Biodiversity effects on yield and unsown species invasion in a temperate
B. E. Frankow-Lindberg1,*, C. Brophy2, R. P. Collins3and J. Connolly4
1Swedish University of Agricultural Sciences, Department of Crop Production Ecology, Box 7043, SE-750 07 Uppsala, Sweden,
2Department of Mathematics, National University of Ireland, Maynooth, Co Kildare, Ireland,3IBERS, Plas Gogerddan,
Aberystwyth SY23 3EB, Wales, UK and4UCD School of Mathematical Sciences, Ecological and Environmental Modelling
Group, University College Dublin, Dublin 4, Ireland
Received: 27 June 2008 Returned for revision: 18 September 2008 Accepted: 3 December 2008Published electronically: 24 January 2009
†Background and Aims Current agricultural practices are based on growing monocultures or binary mixtures over
large areas, with a resultant impoverishing effect on biodiversity at several trophic levels. The effects of increas-
ing the biodiversity of a sward mixture on dry matter yield and unsown species invasion were studied.
†Methods A field experiment involving four grassland species [two grasses – perennial ryegrass (Lolium
perenne) and cocksfoot (Dactylis glomerata) – and two legumes – red clover (Trifolium pratense) and white
clover (Trifolium repens)], grown in monocultures and mixtures in accordance with a simplex design, was
carried out. The legumes were included either as single varieties or as one of two broad genetic-base composites.
The experiment was harvested three times a year over three years; dry matter yield and yield of unsown species
were determined at each harvest. Yields of individual species and interactions between all species present were
estimated through a statistical modelling approach.
†Key Results Species diversity produced a strong positive yield effect that resulted in transgressive over-yielding
in the second and third years. Using broad genetic-base composites of the legumes had a small impact on yield
and species interactions. Invasion by unsown species was strongly reduced by species diversity, but species iden-
tity was also important. Cocksfoot and white clover (with the exception of one broad genetic-base composite)
reduced invasion, while red clover was the most invaded species.
†Conclusions The results show that it is possible to increase, and stabilize, the yield of a grassland crop and
reduce invasion by unsown species by increasing its species diversity.
Key words: Cocksfoot, Dactylis glomerata, diversity effect, invasion, legumes, perennial ryegrass, Lolium
perenne, red clover, Trifolium pratense, simplex design, statistical modelling, transgressive over-yielding,
white clover, T. repens.
The loss of natural ecosystems to agriculture and other human
activities is projected to have a large global impact on biodi-
versity in the future (Chapin et al., 2000; Tilman et al.,
2002). Areas under intensive agriculture, in which applications
of high doses of mineral fertilizers to monoculture crops are a
prominent feature, have a negative impact on biodiversity in
Europe (Reidsma et al., 2006). However, maintaining accepta-
ble levels of agricultural productivity will continue to be a
high priority, mainly because of an increasing global popu-
lation that needs to be fed, but also due to the need for econ-
omic sustainability of the individual farmer. Hence, there is an
urgent need to develop agricultural practices that can deliver
high yields with adequate forage quality, and simultaneously
promote biodiversity. Biodiversity encompasses a broad spec-
trum of biotic scales (Hooper et al., 2005), with components
ranging from genetic diversity within populations through to
the functional diversity of species present and up-scale to eco-
Perennial grasslands occupied around 40 % of the land used
for agriculture by the European Union prior to its recent
enlargement (Rath and Peel, 2005). In productive temperate
grassland systems grass monocultures are often used in order
to simplify management options (e.g. easier to predict the
optimal harvesting time, less variation in feeding quality).
Seed use data for the EU in the 1990s showed that by far
the most commonly sown grassland species was perennial rye-
grass (Kley, 1995) and this remains the case today. However,
large inputs of artificial nitrogen fertilizer are required to maxi-
mize the productivity of such monocultures, and this is
increasingly regarded by policy makers and public opinion
as being environmentally unacceptable. In addition to high
yields, farmers may also value a stable yield level between
years. Multi-species mixtures could provide greater and more
stable primary production through temporal, spatial and
resource niche complementarities (Sanderson et al., 2004;
van Ruijven and Berendse, 2005; Roscher et al., 2008).
Although there is considerable evidence for a strong and
persistent effect of diversity on yield in natural plant systems
(Hooper et al., 2005; Balvanera et al., 2006; Cardinale
et al., 2007; Schmid et al., 2002), few studies have explored
biodiversity effects in a realistic agronomic context (but see
Tracy and Sanderson, 2004a; Sanderson et al., 2005, Picasso
et al., 2008).Apositivediversity effectleadingto
* For correspondence. E-mail: firstname.lastname@example.org
# The Author 2009. Published by Oxford University Press on behalf of the Annals of Botany Company. All rights reserved.
For Permissions, please email: email@example.com
Annals of Botany 103: 913–921, 2009
doi:10.1093/aob/mcp008, available online at www.aob.oxfordjournals.org
transgressive over-yielding (mixture performing better than the
best monoculture) was found in the first year’s results from an
agronomic study (in which some of the data of the present
study were included) at 28 sites in Europe in which mixtures
of four species (two legumes and two grasses) were compared
with the monoculture performance of these species (Kirwan
et al., 2007). Genetic diversity might be expected to provide
additional benefits through improved resistance to biotic (e.g.
Finckh et al., 1999; Cox et al., 2004; Pilet et al., 2006) and
abiotic environmental stresses (Hajjar et al., 2008), and
enhanced niche complementarity (Hooper et al., 2005).
Further, it has been proposed that multi-species mixtures
should be more resistant to invasion by unsown species than
crops composed of fewer species, and this has been confirmed
in some studies, e.g. Hector et al. (2001), Tracy et al. (2004),
Tracy and Sanderson (2004b) Picasso et al. (2008). Although
the evidence from a meta-analysis of 44 biodiversity exper-
iments that manipulated plant species richness is that trans-
gressive over-yielding is not, in general, of wide occurrence
and requires several years to develop (Cardinale et al.,
2007), this conclusion may partly be due to methodological
difficulties inits assessment
Persistence and an increase in a positive diversity effect in
natural plant systems with time have been noted (Hooper and
Dukes, 2004; van Ruijven and Berendse, 2005; Bullock
et al., 2007; Cardinale et al., 2007; Fargione et al., 2007)
but whether the effect persists through time in perennial agro-
nomic grassland systems is of considerable interest.
