Biodivers Conserv

DOI 10.1007/s10531-015-1038-x

ORIGINAL PAPER

Different species traits produce diverse spatial functional

diversity patterns of amphibians

Mariana A. Tsianou1 • Athanasios S. Kallimanis1

Received: 5 July 2015 / Revised: 23 October 2015 / Accepted: 10 December 2015

Ó Springer Science+Business Media Dordrecht 2015

Abstract Trait selection is a critical step of functional diversity (FD) studies and is often
hampered by the limited availability of trait data for large sets of species. Thus, most
studies examine FD using traits determined primarily by the availability of data rather than
a specific function. This practice implicitly assumes that the subset of functional traits
available will be representative of the trait data of interest. Here, we test whether the
different functional trait groups (FTGs-distinct axes of ecological niches of the species e.g.
morphological, reproductive, physiological) show congruent FD spatial patterns and if one
FTG could act as a surrogate for another FTG. We used data on the 85 amphibian species
of Europe and 12 traits to estimate three aspects of FD (richness, evenness and dispersion)
for three different FTGs (Morphological, Reproductive and Habitat-related) and the
complete set of traits. Our results challenge the surrogacy value of one set of FTGs for
another set of FTGs. We further found that this has an impact on identifying hotspots
(different functions will have different hotspots). So the argument that FD is easier to
preserve in less area due to the redundancy of traits may hold true for a specific function,
but will not necessarily be the same for different functions. Finally, our results reveal that,
for any given community, FD is not characterized by a single spatial pattern, but by many
depending on the function analyzed.

Communicated by Dirk Sven Schmeller.

Electronic supplementary material The online version of this article (doi:10.1007/s10531-015-1038-x)

contains supplementary material, which is available to authorized users.

& Mariana A. Tsianou

mtsianou@upatras.gr
Athanasios S. Kallimanis
akallim@upatras.gr
1
Department of Environmental and Natural Resources Management, University of Patras,
30100 Agrinio, Greece

123
 Biodivers Conserv

Keywords Amphibians
Cross-function congruence
Functional diversity
Functional

trait group
Indices

Introduction

A major goal of macroecological studies is to understand spatial patterns of species

richness (Rosenzweig 1995) and the possible mechanisms governing such patterns (Hub-
bell 2001). More recently, ecological research focus shifted towards more functional
aspects of biodiversity with an emphasis on functional diversity (hereafter referred to as
FD). In the last decade, the number of publications regarding FD has increased expo-
nentially with more than 600 articles published till 2015 (papers containing the phrase
‘functional diversity’ in the title in the category ‘Ecology’ in Web of Science). This
increase is due to the acknowledgment of the importance of FD for ecology, since it is
becoming apparent that species richness alone may not be sufficient to fully understand
ecosystem resilience and ecosystem functioning, and that the concept of FD may be more
relevant (Chillo et al. 2011). Furthermore, there are documented cases where FD is linked
to ecosystem processes more strongly than taxonomic diversity (Scherer-Lorenzen 2008;
Philpott et al. 2009; Schmitz 2009; Flynn et al. 2011) but also it has been shown that
multidimensional FD indices are less informative than single traits in modelling ecosystem
process (Dı́az et al. 2007; Pakeman 2014a).
FD measures the variability in the functional traits displayed by species within com-
munities (Dı́az and Cabido 2001; Tilman 2001). Thus, two communities with the same
number of species can have very different FD value depending on how similar or vice
versa dissimilar are the functional traits of the species in each community. A range of
methods have been developed to quantify FD (reviewed in Mouchet et al. 2010) incor-
porating information on a variety of quantitative or qualitative functional traits of species.
Functional traits refer to characteristics of species that influence their performance and
infer a linkage to ecosystem function (McGill et al. 2006). By definition, functional traits
reflect species’ adaptations to different environmental conditions and may be morpho-
logical (e.g. body size), reproductive (e.g. seed or egg size, age of sexual maturity),
physiological (e.g. potential photosynthetic rate), or behavioural (e.g. feeding method)
(Bremner et al. 2003; Dumay et al. 2004; Lepš et al. 2006). So far researchers have used
functional traits that relate species to ecosystem processes, such as productivity or polli-
nation (Hooper et al. 2005), or even traits that link species to their interaction with biotic
and abiotic factors (e.g. resource availability or climatic variations) (Hooper et al. 2002;
Kraft et al. 2015). The first step in any study examining FD is to identify how many and
which functional traits to analyze. Petchey and Gaston (2006) underscore the importance of
trait selection urging researchers to include ‘‘all traits that are important for the function of
interest and no traits that are functionally uninformative’’. However, collecting data on
traits especially in macroecological studies (where many species are included) could be
certainly time and manpower demanding and in some cases it might prove impossible if the
desired traits for the focus species have not been measured. The traits measured and
available in the literature are limited and reflect the interest of the previous researchers.
Generally, body size tends to be the more readily available trait for animals e.g. data on the
body mass of *85 % of bird species are available in Dunning (2007) (see also Jones et al.
2009; Olson et al. 2009; Wilman et al. 2014). Thus body size is the most commonly used

