Impacts of environmental ﬁlters on functional
redundancy in riparian vegetation
*, Cayetano Guti
, David S
and Christer Nilsson
Departamento de Ecolog
ıa e Hidrolog
ıa, Facultad de Biolog
ıa, Universidad de Murcia, Campus de Excelencia
Internacional Regional ‘Campus Mare Nostrum’, 30100 Murcia, Spain;
Catchment Research Group, School of
Biosciences, Cardiff University, The Sir Martin Evans Building, Museum Avenue, Cardiff CF10 3AX, UK;
Departamento de Ecologı
a de Humedales, Estaci
ogica de Do
nana (CSIC), Av. Am
erico Vespuccio, 41092
Institut de Biologia Evolutiva (IBE, CSIC-UPF), Passeig mar
ıtim de la Barceloneta 37-49, 08003
Barcelona, Spain; and
Department of Ecology and Environmental Sciences, Landscape Ecology Group, Ume
University, SE-901 87 Ume
1. Understanding and predicting ecosystem responses to multiple environmental pressures is a
long-standing interest in ecology and environmental management. However, few studies have
examined how the functional features of freshwater biological communities vary along multiple
gradients of environmental stress. Furthermore, modelling these functional features for a whole
river network constitutes a strong potential basis to improve ecosystem management.
2. We explored how functional redundancy of biological communities (FR, a functional fea-
ture related to the stability, resistance and resilience of ecosystems) responds to single and
multiple environmental ﬁlters. We compared these responses with those of functional richness,
evenness and divergence. We used riparian vegetation of a Mediterranea n basin, and three of
the main environmental ﬁlters affecting freshwater communities in such regions, that is
drought, ﬂow regulation and agricultural intensity, thus considering the potential effect of
natural environmental variability. We also assessed the predictability of FR and estimated it
for the entire river network.
3. We found that all functional measures decreased with increasing environmental ﬁlter
intensity. However, FR was more sensitive to single and multiple environmental ﬁlters com-
pared to other functional measures. The best-ﬁtting model explained 59% of the FR variabil-
ity and included agriculture, drought and ﬂow regulation and the pairwise interactions of
agriculture with drought and ﬂow regulation. The parameters of the FR models differed from
null model expectations reﬂecting a non-random decline along stress gradients.
4. Synthesis and applications. We found non- random detrimental effects along environmental
ﬁlters’ gradients for riparian functional redundancy (the most sensitive functional index),
meaning that increased stress could jeopardize stability, resistance and resilience of these sys-
tems. In general, agriculture caused the greatest impact on functional redundancy and func-
tional diversity measures, being the most important stressor for riparian functionality in the
study area. Temporary streams ﬂowing through an agricultural, regulated basin had reduced
values of functional redundancy, whereas the free-ﬂowing medium-sized, perennial water
courses ﬂowing through unaltered sub-basins displayed higher values of functional redun-
dancy and potentially greater stability against human impacts. All these ﬁndings along with
the predicted basin-wide variation of functional redundancy can assist environmental man-
agers in improving monitoring and ecosystem management.
Key-words: drought, ﬂow regulation, functional diversity, functional traits, global change,
habitat ﬁltering, land use, Mediterranean rivers, multiple stressors, plant functional groups
*Correspondence author. E-mail: email@example.com
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society
Journal of Applied Ecology 2016, 53, 846–855 doi: 10.1111/1365-2664.12619
The world’s ecosystems are experiencing an increase in
human impacts causing an unprecedented biodiversity
loss. These changes may alter the functioning of ecosys-
tems and jeopardize the goods and services provided to
humanity (Mouillot et al. 2013). Consequently, predicting
ecosystem responses to multiple human pressures and
interacting natural ﬁlters has become one of the most
challenging tasks for scientists in order to guide conserva-
tion efforts and the management of ecological resources.
Traditionally, ecologists have focused on the response
of the taxonomic community structure to different types
of disturbances. Mitigation of the ecological consequences
of environmental change, however, requires a deeper
understanding of the relationship between biodiversity
and ecosystem functioning (Cardinale et al. 2012). During
the last decade, there has been a growing development of
trait-based approaches to explore the effects of human
activity on ecosystem functioning (Clapcott et al. 2010;
e et al. 2010; Mouillot et al. 2013). Thus, as the
combination of species traits determines the likelihood
that a species can overcome environmental ﬁlters (Kraft,
Godoy & Levine 2015), a non-random species sorting
along environmental gradients is expected (Weiher et al.
2011; Mouillot et al. 2013).
Trait-based approaches allow the estimation of many
components of functional diversity (FD), such as functional
richness, evenness, divergence (Mason et al. 2005 for a
review) and functional redundancy (FR, Fonseca &
Ganade 2001; Rosenfeld 2002; Lalibert
e & Legendre 2010).
Among them, FR is one of the most promising functional
indices since it relates positively to stability, resistance and
resilience of ecosystems (Hooper et al. 2005; Guillemot
et al. 2011). It represents the number of species contribut-
ing similarly to an ecosystem function (Walker 1992; Law-
ton & Brown 1993). Although the notion of redundancy
suggests that functionally similar species may compensate
for the loss or failure of others, there is evidence that
ecosystems need such redundancy to perform their func-
tions efﬁciently and stably over time (Rosenfeld 2002;
Guillemot et al. 2011; Biggs et al. 2012). In fact, a decrease
in FR could be dramatic in non-redundant communities
since the loss or replacement of one species would lead to
loss of unique traits or functions (Hooper et al. 2005),
increasing ecosystem vulnerability (Elmqvist et al. 2003).
We focus especially on the response of FR (but also
considering FD components such as functional richness,
evenness and divergence) to the main environmental
ﬁlters in Mediterranean rivers using riparian trees and
shrubs as model organisms. Riparian vegetation is a key
component in the functioning of freshwater ecosystems
(Hladyz et al. 2011) and provides essential functions,
goods and services such as organic matter supply (Wood-
ward et al. 2012), sediment retention (Tabacchi et al.
2000), and food and shelter for numerous animals (Sabo
& Power 2002). Riparian communities are taxonomically
well studied and species trait information is usually avail-
able, which allows the estimation of functional features.
