Content uploaded by Bernhard Schmid
Author content
All content in this area was uploaded by Bernhard Schmid on Nov 27, 2019
Content may be subject to copyright.
LETTER
Overyielding in experimental grassland
communities – irrespective of species pool
or spatial scale
Christiane Roscher,
1
* Vicky M.
Temperton,
2
Michael Scherer-
Lorenzen,
3
Martin Schmitz,
1,4
Jens Schumacher,
1
Bernhard
Schmid,
4
Nina Buchmann,
3
Wolfgang W. Weisser
1
and
Ernst-Detlef Schulze
2
1
Institute of Ecology, University
of Jena, Dornburger Str. 159,
D-07743 Jena, Germany
2
Max Planck Institute for
Biogeochemistry, PO Box
100164, D-07749 Jena, Germany
3
Institute of Plant Sciences,
ETH-Zurich, Universitaetsstr. 2,
CH-8092 Zurich, Switzerland
4
Institute of Environmental
Sciences, University of Zurich,
Winterthurerstr. 190, CH-8057
Zurich, Switzerland
*Correspondence: E-mail:
christiane.roscher@uni-jena.de
Abstract
In a large integrated biodiversity project (ÔThe Jena ExperimentÕin Germany) we
established two experiments, one with a pool of 60 plant species that ranged broadly
from dominant to subordinate competitors on large 20 ·20 m and small 3.5 ·3.5 m
plots (¼main experiment), and one with a pool of nine potentially dominant species on
small 3.5 ·3.5 m plots (¼dominance experiment). We found identical positive species
richness–aboveground productivity relationships in the main experiment at both scales.
This result suggests that scaling up, at least over the short term, is appropriate in
interpreting the implications of such experiments for larger-scale patterns. The species
richness–productivity relationship was more pronounced in the experiment with
dominant species (46.7 and 82.6% yield increase compared to mean monoculture,
respectively). Additionally, transgressive overyielding occurred more frequently in the
dominance experiment (67.7% of cases) than in the main experiment (23.4% of cases).
Additive partitioning and relative yield total analyses showed that both complementarity
and selection effects contributed to the positive net biodiversity effect.
Keywords
Biodiversity, complementarity effect, dominant species, plant species richness, plot size,
productivity, selection effect, The Jena Experiment.
Ecology Letters (2005) 8: 419–429
INTRODUCTION
A central issue in current ecological research is the potential
influence of biodiversity on ecosystem functioning (Chapin
et al. 1998; Loreau et al. 2001). Environmental conditions,
species interactions and the available species pool all
influence species diversity and ecosystem properties (Leps
ˇ
2004). As such, observational studies cannot provide the
same insights as biodiversity experiments, where either
diversity or ecosystem properties are experimentally mani-
pulated (Pfisterer et al. 2004; Schmid & Hector 2004). For
example, clear causal relationships between species richness
and ecosystem productivity can be examined only with
experimental approaches, keeping environmental conditions
constant among treatments (within-site comparisons). Such
experiments can best be done in ecosystems that allow easy
manipulation of species richness and rapid measurement of
productivity, such as perennial grassland communities
(Loreau et al. 2002). Experiments with this system have
often, but not always found a positive, asymptotic relation-
ship between plant species richness on the x-axis and
aboveground plant biomass production on the y-axis (e.g.
Tilman et al. 1997; Hector et al. 1999; further references in
Schmid et al. 2002b). Initial disputes about the possible
explanations for the positive relationship have been
addressed with the newly developed additive partitioning
analysis method that allows separating a net biodiversity
effect into contributions of complementarity and selection
effects (Loreau & Hector 2001; Hector et al. 2002a).
Complementarity occurs if performance of species in
mixture is on average higher than expected from their
monoculture yields, while the selection effect explains
higher productivity of mixtures by the dominance of
individual, highly productive species.
Despite the advances in experimental design and statis-
tical analysis (Schmid et al. 2002a), major questions remain
about the ability to extrapolate the relationship between
plant species richness and biomass production across time,
space, environments, species pools and other factors.
Analyses of data from the Cedar Creek experiment in
Ecology Letters, (2005) 8: 419–429 doi: 10.1111/j.1461-0248.2005.00736.x
2005 Blackwell Publishing Ltd/CNRS
Minnesota and theoretical considerations suggest that
complementarity effects increase over time whereas positive
selection effects decrease (Tilman et al. 2001; Pacala &
Tilman 2002). The BIODEPTH multi-site experiment has
shown that diversity effects on biomass production can vary
to some extent across localities, but a general pattern was
found if all sites were analysed together (Hector et al.
2002b). Several studies found that increasing CO
2
or
nutrient levels can accentuate diversity effects (Stocker et al.
1999; Reich et al. 2001; Fridley 2002, 2003; He et al. 2002).
Experimental studies often simulate random species loss,
but in natural communities, rare and uncommon species are
subject to higher risk of extinction because of smaller
population sizes (MacArthur & Wilson 1967; Pimm et al.
1988). Recent results of removal experiments reducing
species number in a non-random fashion provided evidence
that dominant species can sometimes control ecosystem
functioning (Smith & Knapp 2003; Zavaleta & Hulvey
2004). It may be that the drop in productivity with random
species loss, observed in experiments, would be less strong
if subordinate species were not included in the species pool
from the beginning (equivalent to subordinate species going
extinct first). If selection effects in such a pool of only
dominant species are the major cause for the relationship,
then productivity would be expected to increase less with
species richness in this case, because averaging only across
highly productive species in monocultures or low-diversity
mixtures will not yield the low values expected with
averaging across monocultures or low-diversity mixtures
of both dominant and subordinate species.