123
Biodivers Conserv

trait in animal FD studies. Body mass variation has been suggested as a suitable proxy for
FD (Fritz and Purvis 2010) since many important traits co-vary with body size. For
example, body size is used as a surrogate of the quantity and size of resources consumed,
organism’s growth rate, life span etc. (Brown 1995). Another limitation is that many
species are rare, endangered or occur in remote locations and thus most of their traits have
not been recorded (Nakagawa and Freckleton 2008). As a result it is time consuming and
expensive to go into the field and measure their traits, not to mention the ethical issues as to
whether we should disturb their few remaining populations in order to measure their
functional traits. So in most cases, a functional trait may have been measured for some
species but not all species in a given community. For example, Trochet et al. (2014)
compiled a database of life-history traits for European amphibians. They collated infor-
mation on the morphological traits of 70–99 % of the species. But, they found that
information for specific traits reflecting key niche characteristics, like metabolism rate, was
available for less than 27 % of the species. Recently, Pakeman (2014b) concluded that
failing to measure the traits of all the species in a community (especially rare species)
affects the value of FD indices and thus may bias any conclusion drawn from such data.
And more generally, it has been argued that deleting missing values may result in mis-
leading inferences or biased estimates of ecological and evolutionary parameters (Hadfield
2008; González-Suárez et al. 2012; Pakeman 2014b).
Trait selection is a critical step in FD studies and a series of questions have to be
addressed in studies measuring FD (Petchey and Gaston 2006): what types of traits?;
Which traits?; How many traits?; How to obtain trait values? According to Petchey and
Gaston (2006) researchers should include all functionally informative traits, since fewer
traits may overestimate the functional redundancy of communities. However, this step is
hampered by the limited availability of data on traits for large sets of species. Thus it is not
surprising that in most relevant studies only few traits are used. We conducted a short
review (searching for papers containing the phrase ‘functional diversity’ or ‘functional
traits’ in the title or text in the category ‘Ecology’ of Web of Science in order to examine
how many traits and which trait categories are used to estimate FD in most relevant studies.
In this short review of 150 published papers on FD which are listed in Supplementary
Material A, we found that 52 % use five or fewer traits, 28.6 % use 6–10 traits, 11.3 % use
11–20 traits, and only 8 % more than 20 traits. In addition, (and even though it is not
explicitly stated) traits used in many cases are determined primarily by availability rather
than for describing a specific function, and in most cases this is the best we can do. So it is
not surprising that most studies include morphological characteristics (e.g. body mass or
body length). And it is often implicitly assumed that three, four or five traits for calculating
multidimensional FD indices is informative enough to reflect the ecological processes
operating in the species assemblage. In other words, this approach of estimating FD using
available traits data relies on an implicit assumption that the FD patterns estimated from
the available functional traits will reflect the FD pattern of the functional traits we are
interested in, but do not have data on. Or alternatively, that there is one FD spatial pattern
(produced if we consider all traits) and that all subsets of traits would produce FD patterns
that reflect this one overall pattern. This implicit assumption is often taken for granted but
has not been tested with empirical data.
As with numerous macroecological analyses of latitudinal gradients in species richness
(e.g. Hawkins et al. 2003) FD does increase with species richness, but this is not neces-
sarily a linear relationship (Petchey and Gaston 2006). The relationship between species
richness and FD has received extensive attention at fine scale studies (e.g. Bady et al. 2005;
Petchey and Gaston 2002). Still our understanding of the spatial patterns of FD at

123
 Biodivers Conserv

continental or global scales is limited, since research only recently has started investigating
the latitudinal patterns of FD and their relationship to other environmental factors (Stevens
et al. 2003; Devictor et al. 2010; Safi et al. 2011). For example, FD of New World bats
increases towards the tropics at a greater rate than the increase in species richness (Stevens
et al. 2003). Safi et al. (2011) demonstrated that mammal communities in the temperate
regions limited functional similarity, the tropical mammal communities consisted of many
more functionally similar species. Unraveling the mechanism on how the functional traits
work (single versus multiple traits or functional trait groups) still is a major challenge.
Recently, novel approaches in conservation science have identified biodiversity hotspots
using functional and phylogenetic diversity instead of species richness (Devictor et al.
2010; Huang et al. 2012; Mazel et al. 2014; Parravicini et al. 2014) considering the high
evolutionary or functional distinctiveness information that might come from species of
conservation interest (Isaac et al. 2007; Mouillot et al. 2013). This multifaceted approach
provided new and useful insights on the congruency between the different facets of
diversity (Mazel et al. 2014). For example, Mazel et al. (2014) concluded that the con-
servation of FD demands less area than that of compositional diversity due to the func-
tional redundancy observed.
Here for the first time, we test the hypothesis that different functional trait groups
(hereafter referred to as FTGs) show congruent FD spatial patterns i.e. that one FTG (a
subset of functional traits e.g. morphological, physiological etc.) could act as a surrogate
for another FTG, and/or that it will produce FD patterns that reflect the FD pattern pro-
duced by the set of all traits. Also we test the hypothesis that the FD patterns produced by
different FTGs would lead to similar inferences regarding the environmental factors that
drive the FD patterns (in our case latitude and species richness). This is the first study
examining this hypothesis with empirical data at coarse scales. We analyzed the FD of the
amphibians of Europe, a well-studied taxonomic group, to test multifaceted patterns of FD
based on different combinations of functional traits and on various FD indices. In this
methodological examination of the congruence of the spatial patterns of multidimensional
FD indices, we used three subsets of traits and the set of all available traits to calculate the
FD indices and compared and contrasted for each index the spatial pattern produced by
each trait group.

Methods

Species and trait data

We used the distribution maps provided by the Red List Assessment (IUCN 2013) for 85
amphibians (49 Anuran and 36 Urodela species) of the European Union to construct a
species richness map with 40 km 9 40 km grid cell size (this resolution according to
Hurlbert and Jetz 2007 is close to the suggested appropriate resolution when using extent
of occurrence maps). Using the species composition of each cell we calculated six FD
indices (functional dendrogram, FRic, FEve, FDis, FDiv and RaoQ; see section below for
description of FD indices) which represent the main aspects of FD (richness, evenness and
divergence) and are calculated using multiple functional traits (for a review see Petchey
and Gaston 2006; Mouchet et al. 2010). We used three main FTGs represented by 12 traits
in total, three traits related to morphology (body length, dorsal main coloration, dorsal
pattern) four to reproduction (number of eggs per clutch, duration of breeding activity,

123
Biodivers Conserv

sexual maturity, life span) and five to habitat use (spawn site, life history habit, time
partitioning, diet and mobility mode; for details see Table 1), to calculate these indices (see
next section for details of calculations). The traits represent many aspects of resource
acquisition and define important niche dimensions of the studied species. Trait information
was collated from various sources such as books, electronic databases for amphibians and
some published papers (Supplementary Material B). The selection of traits was also based
on the completeness of the availability of species’ trait data (check Trochet et al. 2014 for
data deficiency on amphibians’ traits).

