Biodiversity and

Ecosystem
Functioning
Synthesis and Perspectives

EDITED BY

Michel Loreau
Laboratoire d'Ecologie, Ecole Normale SupeÂrieure, Paris, France

Shahid Naeem
Department of Zoology, University of Washington, USA

Pablo Inchausti
Laboratoire d'Ecologie, Ecole Normale SupeÂrieure, Paris, France

1
1
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University's objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Bangkok Buenos Aires Cape Town Chennai
Dar es Salaam Delhi Hong Kong Istanbul Karachi Kolkata
Kuala Lumpur Madrid Melbourne Mexico City Mumbai Nairobi
Sao Paulo Shanghai Taipei Tokyo Toronto
with an associated company in Berlin
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
# Oxford University Press 2002
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2002
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose this same condition on any acquirer
A Catalogue record for this title is available from the British Library
Library of Congress Cataloging in Publication Data
(Data available)
ISBN 0 19 851570 7 (Hbk)
ISBN 0 19 851571 5 (Pbk)
10 9 8 7 6 5 4 3 2 1
Typeset by Newgen Imaging Systems (P) Ltd, Chennai, India
Printed in Great Britain
on acid-free paper by The Bath Press, Avon
 CHAPTER 7

A new look at the relationship

between diversity and stability
M. Loreau, A. Downing, M. Emmerson, A. Gonzalez,
J. Hughes, P. Inchausti, J. Joshi, J. Norberg, and O. Sala

7.1 Introduction
ecosystem functioning. Will ecosystem functional
The relationship between biodiversity and ecosys- properties and services become more variable and
tem functioning has emerged as one of the most less predictable when species diversity is reduced?
exciting and controversial research areas in ecology Are species-rich ecosystems more capable of buf-
over the last decade. Faced with the prospect of a fering environmental variability and maintaining
massive, irreversible loss of biodiversity, ecologists ecosystem processes within acceptable bounds
have begun investigating the potential conse- than species-poor ecosystems?
quences of these dramatic changes in biodiversity To answer these questions appropriately and
for the functioning of natural and managed avoid `reinventing the wheel', it is important to
ecosystems. These investigations have been motiv- realize that they address in a new form a long-
ated by both the scienti®c challenge and the need to standing debate in ecology concerning the rela-
understand better the role biodiversity plays in tionship between the complexity (loosely de®ned
providing sustainable ecological goods and ser- as a combination of species diversity and the
vices for human societies. An increasing amount of number and strength of species interactions) and
evidence suggests that changes in biodiversity stability of ecological systems. The study of this
can have adverse effects on the average rates of relationship has had a long and controversial his-
ecosystem processes such as primary production tory (May 1974; Pimm 1984, 1991; McCann 2000).
and nutrient retention in temperate grassland eco- To delineate the differences between contemporary
systems (see chapters by Hector et al., Chapter 4; issues and the historical debate, we ®rst brie¯y
Tilman et al., Chapter 3; Wardle and van der Putten, revisit the central components of this debate, and
Chapter 14). Most of this evidence, however, comes propose a new, integrated conceptual framework
from relatively short-term experimental studies derived both from lessons from this debate and
(see, however, Petchey et al., Chapter 11) under insights newly arising from current research on
controlled experimental conditions, which are little biodiversity and ecosystem functioning. We then
informative about sustainable functioning. examine, within this framework, how recent the-
The temporal variability of natural environments oretical and experimental work provide new
over a broad spectrum of time scales from days to insights into the complexity±stability debate. Our
centuries (Halley 1996), as well as increasing treatment therefore complements some recent
anthropogenic pressures (Sala et al. 2000), inevit- reviews of the topic (Loreau 2000a; McCann 2000;
ably generate temporal changes in both population Schwartz et al. 2000; Cottingham et al. 2001) which
sizes and ecosystem processes. It is therefore of have not used the framework we present. Finally,
considerable interest to understand how biodivers- we discuss the need to develop new theoretical
ity loss will affect long-term temporal patterns in and methodological approaches and to strengthen