In addition to using different species pools, the spatial
scale of previous biodiversity experiments varied consider-
ably, with plot sizes from 1 to more than 100 m
2
, such that
the experimental effects might have been confounded with
scale effects (Schmid et al. 2002a). Scale effects may be
caused by abiotic factors (e.g. different intensity of exchange
processes with the surrounding environment of large vs.
small stands), biotic factors (e.g. smaller population sizes
and reduced fitness, higher or lower visitation rates of plants
by herbivores or pollinators in small plots), or increasing
environmental heterogeneity with increasing plot size and
other, presently unknown factors. In the literature these
effects are often related to an increasing amount of edge
relative to central parts in smaller plots (see e.g. Groppe
et al. 2001; Fahrig 2003 and references in these).
To examine these issues, we manipulated two factors.
First, we manipulated plot size to examine scale effects.
Second, we manipulated the relative abundance of dominant
species in the species pool to determine the extent to which
the exclusion of minor species changes the richness–
productivity relationships. The experimental set-up of this
integrated project (The Jena Experiment), carried out near
the Saale river in Jena (Germany), comprises a total of more
than 480 plots, arranged on a single large field with four
blocks (Roscher et al. 2004). The main experiment uses a
large species pool of 60 species, on either plots of
20 ·20 m or of 3.5 ·3.5 m. A second experiment
(dominance experiment) uses a species pool of nine
dominant species and plots of 3.5 ·3.5 m. These experi-
ments serve as a platform for a number of studies evaluating
biodiversity effects on ecosystem functioning.
In this paper, we present the first results of The Jena
Experiment obtained by measuring peak biomass at the end
of May in the second year of the project (2003), comparable
to the main measurements analysed in the European
BIODEPTH project (Hector et al. 1999, 2002a,b). First,
we ask if a positive relationship between plant species
richness and productivity (henceforth species richness–
productivity relationship) also applies in The Jena Experi-
ment. Second, we test if the relationship differs between
large and small plots of the main experiment. If the
relationship is independent of plot size, this may indicate
that scaling up from small-plot studies is possible, at least in
the short term. Third, we compare the relationship from the
main experiment with the large species pool with the
dominance experiment.
MATERIALS AND METHODS
Study system
Our study system is a typical Central European mesophilic
grassland community as was traditionally used for haymak-
ing (Ellenberg 1988) before the intensification of agriculture
and the shift to fast rotation, low-diversity seed mixes and
high fertilizer inputs. We selected 60 species on the basis of
their frequent occurrence in the original grassland commu-
nity on floodplains such as along the river Saale near Jena,
Germany (Roscher et al. 2004). For the main experiment
species were divided into four functional groups corres-
ponding to graminoids, legumes, tall herbs, and small herbs,
which were obtained by ordination of 17 species traits (see
Roscher et al. 2004 for details). A subset of nine species
known to become dominant in semi-natural grassland
vegetation (Roscher 1999) and expected to be highly
productive also in monocultures was selected as species
pool for the second experiment: five graminoids (Alopecurus
pratensis,Arrhenatherum elatius,Dactylis glomerata,Phleum
pratense,Poa trivialis), two tall herbs (Anthriscus sylvestris,
Geranium pratense), and two legumes (Trifolium pratense,
T. repens). Two of these species, Anthriscus sylvestris and
Geranium pratense, are known to establish more slowly
(Roberts 1979; Nikolaeva et al. 1985), which should be
considered in the interpretation of the corresponding
experiment. The selection of species for the dominance
experiment was independent of their allocation to the four
420 C. Roscher et al.
2005 Blackwell Publishing Ltd/CNRS
functional groups, because the design of the dominance
experiment focuses on effects of particular species.
Experimental design
Details of the experimental design are given in Roscher et al.
(2004) and summarized in Table 1. For the experiment with
the large species pool (¼main experiment), replications of
the same species richness level comprised all possible
numbers of functional groups. Random selection (with
replacement) of species for mixtures was subject to the
additional constraint that all functional groups are evenly
represented at each level of species richness. Species
mixtures were grown on large 20 ·20 m plots and
identically replicated on smaller 3.5 ·3.5 m plots to test
for the effect of scale. In addition, mixtures composed of the
complete pool of 60 species were established as controls,
with four replicates at both large and small plot sizes. The
experiment with the dominant-species pool (¼dominance
experiment) was established on small 3.5 ·3.5 m plots. In
the dominance experiment species richness levels were more
densely spaced from monocultures to nine-species mixture
(Table 1) and every species and every species pair was
represented with the same frequency at each particular level
of species richness. Furthermore, in the dominance experi-
ment each particular species mixture was replicated on a
second plot of the same size. We grew all species in two
single-species plots of 3.5 ·3.5 m to estimate monoculture
yields. This was necessary for analyses involving expected
relative yields in mixtures.
The field site on the floodplain of the river Saale in Jena
(Thuringia, Germany, 51N, 11E, 135 m a.s.l.) has a mean
annual air temperature of 9.3 C, and average annual
precipitation is 587 mm (Kluge & Mu
¨ller-Westermeier
2000). The experimental area was partitioned into four
blocks, following a gradient of soil characteristics due to
fluvial dynamics of the river Saale. Each block contained an
equal number of large plots on one side with diversity
treatments assigned randomly; the small plots were similarly
arranged along a strip on the other side.
The plots were sown from 11 to 16 May 2002. Seed
material was mixed with groats of soya as a bulking agent to
ensure an even distribution of seed mixtures over the whole
plot in spite of the high variability of seed sizes and shapes
among species. We used total seedling densities of
1000 seeds m
)2
. In all mixtures, species were grown at
maximum evenness. As additional experimental treatments,
sowing density and evenness were manipulated in three of
four quadrats in the small plots of the main experiment
(Roscher et al. 2004), but data are not reported here. We
thus only analyse the data obtained from the normal-density,
maximum-evenness treatment of these small plots. All plots
were weeded regularly, thus maintaining species richness at
the planned levels or slightly below in cases where a species
did not establish. The experimental communities were
mown twice in 2002.