These ecosystems have well-deﬁned, multifunctional and
species-rich vegetation that enables the detection of func-
tional responses even to minor impacts (Nilsson & Sved-
mark 2002; Aguiar et al. 2009). However, very few
studies have examined how the functional features of
freshwater ecosystems vary along gradients of environ-
mental ﬁlters (e.g. Clapcott et al. 2010; Matsuzaki,
Sasaki & Akasaka 2013) and if they can act synergisti-
cally to affect ecosystem resilience and stability (Sasaki
et al. 2015).
We used a data base of woody riparian plants from a
semi-arid Mediterranean catchment (Segura River) to
explore how the main environmental ﬁlters (i.e. pre-
dictable seasonal drought, agriculture and ﬂow regula-
tion), as well as their interactions, may impact the FR
(and other FD indices) of riparian communities. As it is
probable to ﬁnd a relationship between environmental ﬁl-
ters and functional measures simply as a consequence of
an underlying taxonomic richness gradient (Vill
Mason & Mouillot 2008), we also check for non-random-
ness of the empirical response patterns. Finally, we fore-
cast the values of FR for the whole river network as a
basis for ecosystem management. We expect that environ-
mental ﬁlters would reduce the value of measures of func-
tional features and that FR response should be
predictable from large-scale geographical variables. Mod-
elling FR in entire basins in response to environmental ﬁl-
ters could assist decision-makers in setting goals and
designing strategies for conservation and restoration of
Materials and methods
The Segura River basin (SE Spain, Fig. 1) is highly heteroge-
neous (Bruno et al. 2014a), making it ideal to represent other
areas with Mediterranean or semi-arid climates around the world.
The climate ranges from sub-humid in the mountains in the
north-west, where the rivers have relatively stable ﬂows and high
discharges, to semi-arid in the south-eastern lowlands where
streams show more variable ﬂows and lower mean discharge, fea-
turing intermittent streams subject to variable summer drought
(Belmar, Velasco & Martinez-Capel 2011).
The intense expansion of agricultural land (currently 521%
of the entire basin; Fig. 1), especially the parts irrigated during
the last 25 years, resulted in a reduction of natural and semi-
natural areas, now representing 452% of the basin extent. As a
consequence, there is an intense ﬂow regulation, leading to
widespread hydromorphological alterations. In contrast to the
areas currently impacted by agriculture or hydrological alter-
ations, this basin still holds an important number of rivers with
a good ecological status, the study of which allows an assess-
ment of human impacts on biological communities (Bruno et al.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology, 53, 846–855
Impacts of stressors on functional redundancy 847
We selected 71 freshwater river reaches with varying land-use
intensity, ﬂow regulation and ﬂow persistence accounting also for
the natural environmental variability through elevation in the
Segura basin (Fig. 1). Each locality was sampled once between
2010 and 2012 during late spring and summer along 500-m long
reaches at both riversides, as this period is the most suitable for
single surveys (Ferreira & Aguiar 2006). Within these 500-m long
reaches, we noted the presence of woody riparian species, from
the low-water margin up to the natural bankfull limit through
ten transects, thus obtaining a list of species for each locality. We
estimated the species abundance in a semi-quantitative way,
that is three abundance classes according to species dominance
(dominant, frequent and present).
We used a wide set of biological traits in order to capture the entire
range of functions and responses of the riparian plants recorded.
We gathered a total of 30 continuous, semi-continuous and categor-
ical biological effect and response traits to characterize the func-
tional features of the species recorded (Lavorel & Garnier 2002;
Cornelissen et al. 2003; Appendix S1, Supporting information).
Functional effect traits are those biological features that directly
inﬂuence a speciﬁc function of the ecosystem (e.g. primary produc-
tivity, nutrient cycling) while the response traits change according
to the abiotic and biotic environment (e.g. resource availability,
climatic conditions and disturbance regime; D
ıaz & Cabido 2001).
Species-speciﬁc mean trait values were compiled from 59 online
trait data bases and scientiﬁc publications (Table S1.3 in
Appendix S1). The ﬁnal trait data set is found in Appendix S2.
We constructed matrices of taxon counts by site and traits by
taxon to estimate the functional features of riparian communities.
The estimation of the functional components is in continuous
evolution so there are a variety of methodologies to estimate FR
and a lack of consensus about them. Thus, FR was obtained for
each sampling site from two different approaches: (i) considering
FR as the average number of species per functional group (FG;
Rosenfeld 2002; Lalibert
e et al. 2010) and (ii) as the difference
between taxonomic (using the Gini-Simpson diversity index) and
FD (using Rao’s quadratic entropy) (Pillar et al. 2013; data com-
paring both methods are available in Appendix S3). To deﬁne
FGs, we used the approach proposed by D
ıaz & Cabido (2001),
which considers FGs as sets of plants that have traits with similar
functional effects on the dominant ecosystem processes. Thus, the
selection of FGs must represent different life strategies with a
clear ecological signiﬁcance (Naiman, D
ecamps & McClain
2005). First, species were classiﬁed into FGs by means of Ward’s
clustering method based on the effect-trait dissimilarity matrix,
which was estimated using Gower dissimilarity index. Given that
Ward’s clustering method requires a Euclidean distance, we
checked that the Gower effect-trait dissimilarity matrix met this
criterion by ensuring that the eigenvectors of a double-centred
matrix obtained through a principal component analysis were
positive. We deﬁned FGs with a suite of coadapted characteristics
to environmental conditions of channel and riparian zones that
guarantee a minimum number of six species to allow further sta-
tistical analyses. Secondly, after calculating both FR measures
and running the models for them (see below), we focused on the
most sensitive FR approach, which fulﬁlled models’ assumptions
and showed a non-random response to stress. Although similar
qualitative results were obtained when comparing both FR mea-
sures, we retained the FR estimated as the average number of
species per FG since it showed a better performance in response
to environmental ﬁlters (see Appendix S3 for further details).
