Ecology Letters, (2006) 9: 1146–1156 doi: 10.1111/j.1461-0248.2006.00963.

REVIEW AND
SYNTHESIS Quantifying the evidence for biodiversity effects on
ecosystem functioning and services

Abstract
1
Patricia Balvanera, * Andrea B. Concern is growing about the consequences of biodiversity loss for ecosystem
Pfisterer,2 Nina Buchmann,3 functioning, for the provision of ecosystem services, and for human well being.
Jing-Shen He,4 Tohru Experimental evidence for a relationship between biodiversity and ecosystem process
Nakashizuka,5 David Raffaelli6 rates is compelling, but the issue remains contentious. Here, we present the first rigorous
and Bernhard Schmid2 quantitative assessment of this relationship through meta-analysis of experimental work
1
Centro de Investigaciones en
spanning 50 years to June 2004. We analysed 446 measures of biodiversity effects (252 in
Ecosistemas, UNAM, Morelia,
grasslands), 319 of which involved primary producer manipulations or measurements.
Mexico
2
Institute of Environmental
Our analyses show that: biodiversity effects are weaker if biodiversity manipulations are
Sciences, Universität Zürich,
less well controlled; effects of biodiversity change on processes are weaker at the
Zürich, Switzerland ecosystem compared with the community level and are negative at the population level;
3
Institute of Plant Sciences, ETH productivity-related effects decline with increasing number of trophic links between
Zurich, Zurich, Switzerland those elements manipulated and those measured; biodiversity effects on stability
4
Department of Ecology, Peking measures (ÔinsuranceÕ effects) are not stronger than biodiversity effects on performance
University, Peking, China measures. For those ecosystem services which could be assessed here, there is clear
5
Research Institute for evidence that biodiversity has positive effects on most. Whilst such patterns should be
Humanity and Nature, Kyoto, further confirmed, a precautionary approach to biodiversity management would seem
Japan prudent in the meantime.
6
Environment Department,
Keywords
University of York, York, UK
*Correspondence: E-mail:
Biodiversity–ecosystem functioning, diversity manipulations, ecosystem property,
[email protected] ecosystem services, ecosystem type, experimental design, meta-analysis, stability, trophic
level.

Ecology Letters (2006) 9: 1146–1156

a significant role in directly providing goods and services as

INTRODUCTION
well as regulating and modulating ecosystem properties (this
Human needs have been, and continue to be, satisfied at the term is used here to include ÔprocessesÕ and ÔfunctioningÕ)
expense of altered land use, climate, biogeochemical cycles that underpin the delivery of ecosystem services.
and species distributions (MA 2005). As a result, biodiversity Considerable research has gone into teasing out the
is declining a thousand times faster now than at rates found linkages between biodiversity, functioning and services
in the fossil record (MA 2005), raising concerns about (Naeem & Wright 2003), and experimental approaches now
consequences of such loss for ecosystem functioning, the account for 40% of the publications in this area (Fig. 1). Most
provision of ecosystem services and human well being experiments have manipulated diversity or have assembled
(Schläpfer & Schmid 1999; Chapin et al. 2000; Loreau et al. different diversities as a treatment variable and documented
2001; Kinzig et al. 2002; Dı́az et al. 2005; Hooper et al. 2005; the response of ecosystem properties and processes, including
MA 2005; Srivastava & Vellend 2005). Such concerns have modifying effects of environmental factors on such relation-
moved beyond the science community to the global ships (Naeem et al. 1994; Tilman 1996; McGrady-Steed et al.
stakeholder and policy community with the publication of 1997; Hector et al. 1999). The experimental designs used,
the Millennium Assessment (Dı́az et al. 2005; MA 2005). results obtained and interpretations made, have not been
That analysis acknowledges that biodiversity probably plays consistent and the field has been contentious and lively

 2006 Blackwell Publishing Ltd/CNRS

Review and Synthesis Biodiversity and ecosystem functioning/services 1147

100 specifics of experimental designs, the ecosystem properties

Number of publications

measured and the significance and size of reported effects

80
were entered into our database. We did not include duplicate
60 records, for example, the same experiment and same
40
measurement reported in a different publication or measured
in a different year (repeated measures). If, however, the
20 repeated measures were used to derive a new variable such as
0
temporal variation in the ecosystem property, these data were
included. We did not include studies that compared mono-
1995

1996

1997

1998

1999

2000

2001

2002

2003

2004
cultures with mixtures of a single higher diversity level or
Year of publication single-species removal experiments. We used all records that
Figure 1 The number of biodiversity–ecosystem functioning
reported effect sizes, allowing us to calculate correlation
articles published during the last decade is steadily growing (ISI coefficients for the relationship between biodiversity and
Web of Science). Experimental work (filled section) has contribu- ecosystem property, but we excluded studies from our
ted around 40% of the total number of articles (total bar) since the database, which reported only significance.
beginning of this century.
Data analyses