Data collection
The first harvest in 2003 was taken at estimated maximum
biomass during 26 May–5 June, 1 year after sowing. The
plants were cut 3 cm above ground on randomly selected
sample areas of 20 ·50 cm, excluding the outer margin
(50 cm) of the plot. Two samples were harvested in the
dominance experiment and all small-area monocultures. To
account for the expected higher within-plot heterogeneity of
soil conditions in large-area plots, these were sampled with
four replicates per plot, which were combined to yield mean
biomass per plot. All samples were sorted to species. One
sample was taken in small-area plots of the main experi-
ment, and only community biomass was determined without
sorting into species because of time constraints. All samples
were dried (48 h, 70 C) to constant mass and weighed.
Data analyses
The community biomass data were analysed with general
linear models (Schmid et al. 2002a). First, ÔgeographicalÕ
variation was eliminated as a block effect. Second, we fitted
species richness as three contrasts, the first to separate
Table 1 Summary of the experimental design. The main experiment is replicated with identical species mixtures on large and small plots. The
number of plots per species richness level represents replications with different species compositions, except at the 60-species level in the
main experiment and the nine-species level in the dominance experiment where plots have identical species compositions. Plots with mixtures
of 16 and 60 species are not included in the analysis to make species richness ranges of the main and dominance experiment comparable
Experiment
Species
pool Plot size
Species richness
levels
Number of plots per
species richness level
Main experiment
Large plots 60 species 20 ·20 m 1, 2, 4, 8, 16, 60 16, 16, 16, 16, 14, 4
Small plots 60 species 3.5 ·3.5 m 1, 2, 4, 8, 16, 60 16, 16, 16, 16, 14, 4
Dominance experiment 9 species 3.5 ·3.5 m 1, 2, 3, 4, 6, 9 18, 72, 48, 36, 24, 8
Monocultures 60 species 3.5 ·3.5 m 1 120
Overyielding in experimental grassland communities 421
2005 Blackwell Publishing Ltd/CNRS
monocultures from mixtures, the second for the linear
and the third for the quadratic trend with increasing
species number. Then, we fitted dominant vs. large species
pool (¼main experiment vs. dominance experiment), the
interaction between species richness and experiment, small
vs. large plots (within the main experiment), the interaction
of the latter with species richness, the particular species
mixture, the interaction of experiment with mixture, and the
interaction of small vs. large plots with mixture (within the
main experiment). In this overall analysis, we excluded the
16-species mixtures to make the range of species-richness
levels comparable between the main and the dominance
experiment. Furthermore, we sometimes omitted the
monocultures and varied the particular treatment terms
and their sequence to test alternative models. We restricted
our comparative analysis of both experiments to species
richness effects, but individual analysis testing for either
functional group effects (main experiment) or species
identity effects (dominance experiment) is ongoing.
We used different measures to compare the yields of
mixtures relative to their component monocultures. The
additive partitioning method (Loreau & Hector 2001) was
used to calculate complementarity (CE) and selection effects
(SE), along with net biodiversity effects (NE), for both
experiments. Because average yield of monocultures enters
the calculation of the complementarity effect, this measure
of complementarity is sensitively dependent on absolute
yields and overweights the contributions of higher-yielding
species (Loreau & Hector 2001; Fridley 2003). To assess
complementarity also in relative terms, we calculated relative
yield totals (RYT, Hector 1998). The relative yield (RY) of a
species considers its biomass in mixture as a proportion of
its yield in monoculture, and the RYT of the mixture is the
sum of relative yields of all component species (Harper
1977). RYT is directly linked to Ônon-transgressiveÕover-
yielding, where a mixture outperforms the average biomass
of its component monocultures (Fridley 2001; or D
mean
¼
RYT )1 > 0, Loreau 1998). We additionally tested for
ÔtransgressiveÕoveryielding, where a mixture obtains higher
productivity per unit area than its most productive
component monoculture (D
max
> 0, Loreau 1998). These
derived measures were then themselves analysed with
general linear models, although these derived variables have
more complicated theoretical distribution functions than the
normal distribution assumed in general linear models.
RESULTS
Biomass production of individual species in monocultures
and mixtures
Based on aboveground biomass production in monocul-
tures, Onobrychis viciifolia,Bromus erectus,Leucanthemum
vulgare,Centaurea jacea and Arrhenatherum elatius were the
five most productive species in the second year of the
experiment (Fig. 1a). However, in mixtures containing the
complete pool of 60 species, Arrhenatherum elatius reached
the highest productivity and had twice the yield of the
second-most productive species Dactylis glomerata (Fig. 1b).
Species chosen for the dominance experiment had a wide
range of productivities in monoculture, but a consistently
high relative productivity in the 60-species mixture. Five
species (Arrhenatherum elatius,Dactylis glomerata,Phleum
pratense,Poa trivialis,Trifolium pratense) out of the nine
species chosen for the dominance experiment ranked
among the 10 most productive species in the 60-species
mixtures (Fig. 1b), confirming the appropriateness of their
a priori selection.
Aboveground biomass production of mixtures ranged
from 18 to 1096 g m
)2
. The relationship between biomass
production in monoculture and relative yield of a species in
mixtures differed between the two experiments. In the
dominance experiment, all species were located above or
close to the line predicting their biomass in mixture from
their yield in monoculture (Fig. 2). This pattern was
consistent across all diversity levels. Arrhenatherum elatius
showed the greatest relative increase of all species in
mixtures. Species of the main experiment with the large
species pool were expected to include both, species
performing better and species performing worse in mixtures
compared with monocultures. However, the majority of
species had a higher observed than expected relative yield in
mixtures. Even some of the less productive monoculture
species reached higher relative yields in mixtures (e.g.
