectlymanipulated

the

richnessof

resources.Severalcontro

lledexperim

ents

have

provided

counter-e

xamples,sothegeneralityof

this

hypothesisremains

unclear.

Low

to

medium

Andow

(1991),H

illebrand

andCardinale(2004)

8.2.2

Diversity

effectsacross

trophiclevels

8.2.2.1

Top-downeffectsof

consum

erdiversity

oppose

thebotto

m-upeffectsof

resource

diversity.

Holt&Loreau

(2002),T

hébault

andLoreau

(2003,

2005)

8.3.2.2

Recent

meta-analyses

suggestthat

thetop-downeffectsof

consum

erdiversity

arequalitativelydifferent

than

the

botto

m-upeffectsof

resource

diversity.H

owever,these

effectshave

notbeen

opposin

gas

suggestedby

this

hypothesis.

Note,

however,thatfew

studieshave

simultaneously

manipulated

therichnessof

speciesat

adjacent

trophiclevels,

soconclusio

nsaretentative.

Low

Srivastava

etal.(in

press)

8.2.2.2

Diversity

effectson

biom

assproductionand

resource

captureby

agiventro

phiclevela

re

reducedin

thepresence

ofahigher

trophic

level.

HoltandLoreau

(2002),T

hébault

andLoreau

(2003)

8.3.2.1

Thebalanceofevidence

doesnotsupportthishypothesis.

Ofthe

fewexperim

entsthathave

manipulated

speciesrich

nessinthe

presence

vs.absence

ofahigher

trophiclevel,results

are

decid

edlymixed.A

nalysespresentedinthischapterfurther

show

noevidence

that

theeffectsof

plantd

iversityon

plant

biom

assdifferfor

experim

entsperfo

rmed

inthepresence

vs.

absenceof

herbivores.

Low

Mulderet

al.(1999),D

uffyet

al.(2005),W

odjak(2005),

andthischapter

8.2.2.3

Trophiccascades

areweakerin

diverse

communities.

Strong

(1992)

8.3.2.3

Experim

entsanddata

summariesto

datehave

been

equivocal

andcontradictory.At

present,thereisno

clearreason

toaccept

orreject

thishypothesis.

None

Schm

itzetal.(2000),Borere

t

al.(2005),C

ardinale

etal.

(2003,

2006b),W

ilbyet

al.

(2005),Snyderet

al.(2006),

FinkeandDe

nno(2005),

Byrnes

etal.(2006)

Duffy 2002); (2.2) the dilution hypothesis (Ostfeld andLoGiudice 2003), which has also been called theresource concentration hypothesis in the agro-ecologyliterature (Root 1973), suggests that specialist con-sumers become less efficient at finding and attackingtheir resource in a diverse prey assemblage; and (2.3)the balanced diet hypothesis suggests that a morediverse prey assemblage provides a more completenutrition and, as a result, leads to higher consumerbiomass (DeMott 1998). While hypotheses (1) and (2)predict that trophic efficiency will decrease as thediversity of resources increases, (2.3) predicts theopposite.

8.2.2 Diversity effects across trophic levels

8.2.2.1 Top-down effects of consumer diversity opposethe bottom-up effects of resource diversity

C1

–

Cn

R1 Rn

An important, but still unresolved issue is whe-ther the overall impacts of diversity loss atadjacent levels are opposing or reinforcing,antagonistic or synergistic. Hypotheses (2.1) and(2.2) suggest that consumer diversity tends toenhance the flux of resources from lower tohigher trophic levels, whereas resource diversitytends to reduce these fluxes. Collectively, thesetwo hypotheses lead to a third hypothesis: thatextinction of species from adjacent trophic levelswill have opposing impacts on the flux of energyand matter through a food web (Table 8.1). Thisprediction has received some theoretical supportfrom mathematical models showing that simulta-neous changes in diversity from consumers andtheir resource leads to countervailing effects ontotal resource use and biomass production (The-bault and Loreau 2003, Thebault and Loreau 2005,Holt and Loreau 2002). Fox (2004b) provided acounter example in which he used Lotka–Volterramodels to show that the joint response of preybiomass to prey and predator diversity is poten-

tially more complex. While predator diversitygenerally decreases prey biomass, prey diversitycan increase or decrease biomass depending onhow different life-history trade-offs influence thecoexistence of prey.

8.2.2.2 Diversity effects on biomass productionand resource capture by any focal trophic levelare reduced in the presence of higher trophic levels

C1 Cn

+ (0)

R

C�

In their recent review, Duffy et al. (2007) used theterms ‘horizontal’ and ‘vertical’ diversity to distin-guish between the richness of species within atrophic level and the richness of trophic levels thatcomprise a food web. They argued that one of theprimary limitations in merging BEF with food-webtheory is knowing how the impacts of divesitywithin trophic levels depend on the length of foodchains (i.e. how horizontal and vertical diversityinteract). The first step in overcoming this limitationis to ask how the diversity effects of any singletrophic level are altered by the presence or absenceof the next highest trophic level. Holt and Loreau(2002) used simple consumer–resource models toargue that the effects of plant diversity on nutrientuptake and plant biomass production are reducedin the presence of herbivores. This occurs becauseherbivory selects for dominance by poor plantcompetitors that are also the most tolerant toconsumption by herbivores. Subsequent models byThébault and Loreau (2003) also suggested thataddition of higher trophic levels might qualita-tively alter diversity–production relationships atlower levels; however, the direction of theseimpacts depends on both the nature of trade-offsbetween a plant’s competitive ability and ability toresist herbivory, and on the degree of consumerspecialization.