79
80 BIODIVERSITY AND ECOSYSTEM FUNCTIONING

the link between theory and experiments in this was based on randomly constructed model com-
area. We conclude with some implications for policy munities. More realistic food webs incorporating
and management. thermodynamic constraints and observed patterns
of interaction strengths do not necessarily have the
same properties (DeAngelis 1975; de Ruiter et al.
1990). Also, there have been few direct experi-
7.2 Historical and conceptual
mental tests of the theory, and many of the natural
background
patterns that agree with theoretical predictions can
The early view that permeated ecology until the be explained by more parsimonious hypotheses
1960s was that diversity (or complexity) begets such as the trophic cascade model (Cohen and
stability. This view was formalized and theorized Newman 1985). Despite these limitations, the view
by people such as Odum (1953), MacArthur (1955) that diversity and complexity beget instability, not
and Elton (1958) in the 1950s. Odum (1953) and stability, quickly became the new paradigm in the
Elton (1958) observed that simple communities are 1970s and 1980s because of the mathematical rigour
more easily upset than rich ones, i.e. they are more of the theory.
subject to destructive population oscillations and There are other limitations in this theory which
invasions. MacArthur (1955) proposed, using a are critical for the questions that we address here.
heuristic model, that the more pathways there are First, stability is really a meta-concept that covers a
for energy to reach a consumer, the less severe is range of different properties or components. Sum-
the failure of any one pathway. These conclusions marizing the debate, Pimm (1984) recognized a
were based on either intuitive arguments or loose number of these properties and concluded that the
observations, but lacked a strong theoretical and relationship between diversity and each of them
experimental foundation. Probably because they need not be the same. In Table 7.1, we attempt a
represented the conventional wisdom (`don't put classi®cationÐalbeit imperfect, as any classi®ca-
all your eggs in one basket') and the prevailing tionÐof the different components of stability,
philosophical view of the `balance of nature', they which includes more recent notions. Second, each of
became almost universally accepted. these stability properties can be applied to a number
This `conventional wisdom' was seriously chal- of variables of interest at different hierarchical
lenged in the early 1970s by theorists such as levels, such as individual species abundance, com-
Levins (1970), Gardner and Ashby (1970) and May munity species composition, or ecosystem-level
(1972, 1974), who borrowed the formalism of processes or properties (Table 7.1). Again, the rela-
deterministic autonomous dynamical systems from tionship between diversity and any stability prop-
Newtonian physics and showed that, in these erty may be different for different variables (Pimm
model systems, the more complex the system, the 1984). This creates a large matrix of potential
less likely it is to be stable. Stability here was combinations of stability properties and variables
de®ned qualitatively by the fact that the system of interest, of which the new theory concerned only
returns to its equilibrium or steady state after a a small part. Speci®cally, May's (1974) and Pimm's
perturbation. The intuitive explanation for this (1982) theory concerned the qualitative stability and
destabilizing in¯uence of complexity is that the resilience of communities as ensembles of popula-
more diversi®ed and the more connected a system, tions, not ecosystem-level aggregate properties.
the more numerous and the longer the pathways Third, the formalism of autonomous, determin-
along which a perturbation can propagate within istic dynamical systems, which describes a ®xed set
the system, leading to either its collapse or its of variables with time-independent parameters,
explosion. This conclusion was further supported inherently excludes a number of phenomena that
by analyses of one quantitative measure of stability, characterize biological and ecological systems. In
resilience (Table 7.1), in model food webs (Pimm particular, it does not allow for the fact that these
and Lawton 1977; Pimm 1982). This theoretical systems are subject to continuous environmental
work had a number of limitations. In particular, it changes at various temporal scales and have the
 A NEW LOOK AT THE RELATIONSHIP BETWEEN DIVERSITY AND STABILITY 81

Table 7.1 Concepts and de®nitions

Stability property De®nition

Components of stability
Qualitative stability Property of a system that returns to its original state after a perturbation. Generally used for an equilibrium state,
though it could also be applied to systems that return to non-equilibrium trajectories.
Resilience A measure of the speed at which a system returns to its original state after a perturbation* (Webster et al. 1974).
Generally used for an equilibrium state, though it could also be applied to systems that return to
non-equilibrium trajectories.
Resistance A measure of the ability of a system to maintain its original state in the face of an external disruptive force
(Harrison 1979). Generally used for an equilibrium state.
Robustness A measure of the amount of perturbation that a system can tolerate before switching to another state. Closely
related to the concept of ecological resilience sensu Holling* (1973). Can be applied to both equilibrium and
non-equilibrium states.
Ampli®cation envelope Describes how an initial perturbation from an equilibrium state is ampli®ed within a system
(Neubert and Caswell 1997).
Variability A measure of the magnitude of temporal changes in a system property. A phenomenological measure which
does not make any assumption about the existence of an equilibrium or other asymptotic trajectories.
Persistence A measure of the ability of a system to maintain itself through time. Generally used for non-equilibrium or
unstable systems before extinction occurs.

Variables of interest
Individual species abundances
Species composition
Ecosystem processes or properties

Sources of stability/instability
Internal: species interactions, demographic stochasticity
External: environmental changes, biological invasions, extirpations

* Some confusion surrounds the term resilience in the ecological literature. Though the term was ®rst introduced into ecology by Holling (1973),
it has most often been used in the sense de®ned by Webster et al. (1974). We follow here the common usage without any judgement on the
relative merits of the two de®nitions.

ability to react or adapt to these changes through During the golden period of the new paradigm
asynchronous population ¯uctuations, species (the 1970s and 1980s), few dissenting voices were
replacement, phenotypic plasticity and evolutionary heard. Those proposing an alternative viewpoint
changes. By ignoring these features, most of the were ecosystem ecologists emphasizing functional
theory on the complexity and stability of ecological compensation between species as the mechanism
systems has focused on deterministic equilibria that stabilizes ecosystem processes against a back-
and ignored much of the potential for functional ground of wider variability of individual popula-
compensation, both within and between species, tions (Patten 1975; McNaughton 1977, 1993). Though
which is the basis for the stabilization of aggregate often ignored, these ideas are the basis of the new
ecosystem properties. Functional compensation wave of theoretical, experimental, and observa-
between species or types occurs when changes in tional work that developed in the late 1990s. The new
the level of functioning contributed by one type are interest in the functional consequences of bio-
associated with opposite changes in the level of diversity changes in the 1990s has moved the focus
functioning contributed by another, whether these from populations, communities and food webs to
changes be of a dynamical, phenotypic or genetic ecosystems and the interplay between community-
nature. level dynamical processes and ecosystem-level
82 BIODIVERSITY AND ECOSYSTEM FUNCTIONING