Plantago lanceolata,Trifolium pratense). Among the highly
productive monoculture species a notable exception to this
finding was Bromus erectus in four- and eight-species
mixtures.
Species richness–productivity relationships
Statistical analysis of aboveground biomass indicated a
positive relationship between species richness and biomass
production in both experiments (Table 2, Fig. 3a). Alto-
gether, species richness explained 19% of the total variation
of biomass among plots. The contrast between mono-
cultures and mixtures explained 13% of the total variation
(F
1,155
¼36.61, P< 0.001). Within mixtures, the linear
(F
1,155
¼6.80, P¼0.010) and quadratic (F
1,155
¼8.73,
P¼0.004) contrasts of species richness together explained
an additional 6% (¼19–13%) of the total variation, leaving
a negligible amount to any deviations from the second-
degree polynomial. Overall, and as expected, biomass
production was higher in the dominance experiment
(F
1,155
¼26.61; P< 0.001; Table 2), species pool explain-
ing 10% of the total variation. However, contrary to
422 C. Roscher et al.
2005 Blackwell Publishing Ltd/CNRS
expectation, the difference between species pools was larger
for mixtures than for monocultures (F
1,155
¼4.26, P¼
0.041), and this interaction of monoculture vs. mixture
contrast and experiment explained a further 1.5% of the
variation. In total, average yield increase in comparison with
monocultures amounted to 82.6% in the dominance
experiment, and to 46.7% in the main experiment. The
difference in biomass production between 20 ·20 m and
3.5 ·3.5 m plots in the main experiment was extremely
small (F
1,155
¼0.01, P¼0.937). The effect of particular
monoculture species and of particular species compositions
of mixtures was large (Table 2), as observed in previous
studies, accounting for 55% of the total variation. This latter
result is not surprising when the large number of degrees of
freedom is taken into account.
Tests for overyielding and niche complementarity
Using the additive partitioning method of Loreau & Hector
(2001), we analysed the relative contributions of selection
and complementarity effects to the positive net biodiversity
effect (Fig. 3b–d). The net biodiversity effect increased
significantly with species richness (Table 3), and was
stronger in the dominance experiment than in the main
Species rank
Alo pra
Ant syl
Arr ela
Dac glo
Ger pra
Phl pra
Poa tri
Tri pra
Tri rep
Aboveground biomass > 3 cm (g m–2)
0
200
400
600
800
1000
1200
(a)
Species rank
Alo pra
Ant syl
Arr ela
Dac glo
Ger pra
Phl pra
Poa tri
Tri pra
Tri rep
Aboveground biomass > 3 cm (g m–2)
0
20
40
60
80
100
120
140
160
(b)
species in dominance experiment
Figure 1 Rank–dominance relationship of
all 60 plant species used in both experi-
ments. Values are aboveground biomass
mean (+SE). For the monocultures (a),
means were calculated from two identical
replicates in small-area plots, whereas means
of the 60-species mixture (b) were derived
from four identical replicates in the large-
area plots. Grey bars indicate species used in
the dominance experiment. Species abbrevi-
ations are: Alo pra, Alopecurus pratensis; Ant
syl, Anthriscus sylvestris; Arr ela, Arrhenatherum
elatius; Dac glo, Dactylis glomerata; Ger pra,
Geranium pratense; Phl pra, Phleum pratense;
Poa tri, Poa trivialis; Tri pra, Trifolium pratense;
Tri rep, Trifolium repens.
Overyielding in experimental grassland communities 423
2005 Blackwell Publishing Ltd/CNRS
experiment. The complementarity effect was positive across
the whole range of species richness levels (test for overall
mean „0: F
1,132
¼182.64, P< 0.001) but curvilinear
(linear contrast F
1,132
¼2.78, P¼0.098; quadratic contrast
F
1,132
¼7.01, P¼0.009), reaching a maximum at the
richness level of four species in the main experiment and six
species in the dominance experiment (Fig. 3c). The selection
effect was also positive across the whole range of species
richness levels (test for overall mean „0: F
1,132
¼172.06,
P< 0.001) and increased linearly (F
1,132
¼10.54, P¼
0.001; Fig. 3d). Selection and complementarity effects were
significantly larger in the dominance experiment. Further-
more, a large amount of variation in selection and
complementarity effects was due to differences between
particular monoculture species and particular species com-
positions of mixtures.
The comparison of mixtures with the most productive
component monoculture indicated transgressive overyield-
ing for 67.6% of the plots in the dominance experiment,
and a smaller (F
1,132
¼15.53, P< 0.001) proportion of
23.4% of the plots in the main experiment. The mean
values of RYT were >1 in both experiments, supporting our
findings of significant complementarity effects (Fig. 3e). In
both experiments, transgressive overyielding linearly de-
clined with species richness (Fig. 3f, Table 4). In the
dominance experiment, 85.6% (161 of 188 plots) and
in the main experiment with the large species pool 72.9%
(35 of 48 plots) of all plots showed RYT > 1, indicating
4-species mixtures
Aboveground biomass in mixture
0
200
400
600
800
1000
1200
1400
2-species mixtures
Aboveground biomass in mixture
> 3 cm (g m–2)
8-species mixtures
Aboveground biomass in monoculture
> 3 cm (
g
m–2)> 3 cm (
g
m–2)
0 200 400 600 800 1000 1200 1400
Aboveground biomass in mixture
2-species mixtures
4-species mixtures
9-species mixtures
Aboveground biomass in monoculture
0 200 400 600 800 1000 1200 1400
Jena, May 2003
Main experiment Dominance experiment
> 3 cm (g m–2) > 3 cm (g m–2)
0
200
400
600
800
1000
1200
1400
0
200
400
600
800
1000
1200
1400
Figure 2 Species-specific biomass in mono-
cultures and mixtures. Values are mean
(±SD), calculated from two replicates of
small-area monoculture plots, and from
different mixtures per diversity level for
the mixture plots. The line represents the
mixture biomass of species predicted from
their yield in monoculture (monoculture
biomass divided by species richness level).