Quantifying FD

A functional dendrogram index (Petchey and Gaston 2002, 2006; Podani and Schmera
2007) was computed for each grid cell from a matrix of pairwise distances between species
based on their traits and Gower distance was used throughout because it is suitable for
traits not measured on a continuous scale (Gower 1971; Podani and Schmera 2006).
UPGMA clustering was used since it produced the highest cophenetic correlation
(Blackburn et al. 2005) compared with single linkage and complete linkage (c = 0.67).

Table 1 Traits used to calculate functional diversity indices (functional dendrogram index, FRic, FEve,
FDiv, FDis and RaoQ) categorized in three functional trait groups (FTGs): Morphological, Reproductive
and Habitat-related for amphibians of Europe
Functional Trait Type Values Units/categories
trait group
(FTG)

Morphological Body length Continuous Mean Centimetre

Dorsal main Binary 8 Categories Grey, yellow, brown, white, black, red,
coloration green, olive
Dorsal pattern Categorical 3 Categories Spotted, reticular, homogenous
Reproductive Number of Continuous Mean
eggs per
clutch
Duration of Categorical 2 Categories Prolonged, explosive
breeding
activity
Sexual Continuous Mean Years
maturity
Life span Continuous Mean Years
Habitat-related Spawn site Binary 4 Categories Aquatic, terrestrial, arboreal/phytotelms,
parent
Life history Binary 3 Categories Aquatic, terrestrial, arboreal
habit
Time Binary 2 Categories Nocturnal, diurnal
partitioning
Diet Binary 6 Categories Carnivorous, insectivorous,
molluscivorous, cannibalism,
herbivorous, detritivorous
Mobility mode Binary 7 Categories Jumper, walker, runner, climber, swimmer,
crawler, glider

123
 Biodivers Conserv

The cophenetic correlation is a measure of how well the species distances are conserved in
the functional dendrogram compared with the distance matrix (Blackburn et al. 2005).
We also used the following FD indices: FRic (functional richness)—reflecting the
volume of a multidimensional trait space defined by using a convex hull volume, func-
tional evenness (FEve)—reflecting the evenness of abundance distribution in a functional
trait space, FDiv (functional divergence)—reflecting the species deviance from the mean
distance to the centre of gravity of the convex hull, FDis (functional dispersion)—re-
flecting the average distance of individual species to their group centroid in multivariate
trait space that has been defined by an appropriate distance measure and RaoQ (Rao’s
quadratic entropy)—reflecting the sum of pairwise distances between species. FEve, FDiv,
FDis and RaoQ are weighted by the relative abundance of species (Villéger et al. 2008)
which was not available in our study so we used the percentage of the grid cell occupied by
the species range as a proxy to relative abundance.
All five indices were calculated from the distribution of species in a multidimensional
functional trait space following the methods described in Villéger et al. (2008). Principal
coordinates analysis (PCoA) (Gower 1966), was used to describe the distribution of species
within the overall functional trait space defined by all species. The PCoA was used to
ordinate species along axes representing variation in associated sets of traits so that the
distances between species are approximately equal to the dissimilarities in their trait
values. FRic (Villéger et al. 2008) was calculated for each grid cell as the convex hull
volume defining the subset of functional space occupied by centroids for all species present
(but see Podani 2009; that the application of convex hull polygon should be applied with
due caution given to sample sizes). The five indices were calculated for each community
using the FD package in R (Laliberté and Shipley 2011). Each index was calculated firstly
based on the complete set of traits, and secondly for the three FTGs (Morphological,
Reproductive and Habitat-related traits) and thirdly for combinations of FTG (e.g. Mor-
phological and Reproductive).
Correlations between traits might affect reported levels functional redundancy (Petchey
and Gaston 2002). Therefore, we tested for associations among the squared distances
obtained by pairs of variables to check for correlations between the traits (Pavoine et al.
2009). Most of the correlations among the squared distances obtained by pairs of variables
were close to 0.22, with a mean of 0.32 and a standard deviation of 0.06 suggesting low
redundancy between the variables. We also performed Spearman correlations for contin-
uous traits, point biserial correlations between continuous and binary traits, Kruskal–Wallis
test for continuous and categorical trait combinations, phi coefficients for binary traits and
chi squared tests for categorical traits. We only found a significant but weak relationship
(r = 0.09; p \ 0.001) between sexual maturity and life span. Although weakly correlated
we included both traits as they capture different aspects of amphibian functionality.

Statistical analyses

We evaluated the effects of the spatial pattern of each FD index calculated using different
FTGs, by simple correlation analysis. We used the method from Clifford et al. (1989) to
calculate ‘‘effective’’ degrees of freedom and to avoid the overestimation of degrees of
freedom due to spatial autocorrelation. To test the relationship between diversity com-
ponents (species richness and functional richness) and FTGs with latitude, we performed
generalized linear mixed models that accounted for spatial dependence between samples.
These models are generalized linear models with an exponential covariance structure that
accounted for spatial autocorrelation between samples (Pinhero and Douglas 2000).

123
Biodivers Conserv

Fig. 1 The distribution of three functional diversity (FD) indices of functional richness, evenness and
dispersion (FRic, FEve and FDis) calculated using three functional trait groups (FTGs) a Morphological
(traits measured: body length, dorsal main coloration and dorsal pattern), b Reproductive (traits measured:
number of eggs per clutch, duration of breeding activity, sexual maturity and life span) c Habitat-related
(traits measured: spawn site, life-history habit, time partitioning, diet and mobility mode) and d the complete
set of traits for amphibians of European Union at 40 km 9 40 km grid cell resolution

123
 Biodivers Conserv

Therefore, all tests took into account spatial autocorrelation. We also tested the congruency
between the hotspots and coldspots of the different FTGs. In the results presented we
defined hotspots as 5 % of the grid cells with the highest FD values while coldspots as 5 %
of the grid cells with the lowest values. Using different percentages for defining hotspots
and coldspots produced qualitatively similar results (even if the exact numbers differed).
All spatial correlations between FTGs patterns and between the hotspots and coldspots of
the different FTGs were performed using SAM software (Rangel et al. 2010). Generalized
linear models and maps were performed with R statistical software version 3.0.3 (R
Development Core Team 2014) using packages nlme (Pinheiro et al. 2015) and raster
(Hijmans and Etten 2015).