8.2.2.3 Trophic cascades are weaker indiverse communities

C1 Cn

C1�

––

Cn�

R

In his seminal critique of the empirical evidence fortrophic cascades, Strong (1992) argued that cas-cades are ‘a relatively unusual sort of food webmechanics . . . over the full range of ecological com-munities, evidence is that these cascades arerestricted to fairly low-diversity places where greatinfluence can issue from one or a few species’. Hewent on to suggest that trophic cascades are ‘allwet’, meaning they occur primarily in aquatic eco-systems where communities are characterized bylinear, low-diversity food chains. In contrast, heargued that terrestrial food webs are more reticu-late and ‘consumption is so differentiated in spe-ciose systems that its overall effects are buffered’.The idea that diversity modifies the strength oftrophic cascades can be broken down into at leasttwo distinct hypotheses: (1) increasing the diversityof species comprising secondary consumers C’tends to decrease the strength of indirect effects ona basal resource R, and (2) increasing diversity ofprimary consumers C tends to decrease the indirecteffects of C’ on R. This latter hypothesis is verymuch an extension of hypotheses (2.1) and (2.2), asall of these rely on the assumption that anincreasing diversity of resources tends to reduce thetop-down impacts of consumers on food-webdynamics (Table 8.1).

8.3 What do the data say?

8.3.1 Diversity effects within trophic levels

8.3.1.1 Are diversity effects stronger at highertrophic levels? (Hypothesis 2.1)Empirical evidence gathered to date does notappear to support the hypothesis that diversity

effects are stronger at higher trophic levels.Balvanera et al. (2006) reviewed 103 studies inwhich they could examine 400þ correlation coef-ficients relating species richness to a variety ofecological processes. They found no evidence fordifferential correlations between diversity and anyof the response variables at various trophic levels.Similarly, Cardinale et al. (2006a) collated datafrom 111 experiments that have manipulated spe-cies richness and examined how this aspect ofdiversity impacts the capture of resources andproduction of biomass. Their analyses comparedfour trophic groups: (1) microalgal, macroalgal, orherbaceous plants assimilating nutrients or water,(2) protozoan or metazoan herbivores consuminglive algal or herbaceous plant tissue, (3) protozoanor metazoan predators consuming live prey, and(4) bacterial, fungal or metazoan detritivores con-suming dead organic matter. They showed that, onaverage, experimental reduction of species rich-ness decreases the standing stock abundance orbiomass of the focal trophic group, resulting in lesscomplete resource use by that group (Fig. 8.2).However, the standing stock of, and resourcedepletion by, the most diverse polycultures wereindistinguishable from those of species that per-formed best in monoculture. Importantly, theauthors could not detect any statistical differencein the magnitude of diversity effects among thefour trophic groups.Collectively, these meta-analyses suggest there

is considerable generality in the way that thediversity of species impact resource capture andbiomass production in food webs. The fact thatCardinale (2006a) and Balvanera (2006) both foundthat the BEF relationships did not change dra-matically across trophic levels could imply that, ifniche complementarity is the main mechanismdriving these patterns, then the degree of nichecomplementarity could be similar across trophicgroups. Identifying whether the mechanisms thatdictate BEF relationships are the same across dif-ferent levels of biological organization is a keynext step in BEF research (a point we return to inSection 8.4.1). Although studies to date showconsiderable generality in diversity effects acrosstrophic levels, we should emphasize that there stilltend to be fewer absolute numbers of species at


higher trophic levels, and that these species tend tobe disproportionately prone to extinction (a pointwe return to in Section 8.4.2). Thus, it is still rea-sonable to hypothesize that food webs can toleratefewer extinctions at higher trophic levels beforeecosystem functioning is altered.

8.3.1.2 Does resource/prey diversity weakenthe strength of top-down control? (Hypothesis 2.2)Empirical evidence gathered thus far is mostlyconsistent with the hypothesis that increasing preydiversity tends to reduce the impacts of consumerson prey. Andow (1991) tallied the results of 200þstudies of herbivorous arthropods and found thatmore than half of the herbivore species had lowerpopulation sizes on plant polycultures as opposedto monocultures. He argued that the resource con-centration hypothesis, in which specialist con-sumers have a more difficult time finding their

resource in a diverse prey assemblage, bestaccounted for the observed patterns. A summary ofaquatic studies by Hillebrand and Cardinale (2004)tallied results from 172 experimental manipulationsof herbivores and showed that consumption ofalgal biomass generally declined with increasingalgal species richness. Although these patterns areconsistent with hypothesis (2.2), some caution iswarranted when interpreting these summaries,since the studies reviewed did not manipulatespecies diversity directly, and many potentiallyconfounding factors were not controlled for. Thiscaveat is particularly important when consideringthe mixed results from the limited number ofexperiments that have manipulated resourcediversity directly. Several studies do provide evi-dence consistent with the variance-in-edibilityhypothesis (Steiner 2001, Duffy et al. 2005), or for thedilution hypothesis (Keesing et al. 2006, Wilsey and

35, 15, 3

Producers Herbivores Predators Detritivores Producers Herbivores Predators Detritivores3

2 4

2

0

–2

1

0

–1

–2

1

0

LR

mea

n (S

ST)

LR

max

(SST

)L

Rm

ax (R

D)

LR

mea

n (R

D)