(Grime 1997; Wardle et al. 1997; Huston et al. 2000; Lepš Biodiversity effects were measured as simple or multiple
2004). Attempts have been made to provide common correlation coefficients, r. Using r instead of r2 (the
frameworks, identify areas of consensus or future challenges, coefficient of determination) had the advantage that we
as well as potential management and policy implications could assign negative and positive signs to effects. Main-
(Schläpfer & Schmid 1999; Loreau et al. 2001; Kinzig et al. taining negative and positive effects and using a
2002; Schmid et al. 2002; Dı́az et al. 2005; Hooper et al. 2005), Z-transformation (see below) allowed us to test the overall
but these syntheses have taken the form of largely subjective distribution for normality and to obtain normally distributed
assessments through qualitative literature reviews. Such error terms after fitting explanatory terms.
reviews provided an important foundation (in particular Simple correlation coefficients (365 records) were only
Schmid et al. 2002) for us to construct a more complete available where biodiversity was treated as an independent
database using strict selection criteria (Schläpfer & Schmid continuous variable or where a linear or log-linear contrast
1999) for the formal meta-analysis presented here. Specifi- was made for the factor biodiversity. When biodiversity was
cally, we pose the following questions: (i) what are the most analysed as a factor with more than one level (or as a
commonly addressed relationships between biodiversity and polynomial), we calculated multiple correlation coefficients
ecosystem properties? (ii) How do the experimental designs from the entries in the analysis of variance tables (81
used and the ecosystem properties measured affect the records). We used adjusted r2 values to derive correlation
outcomes and interpretation of biodiversity–ecosystem func- coefficients because these correct for the degrees of
tioning relationships? (iii) What can be learnt about biodiver- freedom used to fit a model (Sokal & Rohlf 1995). When
sity–ecosystem service relationships that could be useful for the relationship between the levels of the biodiversity factor
decision makers? and the response variable was generally negative, we gave
the multiple correlation coefficient a minus sign. In addition
to the sign, we also noted the shape of the relationship (see
METHODS below). To simultaneously analyse simple and multiple
Data collection correlation coefficients we normalized them using Fisher’s
z-algorithm (Rosenberg et al. 2000)
One hundred and three publications were included in our  
1þr
database, representing 446 ecosystem property measurements Zr ¼ 0:5  ln ð1Þ
from 1954 to June 2004 (see Appendix S1 and Table S1). 1r
These publications were identified from the ISI Web of and analysed these Zr-values as a new dependent variable. We
Science and Biological Abstracts database using criteria did all analysis with all 446 correlation coefficients and with
previously using the following search terms (Schläpfer & the subset of the 365 simple coefficients. Because the results
Schmid 1999): biodiversity or species richness and stability or ecosystem were the same, we only present those from the full analysis.
function or productivity or yield or food web. Where appropriate, we The common, normalized effects measure allowed us to
contacted authors of publications to obtain additional analyse all data together with a single general-linear
information and additional publications. Information about modelling framework, despite the overwhelming heterogen-

 2006 Blackwell Publishing Ltd/CNRS

1148 P. Balvanera et al. Review and Synthesis

eity of studies. Based on major controversies as well as areas another case, a single publication reported results from two
of consensus identified in previous qualitative synthesis separate experiments. In these two cases, we gave each
(Schläpfer & Schmid 1999; Loreau et al. 2001; Kinzig et al. publication two reference IDs to ensure full nesting. To
2002; Schmid et al. 2002; Dı́az et al. 2005; Hooper et al. avoid weak pseudo-replication due to measurements of
2005), a set of hypothesis were constructed about possible multiple ecosystem properties in single experiments, terms
effects of the specifics of experimental designs and the referring to specifics of experimental design and study site
ecosystem properties measured on the biodiversity effects could be tested against the reference ID instead of the
observed (Table 1). The studies were classified into groups residual mean square as error term. We used this very strict
using a separate explanatory factor for each of the test but list the mean squares in the ANOVA table so that
hypotheses (Table 1). The significance and explanatory readers can calculate the more liberal F-test as well. The
power of these factors and of interactions was then assessed reciprocal of the variance in the individual Zr values, based
in mixed-model analyses of variance (ANOVA). Study site and on the individual study sizes, was used as a weighting factor
reference were random terms in the model. in the ANOVA (Crawley 1993). This ensured that studies with
We compared a small number of alternative models for small sample sizes were not over-rated in comparison with
the fixed terms using adjusted r2 values (which gave the studies with large sample sizes. Throughout the paper, we
same model ranking as Akaike and Bayesian information report result in terms of these weighted average normalized
criteria). The selected final model contained only main effect sizes Zr and their standard errors.
effects but no interactions of fixed terms. Due to Ecosystem properties that could unequivocally be related
correlations between fixed terms, we assessed their explan- to ecosystem services (MA 2003; Dı́az et al. 2005), and thus
atory power in two ways if they were entered: (i) first into that could be assigned a positive (or negative) value for
the model or (ii) in a sequence of decreasing order of their human well being, were further analysed based on mean
F-values when entered first. The random effects were added values and standard errors of effect sizes. Some judgment is
after the fixed effects in the sequence study site/reference, involved in the assignment of positive or negative value,
imposing a nesting of these terms. In one case, a single because a particular ecosystem property may not be seen as
publication reported results from two study sites and in the same benefit by all stakeholders of biodiversity

Table 1 Hypotheses tested in the meta-analysis and corresponding explanatory terms in ANOVA

Explanatory term Null hypothesis

Type of diversity measure H1, biodiversity effects are independent of type of diversity measure used to estimate
relationship (e.g. species vs. functional diversity)
Type of experimental system H2, biodiversity effects are independent of type of experimental system (e.g. bottle, field)
Ecosystem type H3, biodiversity effects are independent of ecosystem type (e.g. grassland, forest)
Main cause of diversity changes H4, biodiversity effects are independent of main cause of diversity changes (direct vs. indirect
manipulation of diversity)
Design for direct species diversity H5, biodiversity effects are the same whether total density is held constant (substitutive
manipulations designs) or not (additive or designs without control of total density)
Type of indirect species diversity gradients H6, biodiversity effects are independent of the type of indirect species diversity gradients
[natural variation vs. gradient (e.g. nitrogen addition)]
Maximum species number H7, biodiversity effects are independent of maximum species number in most diverse
treatment
Trophic-level manipulated H8, biodiversity effects are independent of trophic level manipulated
Trophic level measured H9, biodiversity effects are independent of trophic level measured
Number of trophic links between them H10, biodiversity effects are independent of number of trophic links between level mani-
pulated and level measured
Ecosystem property H11, biodiversity effects are independent of the ecosystem property measured
Organization level of ecosystem property H12, biodiversity effects are independent of the level of organization at which the ecosystem
property was measured (population- vs. community- vs. ecosystem-level)
Biotic vs. abiotic ecosystem properties H13, biodiversity effects are independent of whether ecosystem property is biotic or abiotic
Dominant cycle to which ecosystem H14, biodiversity effects are independent of whether ecosystem property is associated to
property belongs water, nutrient, energy or biotic dynamics
Nature of ecosystem property H15, biodiversity effects are independent of whether ecosystem property is a stock or a rate
Study site H16, biodiversity effects are independent of location of study site

Listed are the null hypotheses we tried to reject.