functional processes (DeAngelis 1992; Jones and among species or phenotypes, and hence more
Lawton 1995; Loreau 2000a). This shift is particu- predictable aggregate community or ecosystem
larly clear in the recent development of theory, properties (Patten 1975; McNaughton 1977). In this
which requires formalization of concepts and hypo- hypothesis, species that are functionally redundant
theses. New approaches explicitly address the link for an ecosystem process at a given time show
between the variability of individual species and temporal complementarity (Loreau 2000a). There
that of aggregate ecosystem properties, and expli- have been a number of variations on this theme
citly incorporate population dynamical responses during the last years (Doak et al. 1998; Naeem 1998;
to environmental ¯uctuations, and even evolu- Tilman et al. 1998; Ives et al. 1999, 2000; Rastetter
tionary adjustments (Ives 1995; Doak et al. 1998; et al. 1999; Tilman 1999a; Walker et al. 1999; Lehman
Yachi and Loreau 1999; Lehman and Tilman 2000; and Tilman 2000).
Norberg et al. 2001). Although the assumptions, degree of generality
These new approaches have generally emphas- and technical approaches differ considerably
ized the potential stabilizing in¯uence of diversity among models, a few generalities do emerge from
on ecosystem properties, in agreement with the con- this recent theoretical work. There is often a tension
ventional wisdom of early ecologists. This refocus- between the destabilizing in¯uence of strong species
ing of the diversity±stability debate does not interactions within the system and the stabilizing
contradict the previous ®ndings of May (1974) and in¯uence of asynchronous species responses to
others, but it does considerably restrict their gen- external forcing on ecosystem properties. As divers-
erality. Previous work focused on qualitative ity increases, the number of interactions may
stability and resilience as the stability properties, on increase, leading to the classical result of decreased
species-level population abundances as the vari- resilience and increased variability of individual
ables of interest, and on deterministic autonomous populations (May 1974; Tilman 1996). This destab-
systems at equilibrium, in which only the internal ilizing effect, however, may be reduced for aggreg-
forces of species interactions came into play. In ate ecosystem properties (May 1974; Tilman 1996;
contrast, new work is focusing on variability as the Hughes and Roughgarden 1998, 2000; Yachi and
main stability property, on ecosystem-level prop- Loreau 1999; Ives et al. 1999, 2000; Lehman and
erties as the variables of interest, and on systems Tilman 2000), and counteracted by decreased mean
subject to environmental ¯uctuations, in which the interaction strength or presence of weak interac-
species' responses to these external ¯uctuations tions (McCann et al. 1998; Ives et al. 2000), which are
interact with the internal forces of species interac- the rule rather than the exception in many natural
tions. The two perspectives are not necessarily con- communities (Paine 1992; Raffaelli and Hall 1994).
tradictory (Tilman 1996; Ives et al. 2000). Another In contrast, variability of ecosystem processes
avenue of research which has received renewed driven by external environmental factors generally
interest concerns the invasibility or invasion resist- decreases as diversity increases because of the
ance of communities or ecosystems (see Levine buffering effect of asynchronous species responses
et al., Chapter 10), which can be interpreted within (Yachi and Loreau 1999; Ives et al. 1999). The net
our conceptual framework (Table 7.1) as the resist- result is generally a smaller variability of aggregate
ance (stability property) of species composition ecosystem properties at a higher diversity (Ives et al.
(variable of interest). 1999; Lehman and Tilman 2000), in agreement with
the insurance hypothesis. Hughes et al. (Chapter 8)
discuss further how variability driven internally by
7.3 What we have learned from theory
species interactions and variability driven externally
The insurance hypothesis (Yachi and Loreau 1999) by environmental ¯uctuations interact to determine
proposes that biodiversity buffers ecosystem pro- ecosystem-level stability. Although most of this new
cesses against environmental changes because dif- theory has been developed for competitive com-
ferent species or phenotypes respond differently to munities, the same conclusions seem to hold for
these changes, leading to functional compensations multi-trophic systems (Ives et al. 2000).
 A NEW LOOK AT THE RELATIONSHIP BETWEEN DIVERSITY AND STABILITY 83

Differences among species or phenotypes in their symmetries amongst species interactions, but Loreau
responses to environmental changes can not only and Behera (1999) found that phenotypic trait
lead to decreased variability, but also to increased diversity generally tends to decrease resilience at
average magnitude of ecosystem processes. When both the population and ecosystem levels. Loreau
selection processes such as competition favour and Behera (1999) also showed that phenotypic
species or phenotypes within a functional group diversity can have a variety of effects on the resist-
that are better adapted to current environmental ance of ecosystem properties. They suggested,
conditions, a higher diversity of types permits a however, that positive ecological selection, by which
greater adaptability of the system, and hence an species with favourable traits become dominant,
enhanced performance, at the functional group should generally yield a positive effect of diversity
level (Yachi and Loreau 1999; Norberg et al. 2001). on the resistance of ecosystem processes at the
Although a high phenotypic trait diversity can lead primary producer level in the case of `negative'
to a lower instantaneous productivity because many perturbations (i.e. perturbations, such as drought,
sub-optimal types are present, a diverse system can that have an intrinsically negative effect on the
have a higher long-term productivity than any production of most species), while the opposite
single type because better adapted types tend to should be true for `positive perturbations' (i.e.
replace less adapted ones. It can even be shown perturbations, such as nitrogen addition, that have
quantitatively that the rate at which succession an intrinsically positive effect on the production of
towards the current optimal type proceeds is pro- most species).
portional to phenotypic diversity (Norberg et al. The effect of species diversity on invasion resist-
2001), which provides an ecological analogue to the ance is another area that has received increased
fundamental theorem of natural selection, and a attention recently, although there have been very
potential approach to de®ning and measuring the few theoretical studies on this issue. It is commonly
ability of ecosystems to adapt to the environment. hypothesized that species-rich communities are
Given this analogy with evolutionary selection, more resistant to invasion than species-poor com-
Loreau (2000a) and Loreau and Hector (2001) have munities because they use resources more com-
coined the term `ecological selection' to describe pletely (Elton 1958; MacArthur 1970; Levine and
changes in dominance and species composition D'Antonio 1999; Tilman 1999). This pattern may be
driven by differences in species traits. This analogy expected when reduced species richness is indeed
allows for employing theoretical approaches devel- accompanied by reduced saturation of niche spaceÐ
oped in evolutionary genetics for disentangling a hypothesis for which there is some experimental
selection from complementarity effects (Loreau and evidence (see below). Otherwise, theory is mixed in
Hector 2001; Hector et al., Chapter 4). The ecolo- its conclusions about species richness as a predictor
gical processes that generate adaptability at the of invasion resistance. The nature of the relationship
ecosystem level also emphasize the importance of between species richness and invasion resistance is
regional species richness for ecosystem functioning expected to depend critically on the coexistence
since external inputs, such as immigration of indi- mechanisms that cause variation in species richness
viduals or propagules, are essential to maintain a (Moore et al. 2001).
wide range of phenotypic traits within an ecosys-
tem, and it is this phenotypic diversity that pro-
7.4 What we have learned
vides adaptability.
from experiments
In contrast to studies on variability, theoretical
studies on resilience and resistance of ecosystem A number of recent experimental studies have
processes after a perturbation have been scarcer. The investigated the relationship between species divers-
results of Hughes and Roughgarden (1998, 2000) ity and various stability properties. Experimental
and Ives et al. (1999, 2000) imply that the resilience manipulations of diversity within a single trophic
of some ecosystem properties may be independ- level have mostly concerned plants in grassland
ent of species richness in systems with special ecosystems (Table 7.2). The studies reviewed in
Table 7.2 Effects of experimental manipulations of species diversity within a single trophic level on ecosystem stability properties