We calculated the three primary components of FD (richness,
evenness and divergence sensu Mason et al. 2005) to compare
their response with those obtained using FR. First, we estimated
three Gower dissimilarity matrices using the all traits by taxon,
response traits by taxon and effect traits by taxon. Functional
richness (FRic) was estimated as the hypervolume enclosing the
functional space ﬁlled by the community (Vill
eger, Mason &
Mouillot 2008). The functional space was built using the six ﬁrst
axes of a principal component analysis based on the all-traits dis-
similarity matrix. The number of axes retained to estimate the
hypervolumes was decided following the method proposed in
Maire et al. (2015). This variable was standardized by its maxi-
mum, ranging from 0 to 1. Functional evenness (FEve) was cal-
culated using the method of the minimum spanning tree in a
functional space based on all-traits dissimilarity matrix (Vill
Mason & Mouillot 2008). Functional divergence (FDis) was mea-
sured as the abundance-weighted functional dispersion of the
response traits (i.e. response diversity). To quantify this metric
for each community, we estimated the weighted mean distance to
the weighted community centroid (Lalibert
e & Legendre 2010).
Fig. 1. Geographical location of the study area showing the 71
sampling sites classiﬁed by elevation (alt), the agricultural area
and the main dams.
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology, 53, 846–855
848 D. Bruno et al.
Changes in land use and ﬂow regulation are globally recognized
as the most important anthropogenic stressors impacting aquatic
and riparian ecosystems (Nilsson & Berggren 2000; Allan 2004),
and particularly in Mediterranean and semi-arid areas as the
study area (Bruno et al. 2014b). We used the percentage of agri-
cultural land at basin scale as a surrogate for land-use intensity. It
was calculated after delineating the catchment for each sampling
point using the A
RCGIS software (v 9.2) (ESRI, Redlands, CA,
USA) and the analysis toolkit NetMap (Benda et al. 2007), taking
as a base the available layers (1 : 25 000) of the Occupation Infor-
mation System of Soil in Spain. The dam regulation index was
estimated using the methodology described in Falcone, Carlisle &
Weber (2010) and adapted from Belmar et al. (2013). This method
uses the number of dams and their regulatory capacity (hm
the drainage area associated with each sampling site. Sites were
assigned from 0 to 8 points for each variable based on their per-
centile value within the data range. Then, those points were added
to provide an index that potentially ranged from 0 (minimum ﬂow
alteration) to 16 (maximum ﬂow alteration).
Drought duration (days per year without water ﬂow) was used
as a surrogate for the natural hydrological stress to which semi-
arid rivers are normally subjected. This ﬁlter sorts the regional spe-
cies pool, leaving species that have developed adaptations to
nuelas, Lloret & Montoya 2001). Drought duration
was estimated at reach scale using the data from the Integrated
System for Rainfall–Runoff Modelling (SIMPA) (Belmar, Velasco
& Martinez-Capel 2011). The SIMPA is a soil moisture balance
model where precipitation, soil and aquifer storages are consid-
ered, used in Spain, for water resources assessment (Ministry for
the Environment 2004) and hydrological classiﬁcations (Bejarano
et al. 2010; Belmar, Velasco & Martinez-Capel 2011). The site-spe-
ciﬁc values of the environmental ﬁlters are shown in Appendix S2.
The relationships between FR and the interacting environmental
ﬁlters were tested using linear mixed-effect models (LME), assum-
ing a Gaussian distribution of the dependent variables. The models
included a ﬁxed part with an intercept and the stressor slopes,
along with a random intercept that accounts for environmental
variability. Accordingly, LMEs produce two R
the marginal R
associated with the ﬁxed effects (those produced by
environmental ﬁlters) and the conditional R
that represents the
ﬁxed effects plus the random effects (those caused by environmen-
tal ﬁlters and variability together). Environmental variability was
considered as a three-level factor representing elevation typology
(high altitude: elevation > 1000 m a.s.l., mid-altitude: 1000 ≥ ele-
vation > 600 m a.s.l., lowlands: elevation ≤ 600 m a.s.l.) since it
summarizes well the natural environmental gradients occurring in
the study area (D
ıaz, Alonso & Guti
errez 2008). We tested the sig-
niﬁcance of simple and quadratic coefﬁcients for each z-standar-
dized (mean = 0, SD = 1) ﬁlter as well as the pairwise-ﬁlter
interaction terms to look for potential combined effects. Before
their standardization, drought duration was log-transformed and
percentage of agricultural land use was arcsine square-root-trans-
formed to improve linearity against response variables. LMEs were
performed using a backward-stepwise procedure retaining the
model that minimizes the Bayesian information criteria (BIC).
Normality, homoscedasticity and spatial autocorrelation of the
model residuals (Moran’s I test; A
RCGIS 9.2) were also assessed.
When either normality or homoscedasticity was not met, alpha
was set to 001. In case neither of these assumptions was met, alpha
was set to 0001. In addition, we examined the statistical relation-
ship between FR and the items that shape it (i.e. species richness
and number of FGs) through ordinary least squares.
A relationship between functional features and environmental
ﬁlters can be found simply as a result of an underlying taxonomic
richness gradient (and its response to stress) due to sampling
eger, Mason & Mouillot 2008) and not due to niche-
based sorting. Thus, we also checked for non-randomness of the
FR model coefﬁcients by using null models. To assess the non-
randomness of the observed trends, empirical parameters should
be distinct from a null distribution of simulated parameters. We
randomly reassigned traits to each species (999 runs) to re-exam-
ine their relationships with the stressors. For randomizations, we
kept the same trait combinations, richness gradient and taxon fre-
quency of occurrence. For each simulation, we used the same
model and procedure as for the empirical data (i.e. we calculated
FR and re-examined its relationship with the same predictors to
obtain the simulated intercepts and slopes for each relationship).
We examined the null model’s statistical signiﬁcance using an
exact two-tailed test to calculate the probability that the observed
value was signiﬁcantly (a = 005) larger or smaller than the simu-
lated distribution. These same analyses (LME and null models)
were also conducted with the three FD measures (FRic, FEve
and FDis). Finally, we followed a similar null model approach to
test whether the relationship between FR and species richness
was different from what is expected by chance.
Finally, using the best-ﬁtting model obtained for FR, we fore-
casted their values for the entire river network. Thus, rivers were
divided into homogeneous reaches characterized by an absence of
tributaries and deﬁned by 409 ﬂuvial nodes in which FR was pre-
dicted. The predictive power of the ﬁnal model was estimated by
a jackknife cross-validation procedure. Thus, the mean error per-
centage of all sampling sites was used as a measure of model reli-
ability. All statistical analyses were performed in the
software (libraries: ‘ade4’, ‘boot’, ‘car’, ‘FD’, ‘MuMIn’, ‘nmle’
and ‘vegan’; R Development Core Team 2013). See R code and
FR functions used in Appendix S2.
