You are on page 1of 7

PERSPECTIVE

PUBLISHED: 22 JUNE 2017 | VOLUME: 1 | ARTICLE NUMBER: 0168

Revisiting the biodiversity–ecosystem


multifunctionality relationship
Lars Gamfeldt†* and Fabian Roger†*
A recent and prominent claim for the value of biodiversity is its importance for sustaining multiple ecosystem functions. The
general idea is intuitively appealing: since all species are to some extent unique, each will be important for a different set of
functions. Therefore, as more functions are considered, a greater diversity of species is necessary to sustain all functions
simultaneously. However, we show here that the relationship between biodiversity and ecosystem functioning does not change
with the number of functions considered. Biodiversity affects the level of multifunctionality via non-additive effects on indi-
vidual functions, and the effect on multifunctionality equals the average effect on single functions. These insights run counter
to messages in the literature. In the light of our simulations, we present limitations and pitfalls with current methods used to
study biodiversity–multifunctionality, which together provide a perspective for future studies.

B
iodiversity is causally linked to ecosystem functioning: changes is a special case within the broader ‘biodiversity begets the level of
in the number of species in a community generally result in multifunctionality’ hypothesis. There is no conceptual difference
changes in a range of functions1–5. Functioning increases with between a species that performs a given function at a very low level
species richness due to niche complementarity, positive interac- and a species that does not perform the function at all. They are both
tions, or the presence of certain influential species. The relationship part of the same continuum. A value of zero will simply decrease the
between richness and functioning can generally be described using average level of multifunctionality slightly more than a low value.
some version of a positive but decelerating function6,7, such that the While the assumption that the slope between biodiversity and
relationship is steep at low richness, and levels off quickly as richness multifunctionality becomes steeper with the number of functions is
increases further. not always mentioned explicitly, it is still an implicit assumption in
Because species differ in their contributions to different func- many papers. It is common for articles to include variations of state-
tions, it has been proposed that biodiversity should be more impor- ments like “the impact of diversity is stronger when multiple func-
tant—and that the relationship should be less saturating—when tions are considered together,” or “more species are needed to sustain
multiple functions are considered. This idea, often referred to as multiple functions than any single function,” or “studies focusing
multifunctionality 8, was originally proposed in a study on seagrass, on individual functions will underestimate levels of biodiversity
where it was observed that a diverse assemblage of small marine required for multifunctionality” (for example, refs 3,8,10,11,24,25,29,30).
grazers simultaneously maximized multiple functions9. Positive The truth of such statements requires that the slope of the diversity–
biodiversity–multifunctionality relationships were later shown ecosystem function relationship changes as we move from single to
for many different systems8,10–23, and a recent meta-analysis sug- multiple functions.
gests that a positive biodiversity–multifunctionality relationship The importance of biodiversity for multifunctionality is
is general24. grounded on a verbal framework that makes intuitive sense. Species
have trade-offs in terms of allocating resources to growth, reproduc-
How biodiversity influences ecosystem multifunctionality tion and survival, so no single species can maximize all ecosystem
Because all species are to some extent functionally unique, high functions9,31,32. A variety of organisms are hence required for an eco-
biodiversity may provide for a high variety of functions. If species system to simultaneously sustain multiple functions. However, this
specialize in different functions, all species are needed to perform intuitive knowledge, blended with vague terminology, has painted a
all of the functions. hazy picture of a causal relationship between biodiversity and mul-
In addition to providing for more functions, increasing bio­ tifunctionality. We argue here that it is time scientists abandon the
diversity is proposed to increase the level of multifunctionality. idea that the effect on multifunctionality is larger than the average
More importantly, it is hypothesized that the importance of bio­ effect on single functions, and that the importance of biodiversity is
diversity increases with the number of functions considered. It is dependent on the number of functions considered. We show that,
this hypothesis that suggests that, as we increase the number of contrary to common belief, increasing the number of functions
functions that we study, the slope between biodiversity and multi­ considered does not by itself change the nature of the relationship
functionality becomes steeper. The literature on biodiversity and between biodiversity and ecosystem multifunctionality.
ecosystem multi­ functionality has mainly focused on this ‘bio­
diversity begets the level of multifunctionality’ hypothesis (for Defining biodiversity and multifunctionality
example, refs 8,10,12,24–26), and, with few exceptions22,27,28, all papers In order to assess the relationship between biodiversity and multi-
on the importance of biodiversity for multifunctionality study sets functionality, both variables first need to be defined. To make our
of functions shared by all focal species. We should point out that discussion easy to follow, we stick with species richness as our meas-
the case in which certain functions are unique to particular species ure of diversity. Species richness is not a particularly good metric

Department of Marine Sciences, University of Gothenburg, Box 461, SE-405 30 Göteborg, Sweden. †These authors contributed equally to this work.
*e-mail: lars.gamfeldt@gu.se; fabian.roger@gu.se

NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol 1


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ECOLOGY & EVOLUTION

Box 1 | Different ways of relating multifunctionality to changes in biodiversity.

