Received: 21 December 2015
|
Revised: 22 March 2017
|
Accepted: 21 April 2017
DOI: 10.1111/geb.12605
RESEARCH PAPER
Linking species interactions with phylogenetic and functional distance in European bird assemblages at broad spatial scales € nkko € nen1 Mikko Mo
|
Vincent Devictor2 | Jukka T. Forsman3 |
Aleksi Lehikoinen4 | Merja Elo1 1 Department of Biological and Environmental Sciences, University of Jyvaskyla, Jyväskylä, Finland 2
Institut des Sciences de l’Evolution de Montpellier, Montpellier, France
3
Department of Ecology and Genetics, University of Oulu, Oulu, Finland 4 The Helsinki Lab of Ornithology, Finnish Museum of Natural History, University of Helsinki, Helsinki, Finland
Correspondence €nkko €nen, Department of Mikko Mo Biological and Environmental Sciences, University of Jyvaskyla, POB 35, Jyväskylä, FI-40014, Finland. Email:
[email protected] Editor: Kathleen Lyons Funding information Academy of Finland, Grant Number 275329 and 275606; Kone Foundation
Abstract Aim: Understanding the relative contribution of different species interactions in shaping community assembly has been a pivotal aim in community ecology. Biotic interactions are acknowledged to be important at local scales, although their signal is assumed to weaken over longer distances. We examine the relationship between positive, neutral and negative pairwise bird abundance distributions and the phylogenetic and functional distance between these pairs after first controlling for habitat associations. Location: France and Finland. Time period: 1984 to 2011 (Finland), 2001 to 2012 (France). Major Taxa studied: Birds. Methods: We used results from French and Finnish land bird monitoring programmes, from which we created three independent datasets (French forests, French farmlands and Finnish forests). Separately for the three datasets, we fitted linear mixed-effects models for pairwise abundance values across years per point count station to infer the association between all common species pairs, while controlling for geographical distribution and habitat associations, and saved pairwise regression coefficients for further analyses. We used a null model approach to infer whether the observed associations (effect sizes) differ from random. Finally, using quantile regression, we analysed the relationships between functional dissimilarity/phylogenetic distance and effect sizes. Results: Our results show both negative and positive species interactions, although negative interactions were twice as common as positive interactions. Closely related species were more likely to show strong associations, both negative and positive, than more distant species across broad spatial scales. For functional dissimilarity, the results varied across datasets. Main conclusions: Our results emphasize the potential of functional and phylogenetic proximity in generating both negative and positive species associations, which can produce pervasive patterns from local to geographical scales. Future assembly studies should refrain from strict dichotomies, such as compensatory dynamics versus environmental forcing, and instead consider the possibility of positive interactions. KEYWORDS
aggregated distribution, functional similarity, heterospecific attraction, interspecific competition, pairwise interactions, phylogenetic distance, segregated distribution
952
|
C 2017 John Wiley & Sons Ltd V
wileyonlinelibrary.com/journal/geb
Global Ecol Biogeogr. 2017;26:952–962.
€ € MONKK ONEN
|
ET AL.
953
1 | INTRODUCTION A fundamental and long-standing goal in community ecology has been to understand the complexity of dependencies among species, and thereby the mechanisms by which communities are assembled. Since the 1950s (Hutchinson, 1959; MacArthur, 1958) interspecific competition and niche partitioning have been the focus of explanations for species coexistence and community assembly patterns (Connor & Simberloff, 1983; Cornell, 1985; Cornell & Lawton, 1992; Diamond, 1975). Experimental evidence also shows that competition is undeniably an important factor for community assembly (Connell, 1983; Goldberg & Barton, 1992; Gurevitch, Morrison, & Hedges, 2000; Schoener, 1983). However, it is not the only one. A growing body of literature demonstrates the importance of facilitative or positive non-trophic interactions (Bertness & Callaway, 1994; Bruno, Stachowicz, & Bertness, 2003; Cardinale, Palmer, & Collins, 2002). Moreover, interactions often result from combinations of positive, negative and neutral rela€ nkko €nen, & Thomson, 2007), potentionships (Seppänen, Forsman, Mo € nkko € nen et al., 1999). tially generating asymmetric interactions (Mo Indeed, Gross (2008) concluded that the joint effects of different interactions may be the most important factor for community assembly, the key question being the relative contribution of each interaction. Species interactions may affect species abundances in communities, leading to patterns in space that are independent of habitat characteristics. Negative interactions, such as competition, are predicted to cause segregated distributions (Gotelli, Graves, & Rahbek, 2010). Positive interactions, predicted to result in aggregated distributions, are well known among plants and sessile animals (Bertness & Callaway, 1994; Bruno et al., 2003), but in mobile animals the prevalence and
Schematic diagram of the relationship between increasing functional dissimilarity/phylogenetic distance (x-axis) and the sign [positive (1), neutral (0), negative (2)] and the strength (increases with increasing symbol size) of the association of a species pair (y-axis). Every black dot represents an imaginary pair of species. Both heterospecific information use (black dashed line) and the limiting similarity principle (continuous line) predict that species associations are most intense among functionally and phylogenetically similar species, but in opposite directions. Heterospecific information use predicts a positive association between functionally/phylogenetically similar species. As the information value decreases with increasing functional dissimilarity/phylogenetic distance, so does the strength of the association. The limiting similarity predicts functionally/ phylogenetically similar species to have a negative association, whereas functionally dissimilar/phylogenetically distant species show neutral associations. The net outcome of an association between similar species depends on the costs of competition and benefits of information use and may thus result in a neutral association, and altogether they form a ‘funnel plot’ FIGURE 1
mechanisms of positive interactions are poorly known. Recent findings
tions (continuum from positive to negative) and concluded that, in par-
about information use in animals imply a likely mechanism. While
ticular, positive species interactions can be manifested from local to
assessing the best site or resources for reproduction, individuals may
larger scales. Indeed, a few empirical studies have revealed both aggre-
use as cues to decide where to settle the presence, behaviour or suc-
gated and segregated distribution patterns among ecologically similar
cess not only of conspecifics (Danchin, Giraldeau, Valone, & Wagner,
species, probably resulting from both competitive and positive interac-
2004), but also of competing heterospecific individuals (Seppänen
tions, independent of habitat characteristics, at broad spatial scales
et al., 2007). Heterospecific information use and attraction to hetero-
(Gotelli et al., 2010; Ricklefs, 2012).