A total of 63 woody riparian species were recorded and
classiﬁed into ﬁve FGs representing different life strategies
and effects with a clear ecological signiﬁcance on ecosys-
tem functioning. Among them, we identiﬁed two groups
of phreatophytes mainly differing in life form (FG1:
shrubby phreatophytes; FG2: arboreal phreatophytes),
both strongly associated with watercourses. Drought-
adapted riparian species showing special leaves, roots and
structural features formed FG3, and riparian evergreen
shrubs formed FG4. Lianas and climbers typical for well-
developed and humid riparian systems shaped FG5 (see
Fig. S1.1 and Table S1.2 in Appendix S1 for details).
Generally, we found that community functional mea-
sures signiﬁcantly decreased with increasing environmental
ﬁlters (Fig. 2). Droughts and especially agriculture caused
the strongest effects on the functional features used. The
© 2016 The Authors. Journal of Applied Ecology © 2016 British Ecological Society, Journal of Applied Ecology, 53, 846–855
Impacts of stressors on functional redundancy 849
responses of FR and functional richness (FRic) were simi-
lar, being the most sensitive indices. The best model for
FR (minimum BIC) showed a higher percentage of
explained deviance (R
= 059) than the best-ﬁtting model
for the FD components (R
< 04, Table 1). The FR
model included agriculture, drought and ﬂow regulation
as well as the interactions of agriculture with the two lat-
ter (Table 1). Conditional R
and marginal R
the same values in the best-ﬁtting mixed-effect models for
all functional indices (i.e. the combination of the different
environmental ﬁlters resulted more explicative). On the
other hand, conditional R
was higher than marginal R
in some mixed-effect models only when considering each
environmental ﬁlter alone (Appendix S4).
Null models revealed that stressors caused non-random
changes in FR. All terms were signiﬁcant (P < 005), and
empirical slopes were signiﬁcantly lower for all environ-
mental ﬁlters, and signiﬁcantly higher for the interactions
among them and the intercept (Fig. S4.2). Nevertheless,
excluding FEve, FD measures did not produce signiﬁcant
null models (P < 005) so their observed responses to
environmental ﬁlters were actually a direct consequence of
taxonomic diversity reduction (see Appendix S4 for
details about the null models).
Species richness was positively related to FR and the
richness of the different FGs, except for FG3 and FG4
that showed a weak, ﬂat response (Fig. 3). Nevertheless,
null model results showed that FR increased more than
expected by chance as taxonomic richness rose
(Appendix S4). More concretely, empirical intercept was
lower in comparison with the simulated distribution ( Z-
score = 278, P = 0005), whilst the empirical slope was
signiﬁcantly higher (i.e. species richness, Z-score = 472,
P = 0003). Finally, species richness and FR showed
humped relationships with the number of functional
groups (FGR, Fig. 3). FGR peaked at medium–high val-
ues of species richness and FR and consequently at low–
moderate anthropogenic stress intensity (Fig. S4.3 in
Appendix S4). The best obtained model was applied to
forecast the FR values for the river stretches of the entire
river network. There was a clear FR gradient, decreasing
from headwaters to lowlands (Fig. 4). The model showed
a mean error percentage of 363% without any geographi-
cal concentration of high residuals.
All environmental ﬁlters led to reductions in FR, thereby
decreasing ecosystem resistance and resilience to future
disturbances. The FD components chieﬂy declined in
response to agriculture, the most important stressor in the
study area. However, only FR showed a non-random
reduction in response to increased stress or species loss,
with the exception of functional evenness, which although
less sensitive, also experienced a non-random decline in
response to stress. Natural environmental variability (i.e.
associated to elevation gradient) exerted an inconspicuous
inﬂuence on the spatial distribution of FD indices in com-
parison with the effect of multiple stressors (Table 1).
Our results for single stressors are similar to those
observed in previous studies, where FR decreased along
single anthropogenic impact gradients for plants (Lalibert
et al. 2010), birds (Huijbers et al. 2015), soil microbes and
invertebrates (Salminen, van Gestel & Oksanen 2001), and
aquatic invertebrates (Guti
anovas et al. 2015).
However, one of the main novelties of this study was in
revealing non-random effects between combined environ-
mental ﬁlters on riparian vegetation. The decrease in FR
following species loss was greater than expected by chance
and particularly evident at high species richness as derived
from the relationships among FR, FG and species richness
(Fig. 3). The reduction in FR could be associated with a
loss of richness within some FGs (mainly FG1, FG2 and
FG5), likely as a consequence of different response diver-
sity or trait combinations of the species within each FG.
Thus, the random loss of one species might not affect
ecosystem functioning in functionally redundant communi-
ties, as its function could be compensated for by the
remaining species of the same FG if they are capable of
expanding to ﬁll the gap (Fonseca & Ganade 2001). In par-
ticular, different responses of functionally equivalent spe-
cies to environmental change increase response diversity
(Elmqvist et al. 2003), enhancing the capacity of ecosys-
tems to resist impacts (Mori, Furukawa & Sasaki 2013).
It is worth noting that FR can be partitioned into intrin-
sic and extrinsic redundancy. We mainly focus here on
intrinsic redundancy, which results from the patterns of
functional similarity among species. On the other hand,
extrinsic redundancy (or lack thereof) can result from non-
random compositional change with respect to functional
traits (Petchey et al. 2007). Thus, although we have
detected that FR increases more than expected by chance
as taxonomic richness rises, further studies will allow us to
explore extrinsic redundancy which is also a key variable in
functional ecology with direct applications to management.
Although both (intrinsic) FR and FD measures
decreased as stressors intensiﬁed, FD components seem to
be less affected by environmental stressors, helping to mit-
igate the effect of stressors in redundant communities (i.e.
the reduction of FD was minimal in redundant places,
Appendix S5). Agriculture, drought and ﬂow regulation
(in order of importance) reduced FR and the FD compo-
nents, suggesting that general functional response is simi-
lar irrespective of whether the impact had natural
(drought) or anthropogenic origin (agriculture and ﬂow
regulation), as found by Guti
anovas et al. (2015).