To understand how biodiversity can be analytically related to


a b 30%
multi­functionality, it is useful to provide a hypothetical example F1
F2 3
(illustrated in the figure). Say that there are three species (X, Y, Z) 1.0

Number of functions ≥ threshold


F3 2
and three functions (F1, F2, F3), with each species providing the Average 1
three functions at different levels (green bars). We here set these
0
three function levels to be 0.2, 0.6 and 1.0. Species X is most

Function level
important for function 1, species Y for function 2, and species 0.6
Z for function 3. The average value for the three functions is the c 70%
same across all species (blue bars). Assuming no niches and equal 3
fitness across species (hence no biodiversity effects), the three- 2
species mixture XYZ has the same average function value as the 0.2 1
single species. If we are concerned with how many functions are 0
provided above a certain threshold of the maximum function lev- X Y Z XYZ 1 3
els, another picture emerges. At a level of 30% of the maximum, a Species composition Species richness
mean of two functions are above the threshold for single species
whereas all three functions are above the threshold for the three- Conceptual example with three species and functions. a, The figure
species mixture; a positive biodiversity effect. At a threshold level shows species X, Y and Z, which perform the functions F1, F2 and F3
of 70%, a mean of one function is above the threshold for single (green bars), and the average of those three functions (blue bars) for
species whereas no function is above the threshold for the mix- monocultures and the mixture. b,c, The number of functions sustained
ture; a negative relationship. This pattern has been described as the above a certain threshold (b, 30%; c, 70%) for monocultures and
jack-of-all-trades effect 23 (see main text). the mixture.

of diversity 33,34 and might not be appropriate under most circum- We consider two scenarios. First we consider a scenario in which
stances, especially if diversity is observed and not manipulated. Yet each individual function is unrelated to biodiversity (Fig. 1a). For
it is the simplest metric and will help us establish our framework, any level of richness, the average function value is the average of
which can be extended to other diversity metrics as desired. all possible species combinations, which is simply the average func-
There are at least four distinct approaches to analyse multi- tion value across all species in the species pool. The mean function
functionality: the species turnover approach8,13, the averaging values for all single functions thus stay the same regardless of spe-
approach16,35, the threshold approach10,12,25, and the multivariate cies richness. Since average multifunctionality is calculated as the
modelling framework26. In this paper, we use the average and mul- average of all standardized single functions, it too does not change
tiple threshold approaches to present the null expectation for bio- with diversity (Fig. 1b). Considering the functions jointly yields the
diversity and multifunctionality. The averaging approach calculates average of all nine functions.
multifunctionality as the average of all standardized function val- If we are interested in the number of functions above a certain
ues. The threshold approach tallies the number of functions that are threshold level (expressed as the percentage of the maximum level
above one or multiple thresholds, usually defined as some percent- of that function across all species combinations and richness lev-
age (often between 1% and 99%) of the maximum observed perfor- els) the expectation depends on the threshold level. The diversity
mance for a given function. The two approaches give complementary effect is positive at low thresholds (below roughly the grand mean of
information, as explained in detail in Box 1. all functions, here 52%), but at thresholds above that, the diversity
effect is negative (Fig. 1c). The patterns in Fig. 1c mirror those in
The biodiversity–multifunctionality relationship Box 1. An alternative way of illustrating the patterns in Fig. 1c is to
To explore the biodiversity multifunctionality–relationship, we sim- plot the slope of the relationship between species richness and the
ulate 15 species (A–O) that perform 9 functions (F1 to F9). The levels number of functions above the full range of thresholds25 (Fig. 1d).
at which each function is performed by each species is drawn from Point p in Fig. 1d shows that the slope is zero at the 51% threshold.
a uniform distribution, U(0,1). Pairwise correlations between func- With the threshold approach, we observe that biodiversity can be
tions vary randomly from negative (–0.5) to positive (0.6) with an both positively and negatively related to multifunctionality without
average correlation close to 0. The nine functions are standardized any non-additive interactions. Non-additive effects include positive
by their maximum (f(x) = x / max(x)). In a second step we simulate a and negative species interactions, which alone or in combination
diversity experiment where we combine the species from the species result in what is referred to as biodiversity effects36. For a scenario
pool in communities with richness 1 to 15. All possible communities with no niches and equal fitness across species, there is a positive
are simulated at each richness level. We use a replacement design, relationship at thresholds below roughly the grand mean across all
that is, with species abundances declining proportionally with the functions and a negative relationship above that. The steepness of
number of species. This means, for instance, that in a ten-species the slope depends on the threshold level (Fig. 1b).
mixture, each species constitutes 10% of the total abundance. This The results presented in Fig. 1a–d are similar to those provided in a
is the most commonly used design in biodiversity ecosystem func- recent paper on forest biodiversity and ecosystem multi­functionality 23.
tioning experiments and simulates a scenario with no niches and The authors propose that these patterns are driven by a mechanism
equal fitness across species (under the assumption that all species referred to as the jack-of-all-trades effect. The English saying ‘jack of
are equally abundant). Such a scenario yields the null expectation all trades, master of none’ is taken as metaphor for the description
used in biodiversity–ecosystem functioning studies when testing for of the pattern that diverse communities can uphold many functions
biodiversity effects. For example, to calculate selection and comple- above low threshold levels, but only few functions above high thresh-
mentarity (sensu ref. 36), the proportion of each species in the species old levels. Species effects on different functions average out in species
mixture, assuming equal abundance, is used to calculate expected mixtures, so that communities with many species will, according to
values of functions, based on species performances in monoculture. the threshold approach, have higher multifunctionality for low levels