Results

Congruence of FD patterns of different trait groups or lack thereof

FD patterns varied across different FTGs and FD indices (Fig. 1 and Fig. 1 on Supple-
mentary Material C). The congruence of FD spatial patterns estimated for the same
communities and for a given facet of FD but based on different functional traits showed
significant correlations but in some cases strong ones (e.g. the correlation between the
Morphological and the Reproductive FTG was 0.492 and 0.047 for FRic and FEve

Table 2 Correlation coefficients (Spearman’s r with corrected degrees of freedom using the method pro-
posed by Clifford et al. 1989) between functional trait groups (FTGs): Morphological, Reproductive,
Habitat-related and the complete set of traits used to calculate the different functional diversity measures
(functional dendrogram index, FRic, FEve, FDiv, FDis and RaoQ). Results are shown for amphibians of
Europe
Functional diversity Functional trait group Reproductive Habitat- Complete
measure (FTG) related set

FRic Morphological 0.492* 0.680*** 0.711***

Reproductive 0.445 0.653*
Habitat-related 0.795***
Functional dendrogram Morphological 0.781*** 0.970*** 0.985***
Reproductive 0.830*** 0.814***
Habitat-related 0.984***
Morphological 0.047 0.143*** 0.135***
FEve Reproductive -0.013*** 0.28**
Habitat-related -0.233**
FDiv Morphological -0.008* -0.332*** 0.214
Reproductive -0.515*** 0.411***
Habitat-related -0.203*
FDis Morphological 0.649*** 0.474*** 0.714***
Reproductive 0.761*** 0.937***
Habitat-related 0.874***
Morphological 0.633*** 0.433*** 0.673***
Reproductive 0.774*** 0.942***
RaoQ Habitat-related 0.878***

* p \ 0.05; ** p \ 0.01; *** p \ 0.001

123
Biodivers Conserv

respectively). Spatial Spearman correlations between the FD of each FTG and the complete
set of traits (Table 2) differed significantly between the different aspects of FD (richness–
evenness-dispersion). FD indices that reflect the volume of functional space occupied
displayed significantly strong correlations in most cases, especially for the functional
dendrogram index. But even in this group, there were weak to non-significant correlations
when comparing between the spatial pattern of FRic estimated with reproductive traits and
all other FRic spatial patterns (Table 2). Significant but of intermediate strength correla-
tions were observed for the indices quantifying functional dispersion (FDis and RaoQ),
with spatial Spearman rank correlation coefficient ranging between 0.433 and 0.942
(Table 2). However in the case of the FD indices quantifying the evenness and divergence
in the distribution of traits (FEve and FDiv) things were very different. In some cases the
correlations were not significant, but more intriguingly in other cases there were significant
but negative correlations (Table 2). For example, the increase of functional divergence
estimated with habitat-related traits was associated with a decrease of functional diver-
gence estimated with reproductive traits (rs = -0.515, p \ 0.05).
We repeated the analysis using combinations of FTGs. The correlations between FD
patterns calculated with these combination and FD pattern calculated with the complete set
of traits differed depending on the aspect of FD analyzed, but generally followed the same
patterns as above (for more detailed presentation of these results see Figs. 1–2 on Sup-
plementary Material D).
In addition to the statistical correlations we examined how well the different FTGs
could lead to the identification of the FD hotspots or coldspots across EU. In general, there
was lower spatial congruence between different FTGs hotspots than the corresponding
congruence of coldspots (Table 1 of Supplementary Material E). Low spatial overlap was
found between Morphological and the complete set of traits but also with Reproductive

Fig. 2 Functional diversity (FD) indices of functional richness and evenness (FRic and FEve) as a function
to species richness, calculated using three functional trait groups (FTGs): Morphological, Reproductive,
Habitat-related and the complete set of traits for amphibians of European Union. Box plots show 10th and
90th percentile, quartile and the median value (black line)

123
 Biodivers Conserv

(2.83 %) FTGs of FRic. Similarly between Morphological and the Reproductive FTGs of
FDiv there was total mismatch (0 %). In contrast, the functional dendrogram index showed
a 74 % hotspot congruency between the Morphological FTG and the complete set of traits
pattern. On average, richness and dispersion functional indices showed low mismatches in
coldspots particularly between Habitat FTG and the complete set of traits (Supplementary
Material E).

Inference regarding drivers of FD patterns estimated from different FTGs

So given the limited congruence of the FD patterns we investigated if the FD patterns of