However, some differences depending on the nature and
source of stress can be observed. In general, agriculture
caused the greatest impact on all functional measures,
probably due to its multiple effects on the riparian com-
munity, such as direct destruction of riparian forest (Allan
& Flecker 1993), higher nutrient loading due to fertiliza-
tion (Monteagudo, Moreno & Picazo 2012), and water
abstraction for irrigation (Belmar et al. 2013). The traits
850 D. Bruno et al.
most disfavoured by agriculture include high species
woodiness, slow growth, large lateral extension, sexual
reproduction and short-term persistent seed bank (Kleyer
1999). Flow regulation modiﬁes abiotic features and
homogenizes habitat conditions (Belmar et al. 2013),
affecting reproduction, recruitment, dispersal opportuni-
ties, succession and fragmentation of the riparian commu-
nity (Jansson, Nilsson & Ren
alt 2000; Nilsson &
Berggren 2000). Drought sorted out woody riparian com-
munities, favouring sclerophyllous and evergreen shrubs
Fig. 2. Plots showing the response of functional redundancy and three measures of functional diversity to single environmental ﬁlters
estimated through mixed-effect models. The solid line represents the ﬁtted models for each single stressor, dashed lines represent the ﬁt-
ted model for the lowland rivers, dotted lines show the ﬁtted model for the mid-altitude rivers, and dashed-dotted lines display the ﬁtted
model for the high-altitude rivers.
Impacts of stressors on functional redundancy 851
(Aguiar & Ferreira 2005). Drought-adapted species usu-
ally have long roots, low seed buoyancy, low canopy, lit-
tle speciﬁc leaf area or small and thick leaves (Cornwell &
Ackerly 2009; Douma et al. 2012). In addition, the regio-
nal persistence of drought could have helped some species
to tolerate ﬂow regulation, which might partially explain
its lower impact on FD measures. In fact, the strong
ﬂow regulation by dams leads to a terrestrialization of
riparian and river communities favouring the occurrence
of opportunistic, terrestrial and drought-adapted species
(Catford et al. 2014). Although these two disturbances
differ in periodicity, timing and origin, both alter the ﬂow
regime and the water supply for riparian vegetation.
The interactions between agriculture and the other
stressors were not surprising since several links exist
among them in the study area. Large agricultural surfaces
in areas with long drought periods produce high water
demands that have triggered massive dam construction
and other hydraulic infrastructures. The combination of
high nutrient loading, clearing of river banks and reduc-
tion of the water-table may favour similarly opportunistic,
drought-adapted and generalist species (e.g. Arundo donax
L., see Quinn & Holt 2008) leading to a simpliﬁcation of
ecosystem structure and function. Given these comple-
mentary effects, the management of anthropogenic pres-
sures should be addressed in a holistic way considering
also the underlying natural stress such as the Mediter-
ranean drought in the study area.
Table 1. Results of mixed-effect models showing the best-ﬁtting model equation, P-values (signiﬁcant coefﬁcients in bold type), marginal
m) and conditional (R
c) goodness-of-ﬁt for the different functional diversity indices
Index Model equation (A)(F)(D) A*FA*DF*DR
FRic y = 00170016A0009D + 001A*D <0001 ns 003 ns 0018 ns 037 038
FDis y = 02530032A <0001 ns ns ns ns ns 021 021
FEve y = 08090074A0097A*F + 0091D*F 0004 ns ns 0002 ns <0001 028 028
y = 2660624A0642D0475F + 0506A*D + 0354A*F <0001 <0001 <0001 0002 0002 ns 059 059
A, agriculture; F, ﬂow regulation; D, drought; ns, non-signiﬁcant coefﬁcient; FR, functional redundancy.
Pairwise interactions are noted with an asterisk.
Fig. 3. Plots relating functional redundancy (FR), taxonomic richness and number of functional groups (FGR). Single results for each
functional group (FG) are also shown.
Fig. 4. Predicted functional redundancy (FR) values for riparian
communities in the entire river network of the Segura basin.
852 D. Bruno et al.
Conservation and biomonitoring efforts have been tra-
ditionally focused on taxonomic features (such as species
presence, abundance and rarity), ignoring other ecosystem
properties (Cadotte, Carscadden & Mirotchnick 2011).
However, functional features are linked to ecosystem
functioning (Hooper et al. 2005) or community assembly
(Weiher et al. 2011), which allows explaining, in some
cases, non-random patterns, as observed here. Their use
has several advantages, such as better intertaxon and
inter-region comparability (McGill et al. 2006). Accord-
ingly, we feel that this kind of measures should be incor-
porated in conservation prioritization and ecosystem
management in order to have a broader perspective of the
response of biological communities to different environ-
mental stressors. In particular, FR informs about the spe-
cies playing similar roles (Lawton & Brown 1993), and
consequently, the likelihood of losing particular ecosystem
functions as a result of biodiversity reduction (Naeem &
Wright 2003). In our case, restoration efforts might be
focused on those river reaches showing slightly or moder-
ately reduced FR values, as a ﬁrst step to recover the
integrity of the riparian functioning, within a context of
cost-effective management (i.e. restoring greatly damaged
places could be less efﬁcient). Particularly, phreatophytes
(both arboreal and shrubby) as well as lianas and climbers
seem to be the most affected FGs, suggesting that their
recovery is essential to reach a better riparian functional-
ity across the study area. Thus, FR could provide addi-
tional and complementary information to taxonomic
diversity on how communities respond to stress. Besides,
quantifying community functional responses to increasing
intensity and frequency of anthropogenic impacts is
required to further evaluate the loss of ecosystem services
associated with biodiversity erosion in the current context
of global change (Cardinale et al. 2012).