2 NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ECOLOGY & EVOLUTION PERSPECTIVE
a e
F1 F2 F3 F1 F2 F3
1.0 1.0
0.5 0.5
0.0
Value of function

Value of function
0.0
F4 F5 F6 F4 F5 F6
1.0 1.0
0.5 0.5
0.0 0.0
F7 F8 F9 F7 F8 F9
1.0 1.0
0.5 0.5
0.0 0.0
4 8 12 4 8 12 4 8 12 4 8 12 4 8 12 4 8 12
Species richness Species richness
b f
Average value of standardized functions

Average value of standardized functions


0.7

0.6
0.6

0.5
0.5

0.4
0.4
1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15
Species richness Species richness
c 10% 20% 30%
g 10% 20% 30%
9 9
6 6
Number of functions ≥ threshold

Number of functions ≥ threshold

3 3
0 0
40% 50% 60% 40% 50% 60%
9 9
6 6
3 3
0 0
70% 80% 90% 70% 80% 90%
9 9
6 6
3 3
0 0
1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15
Species richness Species richness
d h

0.2 0.2
p: 52% p: 62%
Slope
Slope

0.0 0.0

-0.2 –0.2

25 50 75 100 25 50 75 100
Threshold (%) Threshold (%)

Figure 1 | Two scenarios for the relationship between biodiversity and multifunctionality. a–h, The relationship between biodiversity (species richness)
and multifunctionality for a scenario with no species interactions (a-d) and with complementarity for two functions (1 and 6, e–h). Relationships are shown
for individual functions (a,e), average function level (b,f) and for the number of functions above or equal to each of nine thresholds (c,g). Also presented
is how the slope of the relationship between biodiversity and number of functions above threshold changes with threshold level (d,h). In plots a–c and
e–g, point values are rounded to the third decimal and only unique points are shown. Lighter colours represent higher point densities. The regression line
is fitted to all points before rounding (n = 32,767). In plots d and h, the shaded area represents the 95% confidence interval of the slope estimates, point p
indicates the threshold at which the slope turns from positive to negative.

of functioning, but lower multifunctionality if high levels of function- diversity due to the statistical averaging of fluctuations in species’
ing are required. The jack-of-all-trades effect is explained as broadly abundances. While the simulations in the forest multifunctionality
analogous to the averaging effect underlying the portfolio effect 23. The paper 23 and our study reveal similar patterns, we offer a very different
portfolio effect (sensu ref. 37) means that stability rises with species interpretation as we discuss in more detail later.

NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol 3


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ECOLOGY & EVOLUTION

a b c

0.03 0.25 0.25


Slope estimate

Slope estimate
Slope estimate
0.02
0.00 0.00

0.01
–0.25 –0.25

0.00
1 2 3 4 5 6 7 8 9 25 50 75 100 25 50 75 100
Number of functions considered Threshold (%) Threshold (%)
Number of functions 0 3 Number of functions 3 6 Number of species 6 9
with complementarity 6 9 (15 species) 9 (9 functions) 15