different FTGs relationship with environmental drivers led to similar or contrasting con-
clusions. Different aspects of FD displayed different latitudinal gradients. We could group
FD indices regarding their latitudinal gradients in three groups reflecting three aspects of
FD: richness (functional dendrogram index–FRic), evenness (FEve–FDiv) and dispersion
(FDis–RaoQ). Functional richness indices consistently showed negative associations with
latitude, but the strength of the relationship depended on the FTG analyzed, with the
weakest and non significant relationship recorded for the Reproductive FTG of FRic
(-0.002 ± 0.0006; parameter estimate ±standard error, r2 = 0.718, p = 0.093) and the
strongest and significant for the Morphological FTG of FRic (-0.0021 ± 0.002,
r2 = 0.946, p \ 0.05). Latitude was strongly and significantly related to all FTGs of
functional dispersion indices. Although, the strongest relation was recorded for the
Reproductive FTG of FDis (-0.006 ± 0.0012, r2 = 0.969, p = 0.001). Interestingly,
functional evenness displayed contradictory latitudinal gradients, leading to all possible
outcomes. Spatial patterns of FEve based on either the complete set of traits or on the
Reproductive FTG were positively and significantly associated with latitude
(0.0004 ± 0.0004, r2 = 0.821, p \ 0.01 and 0.0006 ± 0.0011, r2 = 0.815, p \ 0.001
respectively). However, FEve patterns based on the Morphological FTG and the Habitat-
related FTG followed the opposite trend showing non-significant but negative correlations
(-0.0003 ± 0.0012, r2 = 0.803, p = 0.833 and -0.0007 ± 0.002, r2 = 0.889, p = 0.668
respectively) (for more details see Table 1 on Supplementary Material F). As a result, there
is a trend for higher functional richness and dispersion at species-rich regions, while
functional evenness and divergence increased at higher latitudes (Fig. 1).
The importance of species richness as a driver of FD patterns again depended strongly
on the aspect of FD analyzed. Generally, relationships with species richness were stronger
for the richness and dispersion aspects of FD (functional dendrogram, FRic, FDis, RaoQ)
while weaker for evenness and divergence (FEve, FDiv). Generalized mixed models with
spatial structure indicated that between the different aspects of FD based on the complete
set of traits and species richness, there was a positive relation which was higher for
functional richness aspect (e.g. for functional dendrogram index: p \ 0.0001, r2 = 0.95,
F = 33326.31; for FRic: p \ 0.0001, r2 = 0.744, F = 2317.12) while for functional
evenness was very low (for FEve: p \ 0.0001, r2 = 0.009, F = 4.56; for FDiv:
p \ 0.0001, r2 = 0.006, F = 1.77). The results showed that 55–60 % of the variation of
the functional dispersion aspect remained unexplained by species richness (for FDis:
p \ 0.0001, r2 = 0.445, F = 972.14; for RaoQ: p \ 0.0001, r2 = 0.414, F = 1084.16).
However for each FD index, the strength and form of the relationship was influenced by the
type of traits used to calculate the index. For the complete set of traits as well as for
morphological and habitat–related traits the relationship between species richness and FRic
is monotonical increase (with r2 ranging between 0.66 and 0.736, p \ 0.0001) but for
reproductive traits the relationship is weaker (r2 = 0.13, p \ 0.0001) and appears to reach

123
Biodivers Conserv

the highest FRic values, even from low species richness value (Fig. 2). Although, all the
FTGs of FEve and FDiv reached their higher values from low species richness (Fig. 2), still
weakly related with species richness (r2 ranging between 0.01 and 0.365, p \ 0.0001).
Interestingly, dispersion indices showed similar patterns with evenness indices (Table 1 on
Supplementary Material G).

Discussion

Our results showed that, in certain cases, different sets of traits used in the calculation of
FD indices lead to different FD patterns for the same FD index and the same species
assemblage. Thus our findings do not support the hypothesis that using one subset of
functional traits to estimate FD patterns will reflect the FD pattern of all other sets of
functional traits. And more interestingly, the spatial pattern produced using more traits
does not necessarily reflect the spatial pattern of the subsets of traits. For FD indices that
are strongly related to species richness different sets of traits produce FD spatial patterns
that are significantly—but only in some cases weakly—correlated, and the FD hotspots for
the different sets of traits are in different geographic locations across European Union. But
more importantly, for FD indices (like evenness) that are not related to species richness the
spatial patterns of FD of different sets of traits are either statistically independent or even
negatively correlated. This has significant impact not only for identifying FD hotspots, but
also for inferring the role of different environmental factors as drivers of FD, and corre-
spondingly on inferring the role of FD as driver of ecosystem services and functions. In
simple words, we found that, for a given taxon, different species traits produce FD patterns
that vary considerably in space.
Different facets of FD display very different patterns and lead to different conclusions
(e.g. Mouchet et al. 2010; Karadimou et al. 2015), something that has been verified in our
study also. However, previous studies use only one set of species traits to estimate FD and
are based on the implicit assumption that there is only one FD pattern and that different
indices capture different aspects of it. Our results indicate that this is not the case. Different
sets of traits used in the calculation of FD indices lead to strikingly different FD patterns
for the same FD index. The non-congruence among these patterns is highly influenced by
the FD index used. The facets of FD that quantify the volume of functional space occupied
(like FRic and functional dendrogram index) are strongly related to species richness (as is
often documented) and display rather strong significant correlations; thus in their case the
common practice of using the set of available traits as a surrogate for other traits is partially
justifiable. But when analyzing other facets of FD (and especially the ones quantifying the
evenness in the distribution of traits, like FEve and FDiv) there is no congruence, and
generalizations about FD patterns should be done with caution.
Previous research highlights the sensitivity of FD indices to measuring the traits of all
the species in the community (Pakeman 2014b). These studies using the same traits dataset
and omitting some rare species of the community document how considerably the value of
FD changed. Here, we kept the species composition intact (and collected trait data for all
species in it, common and rare) but calculated the FD indices for sets of different species
traits, and thus obtained considerable changes. These changes refer not only to the value of
the FD index calculated, but more importantly to the spatial pattern of FD across Europe,
i.e. where the peaks and troughs are located.

123
 Biodivers Conserv

Conservation planning is often based on identifying biodiversity hotspots (Mazel et al.