In this paper results are reported from a field study compris-
ing multi-species mixtures of four grassland species [two
grassesand two legumes:
perenne) and cocksfoot (Dactylis glomerata) and red clover
(Trifolium pratense) and white clover (T. repens)] using the
approach described in Kirwan et al. (2007). The experiment
also included a treatment in which the impact of intraspecific
variation in the legume species was assessed by the use of
single varieties versus broad genetic-base composites, com-
prising mechanical mixtures of seed of many varieties
(Collins et al., 2004). It was hypothesized that: (1) positive
interactions between species would result in larger yields
from mixtures than expected from the performance of the indi-
vidual species sown as monocultures; (2) these benefits would
persist over several years; (3) a broad genetic-base composite
would provide a platform for further and sustained positive
effects of mixing species; and (4) increased species diversity
and the use of broad genetic-base composites would help to
prevent invasion of unsown species into the swards.
(Schmid etal., 2008).
MATERIALS AND METHODS
Site data and plant material
The field experiment was conducted at Svalo ¨v in southern
Sweden (568550N, 13 807 E, 55 m a.s.l.). The climate is cold-
temperate with an annual mean temperature of 7.78C and an
annual precipitation averaging 700 mm. The soil at the site
was a sandy clay loam with a pH of 6.9 and containing 2.12 %
organic matter, 1.3 g nitrogen, 0.36 cmol potassium and
44 mg phosphorus kg21soil. Plots were established in late
June 2003 by drilling, and initial invasion by unsown species
was controlled by spraying with the herbicide bentazon (which
is tolerated by the species sown) shortly after the initial emer-
gence of the crop. The species used were two grasses and two
legumes, of which one species within each functional group
was relatively fast establishing but not persistent (Lolium
perenne L. and Trifolium pratense L.), and the other was rela-
tively slow establishing and persistent (Dactylis glomerata
L. and Trifolium repens L.). Information on the plant material
is presented in Table 1. For each legume species there was one
composites, comprising ‘Broad Central European’ (BC) and
‘Broad Northern European’ (BN) populations. The single var-
ieties of species used (N) had been bred for agricultural use
tained seedof 11commercial varieties bredforusein temperate
regions,plusasmallamount (23 %byweight)ofa‘northernred
clover composite’ population from Norway/Denmark, and
encompassed all categories employed by red clover breeders
(early/late flowering; diploid/tetraploid). The red clover BN
composite contained seed from a population which was
created by the inter-crossing for two generations of 238 acces-
sions comprising the complete red clover collection held in the
Nordic Gene Bank, plus varieties from other parts of the world
(mainly from the former USSR; A´. Helgado ´ttir, Agricultural
University of Iceland, Reykjavik, pers. comm.). The white
clover BC composite contained seed of 13 commercial varieties
bred for use in temperate regions, plus a small amount (5 % by
weight) of ‘gene pool’ material from Central Asia. The white
clover BN composite contained seed from nine varieties bred for
use in northern Scandinavia, plus ten commercial varieties bred
for use in southern Scandinavia (P. Marum, Graminor AS,
Ilseng, Norway, pers.comm.).
Seeding rates of the respective species in monoculture in the
high density treatment (see below) are found in Table 1. These
corresponded to the rates used in official Swedish variety
trials. Low density plots were sown at 60 % of these rates. In
mixtures, the seeding rate of each species was a proportion
(see below) of the monoculture seeding rates.
The experimental layout consisted of 66 communities.
Thirty communities followed the simplex design described in
Kirwan et al. (2007) with two grass and two legume (N)
TABLE 1. Plant material and monoculture seeding rates
rates (kg ha21) for high
Red clover (Trifolium
White clover (Trifolium
Central European, BC
Northern European, BN
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem 914
monocultures and 11 mixtures of these four species sown at
two densities. The 11 mixtures consisted of four mixtures
dominated in turn by each species (sown proportions were
70 % of dominant and 10 % of each other species), six mix-
tures dominated in turn by pairs of species (40% of each of
the two species and 10 % of the other two) and the centroid
community (25% of each species). Eighteen communities
had four monocultures [two grasses (same as before) and
two legumes (BC)] at two densities and five mixtures of
these four species sown at two densities (sown proportions
were 70 % of the dominant and 10 % of each of the other
species, plus the centroid community). Another 18 commu-
nities followed the same design but with BC replaced by
BN. Experimental communities were randomly assigned to
8.8 m2plots. Data from one plot were omitted since an erro-
neous species mixture was sown to this plot. General infor-
mation on simplex designs is available in Cornell (2002).
No harvests were taken in the establishment year. In the fol-
imental plot harvester (Haldrup) 3 times a year, with a stubble
height of approx. 5 cm (normal practice within this climatic
zone). At each harvest biomass yield was determined and
samples for analyses of dry matter (DM) and botanical compo-
bouring plots was prevented by regular spraying of the borders
with glyphosate. No fertilizer nitrogen was added to the exper-
iment. In the final year of the experiment notes were taken of
the unsown species identity present in all plots.
Annual sown DM yield and unsown species DM % for each
year and sown DM yield for each harvest were analysed. The
modelling approach of Kirwan et al. (2007) was used to relate
sown yield (DMyield) for a single harvest or year to variables
seeding rates, overall density and species interaction. Sown
proportions were denoted in the model as G1(ryegrass) and
G2(cocksfoot) for the two grasses and L1N, L1BCand L1BN
for the N, BC and BN genetic treatments, respectively, for
the first legume species (L1, red clover). A similar notation
was used for the second legume species (L2, white clover).