–1

–2

84

2

0

–2

4, 9, 1 4, 1, 1 3, 3, 0 7, 21, 11 3, 11, 0 3, 0, 3 1, 2, 0

5, 9, 0 0, 7, 1 1, 18, 0 16, 13, 0 1, 10, 1 1, 9, 0 0, 16, 0 3, 18, 0

Figure 8.2 Summary of the results of experiments that have manipulated the richness of species in four trophic groups t (producers, herbivores,predators, and detritivores), and examined how richness impacts the standing stock abundance or biomass of t (SST – top graphs) or the fraction ofresources depleted by t (RD – bottom graphs). The y-axes in all graphs give the diversity ‘effect size’, measured using two log ratios. LRmean (left graphs)compares SST and RD in the most diverse polyculture used in a study to the average of all monocultures. LRmax (right graphs) compares SST and RDfrom the most diverse polyculture used in a study to the species having the highest values of SST or RD in monoculture. Each data point is the mean effect sizefor all replicates in an experiment ± 95 per cent CI. Dashed horizontal grey lines give the 95 per cent CI for all experiments combined based on resultsrom a mixed model ANOVA. Numbers below each figure are the number of studies that have shown significantly positive effects of diversity, no effect,or negative effects of diversity. Data are from Cardinale et al. (2006a).


Polley 2002, Otway et al. 2005) where increasingdiversity of resources leads to reduced consumptionby higher trophic levels. Other studies providesupport for the balanced diet hypothesis, showingthat mixed diets of primary producers tend toenhance herbivore growth and biomass accumula-tion (Pfisterer et al. 2003, DeMott 1998). Thus,although the balance of evidence appears consistentwith hypothesis (2.2), these conclusions should beconsidered tentative.

8.3.2 Diversity effects across trophic levels

8.3.2.1 Do top-down effects of diversity differfrom bottom-up effects? (Hypothesis 2.1)To date, studies that have simultaneously manipu-lated the richness of species at adjacent trophiclevels are rare (Fig. 8.1(a)), and it is difficult to drawmany general conclusions about the direction oftop-down versus bottom-up effects of diversity infood webs. However, a recent meta-analysis bySrivastava et al. (2008) suggests that hypothesis (2.1)is not supported in detrital systems. These authorscompiled the results of 90 experiments reported in28 studies of detritivores to ask ‘Do changes inconsumer (i.e. detritivore) diversity have the sameeffect on rates of resource consumption as changesin resource (i.e. detrial) diversity?’. To address thisquestion, they compared the top-down effects ofconsumer (detritivore) diversity on the consump-tion of dead organic matter (decomposition) to the

bottom-up effects of resource (detrital) diversity onconsumption of dead organic matter. Their meta-analysis indicated that reductions in detritivorediversity generally led to reductions in rates ofdecomposition, but changes in the diversity ofdetrital resources led to no detectable change indecomposition. The implication is that consumer,but not resource diversity, impacts consumptionand energy flow in ‘brown’ food webs (detritus-consumer). However, an important point to keep inmind is that the resources studied by Srivastavaet al. (2008) are ‘dead’, meaning they are non-livingresources that have no potential to show dynamiccoupling to their consumers. A number of mathe-matical models suggest that diversity–functionrelationships could be qualitatively different whenresources are ‘living’, such as in ‘green’ food webs(i.e. plant-based systems) where populations havethe potential to respond to changes in the density oftheir consumers (Loreau 2001, Ives et al. 2005). Thepotentially important contrast between systems thathave dynamic (living) vs. non-dynamic (non-living) isan issue that we return to in Section 8.4.1. For now,suffice it to say that we do not know whether theresults of Srivastava et al. (2008) are specific to detritalsystems, or whether they hold more generally.

8.3.2.3 Are diversity effects at one trophic levelaltered by higher levels? (Hypothesis 2.2)Only a handful of experiments have manipulatedthe richness of species in a focal trophic level and

2

Eff

ect o

f pla

nt s

peci

es r

ichn

ess

on p

lant

bio

mas

s, In

(Bp/

Bm

)

1

0

–1

–2

9

29

Yes No Yes

Aquatic Terrestrail

Herbivores present?

No

90 15

FPOFigure 8.3 A summary of the impact of plant species richness on theproduction of plant biomass when herbivores are present or absent inexperimental units. Data were taken from the summaries of Cardinale et al.(2006, 2007). The log ratio of plant biomass in the most diverse polycultureBp to biomass in the average monoculture Bm was analyzed using a mixedmodel ANOVA with herbivores (y/n), ecosystem (aquatic vs. terrestrial), andtheir two-way interaction included as fixed effects, experiment accounted foras a random effect, and observations weighted by the inverse of theirvariance. Analyses indicate that the impacts of plant diversity on plantproduction do not differ when herbivores are absent vs. present (F ¼ 0.01, P¼ 0.92), and that this conclusion is consistent among ecosystems (F ¼0.07, P ¼ 0.80 for interaction). These data and analyses should not betaken as conclusive evidence that herbivores do not impact plantdiversity–biomass relationships since the studies summarized here differ inmany ways that cannot be explicitly accounted for. However, these data canserve as a null hypothesis for experiments that explicitly manipulate plantdiversity in the presence versus absence of higher trophic levels.


then simultaneously manipulated the presence/absence of a higher trophic level. Mulder et al. (1999)varied plant diversity in the presence and absence ofinsect herbivores in a grassland plant assemblage. Inthe absence of herbivores, plant biomass increasedwith plant diversity, whereas when insects werepresent, they fed heavily on species with interme-diate biomass, weakening the impact of plantdiversity and biomass. Conversely, in a seagrasssystem, effects of herbivore richness on plant pro-duction were stronger in the presence of a highertrophic level (crabs) than in their absence (Duffy et al.2005), which presumably occurred because of tra-deoffs between species abilities to compete forresources versus resist predators. In other experi-ments, addition of a higher trophic level changed notonly the magnitude but also the sign of thediversity–function relationship at the prey level (e.g.Hattenschwiler and Gasser 2005, Wojdak 2005).We have been able to further examine hypothesis