424 C. Roscher et al.
2005 Blackwell Publishing Ltd/CNRS
non-transgressive overyielding. Again, values were signifi-
cantly higher in the dominance experiment (F
1,132
¼5.51,
P¼0.020).
DISCUSSION
Our analysis confirms a positive relationship between plant
species richness and biomass productivity in experiments at
different spatial scales and with different species pools.
Both, complementarity and selection effects, had signifi-
cantly positive contributions to the observed net biodiver-
sity effect.
Importance of scale
The analysis of the main experiment with a large species
pool and randomly assembled species mixtures on large-
(20.0 ·20.0 m) and small-area (3.5 ·3.5 m) plots resulted
in no significant difference of biomass production. This
indicates that results from small-scale experiments can be
scaled up and are not biased by effects caused by the small
plot size, at least in the short term. In the long run in which
multiple-generation population dynamics of the different
plant species start to play a role (sexual reproduction, seed
dispersal, seedling recruitment), additional effects could
result from smaller population sizes, reduced fitness of some
plant species and changed visitation rates of associated
animal and fungal species in small plots (see Ouborg et al.
1991; Fischer & Matthies 1998; Groppe et al. 2001 for
examples and further references). By following the popu-
lation dynamics within the different plots we will in future
be able to analyse these longer-term effects.
Most previous experiments that found a positive rela-
tionship between plant diversity and productivity were done
in very different plot sizes, ranging across four orders of
magnitude, from 0.03 m
2
(Naeem et al. 1996), 1 m
2
(van
Ruijven & Berendse 2003), 4 m
2
(Hector et al. 1999) to
169 m
2
(Tilman et al. 1997). The extension to 400 m
2
in our
main experiment still does not indicate a barrier to
extrapolation of this major result. This shows that the
caveat of inappropriate scale of biodiversity experiments for
field-scale predictions may be unwarranted and reinforces
the view that other results from small-scale experiments
should be taken seriously in developing larger-scale ecosys-
tem management applications. Additional effects may start
to play a more important role only if enlarging scale
inevitably leads to the crossing of habitat boundaries (see
Bengtsson et al. 2002).
Importance of species pool
While the species richness–productivity relationships had
similar shapes in the two experiments, one with a large pool
of 60 species and the other with a small pool of nine
potentially dominant species, the dominance experiment did
exhibit a stronger productivity response. This demonstrates
that species richness–productivity relationships can depend
on the selected species pool, in particular the inclusion or
exclusion of subdominant species. Some authors have
proposed that the positive biodiversity–productivity rela-
tionship often found in experiments such as the one
described here, can be explained by the selection effect, in
the sense that the likelihood of including dominant and
therefore productive species rises with increasing number of
species sown in a plot (see Tilman & Lehman 2002 for an
overview). If such a mechanism had been the predominant
cause for the relationships found in the current study, we
would have expected the opposite from the observed
Table 2 Summary of statistical analysis comparing aboveground biomass production in mixtures assembled from a large species pool on
large- and small-area plots or from a pool of potentially dominant species on small-area plots. Model terms were added sequentially and tested
against the species composition term
Source d.f. SS % SS MS FP
Block 3 385 202 2.03 128 401 1.89 0.133
Monoculture vs. mixture 1 2 461 559 13.00 2 461 559 36.31 < 0.001
Species richness (linear) 1 460 737 2.43 460 737 6.80 0.010
Species richness (quadratic) 1 591 755 3.12 591 755 8.73 0.004
Species pool (random vs. dominant) 1 1 803 578 9.52 1 803 578 26.61 < 0.001
Species pool ·monoculture vs. mixture 1 288 569 1.52 288 569 4.26 0.041
Species pool ·species richness (linear) 1 5651 0.03 5651 0.08 0.773
Species pool ·species richness (quadratic) 1 82 887 0.44 82 887 1.22 0.271
Plot size 1 423 0.00 423 0.01 0.937
Plot size ·species richness (linear) 1 6614 0.03 6614 0.10 0.755
Plot size ·species richness (quadratic) 1 22 896 0.12 22 896 0.34 0.562
Species composition 155 10 507 273 55.47 67 789 4.78 < 0.001
Residual 164 2 324 862 12.27 14 176
Overyielding in experimental grassland communities 425
2005 Blackwell Publishing Ltd/CNRS
Species richness
2468
2468
2468 2468
2468
2468
0 200 400 600 800 1000 1200
Species richness
–200 0 200 400 600 800
Net effect (g m–2,
corrected for block)
Species richness
–200 0 200 400 600 800
Selection effect (g m–2,
corrected for block)
Species richness
–200 0 200 400 600 800
Species richness
0.0 0.5 1.0 1.5 2.0 2.5
Species richness
Dmax (corrected for block)
–1.0 –0.5 0.0 0.5 1.0 1.5
(a)
(c)
(e)
(b)
(d)
(f)
Figure 3 Aboveground productivity (a), net biodiversity effect (b), complementarity effect (c), selection effect (d), relative yield total (e), and
transgressive overyielding D
max
(f), as functions of sown species richness. Note that overyielding analyses (b–f) were restricted to the
dominance experiment and the large plots of the main experiment. Symbols are aboveground biomasses for individual plots: () main
experiment on large-area plots (a–f); (•) identical replicates of the main experiment on small-area plots (a); ( ) dominance experiment (a–f).