Figure 2 | The slope between species richness and average multifunctionality in relation to the number of functions and species considered. a, The
figure shows the slopes for the average multifunctionality–diversity relationship (for 1–12 species) for all possible combinations of 9 functions. We present
4 scenarios in which there is complementarity for 0, 3, 6 and 9 functions (out of 9 total). The average slope increases with the number of functions subject
to complementarity, but does not change with the number of functions included. Although individual slopes vary depending on exactly which functions
are included in any one combination of functions, the average slope stays constant over the full range of number of functions. For each scenario and each
number of functions n we plot the slope for all possible combinations of n out of 9 functions, 511 slopes in total per scenario. b, Changes in the effect
of biodiversity as a function of the number of functions using the multi-threshold approach. Curves for 3, 6 and 9 functions are shown holding species
richness constant at 15 species. Maximum and minimum slopes increase as more functions are included. Lines are means ± 95% confidence interval of the
slopes calculated for all possible combinations of n (= 3, 6 or 9) out of 9 functions. c, An equivalent plot to that in b but for three different richness levels,
holding the number of functions constant. It shows that the maximum and minimum slopes decrease with increasing maximum species richness. Lines are
means ± 95% confidence interval of the slopes calculated for all possible combinations of n (= 6, 9 or 15) out of 15 species.

In the second scenario that we simulate, we include a positive functions (compare with Fig. 1e–h for a scenario with two func-
complementarity effect. In our example, we assign complementa- tions with complementarity). As expected, the average slope of the
rity to two of the nine functions (functions 1 and 6, Fig. 1e). We diversity–multifunctionality relationship increases with the number
could have chosen an example including any number of functions of functions subject to complementarity. However, the average slope
subject to complementarity, since our approach is general (as we does not change with the total number of functions considered.
show in ‘Number of functions does not matter’ below). We imple- Although individual slopes vary depending on exactly which func-
ment complementarity by multiplying the single function values by tions are included in any one combination of functions, the average
a complementarity factor (CF), where the complementarity factor is slope stays constant across the full range of number of functions.
a saturating function of species richness (S). The complementarity What does it look like if we instead use the threshold approach
factor is 1 for monocultures and grows to a predefined maximum to study how the effect of biodiversity changes with the number of
(CFmax), which was set to 3 in our example. functions? Figure 2b shows the slope of the relationship between
the number of functions above thresholds and species richness
CF = CFmax × (1 – (1 – 1 / CFmax) × exp(1 – Sr)) against the threshold (as in Fig. 1d,h). We present curves for three,
six and nine functions, holding species richness constant at 15.
where r defines the speed at which CF grows towards CFmax and is Maximum and minimum slopes increase as more functions are
set to 0.5. included. Figure 2c, which presents a similar plot but for three dif-
Complementarity causes the average multifunctionality to be ferent richness levels, holding the number of functions constant,
positively related to species richness (Fig. 1f). The slope of the aver- shows that the maximum and minimum slopes decrease with
age multifunctionality relationship is the average of the seven flat increasing richness. While the patterns in Fig. 2 seem intriguing,
and the two positive individual function slopes. The complemen- they are simply due to the fact that we vary the range on the y-axis
tarity for two functions also causes the upper range of the positive while holding the x-axis constant (panel b—an increasing number
multiple threshold effect to be pushed towards higher threshold lev- of functions with constant richness) or vary the range on the x-axis
els (here around 63%, see Fig. 1g,h). In other words, we observe a while holding the y-axis constant (panel c—an increasing number
positive effect of biodiversity over a wider range of thresholds. A of species with constant number of functions). A change in the
negative effect is observed for thresholds above 68%. maximum effect of diversity does thus not represent any under­lying
biological phenomenon. One way to illustrate this point is to con-
Number of functions does not matter sider a hypothetical example with ten species and nine functions
The scenario in which biodiversity is unrelated to each individual (Supplementary Fig. 1). All functions are perfectly and positively
function demonstrates that biodiversity does not increase the aver- correlated, meaning that for each species the contribution is the
age level of multiple functions (Fig. 1b). Considering more func- same for all functions. This is equivalent to studying a single func-
tions does not affect the relationship between biodiversity and tion. Regardless of how many of the nine functions we subtract or
multifunctionality, as it is often claimed. A direct way of showing add, we still effectively study a single function. Yet, we observe a
this is by plotting the slope between species richness and average change in slopes, depending on how many functions we include.
multifunctionality on the y-axis against the number of functions Furthermore, if Fig. 2b should be taken as evidence for stronger
on the x-axis. Figure 2a shows the slopes for the scenario in which diversity effects at high number of functions we must conclude, by
biodiversity is unrelated to each individual function, as well as for the same logic, that considering more species decreases the effect of
scenarios in which there is complementarity for three, six and nine biodiversity on multifunctionality.