2014; Parravicini et al. 2014). We demonstrated that the use of some traits to define FD
hotspots and coldspots may identify different locations than using different traits and yet
different locations from the complete set of traits. Again the sensitivity of this mismatch
depends on the FD index used, with lower mismatches for functional dendrogram index,
FRic, FDis and RaoQ and greater mismatches for FEve and FDiv. These results are
congruent with those of a recent study revealing important mismatches between taxonomic
and functional richness and also their hotspots (Devictor et al. 2010; Huang et al. 2012;
Mazel et al. 2014). But our results go one step further and show that even for FD the
hotspots are different depending on the species traits analyzed. So we argue that, for the
same taxonomic group, different functions will have different hotspots and would demand
different management plans to preserve different functions. So the argument that FD is
easier to preserve in less area due to the redundancy of traits may hold true for a specific
function, but these areas will not necessarily be the same for different functions. Since
hotspots of different aspects of biodiversity often show low spatial congruence (Orme et al.
2005; Stuart-Smith et al. 2013) we further revealed that different facets of FD but even
more different trait groups in the same FD facet could propose different priority areas. It
has been suggested that identifying areas of high FD could be a way of setting conservation
priorities in order to maintain ecosystem processes and services (Devictor et al. 2010). Our
findings provide interesting questions on which FD hotspots (based on different FTGs)
could be also implemented in monitoring programs or national priorities (such as those
proposed by Schmeller et al. 2014, 2015).
Indeed, it has been assumed that the loss of particular combinations of functional traits
is often related with FD loss, thus might affect ecosystem functioning while the loss of a
given single species might not be as ecologically damaging, as other species which they
persist have functional analogues with (Loreau et al. 2002; Srivastava et al. 2012). Since
hotspots of different aspects of FD often show low spatial congruence, we would rec-
ommend that is necessary to consider a suite of different functional traits to adequately
conserve functional biodiversity, and that this should be done case by case for each
ecosystem function of interest, since different functions may have very different FD
hotspots. Furthermore, we caution that the use of all traits to obtain a single set of FD
hotspots may be ill advised, since it might miss the hotspots of certain functions.
One big issue of biodiversity research is trying to infer the environmental drivers of
diversity patterns. Our results show that this task is more complicated for FD than for
indices of compositional diversity (like species richness). For any given FD index, different
sets of traits led to lesser or greater differences in the correlations with the two factors
examined (in our case, we examined latitude and species richness). Some indices, like FRic
and functional dendrogram index, showed consistent correlations and similar patterns for
all FTGs examined. While for other indices like FEve, we found that the relationship
between FD and latitude may be significantly positive and not significantly negative
depending on the FTG used. This makes inference difficult and generalizations ques-
tionable. While Safi et al. (2011) have proposed that the latitudinal gradient of the
mammalian FD could be partly explained by temperature seasonality here the amphibians
FD based on different sets of traits do not show a clear latitudinal pattern and it may prove
that any single predictor is not possible to explain the latitudinal patterns.
This non-congruence of FD spatial patterns produced by different species traits raises
another important point. There is an ongoing research trying to elucidate the role of FD on
ecosystem functioning and services (among others Dı́az et al. 2007). The reason why no
clear pattern has emerged might be due not to the lack of a relationship, but due to the use

123
Biodivers Conserv

of inappropriate traits to estimate FD. If we use different traits or more interestingly if we

use more traits that are not related to the function of interest this is likely to produced FD
spatial patterns that significantly differ from the FD pattern of interest and obscure the role
of FD on shaping ecosystem services and functions. This pattern might be the reason why
Pakeman (2014a) found that productivity is highly dependent on one leaf trait (leaf dry
matter content) and weakly related to the multidimensional FD indices calculated using
this and other traits.
This is the first study that examines the ‘‘cross-function’’ congruence of FD patterns. So
before generalizations could be made, there is a need to verify these findings for different
taxa, in different biogeographic regions and scales and more importantly for different
functional traits. And even though this analysis is based on only one taxon (amphibians),
we argue that this non-congruence of FD spatial patterns from different traits might prove a
general motive observed in other taxa (e.g. mammals, birds or even plants and microbes)
and scales. In this analysis we used a relatively well known taxon (amphibians) in a well
studied geographic region (Europe) and we still had difficulty to collect many species
traits. Given the results of Pakeman (2014b) about the importance of analyzing all species
in the community, we only used traits that we had information for most if not all of the
species in our study. However, different sets of traits may have different results. Con-
cludingly, we found that different functional traits for the same community may lead to
very different patterns of FD. The assumption that one set of traits could be used as a
surrogate for another is justifiable only for specific sets of traits and for specific FD indices.
But more importantly our results show that FD is not one thing but many different things
depending on the function examined. So what may appear as functional redundancy for a
specific function may not be so for a different function. Also this result may explain the
difficulty to locate the link between biodiversity and ecosystem functioning. Since for
different functions the same biodiversity pattern may result in multiple very different FD
patterns. As a way forward, we suggest that the traits selected for estimating FD should be
those that are already shown to be important response traits for the function of interest (for
example see Pakeman 2011). If that approach is not possible (either because we do not
have data on those traits or do not know which the important response traits), then we
suggest to estimate the FD using different sets of traits and iteratively examine the patterns
produced to find out which traits produce the strongest relation to the phenomenon of
interest and not to be satisfied with only one pattern.

Acknowledgments This work was supported by Grant E.042 from the Research Committee of the
University of Patras (Programme K. Karatheodori 2013). We thank two anonymous reviewers and the editor
for constructive comments on our manuscript.

References
Bady P, Doledec S, Fesl C, Gayraud S, Bacchi M, Scholl F (2005) Use of invertebrate traits for the
biomonitoring of European large rivers: the effects of sampling effort on genus richness and functional
diversity. Freshwater Biol 50:159–173
Blackburn TM, Petchey OL, Cassey P, Gaston KJ (2005) Functional diversity of mammalian predators and
extinction in island birds. Ecology 86:2916–2923
Bremner J, Rogers SI, Frid CLJ (2003) Assessing functional diversity in marine benthic ecosystems: a
comparison of approaches. Mar Ecol Prog Ser 254:11–25
Brown JH (1995) Macroecology. University of Chicago Press, Chicago