 2006 Blackwell Publishing Ltd/CNRS

Review and Synthesis Biodiversity and ecosystem functioning/services 1149

(Srivastava & Vellend 2005). Only those ecosystem prop- nity (15), crop/successional (10) and ruderal/salt marsh
erties for which at least five effect size measurements were (13).
available were included in the analysis.
Trophic level manipulated and trophic level measured
Studies that manipulated diversity and/or measured
Groupings for specifics of experimental design and
diversity effects at different trophic levels were categor-
ecosystem properties (number of records in parentheses)
ized into: primary producer (319 manipulated and 241
Type of diversity measure measured), primary consumer (30 and 91), secondary
These included species richness (393), functional group consumer (four and 13), detritivores (15 and 38),
richness (23), evenness (11) and diversity indices (19). mycorrhiza (47 and 15), multitrophic (31 and five) and
Although we aimed to include diversity effects in the ecosystem level (0 and 43). ÔMultitrophicÕ refers to studies
broadest sense of the word, the majority of studies where diversity was manipulated on more than one
examined species richness effects only. Some studies trophic level or where the ecosystem property involves
reported effects of functional group richness, but only a more than one trophic level (e.g. total macrofaunal
few of these were intentionally designed from the start to biomass). Ecosystem level refers to properties measured
examine effects of varying functional diversity. in the entire ecosystem within the abiotic compartment
(e.g. nutrient loss from the system).
Type of experimental system
System types were bottle (microcosm studies) or pot (111), Number of trophic links
greenhouse, including climate chambers (62) and field (273). We counted the number of trophic links between the
Pot and greenhouse systems differ from field systems in that trophic level manipulated and the level at which the
the latter experience natural climate and light regimes. Field property was measured (Fig. 2).
systems included studies that directly and indirectly mani-
pulated species diversity. Effect form
The shapes of the biodiversity–ecosystem property relation-
Main cause of diversity change ships were classified into negative (40), negative linear (92),
Direct manipulations (398) of diversity were distinguished negative log-linear (41), idiosyncratic (113), positive (70),
from indirect ones (48). Indirect manipulations were found positive linear (56), positive log-linear (34). This classifica-
only in field studies and were further categorized as follows. tion was performed independently of significance or size of
biodiversity effects simply by inspecting results presented in
Type of indirect species diversity gradients the text and figures of the publications analysed. This
Indirect manipulations of diversity were divided into variable is similar to the effect size itself and could be used
natural variation (39) and gradient (9). In the first category, as an alternative dependent variable in log-linear analysis of
naturally varying diversity levels were constructed. In the deviance. We include this variable in the supplementary
second category, a natural (succession) or experimental online material but except for a single case (see below) the
gradient in environmental conditions (nutrient application only reported dependent variable in the present paper is
or multiple factors) generated the differences in diversity effect size per se.
levels.
Ecosystem properties measured
Design of direct species diversity manipulation experiments We included any physical characteristics of the ecosystems,
Direct manipulations of diversity were subdivided into those including process rates of energy and nutrient flow. To
which were set up so that total density remained constant, simplify comparisons, we grouped similar ecosystem pro-
i.e. substitutive experiments (357), and others, mostly perties (EP), which resulted in 28 groups; an additional
additive experiments (41). group was used to collect those measures that could not be
assigned. We distinguished between properties of the
Maximum species number ecosystem and those of an invader (defined as any species
Three levels of maximum diversity were recognized: low added after the establishment of a community) and we also
(£10 species, n ¼ 211), intermediate (11–20 species, n ¼ distinguished between effects on means of properties
104) and high (>20 species, n ¼ 131). measured and those that relate to their variances.

Ecosystem type Organizational level of the ecosystem property measured

These encompassed forest (43), grassland (258), marine (32), We distinguished between population-level properties,
freshwater (68), bacterial microcosm (seven), soil commu- recorded for individual target species, such as density, cover

 2006 Blackwell Publishing Ltd/CNRS

1150 P. Balvanera et al. Review and Synthesis

5
4 1
Multitrophic
2°Consumer

1
7
16

1°Consumer 5

27
12 232
9
5
Mycorrhiza 1°Producer 4 13
25 Figure 2 Number of measurements in pub-
lished biodiversity–ecosystem functioning
3 34 4 experiments for different trophic levels
11 Ecosystem manipulated (base of arrow) and trophic
1
levels measured (end of arrow). A domin-
25
Detritivores ance of measurements and manipulations of
2
primary producers was observed.

or biomass, and their temporal variance; community-level Study site (75)

properties, recorded for multispecies assemblages, such as Generally equivalent to location, this term was used to
density, biomass, consumption, diversity and their temporal distinguish different studies within a single location. Study
variance; and ecosystem-level properties, recorded for site reflects a set of environmental conditions particular to
abiotic components, such as nutrient, water or CO2 and that experiment.
their temporal variance.
Reference ID (105)
Dominant dynamic of ecosystem property This corresponded to individual publications, except where
Properties were assigned to the ecosystem cycle in which they a single publication reported results from more than one
predominate: water, nutrient, energy or biotic dynamics. study, in which case this publication received two reference
IDs. This ID is used to distinguish between groups of
Nature of ecosystem property potentially non-independent measurements in order to
Stock vs. rate measurements of ecosystem properties were avoid pseudo-replication.
distinguished.
RESULTS
Ecosystem service
Ecosystem services are the benefits people obtain from The overall mean of the standardized effect sizes Zr
ecosystems. Our classification followed that of the Millen- (weighted by the reciprocal of the variance of the individual
nium Ecosystem Assessment (MA 2003; Dı́az et al. 2005). A Zr-values) was significantly positive (X ¼ 0.101 ± 0.028,
list of ecosystem properties considered to underpin each t ¼ 3.57, d.f. ¼ 445, P < 0.001), indicating that negative
ecosystem service, as well as the directionality of expected responses of ecosystem properties to biodiversity manipula-
benefits to human well being, is provided below in the tions are less frequent or less strong than positive ones.
Results section. Nevertheless, the reported effect sizes varied greatly,
ranging from )2.71 to 2.39. In the following sections, we
explore the sources of this variation.
Groupings according to place of study and identity of
experiment (number of groups in parentheses)
Effects of specifics of experimental design and study site
Location of study site (60)
Site location of an experiment ranged from a precise Some specifics of the experimental design which we
place to a broad region, depending on the extent of the originally expected to have an influence on effect sizes in
study. fact could not be included in the final analysis model,