4 NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ECOLOGY & EVOLUTION PERSPECTIVE
a
0.4

0.2
Slope estimate

0.0

Scaled by max
–0.2 Scaled (0,1)

25 50 75 100
Threshold (%)

b c
4 f(x) = x – min(x) 6 f(x) = x
max(x) – min(x) max(x)
3
4

Density
Density

2
1

0 0
0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00
Scaled between (0,1) Scaled by maximum

Figure 3 | The effect of standardization choice on the result of the multi-threshold approach. a, The expected result from the multi-threshold approach
for the German site of the BIODEPTH project using two different methods to standardize raw function values. We used real data from the monocultures
to simulate the null expectation for the full richness gradient. We standardized the data to be between 0 and 1 (blue line) or by the maximum (red line).
Both standardization techniques are common in the literature but the choice has a large impact on the resulting slope pattern. The shaded areas represent
the 95% confidence interval of the slope estimates (n = 1,023); the simulation was fully replicated (all possible species composition at each richness
level). b and c show the effect of the two different standardization techniques on the mean and the distribution of the functions, which results in the two
distinct patterns. The dashed lines in b and c mark the grand mean of all standardized function values.

The devil is in the multifunctionality detail such that no function value is close to 0 (for example, mean = 30,
The fact that the relationship between biodiversity and multi­ standard deviation = 6), the standardization by the maximum (or
functionality is sensitive to the number of functions and species, for that matter the equivalent choice of using unstandardized data)
makes it clear that we need to be careful when interpreting the will show a dramatically different pattern compared to a standardi-
results of any one study. But these are not the only factors that com- zation between 0 and 1. The latter will show the now familiar pat-
plicate interpretations. Because, as the saying goes: the devil is in the tern of positive slopes below 50% and negative slopes above. The
detail. For example, the way we prepare our data for analyses will former, however, yields flat slopes (that is, no effect of biodiversity)
have implications for the results. We can use the raw data, standard- up to a threshold of 50%, followed by the typical pattern of positive
ize data by dividing by their maximum (or some top percentage of (50–76%) and negative slopes (>76%, Supplementary Fig. 2). Note
the maximum), standardize them to range from 0 to 1, or convert that in this hypothetical example, the range of threshold values with
data to standard scores. If we compare two ways of standardization the highest positive slopes for the standardization by the maximum
we see that standardizing by the maximum value (f(x) = x / max(x)) is the range of threshold values with the highest negative slopes for
produces a different pattern than if we standardize between 0 and 1 the standardization between 0 and 1. We would like to add that
(f(x) = (x – min(x)) / (max(x) – min(x))). For the Germany site in standardizing functions between 0 and 1 has strong implications for
the BIODEPTH project 8,38,39, the expected slopes for the standardi- the interpretation: if we take the example of a function with values
zation relative to the maximum are generally less steep compared ranging from 90 to 100, then, by standardizing it between 0 and 1, we
to slopes based on the standardization between 0 and 1 (Fig. 3a). effectively say that the species that performs the function at level 90,
The threshold at which the slope turns from positive to negative does not perform the function at all and that the species performing
(compare with point p in Fig. 1d,h) occurs at a higher threshold for the function at level 95 performs it at 50% of its maximum.
the standardization relative to the maximum (around 50%) com- It is clear that the results from any one study on multifunctional-
pared to the standardization between 0 and 1 (around 35%). This ity will be sensitive to which exact approach is used. An important
is because the standardization of each function between 0 and 1 message emerging from our simulations is hence that studies on
decreases the grand mean of all function values. The implications multifunctionality need to be explicit about their methods and null
of such apparently trivial choices have hitherto been overlooked in expectations (for example, given the performance of monoculture).
the field. Distributions for the standardized data for the BIODEPTH Such null expectations are currently absent in the multifunctionality
Germany site are shown in Fig. 3b (standardized between 0 and 1) field. Furthermore, not only will the expected distributions depend
and Fig. 3c (standardized to the maximum). Raw data are available on standardization, but also on the original distributions of each
via the multifunc package25. individual function, as well as on how many functions and species
While we chose to make the point above using experimental are included (Fig. 2). Since the slope of the relationship is sensitive
data from BIODEPTH, other examples can easily be simulated. If, to the number of functions and species, caution should be taken
for example, function values are drawn from a normal distribution when comparing biodiversity effects across studies. The full range

NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol 5


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
PERSPECTIVE NATURE ECOLOGY & EVOLUTION

of null-expectations observed during our simulations goes beyond in summarizing the collective behaviour of multiple ecosystem
what we have space to discuss here. They make it clear, however, that functions and services, and in dealing with trade-offs25. It is quite
the null expectations for the multi-threshold approach vary widely plausible, that when there is a fixed set of functions of interest,
and are non-intuitive. We therefore caution against the use of the with threshold levels that can be specified a priori, and the species
multi-threshold approach as our simulations indicate that a coher- pool is defined, a threshold approach can be used to examine if and
ent interpretation is challenging. As this conclusion does not invali- when biodiversity is important to sustain multifunctionality. There
date the justified criticism of alternative multifunctionality metrics may, for example, be applied contexts in which we know that only
articulated previously 25, we think it is time for the field of multi- a defined small reduction is acceptable for production, whereas we
functionality to reconsider its current use of metrics. We provide an accept much larger decreases in cultural services. However, it is nec-
online simulation tool where a multitude of different scenarios can essary that we are aware that the importance of biodiversity does
be simulated. It is our hope that the online application can aid the not change as we move from single to multiple functions. Hand-
understanding of current and future metrics of multifunctionality waving claims that the value of biodiversity increases with higher
(see ‘Code availability’). dimensions of functioning will only lead research and management
efforts astray.
Concluding remarks
Even though species differ in their importance for different func- Code availability
tions, we demonstrate that the effect on multifunctionality is not The application can be accessed at: https://fabianroger.shinyapps.
larger than the average effect on single functions. Quite the con- io/multifunctionality-simulations/. The source code for the app
trary, it is sobering to realize that in the absence of positive diver- can be found here: https://github.com/FabianRoger/Gamfeldt_
sity effects, multifunctionality can almost always be maximized by Roger_2017_NEE_app and https://doi.org/10.5281/zenodo.496151.
the highest performing monoculture, just as with single functions The R code producing all simulations presented in this article
(Fig. 1b,c). Regressions on the 100% quantile of our simulated data can be found here: https://github.com/FabianRoger/Gamfeldt_
reveal that the relationship between biodiversity and multifunction- Roger_2017_NEE and https://doi.org/10.5281/zenodo.496144. To
ality goes from flat at low thresholds to negative at high thresholds run from within R, execute:
(Supplementary Fig. 3), and is never positive.
As we have demonstrated, claims that considering multiple func- library(shiny)
tions increases the need for, and therefore the value of, biodiversity
are not justified. Many terms in the biodiversity–multifunctionality runGitHub(“FabianRoger/Gamfeldt_Roger_2017_NEE_app”)
literature give fuel to such claims. Examples of terms include “com-
plementarity across multiple functions,”10 “multivariate dominance Received 25 November 2016; accepted 19 April 2017;
effect,”9 and “multivariate complementarity”40. We emphasize that published 22 June 2017
these, and similar, terms are misleading. There can only be selec-
tion and complementarity for single functions, and mechanisms References
1. Schmid, B. et al. in Biodiversity, Ecosystem Functioning, and Human Wellbeing:
behind high levels of multifunctionality must thus also be sought an Ecological and Economic Perspective (eds Naeem, S. et al.) Ch. 2 (Oxford
in single functions. Univ. Press, 2009).
The natural world is, of course, often not characterized by equal 2. Cardinale, B. J. et al. The functional role of producer diversity in ecosystems.