123
 Biodivers Conserv

Chillo V, Anand M, Ojeda RA (2011) Assessing the use of functional diversity as a measure of ecological
resilience in arid rangelands. Ecosystems 14:1168–1177
Clifford P, Richardson S, Hémon D (1989) Assessing the significance of the correlation between two spatial
processes. Biometrics 45:123–134
Devictor V, Mouillot D, Meynard C, Jiguet F, Thuiller W, Mouquet N (2010) Spatial mismatch and
congruence between taxonomic, phylogenetic and functional diversity: the need for integrative con-
servation strategies in a changing world. Ecol Lett 13:1030–1040
Dı́az S, Cabido M (2001) Vive la difference, plant functional diversity matters to ecosystem processes.
Trends Ecol Evol 16:646–655
Dı́az S, Lavorel S, de Bello F, Quétier F, Grigulis K, Robson TM (2007) Incorporating plant functional
diversity effects in ecosystem service assessments. PNAS 104:20684–20689
Dumay O, Tari PS, Tomasini JA, Mouillot D (2004) Functional groups of lagoon fish species in Languedoc
Roussillon, southern France. J Fish Biol 64:970–983
Dunning JB (2007) CRC handbook of avian body masses. Taylor and Francis
Flynn DFB, Mirotchnick N, Jain M, Palmer MI, Naeem S (2011) Functional and phylogenetic diversity as
predictors of biodiversity–ecosystem-function relationships. Ecology 92:1573–1581
Fritz SA, Purvis A (2010) Phylogenetic diversity does not capture body size variation at risk in the world’s
mammals. Proc R Soc Biol Sci Ser B 277:2435–2441
González-Suárez M, Lucas PM, Revilla E (2012) Biases in comparative analyses of extinction risk, mind the
gap. J Anim Ecol 81:1211–1222
Gower JC (1966) Some distance properties of latent root and vector methods used in multivariate analysis.
Biometrika 53:325–338
Gower JC (1971) A general coefficient of similarity and some of its properties. Biometrics 27:857–874
Hadfield JD (2008) Estimating evolutionary parameters when viability selection is operating. Proc R Soc
Biol Sci Ser B 275:723–734
Hawkins BA, Porter EE, Diniz-Filho JaF (2003) Productivity and history as predictors of the latitudinal
diversity gradient of terrestrial birds. Ecology 84:1608–1623
Hijmans RJ, van Etten J (2015) raster: geographic analysis and modeling with raster data. R package version
2. 4–20. https://cran.r-project.org/web/packages/raster/index.html. Accessed 01.01.2015
Hooper DU, Solan M, Symstad AJ, Dı́az S, Gessner MO, Buchmann N, Degrange V, Grime P, Hulot F,
Mermillod-Blondin F, Roy J, Spehn J, van Peer L (2002) Species diversity, functional diversity and
ecosystem functioning. In: Loreau M, Naeem S, Inchausti P (eds) Biodiversity and ecosystem func-
tioning, syntheses and perspectives. Oxford University Press, pp 195–208
Hooper DU, Chapin FS, Ewel JJ, Hector A, Inchausti P, Lavorel S, Lawton JH, Lodge DM, Loreau M,
Naeem S, Schmid B, Setala H, Symstad AJ, Vandermeer J, Wardle DA (2005) Effects of biodiversity
on ecosystem functioning: a consensus of current knowledge. Ecol Monogr 75:3–35
Huang S, Stephens PR, Gittleman JL (2012) Traits, trees and taxa, global dimensions of biodiversity in
mammals. Proc R Soc Biol Sci Ser B 279:4997–5003
Hubbell SP (2001) The unified neutral theory of biodiversity and biogeography. Princeton University Press,
Princeton NJ
Hurlbert AH, Jetz W (2007) Species richness, hotspots, and the scale dependence of range maps in ecology
and conservation. Proc Natl Acad Sci USA 104:13384–13389
Isaac NJB, Turvey ST, Collen B, Waterman C, Baillie JEM (2007) Mammals on the EDGE: conservation
priorities based on threat and phylogeny. PLoS ONE 2:e296
IUCN (2013) The IUCN red list of threatened species. Version 2013. 1 http://www.iucnredlist.org. Accessed
01 August 2013
Jones KE, Bielby J, Cardillo M, Fritz SA, O’Dell J, Orme CDL, Safi K, Sechrest W, Boakes EH, Carbone C,
Connolly C, Cutts MJ, Foster JK, Grenyer R, Habib M, Plaster CA, Price SA, Rigby EA, Rist J,
Teacher A, Bininda-Emonds ORP, Gittleman JL, Mace GM, Purvis A (2009) PanTHERIA: a species-
level database of life history, ecology, and geography of extant and recently extinct mammals. Ecology
90:2648
Karadimou E, Tsiripidis I, Kallimanis AS, Raus T, Dimopoulos P (2015) Functional diversity reveals
complex assembly processes on sea-born volcanic islands. J Veg Sci 26:501–512
Kraft NJB, Godoy O, Levine M (2015) Plant functional traits and the multidimensional nature of species
coexistence. Proc Natl Acad Sci USA 112:797–802
Laliberté E, Shipley B 2011 Package ‘FD’, measuring functional diversity FD from multiple traits, and other
tools for functional ecology. http://cran.r-project.org/web/packages/FD/FD.pdf. Accessed 25 Nov 2014
Lepš J, de Bello F, Lavorel S, Berman S (2006) Quantifying and interpreting functional diversity of natural
communities: practical considerations matter. Preslia 78:481–501