DENS is a variable indicating high or low seeding density,
coded 21 for low and 1 for high. A single variable (evenness)
was used to describe interspecific interaction in a mixture. The
evenness (E) of a mixture as defined in Kirwan et al. (2007)
lies between 0 for a monoculture and 1 for a mixture in
which the sown proportions of all species are equal (the cen-
troid). They used E and E2as explanatory variables to
capture the effect of species interactions. The model fitted was
The coefficients b1and b2represent the monoculture responses
of the grass species and the coefficients b3to b8represent the
monoculture responses of the legume species, at average
density. For example, 3.84, the estimate of b1for year 1 (see
Table S1 in Supplementary data available online), estimates
the monoculture yield of perennial ryegrass for that year. The
evenness coefficients (b10, b11and b12) measure the effects of
evenness for mixtures involving each of the legume popu-
lations. b13is a quadratic effect of evenness that does not
change with legume population. Thus, for mixtures involving
N legumes, the species diversity effect (the contribution of
interaction effects to yield of a mixture at evenness ENover
what could be expected from a proportional mixture of mono-
culture yields) is b10ENþ b13EN
all species), EN¼ 1 and the predicted diversity effect is 5.62
– 2.14 ¼ 3.48 (Table 2). Finally, 1 is a random error term
assumed to be normally and independently distributed with
2estimated by 5.62EN –
2(Table S1). For the centroid (equal sown proportion of
TABLE 2. Mean estimated species diversity effects for sown dry matter yield at the centroid (t ha21) of the three legume population
treatments for each of the three harvest years with tests of significance; also shown for each year are yields (t ha21) at the centroid
predicted both from monoculture performance only, and including the species diversity effect
Harvest year 1Harvest year 2 Harvest year 3
Estimate s.e.P Estimates.e.P Estimates.e.P
Estimated species diversity effect at the centroid
Estimated yield at the centroid based on monoculture performances
BC 7.04 0.175
Estimated yield at the centroid including the species diversity effect
Comparisons among estimates within and across years are presented in the text. N denotes a legume variety, while BC and BN denote Broad Central
European and Broad Northern European genetic-base composites, respectively.
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem 915
mean of zero and constant variance. Differences were tested
between coefficients for grass species (b1vs. b2), for legume
species (b3vs. b6, b4vs. b7, and b5vs. b8) and legume popu-
lations (b3vs. b4vs. b5and b6vs. b7vs. b8), and the evenness
vs. b11vs. b12) and they were tested against zero.
Repeated measures analyses were carried out using the
MIXED procedure and univariate analyses using the GLM
Procedure in the SAS/STAT software, Version 9.1 of the
SAS System for Windows.
Repeated measures analyses of sown species biomass were
performed across harvests within each year and for annual
sown yields across the 3 years of the experiment based on
model (1) augmented with terms for interactions with time,
which were significant. An unstructured covariance matrix
was used to describe the time-dependence among harvests or
years (Verbeke and Molenbergh, 2000). The effects of year
and legume population were tested on the estimated species
diversity effect at the centroid and the average monoculture
yield across years. Transgressive over-yielding was tested for
each legume population using a non-parametric test similar
to that described in Kirwan et al. (2007).
The percentage of total annual unsown species biomass in a
repeated measures analysis across years was analysed using
model (1) and including year effects as just described.
Responses were log-transformed before analysis to reduce
heteroscedasticity. Many mixed plots were free from unsown
species in the two first harvest years (14 in each year), and the
recorded zero values from these plots were replaced by 0.0014,
the lowest recorded non-zero value, before transformation.
There was no interaction between density and the other
model variables. Model (1) (and its extensions for repeated
measures analysis across years) generally emerged as either
the dominant or the best of a number of alternative models
for overall sown yield, for most harvests and years and for
log of unsown species yield (by dominant it is meant that
the explanatory power of the evenness-based species diversity
effects greatly exceeds any refinement introduced by partition-
ing the evenness effect into more coefficients as described
below). Alternative models included a model with separate
interaction coefficients for each pairwise interaction, and a
model that included only two pairwise interaction coefficients,
one for an interaction between any grass and any legume and
(Kirwan et al., 2007). Other models allowed the species diver-
sity effect to vary with legume species. There was some evi-
dence in favour of separate pairwise coefficients for each
interaction for sown yield at some harvests and for a separate
coefficient ? year interaction for the analysis of sown annual
yield. In the latter analysis there was no clear pattern among
the six interaction coefficients across years. Mixtures were
not replicated and the variation around the regression model
was used to estimate the residual mean square and the standard
errors of coefficient estimates (Cornell, 2002).
Estimates of coefficients of model (1) fitted to annual sown
yield as a repeated measure over 3 years show that there was
a considerable and significant decline in monoculture yield
for all species across years (P, 0.05 for each test; Fig. 1,
and Table S1 in Supplementary data, available online). The
estimated DM yield of the cocksfoot monoculture was consist-
ently greater than that of perennial ryegrass (P, 0.01 in each
Dry matter yield (t ha–1)
FIG. 1. Total yield predicted for each of the four monocultures and for three mixtures divided into sown and unsown species for each year. The four mono-
cultures are Dactylis glomerata, Lolium perenne, Trifolium pratense (averaged over N, BC and BN) and Trifolium repens (averaged over N, BC and BN).
The three mixture communities are the Centroid (0.25, 0.25, 0.25, 0.25), Mixture 1 (the average of the mixtures with two species dominant, e.g. 0.4,
0.4, 0.1, 0.1.) and Mixture 2 (the average of the mixtures with one species dominant, e.g. 0.7.01 0.1 0.1). Standard error bars are given for the predicted
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem 916
harvest year; Table S1 in Supplementary data, available
online). White clover consistently outyielded the two grasses
over the 3 years (P , 0.01 for each test except L2BNvs. G2
in year 3), while red clover did so in the first 2 years (P ,
0.001 for each test). There was generally no strong or consist-
ent effect of legume populations in the monoculture perform-
ances of either legume species for any year. Red clover
outyielded white clover in the first year (P, 0.001 for N
and BC populations, P ¼ 0.06 for BN) but underyielded it in
the third year (P, 0.01 for all populations). Density never
affected DM yield.