Reference Diversity Species Ecosystem External Disturbance Time Plot size Diversity Stability Speci®cations to Type of diversity
gradienta comp.b type driversc direction scale (m2) levels property stability property effectd

Berish and Ewel 1988 Succ., F nr Plantation, n; R Neg. 1y 256 1,40,50,60 Resistance r. of ®ne-root bm None
forest succ.
Joshi et al. 20001 Exp., F rr Grassland bi Neg. 4 m (3 y) 0.25 1±32 Resistance r. to loss of above-ground bm " with funct. group no.
Leps et al. 1982 Succ., F nr Grassland n; R Neg. 4y no inf 4±20 Resistance Comparison with pre- and "
post drought y
Mellinger and
McNaughton 1975 Succ., F nr Old ®eld e; ‡ N Pos. 1y 1500 35/50 Resistance r. in bm to N-pulse "
Mulder et al. 2001 Exp., F rr Bryophyte e; R; Neg. 5 d (15 m) 0.24 1±32 Resistance/ Decrease of bm after drought "
community ‡L Resilience compared to control
P®sterer et al. (submitted)1 Exp., F rr Grassland e; ‡ Hi Neg. 2 w (5 y) 0.09 1±32 Resistance r. to loss of above-ground bm "
Tilman and Downing 1994 Nutr., F nr Grassland n; R Neg. 2y 16 1±26 Resistance Decrease of bm in "
drought rel. to normal y
Tilman 1996 Nutr., F nr Grassland n; R Neg. 2y 16 1±26 Resistance Decrease of bm in "
drought rel. to normal y
Brown and Ewel 1987 Succ., F nr Plantation, n Neg. 2y 256 1,40,50,60 Variability v. of herbivory #
forest succ.
Dodd et al. 1994 Nutr., F nr Grassland n Ð 42 y 1000±2000 8±45 Variability v. in bm # (tendency)
Emmerson et al. 2001 Exp., M nr Marine benthic Ð Ð 15 d 41 1±3 Variability v. in nutrient ¯ux #
invertebrates
Tilman 1996 Nutr., F nr Grassland n; R Neg. 8y 16 1±26 Variability v. in bm in non-drought ys #
Leps et al. 1982 Succ., F nr Grassland n; R Neg. 2y no inf. 4±20 Resilience Prop. return during 2 y #
following drought
Tilman and Downing 1994 Nutr., F nr Grassland n; R Neg. 2y 16 1±26 Resilience Deviation 4 y after drought " (optimum)
from pre-d. bm.

a
Exp.: experimentally newly created diversity gradient; Nutr.: gradients created by different nutrient levels; Succ.: gradients created by different successional stages; F: ®eld study; M: microcosm/mesocosm study.
b
r: random mixture; rr: random mixture with restrictions; n: nested design; nr: other non-random mixture.
c
n: natural perturbation; e: experimental perturbation; bi: biological invasion; ‡Hi: addition of an insect herbivore; ‡N: increased N-supply; R: drought. Time scale refers to either the duration of the perturbation (in the case of
experimental perturbationsÐduration of study in brackets) or the duration of the study (in the case of natural perturbations).
d
Presence/absence and direction of the observed ecosystem process. ": positive relationship between diversity and stability property; #: relationship negative; none: no signi®cant relationship; id.: identity (species identity or species-
mixture identity most important).
1
Studies were conducted in the same experimental system.
 A NEW LOOK AT THE RELATIONSHIP BETWEEN DIVERSITY AND STABILITY 85