specifics and potential competitors has been demonstrated from ants
The strength of species associations may depend on the functional
to apes (Seppänen et al., 2007), including the breeding site choices of
and phylogenetic similarity of species involved. Both heterospecific
birds (Kivelä et al., 2014; Loukola, Seppänen, Krams, Torvinen, &
information use and the limiting similarity principle predict that species
€ nkko € nen, Helle, & Soppela, 1990; Seppänen & Forsman, 2013; Mo
associations are most intense, but in opposite directions, among func-
€ nkko €nen, 2003). Another Forsman, 2007; Thomson, Forsman, & Mo
tionally similar species (Figure 1). Heterospecific information use pre-
source producing positive species associations is the benefits of certain
dicts positive associations between species that use similar resources
keystone species, such as beavers and woodpeckers, which free or cre-
(i.e., those that are functionally similar), because the information value
ate new resources for other species and result in positive species spa-
decreases with increasing ecological distance (Seppänen et al., 2007),
tial associations (Belmaker et al., 2015; Heikkinen, Luoto, Virkkala,
while limiting similarity predicts mutual avoidance between similar spe-
€rber, 2007). Pearson, & Ko
cies because of the costs of competition (MacArthur & Levins, 1967).
Although biotic interactions, and particularly competition, have a
However, it is likely that both competition and heterospecific informa-
long history in the study of community patterns, their importance is
tion use are context dependent and that the net outcome of an associ-
often neglected at broad spatial scales, where speciation, extinction
ation between similar species depends on the costs of competition and
and geographical dispersal are expected to be the main driving proc-
€ nkko €nen et al., 1999; Seppänen et al., benefits of information use (Mo
"jo and Rozenfeld esses (Gaston & Blackburn, 2000). Recently, Arau
2007). Therefore, associations between functionally similar species
(2014) modelled spatial consequences of all types of species interac-
may be strong, either positive or negative, and the strength of the
954
|
€ € MONKK ONEN
ET AL.
association may decrease with decreased species similarity. The likeli-
multi-plot monitoring programme of the French avifauna. The pro-
hood of finding such a relationship depends, however, on the traits
gramme followed a standardized protocol from 2001 to 2012 (Jiguet,
considered (Trisos, Petchey, & Tobias, 2014). Using phylogenetic infor-
Devictor, Julliard, & Couvet, 2012) where 2 km 3 2 km sampling plots
mation can be a relevant approach to avoid a priori trait selection. Simi-
are randomly selected within 10-km radius areas, which ensures a rep-
lar patterns are indeed also likely for phylogenetic distance providing
resentative sampling of existing habitats. In each plot, 10 point count
that trait conservatism renders closely related species also ecologically
stations were evenly distributed. At each station, the observer
more similar (Mouquet et al., 2012). In this case, phylogenetically
recorded all birds heard and seen during 5 min, in two sessions during
closely related species should also show both the strongest positive
the breeding season. For each species in each point count station and
and negative associations, and phylogenetically distant species should
each year, the maximal number of individuals recorded during the two
show neutral associations. So far, however, empirical tests of this
sessions is retained as a proxy for the local abundance of that species
hypothesis have yielded mixed results (Godoy, Kraft, & Levine, 2014;
in that plot and year.
Violle, Nemergut, Pu, & Jiang, 2011).
In Finland, point counts have been conducted as part of the
Earlier research on general mechanisms of community assembly
national common bird monitoring scheme between 1984 and 2011
commonly separated the effects of competitive interactions from envi-
(Laaksonen & Lehikoinen, 2013). Each census route included 20 point
ronmental forcing, causing positive covariation among species abun-
count stations located in the habitat that is uniform within a 50-m
dances, and provided support for environmental variation rather than
radius of the station. The habitat of each point was classified into 17
competition driving the variation in species abundances (Houlahan
different habitat categories. Stations within a route were at least
et al., 2007; Mutshinda, O’Hara, & Woiwod, 2009; Ricklefs, 2012).
250 m apart in forested habitats and 350 m apart in open habitats to
However, some of the positive covariation may be attributable to posi-
avoid pseudoreplication. Owing to the very short and synchronized
tive interactions between species, over and above habitat filtering and
breeding season of boreal birds, a route was censused only once per
productivity. Moreover, even if environmental forcing generally prevails
season. At each station, an observer counted, for 5 min, all the
over competitive interactions, negative interactions may not be trivial.
observed land birds during late spring to early summer (between 20
Thus, both negative and positive interactions can leave a signature on
May and 20 June in south-central Finland; between 30 May and 30
community assembly that affects both historical and ecological distri-
June in northern Finland). In both monitoring projects, surveys were
bution patterns. In this article, we study (a) whether there are positive
conducted early in the morning (typically between sunrise and 10:00
or negative associations among bird species in local communities, inde-
hr), which is when the birds were most active, and only on days with
pendently of habitat characteristics, and consequent community pat-
good weather conditions (no rain or heavy wind).
terns at large geographical scales, and (b) whether these positive or negative associations are related to functional similarity and/or phylogenetic distance of the species. We predict that functional dissimilarity/phylogenetic distance and the strength of the associations form a ‘funnel plot’ where the strongest associations, either positive or negative, are between functionally similar or phylogenetically close species, whereas the associations grow weaker with increasing dissimilarity/distance (Figure 1). We use comprehensive bird census data from French forests and farmlands as well as Finnish forests and analyse the three datasets separately to test for the consistency of results across habitat types and geographical areas. Using these three datasets separately offers the possibility of testing the same question on the same group of species originating from different types of landscapes and climatic conditions. Consistent patterns would imply generality and call for further scrutiny of underlying mechanisms.