 2006 Blackwell Publishing Ltd/CNRS

Review and Synthesis Biodiversity and ecosystem functioning/services 1151

suggesting that they need not be a concern when designing effect sizes were close to zero (grassland 0.039 ± 0.038,
future biodiversity experiments. For instance, there was only freshwater )0.010 ± 0.065, marine )0.006 ± 0.109, forest
a weak influence of the type of diversity measure on )0.116 ± 0.076), whereas average effect sizes were larger
measured effect sizes (Table 2). Of particular note is that and positive for the ecosystem types with fewer records
effect sizes were only slightly larger when functional-group (ruderal/salt marsh, 1.058 ± 0.154; bacterial, 0.317 ± 0.095;
rather than species richness was manipulated (adjusted mean crop/successional, 0.245 ± 0.052; soil, 0.094 ± 0.086). This
values ± SE of Zr-values: 0.191 ± 0.103 vs. 0.116 ± 0.030). could imply that the research community’s perception of the
In contrast, the type of experimental system employed magnitude and direction of biodiversity effects may be
(bottle vs. greenhouse vs. field) strongly modified biodiver- biased by the focus to date on relatively few ecosystem types
sity effects (Table 2). More positive effects were found that included measures of negative impacts on properties.
where environmental variables could be controlled best, There was considerable variation among study sites, but this
such as in greenhouses and climate chambers was not significant in the multiway ANOVA using the strict
(0.467 ± 0.084) compared with bottle/pot experiments F-test with reference ID as error term (Table 2). In other
(0.100 ± 0.051) or field experiments (0.007 ± 0.033). words, effect sizes varied as much between references
Effect sizes also varied markedly between different types within study sites as between study sites.
of ecosystem (Table 2). For the four ecosystem types which Although average effect sizes were practically identical for
were represented most frequently in the data set, average studies that manipulated biodiversity directly or indirectly

Table 2 Results from one-way analyses of variance (ANOVA)s in the sequence of decreasing F-values and multiway ANOVA using this sequence
for fitting the corresponding fixed terms (see Methods for details)

H no. Variable d.f. Sum of squares Mean squares F P-value % Explained variance

One-way ANOVA
12 Organization level EP 2 2031.7 1015.9 40.27 <0.001 15.4
5 Type direct manipulations* 2 1802.5 901.2 35.00 <0.001 13.6
7 Maximum species number 2 1319.0 659.3 24.57 <0.001 10.0
2 Experimental system 2 1071.0 535.3 19.54 <0.001 8.1
3 Ecosystem type 7 2255.8 322.3 12.89 <0.001 17.1
11 Ecosystem property 28 3241.7 115.8 4.83 <0.001 24.5
16 Study site 74 6168.6 83.4 4.39 <0.001 46.7
1 Type diversity measure 3 377.2 125.7 4.33 0.005 2.9
15 Nature of EP 1 86.5 86.5 2.92 n.s. 0.7
8 Trophic-level manipulated 5 305.1 61.0 2.08 n.s. 2.3
9 Trophic-level measured 6 295.2 49.2 1.67 n.s. 2.2
10 Number of links 1 37.4 37.4 1.28 n.s. 0.3
14 Cycle type EP 4 143.9 36.0 1.21 n.s. 1.1
13 Biotic vs. abiotic EP 1 27.3 27.3 0.93 n.s. 0.2
6 Type indirect gradient* 2 14.1 7.1 0.24 n.s. 0.1
4 Direct vs. indirect 1 2.2 2.2 0.07 n.s. 0.0
ANOVA for selected model
12 Organization level EP 2 2031.9 1016.0 83.69 <0.001 15.38
5 Type direct manipulations 2 1295.5 647.4 18.19 <0.001à 9.81
7 Maximum species number 2 349.3 174.7 4.91 <0.05à 2.64
2 Experimental system 2 485.0 242.5 6.81 <0.01à 3.67
3 Ecosystem type 7 660.3 94.3 2.65 <0.05à 5.00
11 Ecosystem property 28 1196.6 42.7 3.52 <0.001 9.06
16 Study site 65 2501.7 38.5 1.08 n.s.à 18.94
Reference (within study site) 26 925.5 35.6 2.93 <0.001 7.01
Residual 337 3762.4 12.0 28.49
Total 444 13208.1 29.8 100.00

H no., hypothesis number (see Table 1); n.s., not significant (P > 0.05).
*These two terms include the last term (direct vs. indirect) as a category ÔnoneÕ.
This term includes the term Ôdirect vs. indirectÕ as a category ÔnoneÕ.
àF-test using reference ID as error term.