While the cumulative yield for each year was little affected
by the broad genetic-base composite treatment, the different
legume populations nonetheless showed a different yielding
pattern within each year (see Table S2 in Supplementary
data, available online). Monocultures of the BC red clover
populations had higher estimated yield (b4vs. b3and b4vs.
b5red clover) than the N and BN red clover populations in
the final harvest of each year, significantly so (P , 0.01) in
the first 2 years. Further, monocultures of the BC white
clover populations had higher estimated yield than the BN
white clover populations in the final harvest of each year (b7
vs. b8white clover), significantly so in 2 of the 3 years. In
the other harvests differences were rarely significant.
Evidence of transgressive over-yielding was found in 3-year
accumulated yield in the mixtures using N and BC legumes
(P, 0.05; Table 3). Transgressive over-yielding was observed
in the second and the third years for mixtures using all legume
populations (P, 0.05).
There was a strong and persistent species diversity effect
irrespective of the legume population used. The linear even-
ness effects were always positive and greater in magnitude
than the quadraticevenness
Supplementary data available online), giving a positive
species diversity effect that remained strong over the 3 years
(Table 2). The linear evenness effect varied with both
legume population used (P, 0.001) and year (P, 0.001),
and the quadratic evenness effect varied with year (P,
0.001; Table S1 in Supplementary data, available online).
The maximum species diversity effect occurred at a level of
evenness at or close to 1, the centroid. The species diversity
effect was calculated at the centroid for each genetic diversity
level ? year combination and also the mean yield calculated at
the centroid predicted from the monoculture yields of species,
i.e. with no diversity effect added, and the mean yield includ-
ing the species diversity effect (Table 2). The species diversity
effect at the centroid increased between the first and second
years (P, 0.001; Table 2) and in the third year declined to
levels similar to those in the first year (P , 0.001; Table 2)
for each of them. Yield at the centroid predicted from an
equal mixture of monoculture yields declined strongly over
time, irrespective of the legume population used (P, 0.001
for each test; Table 2). Predicted yield at the centroid including
the species diversity effect also declined over time (P , 0.001
for each test; Table 2) but far less rapidly than when based
solely on monoculture performance. Thus, the species diver-
sity effect on yield relative to yield predicted from the
respective monoculture performances was much higher in
the second and third years than in the first year. Mixtures
with the BC legumes produced a generally stronger species
diversity effect than those with BN legumes (P, 0.05),
while the species diversity effect of the mixtures including
the N legume was not significantly different from that of
either of the two broad genetic-base composites (Table 2).
Invasion by unsown species
The effect of species diversity was very marked and per-
sisted through time (significant species diversity effect each
year P, 0.001), and the predicted level of unsown species
at the centroid was ,2 % even in the third year (Table 4).
The most striking monoculture effects observed were the
ability of cocksfoot and white clover (N and BC) to resist inva-
sions by unsown species (Table 4, and Table S3 in
TABLE 3. Results of tests for transgressive over-yielding
Year HarvestN BC BN
Total for 3 years
The test is carried out for each of the three populations (N, BC and BN),
for each harvest by year combination, for yield accumulated each year and
for yield accumulated over all 3 years. N denotes a legume variety, while BC
and BN denote Broad Central European and Broad Northern European
genetic-base composites, respectively. The value given in the table is the test
statistic*, d, which is the absolute difference between the number of mixture
means better than the best monoculture and worse than the worst
monoculture. Bold indicates significance of the test (P , 0.05). The test is
based on the non-parametric test described in Kirwan et al. (2007).
* Transgressive over-yielding occurs when the yield of a mixture
community exceeds that of the highest-yielding monoculture of its
component species. Here, a permutation test is used to determine the
significance of transgressive over-yielding. There were 11 (4), 5 (4) and 5 (4)
mixture communities (monoculture communities) in the three legume
populations N, BC and BN, respectively. The average across replicates for
each monoculture and mixture community was calculated within each
legume population except for grass monocultures which were averaged across
the legume populations (there was no difference between the grass
monocultures for each legume population), giving 15, 9 and 9 community
values for N, BC and BN respectively. The test statistic, d, is calculated from
these community values as the absolute difference between the number of
mixtures greater than the best monoculture and lower than the worst
monoculture. Significance was tested using a permutation test that compares
the result against that which would be expected by selecting 11 (or 5) values
at random from 15 (or 9) and comparing them with the best and worst of the
remaining 4 (or 4) values. The test is a two-sided test as the mixtures could
be better than the best monoculture or worse than the worst monoculture
(under-yielding) and the direction cannot be prejudged in advance. For more
details see supplementary material in Kirwan et al. (2007).
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem 917
Supplementary data available online). Even in the third year
levels of unsown species were ,6% in these monocultures,
whereas they increased over years to high levels for monocul-
tures of both ryegrass (41 %) and red clover (around 60 %).
Using broad genetic-base composites appeared to have little
effect on the amount of unsown species in red clover swards,
but the BN white clover population contained a larger amount
than the N or the BC populations (P, 0.01; Table 4).
The present study differs from many earlier biodiversity
studies (e.g. Hector et al., 1999; Tilman et al., 2001;
Roscher et al., 2003) with respect to the species used (all of
major agronomic importance in temperate regions) and the
management carried out (aimed at maximizing biomass har-
vested; see Hooper et al., 2005). The design of the experiment
also differed from the above-mentioned studies in establishing
a gradient of evenness with all species present in all mixed
communities, instead of just increasing species richness as a
With respect to the four main hypotheses addressed in this
study it was found that species diversity had a strong, persist-
ent and positive effect on yield and on the ability of the sward
to resist invasion by unsown species. However, using a broad
genetic-base composite in two of the four species in the exper-
iment had little impact on these ecosystem properties.
Diversity effects on yield
It was hypothesized that the mixing of species would result
in larger yields from the mixtures than would be expected from
the individual species monoculture yields. That mixing of
grasses with legumes, with a resulting niche complementarity
for N (Palmborg et al., 2005; Kahmen et al., 2006), results in a
positive yield effect has been shown in other studies (e.g.