abundances of all species and the absence of niches and fitness dif- Am. J. Botany 98, 572–592 (2011).
ferences. Increases in biodiversity, for example, by the addition or 3. Cardinale, B. J. et al. Biodiversity loss and its impact on humanity. Nature
arrival of new species to an ecosystem, often does not translate to 486, 59–67 (2012).
The most-recently published scientific consensus statement on how
a direct and proportional decline in the already present species41 biodiversity influences ecosystem functions and services, including a short
(which is the case in a replacement design). And conversely, loss of forward-looking perspective on biodiversity and multifunctionality.
species may not result in a compensatory and proportional increase 4. Tilman, D., Isbell, F. & Cowles, J. M. Biodiversity and ecosystem functioning.
in the remaining species. In such cases, what we have is a bio­ Ann. Rev. Ecol. Evol. Syst. 45, 471–493 (2014).
diversity effect in terms of niche complementarity for abundance 5. Gamfeldt, L. et al. Marine biodiversity and ecosystem functioning: what’s
known and what’s next? Oikos 124, 252–265 (2015).
and/or biomass production. Such complementarity has strong
6. Cardinale, B. J. et al. Effects of biodiversity on the functioning of trophic groups
potential to drive changes in other functions. For example, higher and ecosystems. Nature 443, 989–992 (2006).
abundance and/or biomass of organisms within and across trophic 7. O’Connor, M. I. et al. A general biodiversity–function relationship is mediated
levels can drive increases in fluxes of material and energy 42,43. by trophic level. Oikos 126, 18–31 (2016).
If we are to better manage ecosystems and the services they pro- 8. Hector, A. & Bagchi, R. Biodiversity and ecosystem multifunctionality. Nature
vide we need a thorough understanding of the role that biodiversity 448, 188–190 (2007).
9. Duffy, J. E., Richardson, J. P. & Canuel, E. A. Grazer diversity effects on
plays. To quote Cardinale et al.44: “the three most important fac-
ecosystem functioning in seagrass beds. Ecol. Lett. 6, 637–645 (2003).
tors that will determine the success of the biodiversity–ecosystem This was the first explicit biodiversity–multifunctionality paper, and it
functioning paradigm will be our ability to identify mechanisms, concluded that a mixture of species simultaneously maximizes multiple
mechanisms, mechanisms!” We argue that this is very true for the functions.
biodiversity–multifunctionality field. In that light, we need to learn 10. Gamfeldt, L., Hillebrand, H. & Jonsson, P. R. Multiple functions increase
more about which functions are driven by biodiversity and how the importance of biodiversity for overall ecosystem functioning. Ecology
89, 1223–1231 (2008).
they in turn affect other functions and multifunctionality. We are
11. He, J.-Z., Ge, Y., Xu, Z. & Chen, C. Linking soil bacterial diversity to ecosystem
not excluding the possibility that there might be a yet undiscovered multifunctionality using backward-elimination boosted trees analysis.
mechanism which directly links diversity and multifunctionality. J. Soils Sediments 9, 547–554 (2009).
However, the burden of proof lies on us ecologists, and we must 12. Zavaleta, E. S., Pasari, J. R., Hulvey, K. B. & Tilman, G. D. Sustaining multiple
stop assuming a mechanism where none has been shown. ecosystem functions in grassland communities requires higher biodiversity.
We emphasize that we do not send the message that multi­ Proc. Natl Acad. Sci. USA 107, 1443–1446 (2010).
13. Isbell, F. et al. High plant diversity is needed to maintain ecosystem services.
functionality should not be studied. It is indeed important to exam-
Nature 477, 199–202 (2011).
ine multiple aspects of functioning associated with biodiversity loss, 14. Mouillot, D., Villeger, S., Scherer-Lorenzen, M. & Mason, N. W. H. Functional
climate change, land use intensification, and other dimensions of structure of biological communities predicts ecosystem multifunctionality.
environmental change45–47. Multifunctionality metrics can be useful PLoS ONE 6, e17476 (2011).