 2006 Blackwell Publishing Ltd/CNRS

1152 P. Balvanera et al. Review and Synthesis

(hypothesis 4), and between versions of indirect manipula- same trophic level (Fig. 3). Plant diversity also appeared to
tions (hypothesis 6), average effect sizes were smaller if enhance belowground plant and microbial biomass (Fig. 3),
direct manipulations maintained total density constant indicating positive biodiversity effects on the regulating
(substitutive designs, 0.031 ± 0.030) than if they did not ecosystem service of erosion control, as large root and
(0.868 ± 0.102) (Table 2). This confirms something which mycorrhizal networks are expected to reduce soil erosion.
has long been known to agricultural scientists and plant Positive biodiversity effects (Fig. 3) were found for most
ecologists using substitutive designs (Harper 1977), the ecosystem properties associated with nutrient cycling
importance of not confounding increasing species richness services. Plant diversity had positive effects on decomposer
and total density in experiments. activity and diversity, and both plant and mycorrhizal
Average effect sizes were positive if the maximum species diversity increased nutrients stored in the plant compart-
richness was larger than 20 species (0.344 ± 0.052) and ment of the ecosystem. It is unclear whether plant or
close to zero for the other two categories (two to 10 species: detritivore diversity has a general effect on soil nutrient
)0.049 ± 0.030; 11–20 species: )0.034 ± 0.081) (Table 2). supply.
Yet only 33 of 105 experiments (reference IDs) employed Increasing the diversity of primary producers contributed
more than 20 species at the highest diversity level. With to a higher diversity of primary consumers, which we
respect to effect form there was an indication that the odds consider here as a supporting service (Fig. 3). Our results
ratio between linear and log-linear-negative or -positive also suggest positive effects of biodiversity on the closely
relationships was greatest in experiments where maximum related regulating service of pest control; higher plant
species richness was lowest (P < 0.05), but even where diversity contributed to lowering plant damage (Fig. 3). The
maximum species richness was high, this ratio was > 1. effects of plant diversity on the performance and diversity
There were no overall effects of trophic level manipu- of predatory insects or other animals that control pests
lated, trophic level measured or number of trophic links require further investigation. In the case of the regulation of
between manipulated and response trophic levels (Table 2). invasive species, a service of economic significance and an
Nevertheless, productivity-related effect sizes did signifi- area of considerable debate (Levine & D’Antonio 1999;
cantly decline with increasing number of trophic links Fargione et al. 2003), we found reduced invader abundance,
(F1,140 ¼ 5.74, P < 0.05). survival, fertility and diversity when plant diversity was
higher (Fig. 3).
Temporal stability is directly linked to reliability of service
Effects of ecosystem properties measured
delivery (Dı́az et al. 2005). Our analysis indicates that more
Biodiversity effects differed significantly among the 29 diverse systems have greater temporal stability, as well as
different groups of ecosystem properties (Table 2). A large greater resistance to external forces such as nutrient
fraction of the variance in effect sizes was explained by perturbations and invading species (Fig. 3). However, this
comparing population-, community- and ecosystem-level was not the case for other stressors such as warming,
measures of ecosystem properties (Organization level EP in drought or a high variance in other environmental condi-
Table 2). Biodiversity negatively affected population-level tions. In contrast to the suggestion of qualitative reviews
measures ()0.332 ± 0.053), but positively affected commu- (e.g. Srivastava & Vellend 2005), portfolio and insurance
nity-level measures (0.270 ± 0.036). Ecosystem-level meas- effects of biodiversity (Tilman 1996; Naeem & Li 1997;
ures showed an intermediate response (0.066 ± 0.046). In Yachi & Loreau 1999), i.e. effects on variances or
contrast, no differences were found between biotic and disturbance responses of ecosystem properties, are not
abiotic ecosystem properties, stocks and rates, nor between more common than performance effects of biodiversity, i.e.
those more related to carbon, nutrient, water or biotic cycles effects on means of ecosystem properties (F1,444 ¼ 0.09,
(terms Ôbiotic vs. abiotic EPÕ, Ônature of EPÕ and Ôcycle type P ¼ 0.75).
EPÕ, respectively, in Table 2).
DISCUSSION
Biodiversity–ecosystem service relationships
The database assembled here clearly contains an over-
Biodiversity effects were explored in more detail by plotting representation of some ecosystem types and ecosystem
mean values and SE for groups of ecosystem properties in properties, especially grasslands and primary production
Fig. 3 and relating these groups to ecosystem services. measures. It is not surprising that experimental grassland
Productivity is a fundamental supporting ecosystem plots are often used as model systems in biodiversity studies,
service that underpins the provision of services such as food because grassland is a widespread system, experiments can
or wood (MA 2003; Dı́az et al. 2005). Generally, increasing be relatively easily set up at constant total density (as
biodiversity at one trophic level increased productivity at the opposed to microcosms with strong population dynamics),

 2006 Blackwell Publishing Ltd/CNRS

Review and Synthesis Biodiversity and ecosystem functioning/services 1153

Responses of ecosystem property Number of

Ecosystem Ecosystem to increasing biodiversity measurements
service property –1 1 0 40 80
0

1° Producer

tertiary productivity
Primary /secondary/
abundance

1° Consumer
abundance

2° Consumer
abundance

Erosion
control
Plant root biomass

Mycorrhiza
abundance

Nutrient cycling Decomposer

activity

Plant nutrient
concentration

Nutrient supply
from soil

1º Consumer diversity
biological diversity

Figure 3 Magnitude and direction of biodi- 1º

1ºConsumers:
Regulation of

(Plant disease severity)

versity effects (shown are mean values and
SE of normalized effect sizes Zr, weighted Decomposer diversity
by the reciprocal of the variance of the
(Invader fitness)
individual Zr-values) and number of meas-
urements available for ecosystem properties
(Invader diversity)
organized into ecosystem services. Coloured
bars show differential effects of trophic level Consumption resistance
manipulated: green, primary producers; blue,
primary consumers; pink, mycorrhiza; Invasion resistance
brown, decomposer; grey, multitrophic
Drought resistance
Stability

(multiple levels simultaneously manipulated).