2 | MATERIAL AND METHODS 2.1 | Data
2.2 | Data handling Original French data included plots sampled for at least 2 years (i.e., 1,914 plots and 19,140 point count stations). Finnish data included 286 routes and 5,760 stations. From these data, we first created three independent datasets by selecting the point count stations situated in forests or in farmlands only (French farmlands, French forests and Finnish forests; Finnish farmland data were too few for the analysis). The aim was to remove most of the variation in bird abundances resulting from habitat structures (habitat filtering). Indeed, one of the obvious sources of species segregation or aggregation is the main habitat type in which a given plot is monitored. A negative association between farmland birds and forest birds would be interpreted as a signal of competition, although those two groups simply do not co-occur. The three datasets contain about 2,900–9,000 point count stations but a rather narrow range of habitats (Table 1). We analysed the three datasets separately to test for the consistency of results across habitat types and geographical areas. We excluded all waterbirds and birds of prey because point count
Data were extracted from two independent datasets from France and
census at the local point count station level provides reliable information
Finland. They are very well suited for the study because together they
on species abundances only for land birds with relatively small home
cover a wide extent (with a relatively small grain size) and consequently
ranges. We also excluded very rare species that were present in < 2%
also a wide spectrum of climatic and environmental conditions, and
of the point count stations. After filtering, our data included 43,000–
they are both of high quality, gathered over multiple years and multiple
206,000 observations for 76–83 species (Table 1), for which we calcu-
sites. The French Breeding Bird Survey is a large-scale, multi-year and
lated average abundance values across years per point count station.
€ € MONKK ONEN
|
ET AL.
955
Biogeographical zones, habitat types and numbers of point count stations, observations and species included in each of the three datasets (French farmlands, French forests and Finnish forests)
TA BL E 1
Dataset
Biogeographical zone
French farmlands
Alpine
Atlantic
Continental
Mediterranean
Number of point count stations
Number of observations
Number of species
Ploughed meadow Unploughed meadow Mixed farmland Open field Permanent crop Ploughed meadow Unploughed meadow Mixed farmland Open field Permanent crop Ploughed meadow Unploughed meadow Mixed farmland Open field Permanent crop Ploughed meadow Unploughed meadow Mixed farmland Open field Permanent crop
60 70 91 122 380 496 595 1,168 1,759 259 506 1,306 859 976 128 37 134 49 6 25 9,026
1,291 1,372 1,963 2,485 7,358 11,930 14,666 28,456 34,626 6,306 12,228 33,795 21,436 19,763 3,125 726 2,637 1,095 177 538 205,973
74 75 76 76 78 81 82 82 82 80 82 82 82 81 79 67 73 70 55 60 83
Deciduous woodland Coniferous woodland Mixed woodland Deciduous woodland Coniferous woodland Mixed woodland Deciduous woodland Coniferous woodland Mixed woodland Deciduous woodland Coniferous woodland Mixed woodland
94 113 161 1,136 216 274 1,519 311 440 181 138 187 4,770
1,543 1,806 2,952 25,670 4,537 6,715 33,458 6,144 9,227 3,054 2,490 3,282 100,878
69 61 67 77 74 74 75 76 75 71 71 75 77
Spruce forest Pine forest Deciduous forest Mixed forest Spruce forest Pine forest Deciduous forest Mixed forest Spruce forest Pine forest Deciduous forest Mixed forest
592 563 320 900 51 134 30 90 30 109 21 77 2,917
9,288 9,322 5,325 15,289 737 1,887 437 1,262 427 1,081 328 936 46,319
76 76 75 75 56 68 55 64 48 50 43 53 76
Habitat type
Total French forests
Alpine
Atlantic
Continental
Mediterranean
Total Finnish forests
Hemi- and south boreal
Mid-boreal
North boreal
Total Note. Total numbers for each dataset are in bold.
The same species occurred in multiple datasets; the number of shared
2006). From these traits, we calculated the Gower distance to repre-
species was 67 in French forest and farmland data, 51 in French and Fin-
sent pairwise trait distances estimated from the species trait matrix
nish forest data, and 45 in French farmland and Finnish forest data.
(Legendre & Legendre, 1998). As we aimed to remove habitat-induced
The matrix of pairwise functional distances was produced from 22
distances between species, we calculated a separate matrix for each
functional traits (Appendix S1 in Supporting Information) using meth-
set of species (French farmlands, French forests and Finnish forests).
ods described by Devictor et al. (2010). These traits encompassed life-
Gower distance accounts for both continuous and qualitative traits and
history traits and feeding habits (Petchey, Evans, Fishburn, & Gaston,
was measured with the function ‘daisy’ of the R package ‘cluster’
2007) and were identified as being important in determining the
(Maechler, Rousseeuw, Struyf, Hubert, & Hornik, 2016). All pairwise
response of bird species to environmental change and in determining
distances were standardized by dividing original distance values by the
the contribution of bird species to ecosystem functions (Sekercioglu,
maximal distance.
956
|
€ € MONKK ONEN
ET AL.
We extracted pairwise phylogenetic distances directly from a
that we analysed forest and farmland data separately (i.e., narrowed
dated molecular phylogenetic tree assembled by Thuiller et al. (2011)
down the variation in habitat structures before analysis) and entered
and then used ultrametric distances from this tree, representing rela-
farmland type (ploughed meadow, unploughed meadow, mixed farm-
tive phylogenetic distances among species, using the function ‘cl_u-
land, open field or permanent crop) or forest type (e.g., deciduous,
trametric’ in the R package ‘clue’ (Hornik, 2005). Phylogenetic
coniferous or mixed forests in France; spruce, pine, deciduous or mixed
information was not available for seven species in Finnish forest data-
forests in Finland) as a random factor, our analysis effectively controls
set, and thus we performed analyses of phylogenetic distance on 69
for fine-scale habitat filtering. Fifth, we entered community size
species.