Table 7.2 were selected as described in SchlaÈpfer resilience (Tilman and Downing 1994) used an
and Schmid (1999) and Schmid et al. (2002); they inadequate measure of resilience (it incorporated
are restricted to those studies that observed effects resistance by ignoring differences in the magnitude
of either experimentally or naturally imposed of the initial effect of the perturbation). The effect
disturbances on the stability properties of commun- disappeared after accounting for confounding fac-
ities differing in diversity within a single trophic tors (Tilman 1996). Thus, overall, the experiments
level. These studies have provided some evidence performed so far provide results that do not con-
that the temporal variability of various ecosystem tradict theory (Loreau and Behera 1999).
properties decreases with increasing diversity, in Experiments that test the effect of species divers-
agreement with the insurance hypothesis (Brown ity at multiple trophic levels on ecosystem stability
and Ewel 1987; Dodd et al. 1994; Tilman 1996; properties (Table 7.3) might re¯ect realistic extinc-
Emmerson et al. 2001; see also Schmid et al. 2002). tion scenarios of complex, highly connected ecosys-
When external perturbations were imposed on the tems which have to face direct but also secondary
system, plant species diversity had a positive effect extinctions (Williams and Martinez 2000). The stud-
on the resistance of above-ground biomass in all ies reviewed in Table 7.3 are restricted to experi-
the studies listed (Leps et al. 1982; Tilman and ments in which species diversity at multiple trophic
Downing 1994; Tilman 1996; Mulder et al. 2001). levels were manipulated and ecosystem stability
The one study that measured the resistance of ®ne- properties were measured (see also SchlaÈpfer and
root biomass, however, did not ®nd a diversity Schmid 1999; Schmid et al. 2002). Two such experi-
effect (Berish and Ewel 1988). The evidence pro- ments found decreasing variability of ecosystem
vided by most of these experiments, however (with properties with increasing diversity (McGrady-
the exception of Mulder et al. 2001), is inconclusive Steed et al. 1997; Naeem and Li 1997), in agreement
because of the presence of potential confounding with theory. The interpretation of these experiments,
factors (Givnish 1994; Huston 1997). For example, however, has been debated because of the presence
in Tilman and Downing's (1994) study, variations of confounding factors: in one study (McGrady-
in diversity resulted from a fertilization gradient Steed et al. 1997), ecosystem variability was con-
with plots receiving the highest fertilization having founded with variability among replicates; in the
the lowest diversity. Fertilization itself could have other (Naeem and Li 1997), variation in species
resulted in the larger response to drought that was diversity was confounded with variation in sim-
observed in the low-diversity treatments, although ilarity among replicates (Wardle 1998).
reanalysis by Tilman (1996) suggests that the effect Resistance of ecosystem processes after a press
of diversity was signi®cant even after controlling perturbation increased with diversity in one study
for fertilization. only (Grif®ths et al. 2000). Studies that measured
All but one study that tested the effects of resistance of community (Petchey et al. 1999) or eco-
external perturbations used negative perturbations system properties after pulse perturbations found
(sensu Loreau and Behera 1999), mainly drought. The either no (Downing, submitted; Petchey et al. 1999)
impact of positive perturbations such as nitrogen or a negative (Hurd and Wolf 1974) relationship
addition was only studied in Mellinger and with increasing diversity. There is no appropriate
McNaughton (1975). Given the anthropogenically theory, however, with which these results can be
induced global change in atmospheric nitrogen compared.
deposition (Vitousek et al. 1997) the effect of species Lastly, a number of experiments have recently
diversity on the resistance of ecosystem and com- been performed on the effects of species diversity
munity properties under positive perturbations within a single trophic level on invasion resistance.
would merit more attention. There have been few Studies reviewed in Table 7.4 were selected based
studies on the in¯uence of species diversity on the on a search on ISI web of science (1988±2001) in
resilience of ecosystem processes (Leps et al. 1982; June 2001 using `biodiversity' and `invasion' as
Tilman and Downing 1994). The one study that search terms (see also Hector et al. 2001a; Levine
found a positive effect of species diversity on et al., Chapter 10, for reviews). The majority of these
Table 7.3 Effects of experimental manipulations of species diversity at multiple trophic levels on ecosystem stability properties

Reference Diversity Species Ecosystem Time Plot Groups with Diversity Number Level of Stability Speci®cations to Type of
gradienta comp.b type scale size diversityc levels troph. gr.d effect property observed variable effecte

Downing submitted Exp., M r Aquatic 300 l p/c1/c2 1±5/1±5/1±5 3 All Resistance r. in resp. rates after Ð
pH-pulse perturbation.
Grif®ths et al. 2000 Rem., M Ð Pasture soil 1y Ð c1/c2/dec/b/f Ð 5 All Resistance r. in decomposition after "
heavy metal press-pert.
Hurd and Wolf 1974 Succ. F nr Old ®eld 6m 1500 m2 p/c1 35, 50 (p) 2 c2 Resistance r. to N pulse-pert. #
Petchey et al. 1999 Exp., M n Aquatic 7w 100 ml p/c1/c2/b 1±5/1,3/0±3/1,3 4 All Resistance Extinction risk due to Ð
temperature elevation
(press-pert.)
Smedes and Hurd 1981 Succ. F nr Marine benthic 2y 0.01 m2 c1/c2/dec 30, 35 (total) 3 Several Resistance r. to predation #
Wardle et al. 2000 Exp., M n Grassland 14 m 0.006 m3 p/c1/c2 1±4/1±2/0±1 3 All Resistance r. to biomass loss and Ð
decomposition after drought
press-pert.
Downing submitted Exp., M r Aquatic 300 l p/c1/c2 1±5/1±5/1±5 3 All Resilience r. in resp. rates after pH "
pulse-pert.
Grif®ths et al. 2000 Rem., M Ð Pasture soil 1y Ð c1/c2/dec/b/f Ð 5 All Resilience r. in decomposition after "
heat pulse-pert.
Smedes and Hurd 1981 Succ., F nr Marine benthic 2y 0.01 m2 c1/c2/dec 30, 35 (total) 3 Several Resilience r. after predation #
McGrady-Steed et al. 1997 Exp., M rr Aquatic 42 d 100 ml p/c1/c2 3±31 4 Several Variability v. in ecosystem respiration #
Naeem and Li 1997 Exp., M r Aquatic 57 d 50 ml p/c1/c2 1±3/1±3/1±3 5 All Variability v. of biomass per trophic #
group

a
Exp.: experimentally newly created diversity gradient; Rem.: diversity gradient created by selective removing of species from existing ecosystems; Nutr.: gradients created by different nutrient levels; Succ.: gradients created by different
successional stages; F: ®eld study; M: microcosm/mesocosm study.
b
r: random mixture; rr: random mixture with restrictions; n: nested design; nr: other non-random mixture.
c
p: primary producer; c1: primary consumer; c2: secondary consumer; dec: decomposer; f: fungivore; b: bacterivore.
d
Number of trophic groups varied.
e
Presence/absence and direction of the observed ecosystem process. ": positive relationship between diversity and stability property; #: relationship negative; none: no signi®cant relationship.
Table 7.4 Effects of experimental manipulations of species diversity within a single trophic level on invasion resistance