(2.2) by collating data from the meta-analyses ofCardinale et al. (2006a, 2007) for studies that havemanipulated the richness of primary producers. Wedivided experiments into those that did versus didnot allow herbivores access to experimental plots orpots, and then compared how plant diversity influ-ence plant biomass between the two types of studies.Although plant species richness generally increasedthe production of plant biomass, we found no evi-dence that herbivores alter the magnitude of plantdiversity effects (Figure 8.3). This was true for studiesperformed in both aquatic as well as terrestrial eco-systems. Although these analyses are far from con-clusive, when taken with the mixed results ofexperiments they suggest that widespread supportfor hypothesis (2.2) is presently lacking.

8.3.2.4 Are trophic cascades weaker indiverse communities? (Hypothesis 2.3)Experiments and data summaries that have addres-sed hypothesis (2.3) to date have been equivocal andcontradictory. Schmitz et al. (2000) performed ameta-analysis of 14 terrestrial experiments thatmanipulated higher predators and found evidencethat the cascading effects of predator removal onplant damage were weaker in systems that hadhigher herbivore diversity. A more comprehensiveanalysis of trophic cascades measured in a variety of

ecosystems found no evidence that variation in thestrength of cascades was related to the richness ofpredators, herbivores, or plants (Borer et al. 2005). Incontrast, a limited number of experiments havemanipulated the diversity of predators at top trophiclevels and shown that diversity can indirectly alterplant biomass by changing rates of herbivory. Cas-cading effects of predator diversity have been dem-onstrated in agricultural (Cardinale et al. 2003, Wilbyet al. 2005, Snyder et al. 2006), salt marsh (Finke andDenno 2005), and kelp forest systems (Byrnes et al.2006), and have been attributed to non-additiveinteractions (Cardinale et al. 2003, Cardinale et al.2006b), omnivory (Bruno and O’Connor 2005), intra-guild predation (Finke and Denno 2005), and chan-ges in herbivore behavior (Finke and Denno 2005,Byrnes et al. 2006). Yet, the magnitude and directionof predator richness impacts on plant biomass andproduction have been inconsistent among studies(see Bruno and Cardinale 2008 for a review). Thus,although predator richness frequently has cascadingimpacts on food-web properties, it is difficult at thispoint in time to predict whether these cascadingeffects generally increase or decrease plant biomass.Therefore, at present, there is no clear evidence thatcan be used to accept or reject Strong’s (1992)hypothesis that trophic cascades are restricted tolow-diversity linear food chains.

8.4 Where do we go from here?

8.4.1 Detailing mechanisms: niche partitioningand life-history tradeoffs

William Dillard, founder and Chairman of Dillard’sdepartment stores, once said that the three mostimportant factors for the success of a business are‘location, location, location’. Similarly, we believethat the three most important factors that willdetermine the success of the BEF paradigm will beour ability to identify mechanisms, mechanisms,mechanisms! Understanding the mechanisms thatunderlie diversity effects essentially requires thatresearchers return to several of ecology’s classicquestions about how niche partitioning and life-history tradeoffs allow species to coexist. Chesson(2000) provided what is perhaps the most elegantlyorganized summary of the mechanisms that allow


coexistence. He showed that, for a wide variety ofmathematical models, coexistence is ultimatelydetermined by the balance of two interacting forces,which he called equalizing and stabilizing. Equalizingforces are those that minimize differences in the fit-ness of species, causing interspecific interactions tohave weaker influence over population dynamics.Hubbel’s (2001) neutral theory of biodiversity is theextreme case of an equalizing force where demo-graphic parameters are assumed to be identicalamong species such that interacting with anotherspecies has the same per capita impact as interactingwith a congener. Equalizing mechanisms are notmathematically stable and cannot allow long-termcoexistence. Rather, equalizing mechanisms onlyserve to slow the inevitable outcome of speciesinteractions. Thus, long-term coexistence requiressome type of stabilizing force that involves nichedifferentiation in space or time. Regardless of whe-ther niche differentiation occurs through partitioningof limited resources, shared predators, or some otherdimension of a species niche, stabilizing forces allshare the feature that they reduce interspecific rela-tive to intraspecific interactions, leading to a percapita growth advantage of a species when rare.

The literature is ripe with models that examinehow reductions in interspecific relative to intra-specific interactions regulate the impacts of speciesdiversity on the production of single trophic-level

systems (Loreau 2004, Tilman et al. 1997c, Ives et al.2005, Cardinale et al. 2004). The discrete timeLotka–Volterra models of competition serve as anexample (Cardinale et al. 2004), where the biomass ofany species i in a local community can be described as

bi tþ 1ð Þ ¼ bi tð Þexp ri 1�biðtÞ þ a

PNj6¼i

bjðtÞ

Ki

0BBB@

1CCCA

26664

37775 ð8:1Þ

Ki is the equilibrium biomass of i in the absence ofcompetitors, ri is the intrinsic rate of increase inbiomass, and a is the ratio of inter- to intra-specificinteraction. If species have similar carrying capaci-ties and symmetric interactions, then all specieshave the same biomass at equilibrium, b(1), andfor any local community

bð1Þ þ aðS� 1ÞbðaÞ ¼ K ð8:2Þ

From this, the total biomass of the community is

Bð1Þ ¼ SK

1þ aðS� 1Þ ð8:3Þ

For the extreme cases of a ¼ 1 or a ¼ 0, eq. 3reduces to B ¼ K and B ¼ SK, respectively, whichshows that community biomass is independent of,