Although FR and other FD indices can be useful
in conservation ecology and environmental manage-
ment, there are some methodological considerations. The
grouping of taxa in FGs may result in a loss of informa-
tion in comparison with continuous measures. On the
other hand, this approach is interesting to explore further
to understand how environmental ﬁlters may modify par-
ticular ecosystem functions and services provided by bio-
logical communities. Thus, each FG and even each
species within the same FG may respond differently to
the same stressor. Otherwise, accounting for intraspeciﬁc
variability enables more accurate measures of multidimen-
sional functional overlap (Guti
anovas et al. 2015),
but gathering such data could be costly in comparison
with the data quality improvement. Finally, functional
measures could depend on the number and nature of
traits used for its computation, as species are more likely
to have non-overlapping functional niches (low FR) in a
functional space when using single or few functional traits
This is one of the ﬁrst studies predicting a whole com-
munity functional measure for an entire administrative
area, which may help to improve ecosystem biomonitor-
ing and management (Devictor et al. 2010; Matsuzaki,
Sasaki & Akasaka 2013; Sasaki et al. 2014). In a changing
environment, this measure provides three major advan-
tages: (i) valuable information on how river ecosystems
respond to human and natural environmental stressors,
which can help managing the current increase of multiple
stressors across the river network, (ii) assessment of stres-
sors’ effects on functional features from the descriptive to
the predictive (being the assessment framework of broad-
scale applicability across ecological domains) and (iii) the
geographical distribution of sites that potentially could
show more stability, resistance and resilience, and vice
Functional redundancy proved to be more sensible than
other FD measures to impacts of the most important
stressors in Mediterranean rivers as well as the interac-
tions between them. FR can be considered as an eco-
logically-sound measure able to detect non-random
responses to single and multiple stressors. According to
the FR gradient found across the catchment, temporary
streams ﬂowing through an agricultural, regulated basin
had reduced values of FR. On the other hand, free-
ﬂowing medium-sized, perennial water courses ﬂowing
through unaltered sub-basins displayed higher values of
FR and potentially greater stability against human
impacts. Thus, undisturbed conditions held more diverse
communities, where redundant species may ensure
ecosystem functioning when response diversity is high.
Our study reveals that the response of FR can be pre-
dicted for entire river networks, constituting a potential
tool to detect more impacted river reaches and improve
their conditions through restoration measures, as well as
to conserve the reaches with better functional
We thank the members of the ‘Landscape Ecology’ (Ume
and ‘Aquatic Ecology’ (University of Murcia) research groups for their
wholehearted support, especially O. Belmar for providing environmental
data, as well as J.A. Carbonell, S. Guareschi and A. Garc
their help and support in the ﬁeld work. D.B. and D.S.-F. were sup-
ported by a predoctoral (FPU) and postdoctoral grant (Juan de la
Cierva programme), respectively, both from the Spanish Ministry of
Economy and Competitiveness. Finally, C.G.-C. was supported by the
MARS project (Managing Aquatic ecosystems and water resources under
multiple stress), funded by the European Union under the 7th Frame-
work Programme, contract no. 603378. J. Bremner as a journal reviewer
and an anonymous reviewer provided insightful comments on the manu-
All data (functional traits and indices, species abundances per site and
environmental variables) and R scripts used to produce this manuscript
are available in Appendix S2.
Impacts of stressors on functional redundancy 853
Aguiar, F.C. & Ferreira, M.T. (2005) Human-disturbed landscapes: effects
on composition and integrity of riparian woody vegetation in the Tagus
River basin, Portugal. Environmental Conservation, 32,30–41.
Aguiar, F.C., Ferreira, M.T., Albuquerque, A., Rodr
alez, P. &
Segurado, P. (2009) Structural and functional responses of riparian veg-
etation to human disturbance: performance and spatial scale-depen-
dence. Fundamental and Applied Limnology, 175, 249–267.
Allan, J.D. (2004) Landscapes and riverscapes: the inﬂuence of land use
on stream ecosystems. Annual Review of Ecology, Evolution, and System-
atics, 35, 257 –284.
Allan, J.D. & Flecker, A.S. (1993) Biodiversity conservation in running
waters. BioScience, 43,32–43.
Bejarano, M.D., Marchamalo, M., Garc
ıa de Jal
on, D. & Gonz
anago, M. (2010) Flow regime patterns and their controlling factors in
the Ebro basin (Spain). Journal of Hydrology, 385, 323–335.
Belmar, O., Velasco, J. & Martinez-Capel, F. (2011) Hydrological classiﬁ-
cation of natural ﬂow regimes to support environmental ﬂow assess-
ments in intensively regulated Mediterranean rivers, Segura River basin
(Spain). Environmental Management, 47, 992 –1004.
Belmar, O., Bruno, D., Mart
ınez-Capel, F., Barqu
ın, J. & Velasco, J.
(2013) Effects of ﬂow regime alteration on ﬂuvial habitats and riparian
quality in a semiarid Mediterranean region. Ecological Indicators, 30,
Benda, L., Miller, D., Andras, K., Bigelow, P., Reeves, G. & Michael, D.
(2007) NetMap: a new tool in support of watershed science and
resource management. Forest Science, 53, 220–238.
Biggs, R., Schl
uter, M., Biggs, D., Bohensky, E.L., BurnSilver, S., Cundill,
G. et al. (2012) Toward principles for enhancing the resilience of ecosys-
tem services. Annual Review of Environment and Resources, 37, 421–448.
Bruno, D., Belmar, O., S
andez, D. & Velasco, J. (2014a) Envi-
ronmental determinants of woody and herbaceous riparian vegetation
patterns in a semi-arid Mediterranean basin. Hydrobiologia, 730,45–57.
Bruno, D., Belmar, O., S
andez, D., Guareschi, S., Mill
& Velasco, J. (2014b) Responses of Mediterranean aquatic and riparian
communities to human pressures at different spatial scales. Ecological
Indicators, 45, 456–464.
Cadotte, M.W., Carscadden, K. & Mirotchnick, N. (2011) Beyond species:
functional diversity and the maintenance of ecological processes and ser-
vices. Journal of Applied Ecology, 48, 1079–1087.
Cardinale, B.J., Duffy, J.E., Gonzalez, A., Hooper, D.U., Perrings, C.,
Venail, P. et al. (2012) Biodiversity loss and its impact on humanity.
Catford, J.A., Morris, W.K., Vesk, P.A., Gippel, C.J. & Downes, B.J.
(2014) Species and environmental characteristics point to ﬂow regulation
and drought as drivers of riparian plant invasion. Diversity and Distribu-
tions, 20, 1084–1096.
Clapcott, J.E., Young, R.G., Goodwin, E.O. & Leathwick, J.R. (2010)
Exploring the response of functional indicators of stream health to
land-use gradients. Freshwater Biology, 55, 2181–2199.