Spehn et al., 2002). However, when a model that partitioned
the diversity effect into a term representing legume ? grass
interaction was fitted, and a term jointly representing grass ?
grass and legume ? legume interactions (Kirwan et al.,
2007) the two coefficients did not differ significantly,
suggesting that the evenness model was a more parsimonious
description of these data. This suggests that the effects of
mixing grasses or mixing legumes had the same impact on
increasing yield as did functional group mixing between
grasses and legumes. Similar results have been reported by
van Ruijven and Berendse (2003, 2005) and Hector et al.
(2007). Thus, it appears that the characteristics of the species
within each functional group were either of a complementary
or a facilitating nature (Hooper et al., 2005). The species diver-
sity effect based on between-species interactions was always
positive. In the centroid mixture the average contribution to
yield of the species diversity effect were 3.55, 4.89 and 3.77
tons DM ha21(þ33 %, þ55 % and þ65 % yield increase
over mean monoculture performances) in the first, second
and third years respectively. This diversity effect is stronger
than the mean effect observed for the first harvest year
across 28 European sites (Kirwan et al., 2007).
The sustained species diversity effect and the declining
monoculture effects suggest that the importance of this diver-
sity contribution to yield increased with time. This was
reflected in the increasing evidence of transgressive over-
yielding for all legume populations as time advanced.
Reports of transgressive over-yielding in biodiversity exper-
iments carried out in the field with grassland species have
been rare (e.g. Hooper and Dukes, 2004; Cardinale et al.,
2007; Lanta and Leps ˇ, 2007; Schmid et al., 2008; but see
Schmid et al., 2002). It is argued (Schmid et al., 2008) that
a wide range in monoculture yields makes it difficult to
studies where transgressive over-yielding has been observed,
the species involved have been high-yielding (Jolliffe and
Wanjau, 1999; Roscher et al., 2005), the soil has been
nutrient-rich (Roscher et al., 2008) and species richness has
been low (Roscher et al., 2005), which was also the case in
the present experiment. This suggests that transgressive over-
yielding should be more common in an agricultural context,
where soil fertility usually is good and the species used are
selected on the basis of their yielding properties, than it is in
natural communities in which low-yielding species frequently
occur. Indeed, in an analysis of 44 plant experiments (39 grass-
land) with species number ranging from 6 to 32 (Cardinale
et al., 2007) the yield of the highest yielding monoculture
was 93 % above the yield of the average monoculture and
transgressive over-yielding was only observed in 12 % of
cases. However, in the present data, the monoculture variation
was lower than this but still of a substantial magnitude (the
highest yielding monoculture exceeded the average monocul-
ture by 51, 51 and 77 % in harvest years 1, 2 and 3, respect-
ively). So, the emergence of transgressive over-yielding over
time here is not due to a relatively small species diversity
effect being compared with rather similar yielding species in
monoculture, but rather to the decline in monoculture yields
over time and the persistence of a relatively large species
The persistent and increasing effect of species diversity on
yield has important practical implications. In farming practice,
TABLE 4. Estimated biomass percentage of unsown species in
the respective monocultures and at the centroid for each of three
harvest years (HY)
Estimated weed content (%)
ParameterHY 1HY 2 HY 3
b3(L1N– red clover)
b4(L1BC– red clover)
b5(L1BN– red clover)
b6(L2N– white clover)
b7(L2BC– white clover)
b8(L2BN– white clover)
N denotes legume varieties, while BC and BN denote Broad Central
European and Broad Northern European genetic-base composites,
respectively. The centroid is the mixture sown with equal proportions of the
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem918
the application of fertilizer nitrogen by European grassland
farmers is a common way to increase and/or maintain yield
over time. Given the need to reduce the load of nitrogen into
the European environment (e.g. Rougoor and van der
Weijden, 2001) the practical benefit of this diversity effect is
obvious, and the use of mixtures could therefore be rec-
ommended as a way of reducing inputs. It should be noted
that no fertilizer nitrogen was added to this experiment.
Using broad genetic-base composite populations of the two
legumes had a relatively small impact on yield in the current
experiment, contrary to hypothesis (3). This could be due to
the fact that traits are more likely to be similar within
species than between different species. Similar results have
been obtained in studies with white clover by Annicchiarico
and Piano (1997) and Williams et al. (2003). Legumes, and
red clover in particular, are susceptible to a vast range of
pathogens that destroy sensitive genotypes (Frame et al.,
1998), and the loss of red clover plants due to diseases was
very high in this experiment. Although there are no data on
disease susceptibility of the red clover populations or pathogen
load in the plots, the fact that yields of all red clover popu-
lations were similar in the final year suggest no difference in
susceptibility to pest and diseases among them. Resistance
of perennial plant species to abiotic stress during winter is
often gained at the expense of late-season growth (e.g. white
clover; Eagles and Othman, 1989). However, in the environ-
ment of southern Sweden the observed differences in
end-of-season growth pattern of the legumes did not appear
to have any impact on the overwintering abilities of the popu-
lations (as judged by monoculture spring yields). Attempts in
the USA to improve the yield of a lucerne (Medicago sativa)
sward by mixing varieties with a different seasonal growth
pattern were short-lived, and this strategy was not rec-
ommended as a way to improve yield (Brummer et al.,
2002). Another explanation for the weak effect of the broad
genetic-base composites might be the occurrence of selection
of the most adapted individuals for the site and the manage-
ment over time (e.g. Frankow-Lindberg, 1999, 2001). Joshi
et al. (2001) found that red clover ecotypes performed best
in their original sites compared with more distant sites, and
their performance was poorer with increased transplanting dis-
tance. The broad genetic-base composites might therefore not
have been well adapted to Swedish conditions. Hypothesis (3)
was thus not confirmed, but it is possible that the use of a
adapted to southern Sweden might have resulted in a more
effect withinthe plant
Diversity effects on unsown species invasion
The species diversity effect on unsown species was very
strong, confirming hypothesis (4). The nature of the relation-
ship between species diversity and resistance to invasion by
unsown species is a hotly debated topic among ecologists
(e.g. Fridley et al., 2007). In this experiment, where no
weeding was carried out, both species diversity and species
identity were important in controlling invasion. The main
unsown species recorded was Taraxacum officinale, which is
a wind-dispersed species that does not form a persistent soil
seed bank. This shows that the soil seed bank was a factor
of little importance for invasion by unsown species at this
site, and that any effect of the herbicide application at the
very beginning of the experiment had little (if any) carry-over
effect on this ecosystem property.