(summed abundances of all species, excluding speciesi and speciesj) as a fixed effect controlling for the possibility that species abundances may
2.3 | Statistical analyses
covary with total community size because of independent responses to
We adopt the pairwise approach to analyse species’ effects on each other’s abundance (i.e., we consider a species pair as the fundamental unit in interactions). The procedure has advantages over the matrix method, where a target of interest is the whole community (i.e., presence–absence matrix), which has been used since the beginning of the studies of co-occurrence patterns (Connor, Collins, & Simberloff, 2013; Diamond, 1975). Most communities contain many potential species pairs, each of which may exhibit positive, negative or random associations. Therefore, single metrics that summarize an entire assemblage can mask the types and strengths of pairwise interactions, and it is therefore instructive to analyse individual pairs of species (Blois et al., 2014; Boulangeat, Gravel, & Thuiller, 2012; Veech, 2013). We relate species abundances against each other to reveal signals of positive, neutral and negative associations, after controlling for variation in bird
productivity. Finally, we added the abundance of speciesj as a fixed effect. We log-transformed [log(n 1 1)] all abundances before analyses and saved pairwise regression coefficients for further analyses. Estimating pairwise associations using a Poisson distribution yielded too many convergence problems and was not feasible because of a very high number of models. Note that using a Poisson distribution should not change the general conclusions derived from our framework based on log-transformed abundances (Ives, 2015). To infer whether the observed associations differ from what could be observed on the basis of randomly distributed individuals, we used a null model approach. First, we defined a regional species pool as all observed species and their abundances, separately for each habitat in each biogeographical zone. Second, we randomly sampled the observed number of individuals from the regional species pool while
abundances attributable to geographical distribution and corollary cli-
preserving the abundance of each species and observed total abun-
matic variation as well as finer habitat associations, beyond the main
dance in each point count station. In other words, we kept the size of
habitat type (i.e., forest and farmland). We fitted a linear mixed-effects
the regional habitat-specific species pools and the size of the local
model for each species pair to infer the association between speciesi
communities fixed, and within these constraints we randomized the
and speciesj while controlling for geographical distribution and finer
composition of local communities at point count stations (for a similar
habitat associations within the main habitat type (forest or farmland;
approach, see Crist, Veech, Gering, & Summerville, 2003). This type of
see Supporting Information Appendix S2 for a detailed flowchart of the
randomization makes a plausible assumption that species habitat asso-
analyses). First, the biogeographical zone was entered into the model
ciations, their relative abundances in regional pools and local commu-
as a random factor to control for large scale variation in climatic condi-
nity size are real ecological properties worth retaining while relaxing
tions. In each dataset, we assigned the point count stations to biogeo-
deterministic pairwise associations. Then we fitted the linear mixed-
graphical zones according to information from the European
effect model described above for the randomized datasets. We
Environmental Agency (2015) for France and Järvinen & Väisänen
repeated this procedure 1,000 times and calculated the standardized
(1980) for Finland (see Table 1 for zones). Second, we included site (a
effect size as the difference between the observed pairwise regression
dummy variable for a point count plot or route), nested within the bio-
coefficient and the mean expected coefficient, divided by the SD. Thus,
geographical zone, to control for small-scale variation in environmental
the effect size measures the direction (positive or negative) and
conditions among plots or routes. Third, we added a second-order
strength of species interactions in the datasets, independent from what
trend
surface
(X 1 Y 1 X2 1 Y2 1 XY,
where
X 5 longitude
and
one expects by chance and sampling artefacts.
Y 5 latitude) to control for the geographical (e.g., temperature and pre-
Given that the effect sizes within species pairs (effect of speciesi
cipitation related) variation in species abundances. Although a minimum
on speciesj, and vice versa) were strongly correlated (Spearman’s
of third-order trend surface is generally recommended (Legendre &
q 5 0.943, n 5 3,403, P < 0.001; Spearman’s q 5 0.956, n 5 2,926,
Legendre, 1998), random effects (zone and site) already control for spa-
P < 0.001; Spearman’s q 5 0.960, n 5 2,850, P < 0.001; in French farm-
tial autocorrelation at the scale of biogeographical zone and at very
lands, French forests and Finnish forests, respectively), we used the
small spatial scales. Moreover, the models with a third-order trend sur-
mean of the effect sizes of each species pair (ESmean) as an observation
face tended to be over-parameterized, leading to model convergence
unit in the analysis. Given that these observations are not independent
problems. Including second-order trend surfaces ensures that our pair-
(every species is represented in multiple species pairs), we used a boot-
wise regression analysis operates on local-scale variation in species
strap method to calculate SEs (Koenker, 2013).
abundances. Fourth, we entered the habitat type of the point count sta-
We considered pairwise species effect sizes ‘strong’ when absolute
tion (Table 1) to control further for species habitat preferences. Given
effect size values were greater than two (i.e., observed association
€ € MONKK ONEN
|
ET AL.
957
deviated more than two SDs from the expected) and ‘weak’ when
the three datasets. In French farmlands there was no relationship in
effect size values were less than two. We predicted a relationship
the lower quantiles but a negative relationship in the upper quantiles,
between the magnitude of the effect sizes, both positive and negative,
whereas in French forests there was only a trend in both lower and
and functional dissimilarity/phylogenetic distance. The magnitude
upper quantiles (Figure 2a,b). In contrast, in Finnish forests there was a
should decrease with increasing dissimilarity/distance. To test this, we
positive relationship in the lower quantiles and a negative relationship
used quantile regression (Cade & Noon, 2003) for all quantiles from 1
in the upper quantiles (Figure 2c). Thus, in French farmlands the level
to 99% quantiles (s ranges from 0.01 to 0.99) at intervals of 1%. We
of aggregation increased with decreasing functional similarity, in Fin-
predicted that the regression between effect size and dissimilarity/dis-
nish forests both the level of segregation and aggregations increased
tance would result in a negative coefficient in the upper (> 50%) quan-
with decreasing functional similarity, whereas in French forests there
tiles
were no relationships.
(reflecting
smaller
positive
effect
sizes
with
increasing
dissimilarity/distance) and a positive coefficient in the lower (< 50%)
Phylogenetic distance and ESmean showed significant relationships
quantiles (reflecting smaller negative effect sizes with increasing dissim-
in both ends of the quantile spectrum in French farmlands and forests
ilarity/distance). If these predictions are verified, the shape of the rela-
(Figure 3a,b). Thus, both aggregated and segregated distributions of
tionship should be a ‘funnel plot’, with higher numbers of positive and
species abundances increased as a function of decreasing phylogenetic
negative associations for lower values of functional dissimilarity or phy-
distance. In Finnish forests, there was a positive relationship between
logenetic distance (Figure 1). To infer whether this is truly the case, we
phylogenetic distance and ESmean in the lower quantiles, but no rela-
plotted the coefficient of each of the quantile regressions as a function
tionship in the upper quantiles (Figure 3c).