Cornelissen, J.H.C., Lavorel, S., Garnier, E., D
ıaz, S., Buchmann, N.,
Gurvich, D.E. et al. (2003) A handbook of protocols for standardised
and easy measurement of plant functional traits worldwide. Australian
Journal of Botany, 51, 335–380.
Cornwell, W.K. & Ackerly, D.D. (2009) Community assembly and shifts
in plant trait distributions across an environmental gradient in coastal
California. Ecological Monographs, 79, 109–126.
Devictor, V., Mouillot, D., Meynard, C., Jiguet, F., Thuiller, W. &
Mouquet, N. (2010) Spatial mismatch and congruence between taxo-
nomic, phylogenetic and functional diversity: the need for integrative
conservation strategies in a changing world. Ecology Letters, 13,
ıaz, A.M., Alonso, M.L.S. & Guti
errez, M.R.V.-A. (2008) Biological
traits of stream macroinvertebrates from a semi-arid catchment: patterns
along complex environmental gradients. Freshwater Biology, 53,1
ıaz, S. & Cabido, M. (2001) Vive la difference: plant functional diversity
matters to ecosystem processes . Trends in Ecology & Evolution, 16, 646–
Douma, J.C., Bar din, V., Bartholomeus, R.P. & Bodegom, P.M. (2012)
Quantifying the functional responses of vegetation to drought and oxy-
gen stress in temperate ecosystems. Functional Ecology, 26, 1355–1365.
Elmqvist, T., Folke, C., Nystrom, M., Peterson, G., Bengtsson, J., Walker,
B. & Norberg, J. (2003) Response diversity, ecosystem change, and resi-
lience. Frontiers in Ecology and the Environment, 1, 488–494.
Falcone, J.A., Carlisle, D.M. & Weber, L.C. (2010) Quantifying human
disturbance in watersheds: variable selection and performance of a GIS-
based disturbance index for predicting the biological condition of peren-
nial streams. Ecological Indicators, 10, 264–273.
Ferreira, M.T. & Aguiar, F.C. (2006) Riparian and aquatic vegetation in
Mediterranean-type streams (western Iberia). Limnetica, 25, 411–424.
Fonseca, C.R. & Ganade, G. (2001) Species functional redundancy, ran-
dom extinctions and the stability of ecosystems. Journal of Ecology, 89,
Guillemot, N., Kulbicki, M., Chabanet, P. & Vigliola, L. (2011) Func-
tional redundancy patterns reveal non-random assembly rules in a spe-
cies-rich marine assemblage. PLoS ONE , 6, e26735.
anovas, C., S
andez, D., Velasco, J., Mill
an, A. &
Bonada, N. (2015) Similarity in the difference: changes in community
functional features along natural and anthropogenic stress gradients.
Ecology, 96, 2458–2466.
ornsson, K., Chauvet, E., Dobson, M., Elosegi, A., Fer-
reira, V. et al. (2011) Stream ecosystem functioning in an agricultural
landscape: the importance of terrestrial-aquatic linkages. Advances in
Ecological Research, 44, 211–276.
Hooper, D.U., Chapin, F.S.I.I.I., Ewel, J.J., Hector, A., Inchausti, P.,
Lavorel, S. et al. (2005) Effects of biodiversity on ecosystem function-
ing: a consensus of current knowledge. Ecological Monographs, 75
Huijbers, C.M., Schlacher, T.A., Schoeman, D.S., Olds, A.D., Weston,
M.A. & Connolly, R.M. (2015) Limited functional redundancy in
vertebrate scavenger guilds fails to compensate for the loss of rap-
tors from urbanized sandy beaches. Diversity and Distributions, 21,
Jansson, R., Nilsson, C. & Ren
alt, B. (2000) Fragmentation of riparian
ﬂoras in rivers with multiple dams. Ecology, 81, 899–903.
Kleyer, M. (1999) Distribution of plant functional types along gradients of
disturbance intensity and resource supply in an agricultural landsc ape.
Journal of Vegetation Science, 10, 697–708.
Kraft, N.J., Godoy, O. & Levine, J.M. (2015) Plant functional traits and
the multidimensional nature of species coexistence. Proceedings of the
National Academy of Sciences of the USA, 112, 797–802.
e, E. & Legendre, P. (2010) A distance-based framework for
measuring functional diversity from multiple traits. Ecology, 91, 299–
e, E., Wells, J.A., DeClerck, F., Metcalfe, D.J., Catterall, C.P.,
Queiroz, C. et al. (2010) Land-use intensiﬁcation reduces functional
redundancy and response diversity in plant communities. Ecology Let-
Lavorel, S. & Garnier, E. (2002) Predicting changes in community compo-
sition and ecosystem functioning from plant traits: revisiting the Holy
Grail. Functional Ecology, 16, 545 –556.
Lawton, J.H. & Brown, V.K. (1993) Redundancy in ecosystems. Biodiver-
sity and Ecosystem Function (eds E.-D. Schulze & H.A. Mooney), pp.
255–270. Springer-Verlag, Berlin.
Maire, E., Grenouillet, G., Brosse, S. & Vill
eger, S. (2015) How many
dimensions are needed to accurately assess functional diversity? A
pragmatic approach for assessing the quality of functional spaces. Glo-
bal Ecology and Biogeography, 24, 728–740.
Mason, N.W.H., Mouillot, D., Lee, W.G. & Wilson, J.B. (2005)
Functional richness, functional evenness and functional divergence:
the primary components of functional diversity. Oikos, 111, 112
Matsuzaki, S.S., Sasaki, T. & Akasaka, M. (2013) Consequences of the
introduction of exotic and translocated species and future extirpations
on the fu nctional diversity of freshwater ﬁsh assemblages. Global Ecol-
ogy and Biogeography, 22, 1071–1082.
McGill, B.J., Enquist, B.J., Weiher, E. & Westoby, M. (2006) Rebuilding
community ecology from functional traits. Trends in Ecology & Evolu-
tion, 21, 178–185.
Ministry for the Environment (2004) Water in Spain. Secretar
ıa de Estado
de Aguas y Costas, Madrid.
Monteagudo, L., Moreno, J.L. & Picazo, F. (2012) River eutrophication:
irrigated vs. non-irrigated agriculture through different spatial scales.
Water Research, 46, 2759–2771.
Mori, A.S., Furukawa, T. & Sasaki, T. (2013) Resp onse diversity determi-
nes the resilience of ecosystems to environmental change. Biological
Reviews, 88, 349–364.