An increased invasion resistance in species-diverse grass-
land communities compared with less diverse communities
has been found in studies where initially unsown species
have been deliberately introduced (Fargione et al., 2003),
where unsown species have been removed over time (Knops
et al., 1999; Hector et al., 2001; Tracy and Sanderson,
2004b), when ingress by unsown species in the harvested
biomass was determined (Picasso et al., 2008), and in
surveys of pastures across a wide geographic area (Tracy and
Sanderson, 2004b). Similar results have also been obtained
for annual crops (Hauggaard-Nielsen et al., 2008). However,
plant functional diversity effects on invasion resistance have
not been extensively studied. In this context, it is of interest
to note that the most common of the unsown species observed
in the experiment belonged to a functional type (deep-rooted
forb) that was not present in the seeding mixtures (Fargione
et al., 2003; Hooper et al., 2005). Species with competitive
traits may, however, play a key role (see below). The causal
mechanism for the increased invasion resistance by species-
diverse communities is likely to be a more complete utilization
of environmental resources (Hooper et al., 2005). Differences
in temporal growth patterns between the grasses and legumes
currently used is, for example, a generally accepted agronomic
fact (see Table S2 in Supplementary data available online).
Of the monocultures, the cocksfoot and two of the white
clover (N and BC) communities were least susceptible to inva-
sion, while all red clover monocultures were the most invaded.
The reason why the BN white clover monoculture did not
conform to the pattern might be that it had a lower growth
rate at the end of the growing season, leaving more ‘space’
for unsown species to enter the sward at this time of the
year. Since an efficient herbicide treatment was applied to
this experiment shortly after sowing, the ingress of unsown
species reflected characteristics of the mature community.
Mwangi et al. (2007) found that grasses exerted a strong nega-
tive effect on invaders, whilst legumes had a positive effect,
which to some extent agrees with the results from this study,
since red clover resisted invasion poorly. Crawley et al.
(1999), reporting a 7-year experiment with four levels of
species richness where unsown species were allowed to
accumulate in the plots, concluded that species identity mat-
tered more than species richness in determining both the
number of invading species and the total biomass of the inva-
ders. Legumes play a vital role in the nitrogen economy of
grassland communities (Spehn et al., 2002; Temperton et al.,
2007), and it has been found that community invasibility is
positively correlated with nitrate availability in the soil
(Dukes, 2001). However, since all white clover monocultures
contained substantially lower amounts of unsown species
than the red clover monocultures, this cannot be the only
mechanism that facilitated invasion of unsown species into
the red clover plots. The fact that red clover plants died
throughout the experimental period, leaving large gaps that
could provide entry points for invading species, was probably
just as important (see, for example, Milbau et al., 2003, 2005).
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem919
One trait common to cocksfoot and white clover is the ability
to spread laterally, cocksfoot both above- (Lorentzen et al.,
2008) and below ground (Personeni and Loiseau, 2004), and
white clover above ground (Frame et al., 1998). This trait
would result in a more complete resource utilization, thus pre-
venting the establishment of unsown species. The present
monoculture results (Table 4) are in accordance with the sug-
gestion made by Richardson and Pys ˇek (2006) that, in future,
more emphasis should be placed on species identities and their
characteristics in the search for an increased understanding of
community invasibility, but in this study the species mixture
results showed an even stronger effect.
It is concluded that increased species diversity within a
forage crop provided clear and persistent yield benefits. This
biodiversity effect could be profitably exploited by grassland
farmers, and has the potential to reduce requirements for ferti-
lizer inputs. Using broad genetic-base composites of the
legumes was of minor importance in terms of yield at this
site, although the use of broad genetic-base populations con-
taining site-adapted germplasm may have produced a different
result. Invasion by unsown species was generally reduced by
increased species diversity of the sward. However, the strong
effect of individual species identity suggests that traits confer-
ring a more complete resource utilization also contributed to
Supplementary data are available online at www.aob.oxford-
journals.org and consist of the following tables. Table S1: esti-
mates of model coefficients for sown dry matter yield (t ha21)
for each of three harvest years. Table S2: estimates of coeffi-
cients for sown dry matter yield (t ha21) for each of the
three harvests within each of the three harvest years. Table
S3: estimatesof model coefficients
species %) for each of the three harvest years.
of log (unsown
We thank SW Seed for providing access to the field site and
excellent technical field assistance, and Mrs Annika Hansson
for doing the botanical analyses. Thanks to An Gehesquiere,
Vibeke Meyer, Petter Marum and Loek van Soest for provid-
ing seed for the broad genetic-base clover populations. This
work was supported by a grant from The Swedish Research
Council for Environment, Agricultural Sciences and Spatial
Planning (22.9/2003-0823) and the EU Commission through
COST Action 852. The Faculty of Natural Resources and
Agricultural Sciences at Swedish University of Agricultural
Sciences provided support to B.E.F.-L. for a stay at
University College Dublin.
Annicchiarico P, Piano E. 1997. Response of white clover genotypes to inter-
genotypic and interspecific interference. Journal of Agricultural Science
Balvanera P, Pfisterer AB, Buchmann N, et al. 2006. Quantifying the evi-
dence for biodiversity effects on ecosystem functioning and services.
Ecology Letters 9: 1146–1156.
Brummer EC, Moore K, Bjork NC. 2002. Agronomic consequences of
dormant-nondormant alfalfa mixtures. Agronomy Journal 94: 782–785.