Ecosystem properties shown in parentheses
Resistance vs.
were considered of negative value for human other disturbances
well being, and thus opposite of effect sizes Natural variation
are shown.

yet they do not require very large areas (as opposed to because finding a significantly negative effect would be just
forests). In addition, primary productivity plays a major role as interesting and just as likely to be reported. Nevertheless,
in delivering a wide range of ecosystem services. Neverthe- there was significant variation between studies in the
less, future biodiversity experiments should embrace a magnitude and direction of biodiversity effects, attributable
broader range of systems, properties and trophic levels if the mainly to specifics of experimental design and the ecosys-
generality of these relationships is to be established. In tem properties measured, as also argued in qualitative
particular, a recent experiment that came to light after our reviews (Hooper et al. 2005).
analysis was carried out (Bell et al. 2005), suggests that
bacterial systems hold great promise for future research of
Specifics of experimental design and ecosystem properties
biodiversity effects on ecosystem functioning.
Notwithstanding this heterogeneity in the database, our A large number of negative effects were associated with
analyses indicate an overall significant positive effect of population-level measures, whilst positive effects were
biodiversity on ecosystem processes. We do not believe that associated with community-level measures. This result
this represents a publication bias towards positive effects, provides perhaps the strongest empirical evidence to date

 2006 Blackwell Publishing Ltd/CNRS

1154 P. Balvanera et al. Review and Synthesis

for the prediction that individual populations are expected are acknowledged to have their own advantages and
to fluctuate more with increasing biodiversity, but the shortcomings; but the present analysis has allowed a formal
community stability and productivity should be enhanced assessment of the degree to which these really are important.
(May 1981; Tilman 1996). Surprisingly, we found no significant differences between
In contrast to the outcomes of qualitative reviews those experiments where diversity was manipulated directly
(Hooper et al. 2005), we could not find a simple dependence and those involving indirect manipulations by altering
of biodiversity effects on the trophic levels manipulated or environmental conditions. However, there was clear evi-
measured. However, we did find productivity-related biodi- dence in favour of substitutive designs with control for
versity effects that declined with increasing number of constant total density of individuals at the start of an
trophic links between those trophic levels which were experiment. If total density is allowed to vary, in most cases
manipulated and those at which the property was measured. in parallel with species richness, larger effects are seen, but
This intuitively compelling result has never been reported one cannot unequivocally attribute them to biodiversity or
before. It is clear that experiments need to be extended density. In other words, such experiments are confounded.
beyond the single trophic level approach to better under- Using a large number of species at the highest diversity
stand such variations in biodiversity effects across an levels of an experiment increases the chances of detecting
ecosystem (Petchey et al. 2002; Raffaelli et al. 2002). biodiversity effects, although this must be weighed up
Variation in biodiversity effects among study sites and against the increased work involved in setting up such an
references suggest that local environmental or specific unrec- experiment. Nevertheless, there is a clear need to include
ognized experimental factors may either increase or decrease higher levels of species richness in experiments. Unfortu-
biodiversity effects. Previous work (Hector et al. 1999) had nately, interesting new simulation and empirical studies
already indicated important influences of location on biodi- which used non-random extinction scenarios (Raffaelli
versity effects. The additional variation among references 2004; Solan et al. 2004; Zavaleta & Hulvey 2004; Bunker
within study sites, which actually made the variation between et al. 2005; Schläpfer et al. 2005; Srivastava & Vellend 2005)
sites non-significant, is reported here for the first time. could not be included in our analysis because they were
Sufficient information is not available to permit analysis published after our analyses were complete.
of biodiversity-modifying factors, such as nutrient levels or An important question when designing a biodiversity–
elevated CO2 (Hooper et al. 2005), but it is clear that ecosystem functioning experiment is what expression of
biodiversity effects are significantly weaker in less-controlled diversity to manipulate: richness, evenness or functional
experimental systems. Indeed, it is much more difficult to groups? The literature is somewhat divided on this issue (Dı́az
maintain diversity treatments on open field plots than in & Cabido 2001; Loreau et al. 2001; Hooper et al. 2005;
closed bottles; environmental heterogeneity, unpredictable Petchey & Gaston 2006; Wright et al. 2006), but the predom-
biotic and abiotic environmental fluctuations and sampling inant view is that functional groups may be more important
variances are greater in the former. Thus, while our results than species richness, consistent with our own findings.
would suggest that further research under controlled
conditions is needed to improve our understanding of
Biodiversity–ecosystem service relationships
biodiversity effects on ecosystem functioning, extrapolation
of those results to the larger landscape scale is likely to be Where ecosystem properties could be related to ecosystem
hindered by the greater environmental heterogeneity and its services (Srivastava & Vellend 2005), clear positive effects of
effects on ecosystem functioning (Loreau et al. 2001; biodiversity were found, for both regulating and supporting
Hooper et al. 2005). In this respect, field experiments are services. Nevertheless, our ability to make these linkages at
likely to be more meaningful for extrapolation to the spatial (landscape) scales relevant to the human enterprise is
landscape scales at which humans impact on biodiversity limited at present (Kremen 2005). There is an urgent need to
and hence service delivery. On the other hand, in a recently extend experimental, observational and theoretical work on
constructed grassland experiment in Jena, Germany, Rosher biodiversity effects for an array of ecosystem functions that
et al. (2005) found a similar plant diversity–productivity can be linked to ecosystem services, such as water quantity
relationship in small plots of 12.25 m2 and in plots more and quality, pollination, regulation of pests and human
than 30 times larger (400 m2). diseases, carbon storage and climate regulation, waste
The effect on our understanding of the relationship management and cultural services, and to evaluate biodi-
between biodiversity and ecosystem functioning of differ- versity–ecosystem service relationships at the larger spatial
ences in the way biodiversity is manipulated, how experi- scales relevant to management (Kremen et al. 2004; Balva-
ments are set up, and how response variables are measured nera et al. 2005).
in such experiments has been much debated (Schmid et al. The role of biodiversity in buffering environmental
2002; Lepš 2004). Different experimental designs and setups variation and thus providing consistent service delivery