of the quantile (s) in question and expected to see a negative relationship. As we did not have a specific hypothesis about the overall rela-
4 | DISCUSSION
tionship in the data, we did not concentrate on the general tendency (i.e., 50% quantile), but on upper versus lower quantiles. We performed
Bird species showed both spatial aggregation and segregation in their
linear mixed-effect models with the package ‘lme4’ (Bates, Maechler,
abundances, independently of habitat structures. Even though overall
Bolker, & Walker, 2014) and quantile regression with ‘quantreg’
mean effect sizes of species associations in the local communities were
(Koenker, 2013) in R Version 3.0.3 (R Development Core Team, 2014),
centred on zero, we found that the majority of species were relatively
and the iterations for the null model approach with Taito supercluster
strongly associated with each other. This highlights the importance of
provided by CSC IT Center for Science Ltd (https://research.csc.fi/
both negative and positive biotic interactions in affecting community
research-home). The R script for calculating the effect sizes is provided
assembly. Our study, using natural communities across a wide geo-
as Supporting Information Appendix S3. Other analyses were per-
graphical area, yielded results that are consistent with the results of
formed with IBM SPSS Statistics 22.0 (IBM, Armonk, NY).
the local manipulative experiments; negative associations are common, but also positive associations occur frequently (Bertness & Callaway,
3 | RESULTS
€nkko € nen, 1994; Gurevitch et al., 2000; Forsman, Seppänen, & Mo 2002; Martorell & Freckleton, 2014; but see Gotelli & Ulrich, 2010).
We observed both positive and negative regression coefficients among
Moreover, our results suggest that high functional and phylogenetic
species pairs tested in all three datasets. The means of the observed
similarities can be important determinants that increase the probability
coefficients were positive (0.02) and the distributions of coefficients
of both negative and positive associations. Also, we found that in bird
were highly similar in the three datasets (see Supporting Information
communities strong negative associations were twice as common as
Appendix S4). Likewise, the effect sizes (ESmean) were centred close to
strong positive associations. This is in line with empirical results of
zero. A large proportion of pairwise species effect sizes can be consid-
Gotelli et al. (2010), who showed strong predominance of spatially seg-
ered strong, because absolute effect size values larger than two (i.e.,
regated over aggregated distributions within foraging and congeneric
observed association deviated more than two SDs from the expected)
guilds in Danish avifauna.
constituted 83, 63 and 41% of all pairwise effect sizes in the French
Species abundance is affected by a multitude of factors, such as
farmland (n 5 6,806), French forest (n 5 5,852) and Finnish forest data-
habitat, productivity and geographical position. Our pairwise approach
sets (n 5 5,700), respectively. In all three datasets, strong negative
controlled for the effects of habitat, first, by restricting the analysis to
associations were approximately twice as common as strong positive
only a limited set of habitat classes (Table 1) and, second, by entering
associations: 57 vs. 26% in French farmland, 39 vs. 23% in French for-
habitat class as a factor in the model. Moreover, we entered biogeo-
est and 27 vs. 15% in Finnish forest data. Thus, birds are not distrib-
graphical zone, sampling site and second-order trend surface in abun-
uted randomly with respect to each other in local communities, when
dances to control for the effects of geographical factors such as
controlling for habitat filtering and productivity, and we found asym-
species geographical distributions and climate related variation. Finally,
metry between positive and negative associations, with prevalence of
we controlled for the community size (i.e., species-independent
the latter.
responses to productivity). Yet our analysis revealed strong signals of
The relationship between functional dissimilarity and the mean of
positive abundance associations. Thus, species abundances were prob-
the effect sizes for species pairs (ESmean) showed differences among
ably also affected by positive biotic interactions between species, and
958
|
€ € MONKK ONEN
ET AL.
F I G U R E 2 Left-hand panels show the relationship between increasing functional dissimilarity and the strength and sign of an association of the abundances of each species pair in (a) French farmlands, (c) French forests and (e) Finnish forests. The strength and sign of an association of the abundances is measured as the mean of the standardized effect sizes (the difference between observed pairwise regression coefficient and mean expected coefficient, divided by the SD) between a species pair (see text for further information). In the event of a statistically significant relationship, the regression line for the quantile regressions in lower (s 5 0.05) and/or upper (s 5 0.95) quantiles are shown. Right-hand panels show the relationship between all quantiles (s) at intervals of 0.01 and the coefficient from the quantile regressions (SEs are shown in grey) in (b) French farmlands, (d) French forests and (f) Finnish forests. Positive coefficients denote a positive relationship in a given quantile between increasing functional dissimilarity and the strength of an association of the abundances, whereas negative coefficients denote a negative relationship
these interactions were strong enough to show up as aggregated distri-
a proportion of this positive covariation is probably attributable to posi-
butions. The recurrent finding that species abundances in natural com-
tive interactions. Therefore, dominance of positive covariation over
munities tend to covary positively rather than negatively has
compensatory dynamics in community dynamics does not necessarily
commonly been attributed to environmental forcing (Houlahan et al.,
imply a low frequency of species interactions. Future studies testing
2007; Mutshinda et al., 2009; Ricklefs, 2012). Our results suggest that
assembly theories should therefore refrain from using strictly
€ € MONKK ONEN
ET AL.
|
959
F I G U R E 3 Left-hand panels show the relationship between increasing relative phylogenetic distance and the strength and sign of an association of the abundances of a species pair in (a) French farmlands, (c) French forests and (e) Finnish forests. The strength and sign of an association of the abundances is measured as the mean of the standardized effect sizes (the difference between observed pairwise regression coefficient and mean expected coefficient, divided by the SD) between a species pair (see text for further information). In the event of statistically significant relationship, the regression line for the quantile regressions in lower (s 5 0.05) and/or upper (s 5 0.95) quantiles are shown. Right-hand panels show the relationship between all quantiles (s) at intervals of 0.01 and the coefficient from the quantile regressions (SEs are shown in grey) in (b) French farmlands, (d) French forests and (f) Finnish forests. Positive coefficients denote a positive relationship in a given quantile between increasing relative phylogenetic distance and the strength of an association of the abundances, whereas negative coefficients denote a negative relationship
dichotomist approaches, such as compensatory dynamics versus envi-
In all three datasets, closer species pairs with respect to phyloge-
ronmental forcing, but should also consider the possibility of positive
netic distances are those showing stronger segregation in their abun-
interactions and underlying processes such as conspecific or heterospe-
dances. Neutral associations, however, were found throughout the
cific information use.
phylogenetic distance spectrum. Thus, a part of phylogenetically close
960
|
€ € MONKK ONEN
ET AL.