Bullock JM, Pywell RF, Burke MJW, Walker KJ. 2007. Restoration of
biodiversity enhances agricultural production. Ecology Letters 4:
Cardinale BJ, Wright JP, Cadotte MW, et al. 2007. Impacts of plant diver-
sity on biomass production increase through time because of species com-
plementarity. Proceedings of the National Academy of Sciences of the
USA 104: 18123–18128.
Chapin III FS, Zavaleta ES, Eviner VT, et al. 2000. Consequences of chan-
ging biodiversity. Nature 405: 234–242.
Collins RP, Connolly J, Porqueddu C. 2004. Effects of legume genetic diver-
sity on the productivity of legume/grass mixtures – COST Action 852.
Grassland Science in Europe 9: 486–488.
Cornell JA. 2002. Experiments with mixtures: designs, models and the analy-
sis of mixture data, 3rd edn. New York, NY: John Wiley and Sons.
Cox CM, Garrett KA, Bowden RL, Fritz AK, Dendy SP. 2004. Cultivar
mixtures for the simultaneous management of multiple diseases: tan
spot band leaf rust of wheat. Phytopathology 94: 961–969.
Crawley MJ, Brown SL, Heard MS, Edwards GR. 1999. Invasion-
resistance in experimental grassland communities: species richness or
species identity? Ecology Letters 2: 140–148.
Dukes JS. 2001. Biodiversity and invasibility in grassland microcosms.
Oecologia 126: 563–568.
Eagles CF, Othman OB. 1989. Yield and persistency of contrasting white
clover populations grown in pure swards and in mixed swards with
S.23 perennial ryegrass. Annals of Applied Biology 114: 545–557.
Fargione J, Brown CS, Tilman D. 2003. Community assembly and invasion:
an experimental test of neutral versus niche processes. Proceedings of the
National Academy of Sciences of the USA 100: 8916–8920.
Fargione J, Tilman D, Dybzinski R, et al. 2007. From selection to comple-
mentarity: shifts in the causes of biodiversity-productivity relationships in
a long-term biodiversity experiment. Proceedings of the Royal Society
Finckh MR, Gacek ES, Czembor HJ, Wolfe MS. 1999. Host frequency and
density effects of powdery mildew and yield in mixtures of barley culti-
vars. Plant Pathology 48: 807–816.
Frame J, Charlton JFL, Laidlaw AS. 1998. Temperate forage legumes.
Wallingford: CAB International.
Frankow-Lindberg BE. 1999. Effects of adaptation to winter stress on
biomass production, growth and morphology of three contrasting white
clover cultivars. Physiologia Plantarum 106: 196–202.
Frankow-Lindberg BE. 2001. Adaptation to winter stress in nine white clover
populations: changes in non-structural carbohydrates during exposure to
simulated winter conditions, and ‘spring’ regrowth potential. Annals of
Botany 88: 745–751.
Friedley JD, Stachowicz JJ, Naeem S, et al. 2007. The invasion paradox:
reconciling pattern and processes in species invasions. Ecology 88: 3–17.
Hajjar R, Jarvis DI, Gemmill-Herren B. 2008. The utility of crop genetic
diversity in maintaining ecosystem services. Agriculture, Ecosystem and
Environment 123: 261–270.
Hauggaard-Nielsen H, Jørnsgard B, Kinane J, Steen Jensen E. 2008. Grain
legume – cereal intercropping: the practical application of diversity, com-
petition and facilitation in arable organic cropping systems. Renewable
Agriculture and Food Systems 23: 3–12.
Hector A, Schmid B, Beierkuhnlein C, et al. 1999. Plant diversity and pro-
ductivity experiments in European grasslands. Science 286: 1123–1127.
Hector A, Dobson K, Minns A, Bazeley-White E, Lawton JH. 2001.
Community diversity and invasion resistance: an experimental test in a
grassland ecosystem and a review of comparable studies. Ecological
Research 16: 819–831.
Hector A, Joshi J, Scherer-Lorenzen M, et al. 2007. Biodiversity and eco-
system functioning: reconciling the results of experimental and observa-
tional studies. Functional Ecology 21: 998–1002.
Hooper DU, Dukes JS. 2004. Overyielding among plant functional groups in
a long-term experiment. Ecology Letters 7: 95–105.
Hooper DU, Chapin III FS, Ewel JJ, et al. 2005. Effects of biodiversity on
ecosystem functioning: a consensus of current knowledge. Ecological
Monographs 75: 3–35.
Jolliffe PA, Wanjau FM. 1999. Competition and productivity in crop mix-
tures: some properties of productive crops. Journal of Agricultural
Science, Cambridge 132: 425–435.
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem 920
Joshi J, Schmid B, Caldeira MC, et al. 2001. Local adaptation enhances per- Download full-text
formance of common plant species. Ecology Letters 4: 536–544.
Kahmen A, Renker C, Unsicker SB, Buchmann N. 2006. Niche comple-
mentarity for nitrogen: an explanation for the biodiversity and ecosystem
functioning relationship? Ecology 87: 1244–1255.
Kirwan L, Sebastia ` MT, Finn JA, et al. 2007. Evenness drives consistent
diversity effects in an intensive grassland system across 28 European
sites. Journal of Ecology 95: 530–539.
Kley G. 1995. Seed production in grass and clover species in Europe.
In: Scho ¨berleinW, Fo ¨rster
International Herbage Seed Conference, 12.
Knops JMH, Tilman D, Haddad NM, et al. 1999. Effects of plant species
richness on invasion dynamics, disease outbreaks, insect abundance and
diversity. Ecology Letters 2: 286–293.
Lanta V, Leps ˇ J. 2007. Effects of species and functional group richness on
production in two fertility environments: an experiment with commu-
nities of perennial plants. Acta Oecologia 32: 93–103.
Lorentzen S, Roscher C, Schumacher J, Schultze E-D, Schmid B. 2008.
Species richness and identity affect the use of aboveground space in
experimental grasslands. Perspectives in Plant Ecology, Evolution and
Systematics 10: 73–87.