Resource dimensionSpecies richness, S

B =SK

a = 0

a = 1

0 < a < 1

1 + a(S – 1)

Com

mun

ity

biom

ass,

B

d

d

w

w

Freq

uenc

y of

uti

lizat

ion

Figure 8.4 (a) Solutions to Lotka–Volterra competition equations showing how species richness affects community biomass production for differinglevels of interaction strength. Note there is a positive, but decelerating relationship between B and c for all 0 < a < 1. This is an inevitable consequence ofniche packing (insets) where the addition of species to a system with finite resource forces the average species to occupy a smaller fraction of resourcespace. Thus, the more species there are, the less each species contributes to resource capture and biomass production, on average.


or a linear function of richness (Fig. 8.4). For allother scenarios where 0 < a < 1, community bio-mass is a positive but decelerating function ofspecies richness. Importantly, the curvilinearity ofthis function has nothing to do with how ‘unique’or ‘redundant’ species are. Rather, the deceleratingrelationship is an inevitable consequence of pack-ing more species into a finite niche axis. Evenwhen all species are specialists with a uniqueniche, the contributions by any single species toresource capture and biomass production mustdecline as a function of richness (i.e. b _ 1/S, (Eqn8.2), causing each increase in diversity to contrib-ute smaller increments to resource capture andbiomass.Equation (8.3) predicts a rather straightforward set

of relationships between species diversity and com-munity biomass for any trophic group that is sup-ported by a non-dynamic resource (e.g. plantsassimilating inorganic resources, detritivores feedingondead organicmatter, etc.). One of the key questionsas we extend BEF theory to multi-trophic systems iswhether this same set of simple relationships holdstrue for systems where the resources are themselvesdynamic. Interestingly, several authors have ana-lyzed Lotka–Volterra models for both dynamic andnon-dynamic resources and found that the effects ofspecies diversity on community biomass are oftenqualitatively similar between one and two trophic-level systems (Thebault and Loreau 2003, Thebaultand Loreau 2006, Ives et al. 2005, Fox 2004b). Thereseems to be just two general instances where newbehaviors emerge in a multi-trophic system. The firstoccurs when dynamic resources, which have thepotential to be overexploited in a multi-trophic sys-tem, are brought to extinction by their consumers.Overexploitation or extinction of resources by adiverse group of generalist consumers can yieldhumped-shaped diversity–biomass relationships inboth predators and prey, which is a BEF relationshipthat is not found in single trophic-level systems(Thebault and Loreau 2003, Thebault and Loreau2006, Ives et al. 2005). Second, there are certain types oflife-history tradeoffs that can alter the shape andmagnitude of a diversity–biomass relationship (The-bault and Loreau 2003, Thebault and Loreau 2006).For example, when resource species exhibit a tradeoffbetween their competitive abilities and their ability to

resist or recover from consumption, this canmoderatecoexistence among prey (Holt et al. 1994) and dictatewhether prey biomass increases or decreases withdiversity (Holt and Loreau 2002, Thebault and Loreau2003). Similarly, the tradeoff between the degree ofresource specialization and assimilation efficiency ofconsumers has important implications for the BEFrelationship. The diversity of consumers that pay nocost to generalism, i.e. that do not trade off theirability to consume a wide diversity of resourcesagainst their efficiency at consuming each of theseresources, typically has a strong destabilizing effecton both population- and ecosystem-levelfluctuations,whereas species diversity has a stabilizing effect onecosystem-level fluctuations when consumers dohave such tradeoffs (Thebault and Loreau 2005).So are the consequences of extinction the same in

single versus multi-trophic systems? Theory pre-dicts that the answer entirely depends on the formof tradeoffs that mediate the coexistence of bothconsumers and their resources, and whether or notresources exhibit density dependent dynamics andoverexploitation by consumers. What we need noware innovative experiments that manipulate thestrength of consumer–resource interactions and/orthe existence of tradeoffs that are presumed tounderlie diversity effects in multi-trophic systems.Although such innovative experiments will nodoubt be challenging, they have the potential toyield some of the most important new insights intothe functioning of food webs.

8.4.2 Realistic scenarios of extinction

It is well established that species extinction is a non-random processes. Throughout both geological andmodern time, certain biological traits such as dis-persal ability, generation time, body size, geo-graphic range, and local density have proven to becorrelated with extinction risk (McKinney 1997,Lawton and May 1995, Purvis et al. 2000a). Trophicposition also appears to be correlated with extinc-tion risk. In marine systems, extinction of fish spe-cies generally proceeds from the top of food websdownward (Pauly et al. 1998), which is partly dueto human preferences for large-bodied fish, andpartly because such fish have low resilience due tolate maturity and slow growth (Myers and Worm


2005). In terrestrial systems, studies similarly reporthigher extinction probabilities for predators thantheir prey (Kruess and Tscharntke 1994, Didhamet al. 1998b).