bird species indeed seem to compete more strongly than distantly
hue, 2002). But more interestingly, our results suggest that close rela-
related species, which plausibly leads to avoidance of such species in
tives and functionally similar species may also aggregate, which is
habitat selection. In Finnish forest bird assemblages, but not in the two
reflected as positive associations among species’ abundances. The
French datasets, the signal of spatial segregation also increased with
likely mechanism is the use of social information (Seppänen et al.,
increasing functional similarity. In concordance with the prediction of
2007, Sebastian-Gonzalez et al., 2010) and/or facilitative interactions
heterospecific information use (Seppänen et al., 2007), the signal of
among species (Bruno et al., 2003). It is evident that both clustering
spatial aggregations was stronger for pairs consisting of similar species
and overdispersion of co-occurring species within communities may
and grew weaker with increasing dissimilarity. However, whether the
occur simultaneously. Our results challenge the implicit assumption
pattern was detected for phylogenetic distance, functional dissimilarity
pertinent to community phylogenetics that assembly through positive
or both showed variation among datasets; in French forests, aggrega-
associations decreases with increasing assembly by competition (see
tion was related to phylogenetic distance, in Finnish forests to func-
Gerhold, Cahill, Winter, Bartish, & Prinzing, 2015). Thus, the relative
tional dissimilarity, and in French farmlands to both.
balance of positive and negative interactions in community assembly
In summary, segregated abundances were related to phylogenetic distance in all of the three cases and to functional dissimilarity in only
cannot be quantified by a single parameter of phylogenetic (or functional) dispersion.
one, whereas aggregated abundances were related to both phyloge-
We acknowledge the fact that only experimental set-ups can truly
netic distance and functional dissimilarity in two cases. Thus, the phylo-
prove the strength and the sign of the interaction between a species
genetic signal in segregated abundances was stronger than that of
pair. It is clear that the data used here do not capture all the small-scale
functional (ecological) similarity. This suggests that the traits we used
habitat characteristics that may affect species aggregation and segrega-
for functional dissimilarity might not be those that affect species’ com-
tion patterns. Indeed, detailed data of environmental conditions might
petitive environment but are more relevant in terms of heterospecific
attenuate the coefficients we found. In contrast, it has been shown
information use, whereas the phylogenetic signal encompasses traits of
that even when modelling forest bird species distributions with very
direct relevance in both respects. This result also raises the issue of
detailed forest structure data, the density of a bird species remains a
trait selection in trait-based analysis. It is likely that a part of the results
significant predictor of the density of a close relative (Kosicki, Stachura,
is dependent on the particular combination of traits used. A finer exam-
Ostrowska, & Rybska, 2015).
ination of pairwise associations (i.e., whether segregated versus aggre-
Social information use and subsequent aggregated distribution in
gated associations are influenced by a specific combination of traits,
local communities result in variation in local species diversity at a given
whether there is any specific trait enhancing coexistence or whether
site that deviates from the diversity predicted by environmental factors
aggregations are phylogenetically clustered) would be an interesting
only, creating both hot- and coldspots of species diversity in the land-
extension of our approach.
scape (Seppänen et al., 2007). Interspecific competition resulting in seg-
Another reason why species pairs consisting of phylogenetically
regated distribution may also create similar deviations from predictions.
closely related species showed strong segregation in their abundances
Our results suggest that such diversity anomalies should carry phyloge-
could result from allopatric speciation. Phylogenetic overdispersion,
netic/functional signal. Species’ interactions may render a proportion
which is often attributed to negative biotic interactions, may instead be
of suitable habitat patches unoccupied by the species also because dis-
consistent with a neutral model of allopatric speciation (Pigot & Eti-
persal among patches in the landscape may be affected by the pres-
enne, 2015). If allospecies rarely co-occurred in the bird assemblages
ence
we studied, and differ somewhat in their functional traits, one could
Consequently, colonization and extinction in fragmented landscapes is
see a pattern where phylogenetic distance is more directly driving neg-
no longer a sole function of landscape patterns and species dispersal
ative abundance associations than functional similarity. Allopatric speci-
abilities but hinges also on other species’ ability to persist in frag-
ation does not, however, provide an explanation for phylogenetic and
mented landscapes. Given that interspecific social information use is
functional patterning of positive abundance associations. It is possible,
widespread (Seppänen et al., 2007), from the point of view of species’
and even probable, that for some species pairs the costs of competition
conservation, it is important to keep in mind that the effect of close rel-
and benefits of information use add up to show as a neutral associa-
atives and ecologically similar species may also be positive.
tion. Hence, it is important to bear in mind that small effect sizes in our study do not necessarily indicate no interaction or weak interactions.
of
close
relatives
and/or
functionally
similar
species.
Predicting species’ responses to various global changes has become crucial because of the ongoing biodiversity crisis. Our results
Our approach, where we simultaneously addressed both negative
"jo & Rozenfeld, 2014) suggesting accompany earlier literature (e.g., Arau
and positive interactions and provided support for both, may help in
that a failure to incorporate species interactions may account for the
understanding why earlier work has found mixed results concerning the role of phylogenetic and ecological distance in species interactions
mixed results of earlier species distribution modelling efforts that ignore interactions.
(Godoy et al., 2014; Violle et al., 2011). Our results support a common expectation in the community assembly literature that owing to com-
ACKNOWLEDG MENTS
petition, close relatives and functionally similar species should show
We are grateful to the hundreds of volunteers involved in the French
segregated patterns in abundance (Webb, Ackerly, McPeek, & Donog-
and Finnish bird censuses. We are also grateful to the Academy of
€ € MONKK ONEN
ET AL.