Milbau A, Nijs I, Van Peer L, Reheul D, De Cauwer B. 2003. Disentangling
invasiveness and invasibility during invasion in synthesized grassland
communities. New Phytologist 159: 657–667.
Milbau A, Nijs I, De Raedemaecker F, Reheul D, De Cauwer B. 2005.
Invasion in grassland gaps: the role of neighbourhood richness, light
availability and species complementarity during two successive years.
Functional Ecology 19: 27–37.
Mwangi PN, Schmitz M, Scherber C, et al. 2007. Niche pre-emption
increases with species richness in experimental communities. Journal
of Ecology 95: 65–78.
Palmborg C, Scherer-Lorentzen
Huss-Danell K, Ho ¨gberg P. 2005. Inorganic soil nitrogen under grass-
land plant communities of different species composition and diversity.
Oikos 110: 271–282.
Personeni E, Loiseau P. 2004. Species strategy and N fluxes in grassland soil:
a question of root litter quality or rhisosphere activity? European Journal
of Agronomy 22: 217–229.
Picasso VD, Brummer EC, Liebman M, Dixon PM, Wilsey BJ. 2008. Crop
species diversity affects productivity and weed suppression in perennial
polycultures under two management strategies. Crop Science 48: 331–342.
Pilet F, Chacon G, Forbes GA, Andrivon D. 2006. Protection of potato cul-
tivars against late blight in mixtures increases with decreasing disease
pressure. Phytopathology 96: 777–783.
Rath M, Peel S. 2005. Grassland in Ireland and the UK. In: McGilloway DA,
ed. Grassland: a global resource. The Netherlands: Wageningen
Academic Publishers, 13–27.
Reidsma P, Tekelenburg T, van den Berg M, Alkemade R. 2006. Impacts of
land-use change on biodiversity: an assessment of agricultural biodiver-
sity in the European Union. Agriculture, Ecosystems and Environment
Richardson DM, Pys ˇek P. 2006. Plant invasions: merging the concepts of
species invasiveness and community invasibility. Progress in Physical
Geography 30: 409–431.
Roscher C, Schumacher J, Baade J, et al. 2003. The role of biodiversity for
element cycling and trophic interactions: an experimental approach in a
grassland community. Basic and Applied Ecology 5: 107–121.
Keds.Proceedingsof the Third
M, Jumpponen A, Carlsson G,
Roscher C, Temperton VM, Scherer-Lorenzen M, et al. 2005. Overyielding
in experimental grassland communities – irrespective of species pool and
scale. Ecology Letters 8: 419–429.
RoscherC, TheinS, Schmid
Complementary nitrogen use among potentially dominant species in a
biodiversity experiment varies between two years. Journal of Ecology
Rougoor C, van der Weijden W. 2001. Towards a European levy on nitro-
gen: a new policy tool for reducing eutrophication, acidification and
climate change. Utrecht: Centre for Agriculture and Environment.
van Ruijven J, Berendse F. 2003. Positive effects of plant species diversity on
productivity in the absence of legumes. Ecology Letters 6: 170–175.
van Ruijven J, Berendse F. 2005. Diversity-productivity relationships: Initial
effects, long-term patterns, and underlying mechanisms. Proceedings of
the National Academy of Science of the USA 102: 695–700.
Sanderson MA, Skinner RH, Barker DJ, Edwards GR, Tracy BF, Wedin
DA. 2004. Plant species diversity and management of temperate forage
and grazing land ecosystem. Crop Science 44: 1132–1144.
Sanderson MA, Soder KJ, Muller LD, Klement KD, Skinner RH, Goslee
SC. 2005. Forage mixture productivity and botanical composition in pas-
tures grazed by dairy cattle. Agronomy Journal 97: 1465–1471.
Schmid B, Joshi J, Schla ¨pfer F. 2002. Empirical evidence for biodiversity –
ecosystem functioning relationships. In: Kinzig A, Tilman D, Pacala S,
eds. Functional consequences of biodiversity: experimental progress
and theoretical extensions. Princeton, NJ: Princeton University Press,
Schmid B, Hector A, Saha P, Loreau M. 2008. Biodiversity effects and
transgressive overyielding. Journal of Plant Ecology 1: 95–102.
Spehn EM, Scherer-Lorenzen M, Schmid B, et al. 2002. The role of
legumes as a component of biodiversity in a cross-European study of
grassland biomass nitrogen. Oikos 98: 205–218.
Temperton VM, Mwangi PN, Scherer-Lorenzen M, Schmid B, Buchmann
N. 2007. Positive interactions between nitrogen-fixing legumes and four
neighbouring species in a biodiversity experiment. Oecologia 151:
Tilman D, Reich PB, Knops J, Wedin D, Mielke T, Lehman C. 2001.
Diversity and productivity in a long-term grassland experiment. Science
Tilman D, Cassman KG, Matson PA, Naylor R, Polasky S. 2002.
Agricultural sustainability and intensive production practices. Nature
Tracy BF, Sanderson MA. 2004a. Productivity and stability relationships in
mowed pasture communities of varying species composition. Crop
Science 44: 2180–2186.
Tracy BF, Sanderson MA. 2004b. Forage productivity, species evenness and
weed invasion in pasture communities. Agriculture, Ecosystems and
Environment 102: 175–183.
Tracy BF, Renne IJ, Gerrish J, Sanderson MA. 2004. Effects of plant diver-
sity on invasion of weed species in experimental pasture communities.
Basic and Applied Ecology 5: 543–550.
Verbeke G, Molenbergh G. 2000. Linear mixed models for longitudinal data.
Springer Series in Statistics. New York, NY: Springer-Verlag.
Williams TA, Abberton MT, Rhodes I. 2003. Performance of white clover
varieties combined in blends and alone when grown with perennial rye-
grass under sheep and cattle grazing. Grass and Forage Science 58:
B,Scherer-Lorentzen M. 2008.
Frankow-Lindberg et al. — Biodiversity effects in a forage ecosystem921