Non-random patterns of extinction can affectdiversity–function relationships in at least two ways:via the functional traits lost, and via changes incommunity interactions. Initially, ecosystem functionmay be most affected by the functional traits of thespecies that preferentially go extinct (Srivastava andVellend 2005, Lavorel and Garnier 2002). Positivecovariance between extinction risk and the magni-tude (Gross and Cardinale 2005) or uniqueness(Petchey and Gaston 2002b) of a species functionaleffects can exacerbate the impacts of species loss onecosystem function (i.e. diversity–function effectsare initially stronger for realistic extinctions than

random extinctions). Predators may have highfunctional importance in food webs, first because ofthe strength of top-down processes in food webs(Duffy 2003), and second because predators mayhave traits that are additionally correlated with highfunctional impact (e.g. body size – Solan et al. 2004).Following extinction of a species, diversity–function

relationships are additionally influenced by the res-ponse of the surviving species to loss of a communitymember. Gross and Cardinale (2005) showed thatthe effect of species interactions amongst survivorsdepends critically on the mechanisms that underliediversity–function relationships: niche partitioning,facilitation or the sampling effect each make verydifferent predictions about how biased extinctionscenarios differ from random extinction scenarios.In food web simulations, Ives and Cardinale (2004)

10–3

1025

1022

1019

1016

1013

1010

107

104

101

10–2

10–5

Size

of s

tud

y sy

stem

(m2

or L

)/m

ean

orga

nism

siz

e (g

)

Spat

ial s

cale

10–1 10–0 101 102

Time-scale

Duration of experiment (d)/mean generation time (d)

103 104 105 10610–2

< 1 generation > 1 generationSm

Cl

Bu

UnTr

MoWo

Figure 8.5 The spatial and temporal scale of biodiversity-ecosystem functioning experiments. The experimental duration (in days) and spatial scale(in m2 or L) of experiments reviewed by Cardinale et al. (2006) were standardized to the mean generation time and body sizes of the focal organisms. Datawere divided into four trophic groups: Plants ¼ green circles, Herbivores ¼ blue triangles, Predators ¼ red squares, Detritivores ¼ brown diamonds.The scale of each individual study is given by smaller symbols while the medians for each trophic group are shown as larger symbols. The box denoted by thedashed line gives the 10th and 90th percentiles for the scale of all experiments. For comparative purposes we show the scale of several natural extinctions:Wo ¼ wolves from Yellowstone National Park, USA; Mo ¼ Moa from New Zealand; Tr ¼ Trout from Lake Superior, USA; Un ¼ Unionid mussels fromthe lower Mississippi River, USA; Bu ¼ Various species of butterflies in Europe; Cl ¼ Loss of certain cladoceran zooplankton from Lake Superior, USA;Sm ¼ Global eradication of the small pox virus.


showed that the coupling of directional extinctionwith species interactions can lead to unexpectedchanges in the functional importance of species.Although it is clear that non-random patterns ofextinction can have very different implications thanthe random extinctions that commonly simulated inexperiments, our ability to predict the functionalchanges that stem from non-random extinction –

particularly the top-down effects of species loss onecosystem function – is still in its infancy. After ourneed to characterize interaction strengths and inter-specific tradeoffs (Section 8.4.1), our single biggest gapof knowledge stems from a lack of information aboutlevels of covariance between extinction risk and spe-cies-specific impacts on rates of ecological processes atvarious trophic levels.

8.4.3 Environmental heterogeneity, patchdynamics, and scale

The typical biodiversity experiment performed to datehas taken place in experimental units slightly largerthan a five-gallon bucket, and has run for less than onegeneration of the focal organisms (Fig. 8.5). Whilethere are noteworthy exceptions (Tilman et al. 2001,Hector et al. 1999), it seems safe to say that most of ourinferences about biodiversity stem from experimentsperformed at spatial scales much smaller, and tem-poral scales much shorter than those at which speciesextinctions actuallymatter (also seeNaeem 2001a for amore complete review). Overcoming this mismatch inscale is a daunting task, and the difficulties of per-forming large-scale, long-term experiments are whyecologists use simplified model systems in the firstplace (Srivastava et al. 2004). Nevertheless, ecologistshave begun to make progress on these issues byincorporating the important ecological factors thatco-vary with scale into their experimental designs(Cardinale and Palmer 2002, Dimitrakopoulos andSchmid 2004, Mulder et al. 2001) and accounting forthem in meta-analyses of experiments performed atdifferent scales (Cardinale et al. 2007).The issue of scale is by no means unique to BEF

research, nor is it specific to multi-trophic systems.There are, however, certain characteristics of multi-trophic systems that make it especially important thatwe deal more directly with the issue. Namely, dis-

persal as a process affecting species coexistencebecomes particularly prominent at higher trophiclevels where organisms are typically more mobile (atleast, on the shorter time-scales of most experiments)and, therefore, have the ability to integrate informa-tion across a landscape and aggregate in response tothe density of their prey. This is important becausedispersal and aggregation across spatially distinctpatches or habitat boundaries can translate intovarious forms of niche partitioning that stabilizecompetitive interactions and consumer–resourcedynamics (Armstrong 1976, McCann et al. 2005). As itmodifies coexistence, dispersal across patches orhabitat boundaries can also qualitatively alter the BEFrelationship (Mouquet et al. 2002).Although most of the work that has examined how

dispersal affects BEF relationships has focused onsingle trophic level systems, it is useful to quicklyreview here and then consider how these predictionsmight be extended to systems with dynamicsresources. A wide variety of ecological models havehighlighted the important role that dispersal plays inmaintaining the diversity of communities (e.g. IslandBiogeography Theory –MacArthur and Wilson 1967,‘mass’ effects – Shmida and Wilson 1985, ‘rescue’effects – Brown and Kodricbrown 1977). Historically,models of dispersal have been phenomenological,meaning they did not explain the existence of diver-sity based on first principles. Instead, these modelsassumed therewas some ‘magical’pool of species thatcoexisted at large scales via some unknown mecha-nism(s), and these species generated propagules thatcould subsidize local populations. The emergence ofmeta-community theory (Leibold et al. 2004) repre-sented a major advance because these modelsacknowledged that everything in a propagule poolmust ultimately come from the collection of patchesor habitats that span a species range. Based on firstprinciples, meta-community models predict both thecauses and consequences of diversity at ‘local’(organisms interacting as communities within pat-ches) and ‘regional’ scales (patches of communitiesconnected by dispersal).One common form of meta-community models

assumes that species coexist through tradeoffs intheir abilities to compete in patches that have dif-fering types or supply rates of resources (i.e. whatLeibold et al. 2004 call ‘species-sorting’ models).