Finland (project no. 275329 to M.M. and project no. 275606 to A.L.) and Kone Foundation (to M.E. and J.T.F.) for funding. We thank M. Puurtinen and A. Kahilainen for a useful piece of statistical advice as well as the Editor of Global Ecology and Biogeography and two anonymous referees for their comments, which significantly helped to improve the paper. K. Eyvindson kindly edited the text. This paper was initially submitted, reviewed and revised in Peerage of Science (http://www.peerageofscience.org/), and we are grateful to peers #1093, #1096, #1101 and #1111 for very constructive comments. M.M. is grateful to CEFE/CNRS, and particularly to prof. D. Grémillet, for hospitality. RE FE RE NCE S "jo, M. B., & Rozenfeld, A. (2014). The geographic scaling of biotic Arau interactions. Ecography, 37, 406–415. Bates, D., Maechler, M., Bolker, B., & Walker, S. (2014). lme4: Linear mixed-effects models using Eigen and S4. R package version 1.1-7. Retrieved from http://CRAN.R-project.org/package5lme4 Belmaker, J., Zarnetske, P., Tuanmu, M.-N., Zonneveld, S., Record, S., Strecker, A., & Beaudrot, L. (2015). Empirical evidence for the scale dependence of biotic interactions. Global Ecology and Biogeography, 24, 750–761. Bertness, M. D., & Callaway, R. M. (1994). Positive interactions in communities. Trends in Ecology and Evolution, 9, 191–193. Blois, J. L., Gotelli, N. J., Behrensmeyer, A. K., Faith, J. T., Lyons, S. K., Williams, J. W., . . . Wing, S. (2014). A framework for evaluating the influence of climate, dispersal limitation, and biotic interactions using fossil pollen associations across the late Quaternary. Ecography, 37, 1095–1108. Boulangeat, I., Gravel, D., & Thuiller, W. (2012). Accounting for dispersal and biotic interactions to disentangle the drivers of species distributions and their abundances. Ecology Letters, 15, 584–593. Bruno, J. F., Stachowicz, J. J., & Bertness, M. D. (2003). Inclusion of facilitation into ecological theory. Trends in Ecology and Evolution, 18, 119–125. Cade, B. S., & Noon, B. R. (2003). A gentle introduction to quantile regression for ecologists. Frontiers in Ecology and the Environment, 1, 412–420. Cardinale, B. J., Palmer, M. A., & Collins, S. L. (2002). Species diversity enhances ecosystem functioning through interspecific facilitation. Nature, 415, 426–429. Connell, J. H. (1983). On the prevalence and relative importance of interspecific competition: Evidence from field experiments. The American Naturalist, 122, 661–696. Connor, E. F., Collins, M. D., & Simberloff, D. (2013). The checkered history of checkerboard distributions: Reply. Ecology, 94, 2403–2414. Connor, E. F., & Simberloff, D. (1983). Interspecific competition and species co-occurrence patterns on islands: Null models and the evaluation of evidence. Oikos, 41, 455–465.
|
961
Danchin, E., Giraldeau, L.-A., Valone, T. J., & Wagner, R. H. (2004). Public information: From nosy neighbors to cultural evolution. Science, 305, 487–491. 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 conservation strategies in a changing world. Ecology Letters, 13, 1030–1040. Diamond, J. M. (1975). Assembly of species communities. In M. L. Cody & J. M. Diamond (Eds.), Ecology and evolution of communities (pp. 342–444). Cambridge, MA: Belknapp Press. European Environmental Agency. (2015). The European environment: State and outlook 2015. Data and Maps. Retrieved from https:// www.eea.europa.eu/data-and-maps € nkko € nen, M. (2002). Positive fitness Forsman, J. T., Seppänen, J.-T., & Mo consequences of interspecific interaction with a potential competitor. Proceedings of the Royal Society B: Biological Sciences, 269, 1619– 1623. Gaston, K. J., & Blackburn, T. M. (2000). Pattern and process in macroecology. Oxford, UK: Blackwell. Gerhold, P., Cahill, J. F., Winter, M., Bartish, I. V., & Prinzing, A. (2015). Phylogenetic patterns are not proxies of community assembly mechanisms (they are far better). Functional Ecology, 29, 600–614. Godoy, O., Kraft, N. J. B., & Levine, J. M. (2014). Phylogenetic relatedness and the determinants of competitive outcomes. Ecology Letters, 17, 836–844. Goldberg, D. E., & Barton, A. M. (1992). Patterns and consequences of interspecific competition in natural communities: A review of field experiments with plants. The American Naturalist, 139, 771–801. Gotelli, N. J., Graves, G. R., & Rahbek, C. (2010). Macroecological signals of species interactions in the Danish avifauna. Proceedings of the National Academy of Sciences USA, 107, 5030–5035. Gotelli, N. J., & Ulrich, W. (2010). The empirical Bayes approach as a tool to identify non-random species associations. Oecologia, 162, 463–477. Gross, K. (2008). Positive interactions among competitors can produce species-rich communities. Ecology Letters, 11, 929–936. Gurevitch, J., Morrison, J. A., & Hedges, L. V. (2000). The interaction between competition and predation: A meta-analysis of field experiments. The American Naturalist, 155, 435–453. €rber, J. H. Heikkinen, R. K., Luoto, M., Virkkala, R., Pearson, R. G., & Ko (2007). Biotic interactions improve prediction of boreal bird distributions at macro-scales. Global Ecology and Biogeography, 16, 754–763. Hornik, K. (2005). A CLUE for CLUster ensembles. Journal of Statistical Software, 14, 1–25. Houlahan, J. E., Currie, D. J., Cottenie, K., Cumming, G. S., Ernest, S. K. M., Findlay, C. S., . . . Wondzell, S. M. (2007). Compensatory dynamics are rare in natural ecological communities. Proceedings of the National Academy of Sciences USA, 104, 3273–3277. Hutchinson, G. E. (1959). Homage to Santa-Rosalia or why are there so many kinds of animals. The American Naturalist, 93, 145–159.
Cornell, H. V. (1985). Local and regional richness of cynipine gall wasps on California oaks. Ecology, 66, 1247–1260.
Ives, A. R. (2015). For testing the significance of regression coefficients, go ahead and log-transform count data. Methods in Ecology and Evolution, 6, 828–835.
Cornell, H. V., & Lawton, J. H. (1992). Species interactions, local and regional processes, and limits to the richness of ecological communities: A theoretical perspective. Journal of Animal Ecology, 61, 1–12.
Järvinen, O., & Väisänen, R. A. (1980). Quantitative biogeography of Finnish land birds as compared with regionality in other taxa. Annales Zoologici Fennici, 17, 67–85.
Crist, T. O., Veech, J. A., Gering, J. C., & Summerville, K. S. (2003). Partitioning species diversity across landscapes and regions: A hierarchical analysis of a, b, and g diversity. The American Naturalist, 162, 734–743.