6 NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
NATURE ECOLOGY & EVOLUTION PERSPECTIVE
15. Peter, H. et al. Multifunctionality and diversity in bacterial biofilms. PLoS ONE 37. Doak, D. F. et al. The statistical inevitability of stability–diversity relationships
6, e23225 (2011). in community ecology. Am. Nat. 151, 264–276 (1998).
16. Maestre, F. T. et al. plant species richness and ecosystem multifunctionality in 38. Hector, A. et al. Plant diversity and productivity experiments in European
global drylands. Science 335, 214–218 (2012). grasslands. Science 286, 1123–1127 (1999).
17. Bradford, M. A. et al. Discontinuity in the responses of ecosystem 39. Spehn, E. M. et al. Ecosystem effects of biodiversity manipulations in European
processes and multifunctionality to altered soil community composition. grasslands. Ecol. Monogr. 75, 37–63 (2005).
Proc. Natl Acad. Sci. USA 111, 14478–14483 (2014). 40. Stachowicz, J. J., Bruno, J. F. & Duffy, J. E. Understanding the effects of marine
18. Wagg, C., Bender, S. F., Widmer, F. & van der Heijden, M. G. A. Soil biodiversity biodiversity on communities and ecosystems. Ann. Rev. Ecol. Evol. Syst.
and soil community composition determine ecosystem multifunctionality. 38, 739–766 (2007).
Proc. Natl Acad. Sci. USA 111, 5266–5270 (2014). 41. Vilà, M. et al. Ecological impacts of invasive alien plants: a meta-analysis
19. Perkins, D. M. et al. Higher biodiversity is required to sustain multiple ecosystem of their effects on species, communities and ecosystems. Ecol. Lett.
processes across temperature regimes. Global Change Biol. 21, 396–406 (2015). 14, 702–708 (2011).
20. Lohbeck, M., Bongers, F., Martinez-Ramos, M. & Poorter, L. The importance 42. Grime, J. P. Benefits of plant diversity to ecosystems: immediate, filter and
of biodiversity and dominance for multiple ecosystem functions in a human- founder effects. J. Ecol. 86, 902–910 (1998).
modified tropical landscape. Ecology 97, 2772–2779 (2016). 43. Lohrer, A. M., Thrush, S. F. & Gibbs, M. M. Bioturbators enhance
21. Mori, A. S. et al. Low multifunctional redundancy of soil fungal diversity at ecosystem function through complex biogeochemical interactions. Nature
multiple scales. Ecol. Lett. 19, 249–259 (2016). 431, 1092–1095 (2004).
22. Soliveres, S. et al. Locally rare species influence grassland ecosystem 44. Cardinale, B. J. et al. in Biodiversity, Ecosystem Functioning, and Human
multifunctionality. Phil. Trans. R. Soc. B 371, 20150269 (2016). Wellbeing: an Ecological and Economic Perspective (eds Naeem, S. et al.) Ch. 8
23. van der Plas, F. et al. Jack-of-all-trades effects drive biodiversity-ecosystem (Oxford Univ. Press, 2009).
multifunctionality relationships in European forests. Nat. Commun. 45. Alsterberg, C., Sundbäck, K. & Gamfeldt, L. Multiple stressors and
7, 11109 (2016). multifunctionality: limited effects on an illuminated benthic system. Biol. Lett.
24. Lefcheck, J. S. et al. Biodiversity enhances ecosystem multifunctionality across 10, 20140640 (2014).
trophic levels and habitats. Nat. Commun. 6, 6936 (2015). 46. Allan, E. et al. Land use intensification alters ecosystem multifunctionality
This is to date the only meta-analysis of biodiversity–multifunctionality via loss of biodiversity and changes to functional composition. Ecol. Lett.
studies. 18, 834–843 (2015).
25. Byrnes, J. E. K. et al. Investigating the relationship between biodiversity and 47. Jing, X. et al. The links between ecosystem multifunctionality and above-
ecosystem multifunctionality: challenges and solutions. Methods Ecol. Evol. and belowground biodiversity are mediated by climate. Nat. Commun.
5, 111–124 (2014). 6, 8159 (2015).
This is the first methods and review paper about the different approaches
available to analyse biodiversity and multifunctionality.
26. Dooley, A. et al. Testing the effects of diversity on ecosystem multifunctionality Acknowledgements
We would like to thank M. Jonsson, J. Griffin and J. Bengtsson for providing valuable
using a multivariate model. Ecol. Lett. 18, 1242–1251 (2015).
feedback and comments on earlier drafts. This work was supported by grant 621-2009-
27. Hensel, M. J. S. & Silliman, B. R. Consumer diversity across kingdoms
5457 from the Swedish research council VR to L.G.
supports multiple functions in a coastal ecosystem. Proc. Natl Acad. Sci. USA
110, 20621–20626 (2013).
28. Soliveres, S. et al. Biodiversity at multiple trophic levels is needed for ecosystem
Author contributions
multifunctionality. Nature 536, 456–459 (2016). Study idea by L.G.; L.G. and F.R. designed the study; F.R. wrote the script for the
29. Duffy, J. E. Why biodiversity is important to the functioning of real-world analyses; L.G. and F.R. interpreted the analyses and wrote the paper.
ecosystems. Front. Ecol. Environ. 7, 437–444 (2009).
30. van der Plas, F. et al. Biotic homogenization can decrease landscape-scale forest
multifunctionality. Proc. Natl Acad. Sci. USA 113, 3557–3562 (2016). Additional information
31. Gamfeldt, L. et al. Higher levels of multiple ecosystem services are found in Supplementary information is available for this paper.
forests with more tree species. Nat. Commun. 4, 1340 (2013).
Reprints and permissions information is available at www.nature.com/reprints.
32. Díaz, S. et al. The global spectrum of plant form and function. Nature
529, 167–171 (2016). Correspondence should be addressed to L.G. and F.R.
33. Jost, L. Entropy and diversity. Oikos 113, 363–375 (2006). How to cite this article: Gamfeldt, L. & Roger, F. Revisiting the biodiversity–ecosystem
34. Hillebrand, H., Bennett, D. M. & Cadotte, M. W. The consequences of multifunctionality relationship. Nat. Ecol. Evol. 1, 0168 (2017).
dominance: A review of the effects of evenness on local and regional ecosystem Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
processes. Ecology 89, 1510–1520 (2008). published maps and institutional affiliations.
35. Hooper, D. U. & Vitousek, P. M. Effects of plant composition and diversity on
nutrient cycling. Ecol. Monogr. 68, 121–149 (1998).
36. Loreau, M. & Hector, A. Partitioning selection and complementarity in Competing interests
biodiversity experiments. Nature 412, 72–75 (2001). The authors declare no competing financial interests.

NATURE ECOLOGY & EVOLUTION 1, 0168 (2017) | DOI: 10.1038/s41559-017-0168 | www.nature.com/natecolevol 7


©
2
0
1
7
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
,
p
a
r
t
o
f
S
p
r
i
n
g
e
r
N
a
t
u
r
e
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.

You might also like