Reference Type of Species Ecosystem External Time Plot size Diversity Species Stability Speci®cations to stability Type of diversity
exp.a comp.b type driversc scale (m2) levels pool property property effectd

Crawley et al. 1999 F nr Grassland bi 7 y (8 y) 9.0 1±4, 80 4, 80 Resistance Weed invasion resistance None/identity
Dukes 2001 M rr Med. grassl. bi 1y 0.03 1±16 16 Resistance Weed invasion resistance " with funct. group no.
Hector et al. 2001a F rr Grassland bi 4y 4.0 1±11 47 Resistance Weed invasion resistance "
Joshi et al. 20001 F rr Grassland bi 1 y (3 y) 0.25 1±32 48 Resistance Weed invasion resistance " with funct. group no.
Knops et al. 19992 F r Grassland bi 2 y (4 y) 9.0 1±24 24 Resistance Weed invasion resistance "
Lavorel et al. 19993 F nr Med. grassl. bi 1 m (8 m) 4.0 3,6,18 18 Resistance Weed invasion resistance None
Palmer and Maurer 1997 F nr Crops bi 4m 5.0 1,5 5 Resistance Weed invasion resistance #/none
van der Putten et al. 2000 F rr/n Grassland bi 2y 100 4,15 37 Resistance Weed invasion resistance " (id.)
Troumbis et al. submitted F rr Med. grassl. bi 1 y (4 y) 1.0 1±18 23 Resistance Weed invasion resistance "
Diemer and Schmid 20011 F rr Grassland bi 2 y (4 y) 4.0 1±32 48 Resistance Weed invasion resistance (phytometer "
study)
Levine 2000 F r Med. riparian bi 1 y (2 y) 0.04 1±9 9 Resistance (sown) Weed invasion resistance " (id.)
primary succ.
McGrady-Steed et al. 1997 M rr Aquatic bi 2 w (8 w) 100 ml 3±16 27 Resistance Invasion resistance (controlled addition " (id.)
of Euplotes sp. (protozoa))
Naeem et al. 20002 F r Grassland bi 1 y (3 y) 9.0 1±24 24 Resistance Weed invasion resistance (phytometer "
study)
Prieur-Richard et al. 20003 F nr Med. grassl. bi 7 m (1 y) 4.0 3,6,18 18 Resistance Weed invasion resistance (phytometer (") /funct. gr. identity.
study)
Stachowicz et al. 1999 F r Marine benthic bi 65 d (Ð) 0.01 1±4 4 Resistance Invasion resistance (controlled addition "
of an exotic ascidian species)
Symstad 2000 F n Grassland bi 2 y (4 y) 32 1±3 funct. groups Ð Resistance (sown) Weed invasion resistance " with funct. group no.

a
Exp.: experimentally newly created diversity gradient; rem.: diversity gradient created by selective removing of species from existing ecosystems; nutr.: gradients created by different nutrient levels; succ.: gradients created by different
successional stages; P: phytotron; F: ®eld study; M: microcosm/mesocosm study.
b
r: random mixture; rr: random mixture with restrictions; n: nested design; nr: other non-random mixture.
c
bi: biological invasion. Time scale refers to either the duration of the perturbation (in the case of experimental perturbationsÐduration of study in brackets) or the duration of the study (in the case of natural perturbations).
d
Presence/absence and direction of the observed ecosystem process. ": positive relationship between diversity and stability property; #: relationship negative; none: no signi®cant relationship; id.: identity (species identity or species-
mixture identity most important).
1,2,3
Studies were conducted in the same experimental system.
88 BIODIVERSITY AND ECOSYSTEM FUNCTIONING