Jiguet, F., Devictor, V., Julliard, R., & Couvet, D. (2012). French citizens monitoring ordinary birds provide tools for conservation and ecological sciences. Acta Oecologica, 44, 58–66.
962
|
Kivelä, S. M., Seppänen, J.-T., Ovaskainen, O., Doligez, B., Gustafsson, L., € nkko € nen, M., & Forsman, J. T. (2014). The past and the present in Mo decision-making: The use of con- and heterospecific cues in nest-site selection. Ecology, 95, 3428–3439. Koenker, R. (2013). quantreg: Quantile Regression. R package version 5.05. Retrieved from http://CRAN.R-project.org/package5quantreg Kosicki, J. Z., Stachura, K., Ostrowska, M., & Rybska, E. (2015). Complex species distribution models of Goldcrests and Firecrests densities in Poland: Are remote sensing-based predictors sufficient? Ecological Research, 30, 625–638. Laaksonen, T., & Lehikoinen, A. (2013). Population trends in boreal birds: Continuing declines in agricultural, northern, and long-distance migrant species. Biological Conservation, 168, 99–107. Legendre, P., & Legendre, L. (1998). Numerical ecology. Amsterdam, The Netherlands: Elsevier Science. Loukola, O. J., Seppänen, J.-T., Krams, I., Torvinen, S. S., & Forsman, J. T. (2013). Observed fitness may affect niche overlap in competing species via selective social information use. The American Naturalist, 182, 474–483. MacArthur, R. H. (1958). Population ecology of some warblers of northeastern coniferous forests. Ecology. 39, 599–619. MacArthur, R., & Levins, R. (1967). The limiting similarity, convergence, and divergence of coexisting species. The American Naturalist, 101, 377–385. Maechler, M., Rousseeuw, P., Struyf, A., Hubert, M., & Hornik, K. (2016). cluster: Cluster analysis basics and extensions. R package version 2.0.5. Retrieved from https://cran.r-project.org/web/packages/cluster/index. html Martorell, C., & Freckleton, R. P. (2014). Testing the roles of competition, facilitation and stochasticity on community structure in a species-rich assemblage. Journal of Ecology, 102, 74–85. € nkko € nen, M., Härdling, R., Forsman, J. T., Tuomi, J., Hardling, R., ForsMo man, J. T., & Tuomi, J. (1999). Evolution of heterospecific attraction: Using other species as cues in habitat selection. Evolutionary Ecology, 13, 91–104. € nkko € nen, M., Helle, P., & Soppela, K. (1990). Numerical and behaviouMo ral responses of migrant passerines to experimental manipulation of resident tits (Parus spp.): Heterospecific attraction in northern breeding bird communites? Oecologia, 85, 218–225. Mouquet, N., Devictor, V., Meynard, C. N., Munoz, F., Bersier, L. F., Chave, J., . . . Thuiller, W. (2012). Ecophylogenetics: Advances and perspectives. Biological Reviews, 87, 769–785. Mutshinda, C. M., O’Hara, R. B., & Woiwod, I. P. (2009). What drives community dynamics? Proceedings of the Royal Society B: Biological Sciences, 276, 2923–2929.
€ € MONKK ONEN
ET AL.
Ricklefs, R. E. (2012). Habitat-independent spatial structure in populations of some forest birds in eastern North America. Journal of Animal Ecology, 82, 145–154. Schoener, T. W. (1983). Field experiments on interspecific competition. The American Naturalist, 122, 240–285. Sebastia"n-Gonz"alez, E., S"anchez-Zapata, J. A., Botella, F., & Ovaskainen, O. (2010). Testing the heterospecific attraction hypothesis with timeseries data on species co-occurrence. Proceedings of the Royal Society B: Biological Sciences, 277, 2983–2990. Sekercioglu, C. H. (2006). Increasing awareness of avian ecological function. Trends in Ecology and Evolution, 21, 464–471. Seppänen, J.-T., & Forsman, J. T. (2007). Interspecific social learning: Novel preference can be acquired from a competing species. Current Biology, 17, 1248–1252. € nkko € nen, M., & Thomson, R. L. (2007). Seppänen, J.-T., Forsman, J. T., Mo Social information use is a process across time, space, and ecology, reaching heterospecifics. Ecology, 88, 1622–1633. € nkko € nen, M. (2003). Positive interThomson, R. L., Forsman, J. T., & Mo actions between migrant and resident birds: Testing the heterospecific attraction hypothesis. Oecologia, 134, 431–438. Thuiller, W., Lavergne, S., Roquet, C., Boulangeat, I., Lafourcade, B., & Araujo, M. B. (2011). Consequences of climate change on the tree of life in Europe. Nature, 470, 531–534. Trisos, C. H., Petchey, O. L., & Tobias, J. A. (2014). Unraveling the interplay of community assembly processes acting on multiple niche axes across spatial scales. The American Naturalist, 184, 593–608. Veech, J. A. (2013). A probabilistic model for analysing species co-occurrence. Global Ecology and Biogeography, 22, 252–260. Violle, C., Nemergut, D. R., Pu, Z., & Jiang, L. (2011). Phylogenetic limiting similarity and competitive exclusion. Ecology Letters, 14, 782–787. Webb, C. O., Ackerly, D. D., McPeek, M. A., & Donoghue, M. J. (2002). Phylogenies and community ecology. Annual Review of Ecology and Systematics, 33, 475–505.
BIOSK ET CH € NEN is a professor in applied ecology at the University of € NKKo MIKKO Mo Jyväskylä, Finland. His main research interests are community and landscape ecology, conservation biology and environmental economics. SUPPORTING INFORMATION Additional Supporting Information may be found online in the supporting information tab for this article.
Petchey, O. L., Evans, K. L., Fishburn, I. S., & Gaston, K. J. (2007). Low functional diversity and no redundancy in British avian assemblages. The Journal of Animal Ecology, 76, 977–985.
€ nkko €nen M, Devictor V, Forsman How to cite this article: Mo
Pigot, A. L., & Etienne, R. S. (2015). A new dynamic null model for phylogenetic community structure. Ecology Letters, 18, 153–163.
genetic and functional distance in European bird assemblages at
R Development Core Team. (2014). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/
JT, Lehikoinen A, Elo M. Linking species interactions with phylobroad spatial scales. Global Ecol Biogeogr. 2017;26:952–962. https://doi.org/10.1111/geb.12605