Mouillot, D., Graham, N.A.J., Vill
eger, S., Mason, N.W.H. & Bellwood,
D.R. (2013) A functional approach reveals community responses to dis-
turbances. Trends in Ecology and Evolution, 28, 167–177.
854 D. Bruno et al.
Naeem, S. & Wright, J.P. (2003) Disentangling biodiversity effects on
ecosystem functioning: deriving solutions to a seemingly insurmountable
problem. Ecology Letters, 6, 567–579.
Naiman, R.J., D
ecamps, H. & McClain, M.E. (2005) Riparia: Ecology,
Conservation, and Management of Stream Side Communities. Elsevier
Nilsson, C. & Berggren, K. (2000) Alterations of riparian ecosystems
caused by river regulation. BioScience, 50, 783–792.
Nilsson, C. & Svedmark, M. (2002) Basic princi ples and ecological conse-
quences of changing water regimes: riparian plant communities. Envi-
ronmental Management, 30, 468–480.
nuelas, J., Lloret, F. & Montoya, R. (2001) Severe drought effects on
Mediterranean woody ﬂora in Spain. Forest Science, 47, 214–218.
Petchey, O.L., Evans, K.L., Fishburn, I.S. & Gaston, K.J. (2007) Low
functional diversity and no redundancy in British avian assemblages.
Journal of Animal Ecology, 76, 977–985.
Pillar, V.D., Blanco, C.C., M
uller, S.C., Sosinski, E.E., Joner, F. &
Duarte, L.D. (2013) Functional redundancy and stability in plant com-
munities. Journal of Vegetation Science, 24, 963–974.
Quinn, L.D. & Holt, J.S. (2008) Ecological correlates of invasion by
Arundo donax in three southern California riparian habitats. Biological
Invasions, 10, 591–601.
R Development Core Team (2013) R: A Language and Environment for
Statistical Computing. R Foundation for Statistical Computing, Vienna.
Rosenfeld, J.S. (2002) Functional redundancy in ecology and conservation.
Oikos, 98, 156–162.
Sabo, J.L. & Power, M.E. (2002) River-watershed exchange: effects of
riverine subsidies on riparian lizards and their terrestrial prey. Ecology ,
Salminen, J., van Gestel, C.A.M. & Oksanen, J. (2001) Poll ution-induced
community tolerance and functional redundancy in a decomposer food
web in metal-stressed soil. Environmental Toxicology and Chemistry, 20,
Sasaki, T., Katabuchi, M., Kamiyama, C., Shimazaki, M., Nakashizuka,
T. & Hikosaka, K. (2014) Vulnerability of moorland plant communities
to environmental change: consequences of realistic species loss on func-
tional diversity. Journal of Applied Ecology,
Sasaki, T., Furukawa, T., Iwasaki, Y., Seto, M. & Mori, A.S. (2015) Per-
spectives for ecosystem management based on ecosystem resilience and
ecological thresholds against multiple and stochastic disturbances. Eco-
logical Indicators, 57, 395–408.
Tabacchi, E., Lambs, L., Guilloy, H., Planty-Tabacchi, A.M., Muller, E.
& Decamps, H. (2000) Impacts of riparian vegetation on hydrological
processes. Hydrological Processes, 14, 2959–2976.
eger, S., Mason, N.W. & Mouillot, D. (2008) New multidimensional
functional diversity indices for a multifaceted framework in functional
ecology. Ecology, 89, 2290 –2301.
Walker, B.H. (1992) Biodiversity and ecological redundancy. Conservation
Weiher, E., Freund, D., Bunton, T., Stefanski, A., Lee, T. & Bentivenga,
S. (2011) Advances, challenges and a developing synthesis of ecological
community assembly theory. Philosophical Transactions of the Royal
Society of London B: Biological Sciences, 366, 2403–2413.
Woodward, G., Gessner, M.O., Giller, P.S., Gulis, V., Hladyz, S., Lecerf,
A. et al. (2012) Continental-scale effects of nutrient pollution on stream
ecosystem functioning. Science, 336, 1438– 1440.
Received 11 September 2015; accepted 28 January 2016
Handling Editor: Jin-Tian Li
Additional Supporting Information may be found in the online version
of this article.
Appendix S1. Effect and response functional traits information.
Table S1.1. Effect and response traits considered to characterize the
functional features of the woody riparian species.
Table S1.2. Functional group description based on distinctive
functional effect traits.
Table S1.3. Data sources used to obtain the functional trait values.
Fig. S1.1. Dendrogram resulting from classifying riparian species
according to their similarity in the functional effect traits.
Appendix S2. R code and data ﬁles.
Appendix S3. Details about functional redundancy estimations.
Table S3.1. Results of the linear mixed-effect models relating the
divisive and additive estimations of FR with individual stressors.
Table S3.2. Results of the null models for the models relating the
divisive and additive estimations of FR with individual stressors.
Table S3.3. Results of the linear mixed-effect models for the best-
ﬁtting models relating the divisive and additive estimations of FR
with multiple stressors.
Table S3.4. Results of the null models for the best-ﬁtting models
relating the divisive and additive estimations of FR with multiple
Fig. S3.1. Values of the divisive (FRa) and additive (FRb)
estimations of FR across the study area.
Fig. S3.2. Plots showing the response of FRa and FRb to single
Appendix S4. Linear mixed-effect models, null models and
Table S4.1. Results of linear mixed-effects models for single
Table S4.2. Null model results for the individual signiﬁcant
stressors for each functional index.
Table S4.3. Null model results for the best-ﬁtting models for each
Fig. S4.1. Plots showing the residuals’ normality and homoscedas-
ticity of the best-ﬁtting model for FR.
Fig. S4.2. Results of null models for each model parameter of
functional redundancy’s best-ﬁtting model.
Fig. S4.3. Best-ﬁtting model for FGR in response to environmental
Appendix S5. Spatial pattern of the functional diversity and
redundancy indices and their pairwise relationships.
Table S5.1. Spearman correlations between functional indices.
Fig. S5.1. Spatial pattern of FR and FRic.
Fig. S5.2. Spatial pattern of FEve and FDis.
Fig. S5.3. Relationships between functional indices.
Impacts of stressors on functional redundancy 855