 2006 Blackwell Publishing Ltd/CNRS

Review and Synthesis Biodiversity and ecosystem functioning/services 1155

has received extensive theoretical treatment (Tilman 1996; expected to compromise service delivery. Secondly, whilst
Yachi & Loreau 1999; Hooper et al. 2005). In general, a further research is needed to confirm such linkages, in
positive effect of biodiversity is expected on the stability of particular to extend the work to a broader range of systems
ecosystem properties (Tilman 1996; Naeem & Li 1997; and properties, society in the meantime should proceed in a
Yachi & Loreau 1999; Hooper et al. 2005), and qualitative precautionary manner in its use and management of
reviews have suggested that such effects on the variance in biodiversity.
processes (stability) may be stronger than the effects on
means (stocks and fluxes; Srivastava & Vellend 2005). The
ACKNOWLEDGEMENTS
quantitative results from our meta-analysis do not support
this view, rather indicating that biodiversity effects on We thank the Swiss Biodiversity Forum for administrative
disturbance buffering are dependent on the nature of the support and guidance and SCOPE, with Chris Field, Carlo
disturbance. Thus, while biodiversity effects on buffering of Heip and Osvaldo Sala for suggestions about the project
nutrient perturbations and invading species were positive, design. We acknowledge helpful comments from three
biodiversity effects on buffering influences of warming, anonymous referees and S. Naeem which improved the
drought or high environmental variance were neutral or manuscript. We thank the Swiss Agency for the Environ-
slightly negative. ment, Forests and Landscape (SAEFL) and the University
of Zurich for financial support. We thank Manuel Maass for
suggestions and Alberto Valencia and Heberto Ferreira for
CONCLUSIONS
technical support.
Whilst there are many qualitative reviews and position
statements about the effects of biodiversity on ecosystem
REFERENCES
properties and services, our analysis provides the first
extensive quantitative meta-analysis of this relationship. This Balvanera, P., Kremen, C. & Martı́nez-Ramos, M. (2005). Applying
analysis suggests that simple generalizations among ecosys- community structure analysis to ecosystem function: examples
from pollination and carbon storage. Ecol. Appl., 15, 360–375.
tem types, ecosystem properties or trophic level manipula-
Bell, T., Newman, J.A., Silverman, B.W., Turner, S.L. & Liley, A.K.
ted or measured will be difficult to sustain. Considerations (2005). The contribution of species richness and composition to
of the way in which biodiversity is defined and manipulated, bacterial services. Nature, 436, 1157–1160.
and disentangling the many separate effects and the Bunker, D.E., DeClerck, F., Bradford, J.C., Colwell, R.K., Perfecto,
interactions between them, as well as those with environ- I., Phillips, O.L. et al. (2005). Species loss and aboveground
mental heterogeneity, will be a major challenge for the next carbon storage in a tropical forest. Science, 310, 1029–1031.
generation of experiments. We offer our database (Supple- Chapin, F.S., Zavaleta, E.S., Eviner, V.T., Naylor, R.L., Vitousek,
P.M., Reynolds, H.L. et al. (2000). Consequences of changing
mentary material) as a building block for continued synthesis
biodiversity. Nature, 405, 234–242.
attempts. The advantages of a formal meta-analysis are Crawley, M.J. (1993). Glim for Ecologists. Blackwell, Oxford.
illustrated by the following novel contributions we have Dı́az, S. & Cabido, M. (2001). Vive la differénce: plant functional
been able to bring to the synthesis: (i) biodiversity effects are diversity matters to ecosystem functioning (review article). Trends
weaker if biodiversity manipulations are less well controlled Ecol. Evol., 16, 646–655.
(e.g. field vs. greenhouse or climate chamber); (ii) bio- Dı́az, S., Tilman, D., Fargione, J., Chapin, F.S. III, Dirzo, R.,
diversity effects are weaker if the highest diversity levels Ktzberber, T. et al. (2005). Biodiversity regulation of ecosystem
services. In: Trends and Conditions (ed. MA). Island Press,
in an experiment are lower (e.g. £ 10 vs. > 10 species);
Washington, DC, pp. 279–329.
(iii) biodiversity experiments should avoid confounding Fargione, J., Brown, C.S. & Tilman, D. (2003). Community
diversity and total density (they should use a substitutive assembly and invasion: an experimental test of neutral versus
design); (iv) biodiversity effects are weaker at the ecosystem niche processes. Proc. Natl Acad. Sci. USA, 100, 8916–8920.
than the community level and negative at the population Grime, J.P. (1997). Biodiversity and ecosystem function: the debate
level; (v) productivity-related biodiversity effects decline deepens. Science, 277, 1260–1261.
with increasing number of trophic links between level Harper, J.L. (1977). Plant Population Biology. Academic Press, Lon-
don.
manipulated and level measured; (vi) biodiversity effects on
Hector, A., Schmid, B., Beierkuhnlein, C., Caldeira, M.C., Diemer,
stability measures are not obviously stronger than biodiver- M., Dimitrakopoulos, P.G. et al. (1999). Plant diversity and
sity effects on performance measures. productivity experiments in European grasslands. Science, 286,
There are clear messages for policy makers from these 1123–1127.
analyses. First, for those ecosystem services that could be Hooper, D.U., Chapin, F.S., Ewell, J.J., Hector, A., Inchausti, P.,
assessed in the present study, there is clear evidence that Lavorel, S. et al. (2005). Effects of biodiversity on ecosystem
biodiversity has positive effects on the provision of those functioning: a consensus of current knowledge. Ecol. Monogr., 75,
3–35.
services and that further biodiversity loss can only be