Lines show predicted values from the regression model: (solid line) main experiment on large-area plots; (dashed line) main experiment on
small-area plots; (dash-dot line) dominance experiment.
426 C. Roscher et al.
2005 Blackwell Publishing Ltd/CNRS
results, i.e. a less pronounced increase, starting at higher
values in the dominance experiment. This is because
averaging across highly productive species in monocultures
or low-diversity mixtures will produce much higher values
than averaging across monocultures of both dominant and
subdominant species. An explanation for the stronger
biodiversity effect could be a greater degree of niche
complementarity among dominant species (Fargione et al.
2003) than among subdominant ones. Subdominant species
may be able to coexist due to competition/colonization
trade-offs (Levine & Rees 2002) or due to stochastic
processes (Hubbell 2001); these are mechanisms that are less
likely than niche complementarity to result in increased
productivity of high-diversity mixtures.
Complementary resource use and facilitation (combined
under the term complementarity effects in the additive
partitioning method of Loreau & Hector 2001) are often
considered as primary mechanisms behind overyielding
accompanied by varying contributions of selection effects
ranging from predominantly negative (e.g. van Ruijven &
Berendse 2003; Hooper & Dukes 2004) to positive (e.g.
Dimitrakopoulos & Schmid 2004) or being of minor
importance (Loreau & Hector 2001). In this study, we
found a positive net biodiversity effect comprising positive
selection and complementarity effects in both experiments.
In the dominance experiment, however, these measures
were on average higher than in the main experiment with
the large species pool, and the majority of mixtures
overyielded both transgressively and non-transgressively in
the dominance experiment. One reason for this can be seen
by simple visual comparison of monoculture vs. mixture
biomasses (Fig. 2). Even considering the delayed establish-
ment of two species (Anthriscus sylvestris,Geranium pratense)it
remains obvious that the dominant species, especially
Table 3 Summary of statistical analysis comparing net biodiversity effect (NE), selection effect (SE) and complementarity effect (CE)
calculated for aboveground biomass production in mixtures assembled from a large species pool on large-area plots or from a pool of
potentially dominant species on small-area plots. Model terms were added sequentially and tested against the species composition term
Source d.f.
NE SE CE
MS % SS FP MS % SS FPMS % SS FP
Block 3 56 353 2.43 2.04 0.111 10 116 0.69 0.39 0.758 75 153 5.96 4.21 0.007
Species richness (linear) 1 552 746 7.95 20.02 < 0.001 270 998 6.15 10.54 0.001 49 682 1.31 2.78 0.098
Species richness (quadratic) 1 279 232 4.01 10.12 0.002 30 452 0.69 1.18 0.279 125 259 3.31 7.01 0.009
Species pool (random vs.
dominant)
1 718 515 10.33 26.03 < 0.001 220 343 5.00 8.57 0.004 143 070 3.78 8.01 0.005
Species pool ·species
richness (linear)
1 16 572 0.24 0.60 0.440 3751 0.09 0.15 0.703 4555 0.12 0.25 0.614
Species pool ·species
richness (quadratic)
1 10 504 0.15 0.38 0.538 9176 0.21 0.36 0.551 39 315 1.04 2.20 0.140
Species composition 132 27 605 52.38 1.66 0.005 25 718 77.10 5.45 0.000 17 865 62.31 2.00 < 0.001
Residual 94 16 666 22.52 4716 10.07 8928 22.17
Table 4 Summary of statistical analysis comparing relative yield total (RYT) or non-transgressive overyielding (D
mean
¼RYT )1), and
transgressive overyielding (D
max
) calculated for aboveground biomass production in mixtures assembled from a large species pool on large-
area plots or from a pool of potentially dominant species on small-area plots. Model terms were added sequentially and tested against the
species composition term
Source d.f.
RYT D
max
MS % SS FP MS % SS FP
Block 3 0.492 3.53 2.12 0.101 0.108 0.99 0.67 0.570
Species richness (linear) 1 0.150 0.36 0.65 0.423 1.524 4.65 9.50 0.003
Species richness (quadratic) 1 0.818 1.96 3.53 0.063 0.000 0.00 0.00 0.964
Species pool (random vs. dominant) 1 0.948 2.27 4.08 0.045 2.492 7.60 15.53 < 0.001
Species pool ·species richness (linear) 1 0.050 0.12 0.22 0.643 0.277 0.85 1.73 0.191
Species pool ·species richness (quadratic) 1 0.190 0.45 0.82 0.368 0.013 0.04 0.08 0.776
Species composition 132 0.232 73.33 2.90 < 0.001 0.160 64.61 2.16 < 0.001
Residual 94 0.080 17.98 0.074 21.27
Overyielding in experimental grassland communities 427
2005 Blackwell Publishing Ltd/CNRS
Arrhenatherum elatius and Dactylis glomerata, often increased in
relative yields in mixture compared to monoculture,
indicating stronger intraspecific than interspecific com-
petition.
Our findings that complementarity effects reach a
maximum at lower diversity levels, whereas selection effects
increase linearly with species richness, support a hypothesis
that needs to be tested in further analyses. The prevailing
importance of reduced intraspecific competition, which
increases the likelihood of complementarity, seems to reach
a limit beyond which further species additions do not
increase the total niche space taken up by the community
(Dimitrakopoulos & Schmid 2004), but rather lead to
suppression of some species by others as measured by the
selection effect. Furthermore, adding more and more
species reduces the proportional densities of all species,
including the potentially high-yielding ones, perhaps to a
level where some of them require considerable time to
establish dominance.