studies showed a positive relationship between experiments. For example, few studies replicated
plant species or functional-group richness and the highest diversity level with different species
resistance against naturally invading weeds. In addi- mixtures. Another limitation of these studies is that
tion, all studies investigating the impact of divers- they have typically used `invaders' from the extant
ity within a trophic level on the performance of regional ¯ora which have coevolved with those spe-
experimentally added invaders showed increased cies that constitute `invaded' communities. Invasion
invasion resistance with community diversity. Only by new exotic species is likely to follow different
one experiment found the opposite effect (Palmer dynamics.
and Maurer 1997); this experiment investigated
weed invasion in crop monocultures and ®ve-
species mixtures and found that the more diverse
7.5 Strengthening the link between
crop mixtures harboured a more species-rich
theory and experiments
(mostly annual) weed community. Weed invasion
resistance in terms of weed biomass, however, was Perhaps for the ®rst time in the history of the
not affected by diversity. Another work by Lavorel diversity±stability debate, we now have two essen-
et al. (1999) found no diversity effect on invasion tial ingredients for scienti®c progress: ®rst, a con-
resistance in Mediterranean grassland commun- ceptual framework that is suf®ciently broad and
ities. In contrast to the other studies in Table 7.4, clearÐalbeit certainly improvableÐto avoid con-
these two studies were relatively short-term fusion and sweeping generalizations, and, second,
experiments in which weed species were estab- a convergence of observational, experimental and
lished at the same time as target communities. No theoretical approaches towards common objectives
predictable effect of diversity on invasion resist- and questions. It must be borne in mind, however,
ance was found either in an experiment by Crawley that the current work focused on biodiversity and
et al. (1999), in which species composition was non- ecosystem functioning is addressing only part of
random, just as in Palmer and Maurer (1997) and the original debate, several aspects of which
Lavorel et al. (1999). From the currently available remain untested.
evidence, it seems that more diverse communities Theory has historically been prominent in the
are harder to invade in fully established commun- diversity±stability debate. But profusion of theory
ities with random, restricted random, or nested is no guarantee of clarity and relevance. As experi-
(one study) designs, with true replication of mental and observational evidence accumulates,
diversity levels and experimental introduction of the weaknesses of past, abstract theories have
invading species. become more apparent. Theory needs to evolve to
The mechanisms behind the positive relationship provide better guidance for experiments. Most of
between diversity and invasion resistance in these the classical equilibrium approaches based on auto-
small-scale experiments probably involve both better nomous dynamical systems may be inadequate to
resource utilization in more diverse communities understand stability properties such as variability,
(Hector et al. 2001a; Levine et al., Chapter 10) and resilience and resistance at the ecosystem level.
selection processes since more diverse commun- Here, we have argued that, to understand func-
ities have a higher chance to contain species that tional compensations inecosystems, new approaches
bene®t from altered environmental conditions should be developed that take into account the
(Schmid et al. 2002). Strong effects of species iden- dynamics of diversity and the potential for adapt-
tity or species composition were detected in four ive changes through asynchronous species ¯uc-
of the 13 studies that found a positive effect of tuations, species replacement, phenotypic plasticity
diversity on invasion resistance. Disentangling and evolutionary change. In other words, ecosys-
species diversity and species identity effects, how- tems must be fully treated as complex adaptive
ever, requires true replication of diversity treatments systems, as proposed by such scientists as Holling
with different species assemblages (Allison 1999), (1986) and Levin (1999). Most of the current theory
a requirement which was not ful®lled in all is also borrowed from community ecology, with
 A NEW LOOK AT THE RELATIONSHIP BETWEEN DIVERSITY AND STABILITY 89

an emphasis on total plant biomass or primary de®ne the subset of environmental and ecological
production as the ecosystem properties invest- conditions conducive to the establishment of such a
igated. Total plant biomass and primary produc- mode of ecosystem functioning. Perhaps the main
tion are easily related to individual plant or dif®culty here is the production under experimental
population-level properties by simple aggregation, conditions of realistic environmental ¯uctuations
but this might be less straightforward for other with a controlled frequency structure. Methodo-
ecosystem processes. The historical separation bet- logical advances in this direction have been made
ween community ecology, which is demography (Cohen et al. 1998) and key microcosm experiments
oriented, and ecosystem ecology, which focuses on are starting to be conducted. Clearly, there is still
whole-ecosystem functional processes, demands scope for many innovative ideas in the design of
new approaches to lay a bridge between these experiments in this area.
different perspectives (Loreau 2000a). There have To strengthen the link between theory and experi-
been very few attempts to explore the effects of ments, theoretical and experimental studies should
biodiversity on the functioning of full ecosystems attempt to adopt similar measurements of stability.
comprising higher trophic levels, decomposers and Many past theoretical developments and predic-
nutrient cycling (Loreau 1996, 2001), and none as tions are dif®cult to directly test experimentally
yet has considered stability explicitly. because equivalent measures of stability often do
It is very encouraging that experiments have not exist in experimental systems. For example, there
started to test new ideas on the relationships is not a straightforward experimental equivalent to
between the diversity and various aspects of stab- an eigenvalue. Experimental approaches, in turn,
ility of ecosystems. A number of these experiments, must consider relevant theoretical work when
however, have been debated because of the poten- designing and interpreting results. Experimental
tial presence of confounding factors, which now response variables could be chosen to correspond
need to be addressed by new experimental designs. more closely to theoretical stability estimates. In
To date, no experimental studies have directly addition, care must be taken to not misapply the-
manipulated long-term environmental variability oretical results to experimental results, particularly
to test the potentially important role that environ- when the de®nitions of stability differ.
mental ¯uctuations may play as both the creator One of the dif®culties of measuring stability
and driver of the conditions necessary for the in natural ecosystems is that natural ecosystems
existence of compensatory dynamics. Several show a variety of complex dynamics. Many eco-
empirical studies suggest that this may be so. Some systems experience predictable variations, such
microcosm studies have demonstrated that certain as the seasonal changes in the pelagic community
types and frequency of environmental ¯uctuation of temperate lakes or succession in forests, or
may set levels of species richness and affect com- react to disturbance in a fairly predictable man-
munity stability (Eddison and Ollason 1978; Ollason ner; algal blooms following eutrophication or
1977; Rashit and Bazin 1987). More directly, the re-establishment of forests after local clear-cuts are
work reported by Frost et al. (1995) and Klug et al. examples. Under normal environmental ¯uctua-
(2000) provides some short-term evidence demon- tions, ecosystems often develop along a trajectory
strating the operation of compensatory dynamics that is an environmentally determined dynamical
in lake communities in response to pH perturba- attractor; systems starting out with different initial
tions, and Morgan-Ernest and Brown (2001) provide conditions then converge over time. Such non-
long-term evidence for the existence of similar com- equilibrium systems are stable and return to their
pensatory dynamics in arid grassland communities. attractor following a perturbation. In this case, tests
Establishing the general importance of the insur- of stability properties following perturbation could
ance hypothesis would require the demonstration use the deviation between a perturbed system and
of the assembly or evolution of an ecosystem func- a control system as a measure (e.g. Wardle et al.
tioning in this manner under controlled environ- 2000a) or, alternatively, the relative difference in
mental ¯uctuations. Experiments of this kind would disturbance effects along a gradient of diversity
90 BIODIVERSITY AND ECOSYSTEM FUNCTIONING