 2006 Blackwell Publishing Ltd/CNRS

1156 P. Balvanera et al. Review and Synthesis

Huston, M.A., Aarssen, L.W., Austin, M.P., Cade, B.S., Fridley, Schläpfer, F., Pfisterer, A.B. & Schmid, B. (2005). Non-random
J.D., Garnier, E. et al. (2000). No consistent effect of plant species extinction and plant production: implications for eco-
diversity on productivity. Science, 289, 1255. system functioning. J. Appl. Ecol., 42, 13–24.
Kinzig, A., Pacala, S.W. & Tilman, D. (2002). Functional Consequences Schmid, B., Joshi, J. & Schläpfer, F. (2002). Empirical evidence for
of Biodiversity: Empirical Progress and Theoretical Extensions. Princeton biodiversity–ecosystem functioning relationships. In: Functional
University Press, Princeton, NJ. Consequences of Biodiversity: Experimental Progress and Theoretical
Kremen, C. (2005). Managing ecosystem services: what do we need Extensions (eds Kinzig, A., Tilman, D. & Pacala, P.). Princeton
to know about their ecology? Ecol. Lett., 8, 468–479. University Press, Princeton, NJ, pp. 120–150.
Kremen, C., Williams, N.M., Bugg, R.L., Fay, J.P. & Thorp, R.W. Sokal, R.R. & Rohlf, F.J. (1995). Biometry. The Principles and Practice of
(2004). The area requirements of an ecosystem service: crop Statistics in Biological Research. Freeman, New York.
pollination by native bee communities in California. Ecol. Lett., 7, Solan, M., Cardinale, B.J., Downing, A.L., Engelhardt, K.A.M.,
1109–1119. Ruesink, J.L. & Srivastava, D.S. (2004). Extinction and eco-
Lepš, J. (2004). Diversity and ecosystem function. In: Vegetation system function in the marine benthos. Science, 306, 1177–
Ecology (ed. Maarel, E.v.d.). Blackwell, Oxford, pp. 199–237. 1180.
Levine, J.M. & D’Antonio, C.M. (1999). Elton revisited: a review of Srivastava, D.S. & Vellend, M. (2005). Biodiversity–ecosystem
evidence linking diversity and invasibility. Oikos, 87, 15–26. function research: is it relevant to conservation? Annu. Rev. Ecol.
Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J.P., Evol. Syst., 36, 267–294.
Hector, A. et al. (2001). Biodiversity and ecosystem functioning: Tilman, D. (1996). Biodiversity: population versus ecosystem
current knowledge and future challenges. Science, 294, 804–808. stability. Ecology, 77, 350–363.
MA (2003). Ecosystems and Human Well-being: A Framework for Wardle, D.A., Booner, K.I. & Nicholson, K.S. (1997). Biodiversity
Assessment. Island Press, Washington, DC. and plant litter: experimental evidence which does not support
MA (2005). Ecosystems and Human Well-Being: Synthesis. Island Press, the view that enhanced species richness improves ecosystem
Washington, DC. function. Oikos, 79, 247–258.
May, R.M. (1981). Theoretical Ecology: Principles and Applications, 2nd Wright, J.P., Naeem, S., Hector, A., Lehman, C., Reich, P.B.,
edn. Blackwell Scientific Publications, Oxford. Schmid, B. et al. (2006). Conventional functional classification
McGrady-Steed, J., Harris, P.M. & Morin, P.J. (1997). Biodiversity schemes underestimate the relationship with ecosystem func-
regulates ecosystem predictability. Nature, 390, 162–163. tioning. Ecol. Lett., 9, 111–120.
Naeem, S. & Li, S. (1997). Biodiversity enhances ecosystem relia- Yachi, S. & Loreau, M. (1999). Biodiversity and ecosystem
bility. Nature, 390, 507–509. functioning in a fluctuating environment: the insurance
Naeem, S. & Wright, J.P. (2003). Disentangling biodiversity effects hypothesis. Proc. Natl Acad. Sci. USA, 96, 1463–1468.
on ecosystem functioning: deriving solutions to a seemingly Zavaleta, E.S. & Hulvey, K.B. (2004). Realistic species losses
insurmountable problem. Ecol. Lett., 6, 567–579. disproportionately reduce grassland resistance to biological
Naeem, S., Thompson, L.J., Lawler, S.P., Lawton, J.H. & Woodfin, invaders. Proc. Natl Acad. Sci. USA, 306, 1175–1177.
R.M. (1994). Declining biodiversity can alter the performance of
ecosystems. Nature, 368, 734–736.
Petchey, O.L. & Gaston, K.J. (2006). Functional diversity: back to SUPPLEMENTARY MATERIAL
basics and looking forward. Ecol. Lett., 9, 741–758.
Petchey, O., Morin, P.J., Hulot, F.D., Loreau, M., McGrady-Steed,
The following supplementary material is available for this
J. & Naeem, S. (2002). Contributions of aquatic model systems article:
to our understanding of biodiversity and ecosystem functioning. Appendix S1 Literature used for the meta-analysis.
In: Biodiversity and Ecosystem Functioning. Synthesis and Perspectives
Table S1 Database used in the meta-analysis.
(eds Loreau, M., Naeem, S. & Inchausti, P.). Oxford University
Press, Oxford, pp. 127–138. This material is available as part of the online article from:
Raffaelli, D. (2004). How extinction pattern affect ecosystems. http://www.blackwell-synergy.com/doi/full/10.1111/
Science, 306, 1141–1142.
Raffaelli, D., van der Putten, W.H., Persson, L., Wardle, D.A.,
j.1461-0248.2006.00963.x
Petchey, O., Koricheva, J. et al. (2002). Multi-trophic dynamics Please note: Blackwell Publishing are not responsible for the
and ecosystem processes. In: Biodiversity and Ecosystem Functioning: content or functionally of any supplementary materials
Synthesis and Perspectives (eds Loreau, M., Naeem, S. & Inchausti,
P.). Oxford University Press, Oxford, pp. 147–154.
supplied by the authors. Any queries (other than missing
Rosenberg, M.J., Adams, D.C. & Gurevitch, J. (2000). Metawin 2.0 material) should be directed to the corresponding author for
User’s Manual: Statistical Software for Meta-Analysis. Sinauer Asso- the article.
ciates, Sunderland, MA.
Rosher, C., Temperton, V.M., Scherer-Lorenzen, M., Schmitz, M.,
Schumacher, J., Schmid, B. et al. (2005). Overyielding in
experimental grassland communities- irrespective of species pool
or spatial scale. Ecol. Lett., 8, 419–429. Editor, Shahid Naeem
Schläpfer, F. & Schmid, B. (1999). Ecosystem effects of biodi- Manuscript received 22 March 2006
versity – a classification of hypotheses and exploration of First decision made 15 May 2006
empirical results. Ecol. Appl., 9, 893–912. Manuscript accepted 3 July 2006

 2006 Blackwell Publishing Ltd/CNRS