These models predict that at the scale of any localcommunity, increasing the number of species in themeta-community serves only to ensure that speciesbest adapted to a given patch will colonize anddominate that patch. This is the typical ‘selectioneffect’ of diversity (Loreau and Hector 2001, Huston1997), which has been formalized as follows:assume that species can be ranked by their carryingcapacities such that K(m) represents the specieshaving the highest carrying capacity in any singlepatch, K(m-1) is the next highest, and so on. If com-petition among species is strong (a ¼ 1 in Eqn 8.1),only one species will be present in a patch atequilibrium, and the biomass in a patch will be

Bð1Þ ¼ Ncol

gKðmÞ þ 1�Ncol

g

� �Ncol

g� 1

� �Kðm�1Þ

þ 1�Ncol

g

� �1� Ncol

g� 1

� �Ncol

g� 2

� �Kðm�2Þ þ :::

ð8:4Þ

Equation (8.4) says that the amount of biomassproduced in a patch at equilibrium is proportionalto the probability, Ncol/g, that the species with thehighest carrying capacity, K(m), will colonize thepatch. If a patch is not colonized by the mostproductive species, then the probability that thesecond most productive species, K(m–1), will colo-nize and dominate the patch is 1� Ncol

g

� �Ncol

g�1

� �.

Note that as the number of species colonizing apatch increases, the probability that a patchbecomes dominated by the most productive spe-cies in the regional species pool approaches unity.However, one key point is that for the selectioneffect to operate in the first place, species diversitymust first exist in the regional colonist pool (i.e. gmust exist at the scale of a meta-community). Butin order for diversity to be maintained in theregional colonist pool, species must exhibit someform of tradeoff that ensures they use resources inways that are complementary across patches. Thissuggests that the same mechanisms that ensurecomplementary use of resources across patches ina region also produce species-specific selectioneffects at the scale of a local community (Cardinaleet al. 2004).

Loreau et al. (2003) similarly showed that coexis-tence of species at a regional scale could maximizebiological production at a local scale, and called thisthe ‘spatial insurance’ hypothesis of diversity (also seeChapter 10, where Gonzalez treats the issue exten-

sively.). The general idea of the spatial insurancehypothesis is that while one species may be sufficientto maximize production in any local community, themaximization of productivity across all patches in anyheterogeneous landscape requires that a diversity ofspecies exhibit niche differences at a regional scale.Meta-community models like that used to generatethe spatial insurance hypothesis are importantbecause they serve as a springboard from which wecan address more pressing issues within the field ofBEF research. From the perspective of basic theory,we need to extend meta-community models to con-sider how species diversity impacts the production ofcommunity biomass when consumers and theirresources both move across a spatially heterogeneouslandscape. We need to know what happens to BEFrelationships when (1) resources have a spatial refugefrom their consumers, (2) consumers and resourcesdisperse at similar versus different rates, or (3) speciesexhibit spatially mediated tradeoffs, such as in theirdispersal versus competitive abilities, or dispersalversus ability to resist consumption. At the same time,we need experiments that explicitly mimic theassumptions of different meta-community models,and then examine how diversity impacts the pro-duction of local and regional biomass for variousmechanisms that allow consumer–resource coexis-tence. These advances are essential if we expect topredict the ecological consequences of extinctionfrom real food webs where the norm is that speciesmove across habitat boundaries and make choicesabout where to spend their time in order to maximizefitness.

8.4.4 Socio-economic impacts of food webdiversity

After several decades of research, it has becomeapparent that loss of diversity from an ecosystem canhave impacts on ecological processes that rival, if notexceed, many other forms of environmental change.Ecologists are now in a position to estimate thenumber of species required to maximize the removalof greenhouses gasses like CO2 from the atmosphere,remove nutrient pollutants from streams and lakesthat serve as drinking water, or to produce crops andfisheries. Indeed, it is now possible to make reason-ably educated estimates of how diversity loss


translates into societally meaningful units – whetherthat be in dollars, health risks, carbon credits, orotherwise.The socio-economic implications of biodiversity

are perhaps most obvious from studies of higher-trophic levels, including those of pollinators, andof natural enemies that control pest populations.

Invertebrate predators, parasitoids, and pathogenscan be important promoters of top-down control interrestrial food webs, helping to keep pests beloweconomically damaging levels. This natural bio-logical control of pests represents a valuable eco-system service that is essential to sustainableproduction of food and fibre. Recent economic

Box 8.1 Socioeconomic impacts of predator diversity

One of the primary services that ecosystems provide tosociety is the biological control of insect pests. Thisservice is estimated to be worth US$400 billion per yearglobally (Costanza et al. 1997). Although it has long beenassumed that effective pest management requires adiversity of predators, parasites, and pathogens (collectivelycalled ‘natural enemies’), experiments designed toexplicitly test this hypothesis have only recently begun.Two case studies highlight the range of results observedthus far.