123
Biodivers Conserv

Loreau M, Naeem S, Inchausti P (2002) Biodiversity and ecosystem functioning: synthesis and perspectives.
Oxford University Press, New York
Mazel F, Guilhaumon F, Mouquet N, Devictor V, Gravel G, Renaud J, Cianciaruso MV, Loyola R, Diniz-
Filho JAF, Mouillot D, Thuiller W (2014) Multifaceted diversity-area relationships reveal global
hotspots of mammalian species, trait and lineage diversity. Glob Ecol Biogeogr 23:836–847
McGill BJ, Enquist BJ, Weiher E, Westoby M (2006) Rebuilding community ecology from functional traits.
Trends Ecol Evol 21:178–185
Mouchet MA, Villéger S, Mason NWH, Mouillot D (2010) Functional diversity measures, an overview of
their redundancy and their ability to discriminate community assembly rules. Funct Ecol 24:867–876
Mouillot D, Bellwood DR, Baraloto C, Chave J, Galzin R, Harmelin-vivien M, Kulbicki M, Lavergne S,
Lavorel S, Mouquet N, Paine TCE, Renaud J, Thuiller W (2013) Rare species support vulnerable
functions in high-diversity ecosystems. PLoS Biol 11(5):e1001569
Nakagawa S, Freckleton RP (2008) Missing inaction, the dangers of ignoring missing data. Trends Ecol
Evol 23:592–596
Olson VA, Davies RG, Orme CDL, Thomas GH, Meiri S, Blackburn TM, Gaston KJ, Owens IPF, Bennett
PM (2009) Global biogeography and ecology of body size in birds. Ecol Lett 12:249–259
Orme CDL, Davies RG, Burgess M, Eigenbrod F, Pickup N, Olson VA, Webster AJ, Ding TS, Rasmussen
PC, Ridgely RS, Stattersfield AJ, Bennett PM, Blackburn TM, Gaston KJ, Owens IPF (2005) Global
hotspots of species richness are not congruent with endemism or threat. Nature 436:1016–1019
Pakeman RJ (2011) Functional diversity indices reveal the impacts of land use intensification on plant
community assembly. J Ecol 99:1143–1151
Pakeman RJ (2014a) Leaf dry matter content predicts herbivore productivity, but its functional diversity is
positively related to resilience in grasslands. PLoS ONE 9(7):e101876
Pakeman RJ (2014b) Functional trait metrics are sensitive to the completeness of the species’ trait data.
Methods Ecol Evol 5:9–15
Parravicini V, Villéger S, McClanahan TR, Arias-González JE, Bellwood DR, Belmaker J, Chabanet P,
Floeter SR, Friedlander AM, Guilhaumon F, Vigliola L, Kulbicki M, Mouillot D (2014) Global
mismatch between species richness and vulnerability of reef fish assemblages. Ecol Lett 17:1101–1110
Pavoine S, Vallet J, Dufour AB, Gachet S, Daniel H (2009) On the challenge of treating various types of
variables, application for improving the measurement of functional diversity. Oikos 118:391–402
Petchey OL, Gaston KJ (2002) Functional diversity (FD), species richness and community composition.
Ecol Lett 5:402–411
Petchey OL, Gaston KJ (2006) Functional diversity: back to basics and looking forward. Ecol Lett
9:741–758
Philpott SM, Soong O, Lowenstein JH, Pulido AL, Lopez DT, Flynn DFB, Declerck F (2009) Functional
richness and ecosystem services, bird predation on arthropods in tropical agroecosystems. Ecol Appl
19:1858–1867
Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team (2015) nlme: linear and nonlinear mixed effects
models. R package version 3.1–122. https://cran.r-project.org/web/packages/nlme/index.html. Acces-
sed 01.01.2015
Pinhero JC, Douglas MB (2000) Mixed-effects models in S and S-PLUS. Springer, New York
Podani J (2009) Convex hulls, habitat filtering and functional diversity: mathematical elegance versus
ecological interpretability. Community Ecol 10:244–250
Podani J, Schmera D (2006) On dendrogram-based measures of functional diversity. Oikos 115:179–185
Podani J, Schmera D (2007) How should a dendrogrambased measure of functional diversity function? A
rejoinder to Petchey and Gaston. Oikos 116:1427–1430
R Development Core Team 2014 R: a language and environment for statistical computing-version 3.0.3. R
foundation for statistical computing, Austria. http://www.r-project.org/. Accessed 25 Nov 2014
Rangel TFLVB, Diniz-Filho JAF, Bini LM (2010) SAM: a comprehensive application for spatial analysis in
macroecology. Ecography 33:46–50
Rosenzweig ML (1995) Species diversity in space and time. Cambridge University Press, Cambridge
Safi K, Cianciaruso MV, Loyola RD, Brito D, Armour-Marshall K, Diniz-Filho JAF (2011) Understanding
global patterns of mammalian functional and phylogenetic diversity. Proc R Soc Biol Sci Ser B
366:2536–2544
Scherer-Lorenzen M (2008) Functional diversity affects decomposition processes in experimental grass-
lands. Funct Ecol 22:547–555
Schmeller DS, Evans D, Lin Y-P, Henle K (2014) The national responsibility approach to setting conser-
vation priorities—recommendations for its use. J Nat Conserv 22:349–357
Schmeller DS, Juillard R, Bellingham PJ, Böhm M, Brummitt N, Chiarucci A, Couvet D, Elmendorf S,
Forsyth DM, Garcia-Moreno J, Gregory RD, Magnusson WE, Martin LJ, McGeoch MA, Mihoub J-B,

123
 Biodivers Conserv

Pereira HM, Proença V, Van Swaay C, Yahara T, Belnap J (2015) Towards a global terrestrial species
monitoring program. J Nat Conserv 25:51–57
Schmitz OJ (2009) Effects of predator functional diversity on grassland ecosystem function. Ecology
90:2339–2345
Srivastava DS, Cadotte MW, MacDonald AAM, Marushia RG, Mirotchnick N (2012) Phylogenetic
diversity and the functioning of ecosystems. Ecol Lett 15:637–648
Stevens RD, Cox SB, Strauss RE, Willig MR (2003) Patterns of functional diversity across an extensive
environmental gradient: vertebrate consumers, hidden treatments and latitudinal trends. Ecol Lett
6:1099–1108
Stuart-Smith RD, Bates AE, Lefcheck JS, Duffy JE, Baker SC, Thomson RJ, Stuart-Smith JF, Hill NA,
Kininmonth SJ, Airoldi L, Becerro MA, Campbell SJ, Dawson TP, Navarrete SA, Soler GA, Strain
EMA, Willis TJ, Edgar GJ (2013) Integrating abundance and functional traits reveals new global
hotspots of fish diversity. Nature 501:539–542
Tilman D (2001) Functional diversity. In: Levin SA (ed) Encyclopaedia of biodiversity, 2nd edn. Academic
press, pp. 109–120
Trochet A, Moulherat S, Calvez O, Stevens V, Clobert J, Schmeller D (2014) A database of life-history traits
of European amphibians. Biodivers Data J 2:e4123
Villéger S, Mason NWH, Mouillot D (2008) New multidimensional functional diversity indices for a
multifaceted framework in functional ecology. Ecology 89:2290–2301
Wilman H, Belmaker J, Simpson J, de la Rosa C, Rivadeneira MM, Jetz W (2014) EltonTraits 1.0: species-
level foraging attributes of the world’s birds and mammals. Ecology 95:2027–2027

123