(e.g. van der Putten et al. 2000). A caveat, however, ural ecosystems will be to untangle the effects of
is that perturbations may be initially ampli®ed environmental factors that drive natural variations
before returning to the original state (Neubert and in diversity and of diversity itself, and to develop
Caswell 1997), in which case a suf®ciently long new theory that integrates the mutual interactions
experimental time period is necessary to ensure among biodiversity changes, ecosystem function-
that the system does converge. ing and abiotic factors into a single, uni®ed picture
Ecosystems that exhibit more complex dynamics (Loreau et al. 2001).
or ¯ip between alternative stable states (Scheffer
et al. 2001) will be harder to analyse because there is
no single reference system. A disturbance can switch
7.6 Implications for policy and
a system into a different con®guration such that
management
the `recovered' system is vastly different than the
unperturbed control. Such systems require a focus As human impact on ecological systems increases,
on what Holling (1973) termed `ecological resili- scientists are increasingly challenged to commun-
ence', or what we here call `robustness' (Table 7.1). icate new knowledge to policy- and decision-
An appropriate experimental design would employ makers (Lubchenco 1998). Does our current state of
a range of disturbance magnitudes, which allows knowledge of diversity±stability relationships pro-
de®ning the domain within which disturbed sys- vide speci®c information for policy and manage-
tems will not loose critical functional groups or ment decisions?
processes. Complete similarity with the undis- Society depends on the steady and predictable
turbed system may not be necessary as a criterion inputs of ecosystem services (Daily 1997). Current
for recovery; one might be interested in maintain- evidence suggests that higher diversity may provide
ing the same ecological processes, such as top- greater reliability in the production of ecosystem
down or bottom-up control. Thus, different types services such as food and ®bre production, pollina-
of systems may require different kinds of concepts, tion levels, and nutrient cycling. Diversity may also
measures and experimental design. decrease the probability of successful invasions of
A future challenge will be to recognize various non-native species, many of which have had sub-
types of complex dynamics in natural ecosystems, stantial economic, conservation, and societal con-
and to incorporate them in theoretical work sequences (Mooney and Drake 1986; Drake and
attempting to study the relationship between Mooney 1989). Extinctions of native species may
stability and diversity. Such theory would hopefully lead to a further decrease in stability that causes a
produce realistic patterns of diversity±stability cascade of other extinctions, accelerating the rate of
relationships, provide testable mechanisms, and community change (Pimm et al. 1988; Borrvall et al.
help to sharpen and focus experiments designed to 2000). Finally, declines in ecosystem stability may
explore diversity±stability relationships. As theor- reduce our ability to predict or detect future envir-
etical and experimental foundations become more onmental changes in a background of higher eco-
solid, there will also be an increasing need for system variability, including the in¯uence of slow
long-term empirical data in the ®eld. Long-term processes such as climate change (Cottingham et al.
monitoring of both biodiversity and ecosystems 2000). Thus, the impact of biodiversity on ecosys-
processes is critical to apply our basic scienti®c tem stability appears to be a relevant feature to
understanding to real ecosystems, both natural and consider in policy and management decisions.
managed. Such data will need to be scaled to the We have shown, however, that there are a number
turnover time of the ecosystem processes being of components to stability, and that changes in
considered if we are to understand their implica- diversity may alter ecosystem stability in a variety
tions and relevance in the context of the natural of ways. Stability at one level may require change
functioning of ecosystems. This means that in some at another level; for instance, we have discussed
ecosystems, such as forests, very long time series how increased average magnitude and decreased
will be necessary. The critical challenges with nat- variability of ecosystem processes come about
 A NEW LOOK AT THE RELATIONSHIP BETWEEN DIVERSITY AND STABILITY 91

through changes at the species level. Despite pro®tably focused on economically important
progress in our understanding of diversity±stabil- ecosystems, such as agricultural ecosystems, and
ity relationships, current research is still largely on ecosystem goods and services of importance to
unable to provide speci®c policy recommendations society, including reliable supplies of clean fresh
due to the lack of abundant, consistent, and rele- water, and crop and ®sheries production.
vant long-term data on ecosystem processes for Given what we currently understand about the
most biomes (Schwartz et al. 2000; Cottingham et al. potential of diversity to buffer ecosystems against
2001; Hector et al. 2001b; Lawler et al. 2002). environmental ¯uctuations, future management
Diversity appears to play some role in maintaining efforts should look towards preserving the already
stability of certain processes in a handful of eco- `built-in' capacity of ecosystems to adapt to envir-
systems, but more research is needed before we onmental perturbations. This approach would
can con®dently justify biodiversity conservation on require an emphasis on preserving regional species
the basis of its ability to enhance or protect eco- diversity and the necessary habitat connectivity
system stability. Future work should begin to focus required for the assembly of local communities in
on scenarios that are immediately relevant to order to maintain the potential for high local
human society. For example, human impact causes diversity. Until we have a better understanding of
non-random changes in diversity or composition how diversity relates to stability, management
(Petchey et al. 1999). Exploring the consequences of strategies aimed at preserving diversity will at the
these non-random diversity changes for stability very least increase the potential for ecosystems to
will become important. Research could also be respond to future, changing environments.