FPO

Case study 1: Predator diversity decreases pestpopulations

In a field experiment performed in Wisconsin, USA,Cardinale et al. (2003) manipulated the richness of threenatural enemies of aphids (pea and cowpea) that areherbivorous pests of alfalfa. Two of the enemies – aladybeetle and an assassin bug – were generalist predatorsthat fed on both aphid species. The third was a specialistparasitoid wasp that attacks only pea aphids. They foundthat as generalist predators reduced the density of both

aphids, the parasitoid wasp became more efficient atattacking the pea aphid. As a result, when all threeenemies were together they reduced aphid populations toone-half of that achieved by any enemy species alone.This translated to a 51 per cent increase in the yield ofalfalfa. Alfalfa is the fourth most widely grown crop in theUSA with an estimated annual value of US$11.7 billion(source: US Department of Agriculture). In 2003 whenthis study was performed, alfalfa was selling for $150 peracre. The state of Wisconsin dedicates 3.5 million acres tothe production of alfalfa. Assuming the results of thisexperiment can be generalized to Wisconsin, theeconomic benefit of predator diversity would be roughly US$525 million during a single harvesting cycle. In a typicalyear in the midwestern USA, alfalfa is harvested 3· persummer.

Case study 2: Predator diversity increases pestpopulations

In a second field experiment, Cardinale et al. (2006)manipulated the diversity of a different group of aphidpredators, this time focusing on three species ofladybeetles that are all generalist predators. When theladybeetles were placed together in field enclosures, theytended to compete with each other in a way thatreduced their individual ability to capture prey. As a result,more diverse predator assemblages were roughly 60 percent less efficient at controlling aphid populations thanexpected based on how each ladybeetle performedwhen alone. In this case, the antagonistic interactionsamong the predators led to a 17 per cent decrease inalfalfa yield. This result emphasizes that predator speciescan interact in ways that may have economic costs. A keychallenge for ecologists is to determine the frequency ofpositive and negative interactions among predators thatmight help us evaluate the costs versus benefits ofbiodiversity.

AQ: Pleaseprovidecaption forthis figure.


valuation of the services provided by insects sug-gests the value of biological control of native pestsby natural enemies is $4.49 billion per year in theUSA alone (Losey and Vaughan 2006), and > US$400 billion per year at a global scale (Costanzaet al. 1997). While classical biological control tendsto focus on the contribution of individual species ofnatural enemies, a growing number of studies sug-gest that the efficiency of biocontrol is often a func-tion of non-additive interactions among multiplepredators, parasitoids, and pathogens (Rosenheim2007, Losey and Denno 1998, Snyder and Ives 2003,Snyder et al. 2006, Finke and Denno 2005, Cardinaleet al. 2003, Cardinale et al. 2006b). Although it is notyet clear whether these interactions among enemiesgenerally increase or decrease prey populations, it isclear that the economic impacts of natural enemydiversity can be substantial (Box 8.1).

Crop pollination is another ecosystem servicecentred on interactions across trophic levels. Theglobal value of pollination services have been esti-mated at US$117 billion per year (Costanza et al.1997) and in a recent review, Klein et al. (2007)concluded that fruit, vegetable, or seed biodiversity(i.e. richness, abundance, and distribution of mul-tiple species of pollinators) in delivering this eco-system service are often poorly quantified. Similarto evaluations of classical biocontrol, where thefocus is on the action of one or few natural enemiesrather than diversity per se, many of the economicvaluations of pollination services consider the con-tribution of honey bees alone. Klein et al. (2007)report case studies for nine crops on four continentsimplicating a diversity of pollinators and revealingthat agricultural intensification jeopardizes wild beecommunities and their stabilizing effect on polli-nation services at the landscape scale. At the indi-vidual farm level, such natural pollination servicescan contribute significantly to annual income; astudy from a coffee plantation in Costa Rica, forexample, indicated native bee species account for$62,000, or 7 per cent of the farm’s annual income

(Ricketts et al. 2004). At a more regional level, Loseyand Vaughan (2006) calculate that native pollinators(mostly bees) may be responsible for > $3 billion offruit and vegetables produced in the USA.Although often less direct, changes in biodiver-

sity and associated trophic structure have majorimplications for issues such as disease risk, withassociated impacts on economics and human wellbeing. For example, top predators are often the firstspecies to disappear as habitat is destroyed andfragmented. As elaborated in Chapter 15, whenpredators are lost to ecosystems, their prey mayincrease in abundance, leading to increased trans-mission efficiency of zoonotic diseases such asLyme disease (Ostfeld and Holt 2004, Dobson et al.2006). While quantifying the benefit of biodiversityin terms of disease regulation and infected casesaverted is clearly complex, many diseases such asmalaria, tick-borne encephalitis, and West Nilefever have been shown to increase as biodiversityfalls (Dobson et al. 2006, and Chapter 15).

8.5 Summary

The emerging paradigm of Biodiversity Effects onEcosystem Functioning has shown great potentialto augment ecology’s historical focus on the causesof biodiversity with a much more contemporaryunderstanding of its ecological consequences. Evenso, BEF studies have, thus far, been limited tohighly simplified ‘model’ communities that arenowhere near the trophic complexity of real com-munities. To overcome this limitation, it is nowimperative that ecologists begin to merge the BEFparadigm with more classic ideas in food webecology that detail how interactions among trophiclevels that play out in space and time can constrainfluxes of energy and matter. Most hypotheses aboutthe functional role of diversity within and acrosstrophic levels are in their infancy, and they repre-sent a rich opportunity for new work during thesecond generation of BEF experiments.


Towards a food web perspective on biodiversity and ecosystem

Documents