To summarize, with our experimental approach we found
a positive within-site relationship between plant species
richness and aboveground biomass production. This rela-
tionship was very robust, independent of spatial scales or
species pools. In addition, in our experimental temperate
grasslands, the complementarity effect seems to operate
most strongly between dominant species and at low species
richness, where it is the prominent driver for the observed
increase in ecosystem functioning with increasing plant
diversity.
ACKNOWLEDGEMENTS
We thank S. Naeem, D.U. Hooper and two anonymous
referees for critical comments that helped to improve the
manuscript. The Jena Experiment is funded by the
Deutsche Forschungsgemeinschaft (DFG, FOR 456), with
additional support from the Friedrich Schiller University of
Jena and the Max Planck Society. We are grateful to the
many people who helped with the management of
the experiment, in particular the gardeners S. Eismann,
S. Junghans, B. Lenk, H. Scheffler and U. Wehmeier, and
many student helpers, especially M. Ba
¨rwolff, J. Janec
ˇek,
E. Machalett, N. Mitschunas, C. Mo
¨ller, A. Rinck, F. Walther
and K. Wu
¨rfel, assisting in the biomass harvests. Thanks
also to all the helpers during the weeding campaigns.
REFERENCES
Bengtsson, J., Engelhardt, K., Giller, P., Hobbie, S., Lawrence, D.,
Levine, J. et al. (2002). SlippinÕand slidinÕbetween the scales: the
scaling components of biodiversity–ecosystem functioning
relations. In: Biodiversity and Ecosystem Functioning: Synthesis and
Perspectives (ed. Loreau, M., Naeem, S. & Inchausti, P.). Oxford
University Press, Oxford, pp. 209–220.
Chapin, F.S., III, Sala, O.E., Burke, I.C., Grime, J.P., Hooper,
D.U., Lauenroth, W.K. et al. (1998). Ecosystem consequences of
changing biodiversity. BioScience, 48, 45–52.
Dimitrakopoulos, P.G. & Schmid, B. (2004). Biodiversity effects
increase linearly with biotope space. Ecol. Lett., 7, 574–583.
Ellenberg, H. (1988). Vegetation Ecology of Central Europe. Cambridge
University Press, Cambridge.
Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity.
Ann. Rev. Ecol. Evol. Syst., 34, 487–515.
Fargione, J., Brown, C.S. & Tilman, D. (2003). Community
assembly and invasion: an experimental test of neutral versus
niche processes. Proc. Natl. Acad. Sci. USA, 100, 8916–8920.
Fischer, M. & Matthies, D. (1998). Effects of population size on
performance in the rare plant Gentianella germanica.J. Ecol., 86,
195–204.
Fridley, J.D. (2001). The influence of species diversity on ecosys-
tem productivity: how, where, and why? Oikos, 93, 514–526.
Fridley, J.D. (2002). Resource availability dominates and alters the
relationship between species diversity and ecosystem produc-
tivity in experimental plant communities. Oecologia, 132, 271–277.
Fridley, J.D. (2003). Diversity effects on production in different
light and fertility environments: an experiment with communi-
ties of annual plants. J. Ecol., 91, 396–406.
Groppe, K., Steinger, T., Schmid, B., Baur, B. & Boller, T. (2001).
Effects of habitat fragmentation on choke disease (Epichloe
¨
bromicola) in the grass Bromus erectus.J. Ecol., 89, 247–255.
Harper, J.L. (1977). The Population Biology of Plants. Academic Press,
London.
He, J.-S., Bazzaz, F.A. & Schmid, B. (2002). Interactive effects of
diversity, nutrients and elevated CO
2
on experimental plant
communities. Oikos, 97, 337–348.
Hector, A. (1998). The effect of diversity on productivity: detecting
the role of species complementarity. Oikos, 82, 597–599.
Hector, A., Schmid, B., Beierkuhnlein, C., Caldeira, M.C., Diemer,
M., Dimitrakopoulos, P.G. et al. (1999). Plant diversity and
productivity experiments in European grasslands. Science, 286,
1123–1127.
Hector, A., Bazeley-White, E., Loreau, M., Otway, S. & Schmid, B.
(2002a). Overyielding in grassland communities: testing the
sampling effect hypothesis with replicated biodiversity experi-
ments. Ecol. Lett., 5, 502–511.
Hector, A., Loreau, M., Schmid, B. & The BIODEPTH Project
(2002b). Biodiversity manipulation experiments: studies
replicated at multiple sites. In: Biodiversity and Ecosystem Function-
ing: Synthesis and Perspectives (ed. Loreau, M., Naeem, S. &
Inchausti, P.). Oxford University Press, Oxford, pp. 36–46.
Hooper, D.U. & Dukes, J.S. (2004). Overyielding among plant
functional groups in a long-term experiment. Ecol. Lett., 7, 95–
105.
Hubbell, S.P. (2001). The Unified Neutral Theory of Biodiversity and
Biogeography. Princeton University Press, Princeton.
Kluge, G. & Mu
¨ller-Westermeier, G. (2000). Das Klima ausge-
wa
¨hlter Orte der Bundesrepublik Deutschland: Jena. Berichte des
Deutschen Wetterdienstes, 213–290.
Leps
ˇ, J. (2004). What do the biodiversity experiments tell us about
consequences of plant species loss in the real world? Basic Appl.
Ecol., 5, 529–534.
428 C. Roscher et al.
2005 Blackwell Publishing Ltd/CNRS
Levine, J.M. & Rees, M. (2002). Coexistence and relative abun-
dance in annual plant assemblages: the roles of competition and
colonization. Am. Nat., 160, 452–467.
Loreau, M. (1998). Separating sampling and other effects in bio-
diversity experiments. Oikos, 82, 600–602.
Loreau, M. & Hector, A. (2001). Partitioning selection and com-
plementarity in biodiversity experiments. Nature, 412, 72–76
(Erratum 413, 548).
Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J.P.,
Hector, A. et al. (2001). Biodiversity and ecosystem functioning:
current knowledge and future challenges. Science, 294, 804–808.
Loreau, M., Naeem, S. & Inchausti, P. (2002). Biodiversity and Eco-
system Functioning: Synthesis and Perspectives. Oxford University
Press, Oxford.
MacArthur, R.H. & Wilson, E.O. (1967). The Theory of Island Bio-
geography. Princeton University Press, Princeton.
Naeem, S., Ha˚kansson, K., Lawton, J.H., Crawley, M.J. &
Thompson, L.J. (1996). Biodiversity and plant productivity in a
model assemblage of plant species. Oikos, 76, 259–264.
Nikolaeva, M.G., Razumova, M.V. & Gladkova, V.N. (1985).
Spravocnik po prorascivaniju pokojascich semjan. Izd. Nauka, Lenin-
grad.
Ouborg, N.J., van Treuren, R. & van Damme, J.M.M. (1991). The
significance of genetic erosion in the process of extinction. II.
Morphological variation and fitness components in populations
of varying size of Salvia pratensis L. and Scabiosa columbaria L.
Oecologia, 86, 359–367.
Pacala, S. & Tilman, D. (2002). The transition from sampling to
complementarity. In: Functional Consequences of Biodiversity:
Experimental Progress and Theoretical Extensions (ed. Kinzig, A.,
Tilman, D. & Pacala, S.). Princeton University Press, Princeton,
pp. 151–166.
Pfisterer, A.B., Joshi, J., Schmid, B. & Fischer, M. (2004). Rapid
decay of diversity–productivity relationships after invasion in
experimental plant communities. Basic Appl. Ecol., 5, 5–14.
Pimm, S.L., Jones, H.L. & Diamond, J. (1988). On the risk of
extinction. Am. Nat., 132, 734–737.
Reich, P.B., Knops, J., Tilman, D., Craine, J., Ellsworth, D.,
Tjoelker, M. et al. (2001). Plant diversity enhances ecosystem
responses to elevated CO
2
and nitrogen deposition. Nature, 410,
809–810.
Roberts, H.A. (1979). Periodicity of seedling emergence and seed
survival in some Umbelliferae. J. Appl. Ecol., 16, 195–201.
Roscher, C. (1999). Die Graslandvegetation des Land-
schaftsschutzgebietes ÔMittleres IlmtalÕbei Weimar (Thu
¨ringen).
Gleditschia, 27, 57–77.
Roscher, C., Schumacher, J., Baade, J., Wilcke, W., Gleixner, G.,
Weisser, W.W. et al. (2004). The role of biodiversity for element
cycling and trophic interactions: an experimental approach in a
grassland community. Basic Appl. Ecol., 5, 107–121.
van Ruijven, J. & Berendse, F. (2003). Positive effects of plant
species diversity on productivity in the absence of legumes. Ecol.
Lett., 6, 170–175.
Schmid, B. & Hector, A. (2004). The value of biodiversity
experiments. Basic Appl. Ecol., 5, 535–542.
Schmid, B., Hector, A., Huston, M., Inchausti, P., Nijs, I., Leadley,
P. et al. (2002a). The design and analysis of biodiversity experi-
ments. In: Biodiversity and Ecosystem Functioning: Synthesis and
Perspectives (ed. Loreau, M., Naeem, S. & Inchausti, P.). Oxford
University Press, Oxford, pp. 61–78.
Schmid, B., Joshi, J. & Schla
¨pfer, F. (2002b). Empirical evidence
for biodiversity–ecosystem functioning relationships. In: Func-
tional Consequences of Biodiversity: Experimental Progress and Theoretical
Extensions (ed. Kinzig, A., Tilman, D. & Pacala, P.). Princeton
University Press, Princeton, pp. 120–150.
Smith, M.D. & Knapp, A.K. (2003). Dominant species maintain
ecosystem function with non-random species loss. Ecol. Lett.,6,
509–517.
Stocker, R., Ko
¨rner, C., Schmid, B., Niklaus, P.A. & Leadley, P.W.
(1999). A field study of the effects of elevated CO
2
and plant
species diversity on ecosystem-level gas exchange in a planted
calcareous grassland. Global Change Biol., 5, 95–105.
Tilman, D. & Lehman, C. (2002). Biodiversity, composition, and
ecosystem processes: theory and concepts. In: The Functional
Consequences of Biodiversity: Empirical Progress and Theoretical Exten-
sions (ed. Kinzig, A.P., Pacala, S.W. & Tilman, D.). Princeton
University Press, Princeton, Oxford, pp. 9–41.
Tilman, D., Knops, J.M.H., Wedin, D., Reich, P., Ritchie, M. &
Siemann, E. (1997). The influence of functional diversity and
composition on ecosystem processes. Science, 277, 1300–1302.
Tilman, D., Reich, P.B., Knops, J., Wedin, D., Mielke, T. &
Lehman, C. (2001). Diversity and productivity in a long-term
grassland experiment. Science, 294, 843–845.
Zavaleta, E.S. & Hulvey, K.B. (2004). Realistic species losses
disproportionately reduce grassland resistance to biological
invaders. Nature, 306, 1175–1177.
Editor, Shahid Naeem
Manuscript received 9 November 2004
First decision made 16 December 2004
Second decision made 10 January 2005
Manuscript accepted 13 January 2005
Overyielding in experimental grassland communities 429
2005 Blackwell Publishing Ltd/CNRS